Financial Accounting and Equity Markets : Selected Essays of Philip Brown 9781135077587, 9780415814614

Citation preview

Financial Accounting and Equity Markets Philip Brown is one of the most admired and respected accounting academics alive today. He was a pioneer in capital markets research in accounting, and his 1968 article, coauthored with Ray Ball, “An Empirical Evaluation of Accounting Income Numbers,” arguably had a greater impact on the course of accounting research, directly and indirectly, than any other article written during the second half of the twentieth century. Since that time, his innovative research has focused on issues that bridge accounting and finance, including the relationships between net profit reports and the stock market, the long-run performance of acquiring firms, statutory sanctions and voluntary corporate disclosure, and the politics and future of national accounting standards, to name a few. This volume brings together the greatest hits of Brown’s career, including several articles published in out-of-the-way places, for easier use by students and researchers in the field. With a foreword written by Stephen A. Zeff, and an introduction that discusses the evolution of Brown’s research interests and explains the context for each of the essays included in the volume, this book offers the reader a unique look inside this remarkable 50-year career. Philip Brown is Professor Emeritus of Accounting and Finance at the University of Western Australia and he holds a joint professorial appointment at the University of New South Wales, Australia. In 1991 Philip Brown was the American Accounting Association’s

2

Distinguished International Visiting Lecturer, and in 1991/1992 he was the inaugural Coopers and Lybrand-Accounting Association of Australia and New Zealand Visiting Research Professor in Australasia. In 1986 Philip Brown and his University of Chicago colleague, Ray Ball, received the inaugural American Accounting Association’s Seminal Contribution to Accounting Literature Award for the paper they published in 1968.

3

Routledge Historical Perspectives in Accounting Edited by Stephen A. Zeff, Rice University, USA 1 Profitability, Accounting Theory and Methodology The Selected Essays of Geoffrey Whittington Geoffrey Whittington 2 Financial Reporting in the UK A History of the Accounting Standards Committee, 1969–1990 Brian A. Rutherford 3 Insights from Accounting History Selected Writings of Stephen Zeff Stephen A. Zeff 4 Financial Accounting and Equity Markets The Selected Essays of Philip Brown Philip Brown

4

Financial Accounting and Equity Markets The Selected Essays of Philip Brown Philip Brown

5

First published 2013 by Routledge 711 Third Avenue, New York, NY 10017 Simultaneously published in the UK by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN Routledge is an imprint of the Taylor & Francis Group, an informa business © 2013 Taylor & Francis The right of Philip Brown to be identified as author of this work has been asserted by him in accordance with sections 77 and 78 of the Copyright, Designs and Patents Act 1988. All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. 6

Library of Congress Cataloging-in-Publication Data Brown, Philip (Philip Ronald), 1940– Financial accounting and equity markets : the selected essays of Philip Brown / by Philip Brown. p. cm. — (Routledge historical perspectives in accounting ; 4) Includes bibliographical references and index. 1. Finance. 2. Accounting. Ronald), 1940– I. Title. HG173.B766

3. Brown, Philip (Philip

2012

657—dc23 2012033045 ISBN: 978-0-415-81461-4 (hbk) ISBN: 978-0-203-06702-4 (ebk)

7

Contents P. Brown: Publications Acknowledgments Foreword Introduction PART I Returns and Earnings 1 (with R. Ball) “An empirical evaluation of accounting income numbers”, Journal of Accounting Research, 6(2), 1968, 159–178. 2 “Invited Remarks: Ball and Brown [1968]”, Journal of Accounting Research, 27 (Supplement), 1989, 202–217. 3 “The impact of the annual net profit report on the stock market”, The Australian Accountant, 40(July), 1970, 277–283. 4 (with J. W. Kennelly) “The informational content of quarterly earnings: an extension and some further evidence”, The Journal of Business, 45(3), 1972, 403–415. 5 (with B. Howitt and M. Wee) “Order flow and price effects surrounding an ASX announcement”, 2005 AFAANZ Conference Proceedings, Accounting and Finance Association of Australia and New Zealand (July 2005), CD-ROM. PART II 8

Miscellaneous Issues 6 (with R. Ball) “Identifying some issues in C.C.A.”, The Australian Accountant, 47(August), 1977, 413–414 and 419–421. 7 (with T. S. Walter) “Sharemarket efficiency and the experts: some Australian findings”, Australian Journal of Management, 7(1), 1982, 19–31. 8 (with G. Gallery and O. Goei) “Does market misvaluation help explain share market long-run underperformance following a seasoned equity issue?” Accounting and Finance, 46(2), 2006, 191–219. 9 (with I. Dunlop) “A case of reporting form over substance”, Australian Accounting Review, 1(2), 1991, 40–47. 10 (with H. Y. Izan and A. L. Loh) “Fixed asset revaluations and managerial incentives”, Abacus, 28(1), 1992, 36–57. PART III Standard-setting and Regulation 11 (with S. L. Taylor and T. S. Walter) “The impact of statutory sanctions on the level and information content of voluntary corporate disclosure”, Abacus, 35(2), 1999, 138–162. 12 (with W. Beekes) “Do better-governed Australian firms make more informative disclosures?” Journal of Business Finance & Accounting, 33(3–4), 2006, 422–450.

9

13 (with B. Howieson) “Capital markets research and accounting standard setting”, Accounting and Finance, 38(1), 1998, 5–28. 14 (with G. Clinch) “Global harmonisation of accounting standards: what research into capital markets tells us”, Australian Accounting Review, 8(1), 1998, 21–29. 15 “International Financial Reporting Standards: what are the benefits?” Accounting and Business Research, 41(3), 2011, 269–285. 16 (with A. Tarca) “Politics, processes and the future of Australian accounting standards”, Abacus, 37(3), 2001, 267–296. 17 (with A. Tarca) “Achieving high quality, comparable financial reporting: a review of independent enforcement bodies in Australia and the United Kingdom”, Abacus, 43(4), 2007, 438–473. PART IV In Theory 18 “A note on the inverse (reverse) sum-of-the-years’-digits method and other ways to amortise goodwill”, Australian Accounting Review, 5(1), 1995, 17–21. 19 (with A. Szimayer) “Valuing executive stock options: performance hurdles, early exercise and stochastic volatility”, Accounting and Finance, 48(3), 2008, 363–389. Index 10

P. Brown: Publications (to October 2011) Author’s note: papers included in the anthology are marked by *[N], where N is the sequence number of the paper in the anthology (N = 1, …, 19). I Books and Monographs (with G. Foster and E. Noreen) Security analyst multi-year earnings forecasts and the capital market, Studies in Accounting Research #21, American Accounting Association, Sarasota, 1985. (with R. Ball, F. Finn, and R. Officer) (eds) Sharemarkets and Portfolio Theory, University of Queensland Press, 1980 (2nd edn. 1989). (with R. Mathews, chairman, and M. Jackson) Report of the Committee of Review of the Accounting Discipline in Higher Education (3 Volumes), AGPS, Canberra, September 1990. (with G. Clinch and G. Foster) Market microstructure and capital market information content research, Studies in Accounting Research #32, American Accounting Association, Sarasota, 1992. “Capital Markets-Based Research in Accounting: an Introduction”, Coopers & Ly-brand Accounting Research Methodology Monograph #1, Coopers & Lybrand and Accounting Association of Australia and New Zealand, Melbourne Australia, 1994, vii + 182 (translated into Chinese, Japanese, and Korean and published in separate editions). II Pamphlets and Chapters in Books 11

“Those half-yearly reports”, Australian Society of Accountants Research Bulletin Number 13, June 1972. (with P. Hancock) “Profit reports and the sharemarket”, in I. Tilley and P. Jubb (eds), Capital, Income and Decision Making (Sydney: Holt, Rinehart and Winston), 1977. (with F. J. Finn) “Asset revaluations and stock prices: alternative interpretations of a study by Sharpe and Walker”, in R. Ball, P. Brown, F. J. Finn, and R. R. Officer (eds), Share Markets and Portfolio Theory, University of Queensland Press, 1988. III Papers and Notes in Refereed Academic Journals (with R. Ball) “Some preliminary findings on the association between the earnings of a firm, its industry, and the economy”, Journal of Accounting Research, 5 (Supplement), 1967, 55–77. *[1] (with R. Ball) “An empirical evaluation of accounting income numbers”, Journal of Accounting Research, 6(2), 1968, 159–178. (with V. Niederhoffer) “The predictive content of quarterly earnings”, The Journal of Business, 41(4), 1968, 488–497. (with R. Ball) “Portfolio theory and accounting”, Journal of Accounting Research, 7(2), 1969, 300–323. *[4] (with J. W. Kennelly) “The informational content of quarterly earnings: an extension and some further evidence”, The Journal of Business, 45(3), 1972, 403–415.

12

(with R. Ball and R. R. Officer) “Asset pricing in the Australian industrial equity market”, Australian Journal of Management, 1(1), 1976a, 1–32. (with R. Ball and F. Finn), “Share capitalisation changes, information, and the Australian equity market”, Australian Journal of Management, 2(1), 1977, 127–147. (with F. J. Finn and P. Hancock) “Dividend changes, earnings reports, and share prices: some Australian findings”, Australian Journal of Management, 2(2), 1977, 127–148. (with R. Ball and R. R. Officer) “Asset pricing in the Australian industrial equity market: reply”, Australian Journal of Management, 2(2), 1977, 195–199. (with R. Ball) “Market efficiency, random walks and seasonals in Australian equity prices”, Accounting Education, 18(1), 1978, 1–17. (with R. Ball, F. J. Finn, and R. R. Officer) “Dividends and the value of the firm: evidence from the Australian equity market”, Australian Journal of Management, 4(1), 1979, 13–26. (with R. Ball) “Risk and return from equity investments in the Australian mining industry: January 1958–February 1979”, Australian Journal of Management, 5(1&2), 1980a, 45–66. *[7] (with T. S. Walter) “Sharemarket efficiency and the experts: some Australian findings”, Australian Journal of Management, 7(1), 1982, 19–31.

13

(with A. W. Kleidon and T. A. Marsh) “New evidence on the nature of size-related anomalies in stock prices”, Journal of Financial Economics, 12(1), 1983, 33–56. (with A. W. Kleidon, D. B. Keim, and T. A. Marsh) “Stock return seasonalities and the tax-loss selling hypothesis”, Journal of Financial Economics, 12(1), 1983, 105–127. (with M. Bazley and H. Y. Izan) “An analysis of lease disclosures by Australian companies”, Abacus, 21(1), 1985, 44–62. (with A. Horin) “Assessing competition in the market for corporate control: Australian evidence”, Australian Journal of Management, 11(1), 1986, 23–50. (with T. S. Walter) “Ex-dividend day behaviour of Australian share prices”, Australian Journal of Management, 11(2), 1986, 139–152. (with T. S. Walter) “Ex-dividend day behaviour of Australian share prices (erratum)”, Australian Journal of Management, 12(1), 1987, 145–148. *[2] “Invited Remarks: Ball and Brown [1968]”, Journal of Accounting Research, 27 (Supplement), 1989, 202–217. *[9] (with I. Dunlop) “A case of reporting form over substance”, Australian Accounting Review, 1(2), 1991, 40–47. *[10] (with H. Y. Izan and A. L. Loh) “Fixed asset revaluations and managerial incentives”, Abacus, 28(1), 1992, 36–57.

14

(with A. Clarke) “The ex-dividend day behaviour of Australian share prices before and after dividend imputation”, Australian Journal of Managemen t, 18(1), 1993, 1–40. (with B. Howieson) “Accounting for employee share options”, Australian Accounting Review, 4(2), 1994, 22–34. (with I. Dunlop) “Valuing employee share options: four Australian case studies”, Australian Accounting Review, 4(2), 1994, 35–41. *[18] “A note on the inverse (reverse) sum-of-the-years’-digits method and other ways to amortise goodwill”, Australian Accounting Review, 5(1), 1995, 17–21. (with M. Aitken, H. Y. Izan, and T. S. Walter) “An intraday analysis of the probability of trading on the ASX at the asking price”, Australian Journal of Management, 20(2), 1995, 115–154. (with M. Aitken, C. Buckland, H. Y. Izan, and T. S. Walter) “Price clustering on the Australian stock exchange”, Pacific-Basin Finance Journal, 4(2–3), 1996, 297–314. (with P. Goldschmidt) “ALCOD IDSS: assisting the Australian stock market surveillance team’s review process”, Applied Artificial Intelligence, 10(6), 1996, 625–641. (with D. Walsh and A. Yuen) “The interaction between order imbalance and stock price”, Pacific-Basin Finance Journal, 5(5), 1997, 539–557. 15

*[13] (with B. Howieson) “Capital markets research and accounting standard setting”, Accounting and Finance, 38(1), 1998, 5–28. *[14] (with G. Clinch) “Global harmonisation of accounting standards: what research into capital markets tells us”, Australian Accounting Review, 8(1), 1998, 21–29. (with R. da Silva Rosa) “Australia’s corporate law reform and the market for corporate control”, Agenda, 5(2), 1998a, 179–188. (with R. da Silva Rosa) “Research method and the long-run performance of acquiring firms”, Australian Journal of Management, 23(1), 1998b, 23–38. *[11] (with S. L. Taylor and T. S. Walter) “The impact of statutory sanctions on the level and information content of voluntary corporate disclosure”, Abacus, 35(2), 1999, 138–162. (with N. Thomson and D. Walsh) “Characteristics of the order flow through an electronic open limit order book”, Journal of International Financial Markets, Institutions and Money, 9(4), 1999, 335–357. “Financial markets and financial accounting”, China Accounting and Finance Review, 1(2), 1999, 16–33. “Future disclosures”, Pacific Accounting Review, 11(2), 1999, 49–54. (with A. Clarke, J.C.Y. How, and K.J.P. Lim) “The accuracy of management dividend forecasts in

16

Australia”, Pacific-Basin Finance Journal, 8(3–4), 2000, 309–331. *[16] (with A. Tarca) “Politics, processes and the future of Australian accounting standards”, Abacus, 37(3), 2001, 267–296. (with E. Yew) “How do investors regard ESOs?” Australian Accounting Review, 12(1), 2002, 36–42. (with A. Chua and J. Mitchell) “The influence of cultural factors on price clustering: evidence from Asia–Pacific stock markets”, Pacific-Basin Finance Journal, 10(3), 2002, 307–332. (with A. Clarke, J.C.Y. How, and K.J.P. Lim) “Analysts’ dividend forecasts”, Pacific-Basin Finance Journal, 10(4), 2002, 371–391. (with M. Foo and I. Watson) “Trading by insiders in Australia: evidence on the profitability of directors’ trades”, Company and Securities Law Journal, 21(4), 2003, 248–261. (with A. Tarca) “A commentary on issues relating to the enforcement of international financial reporting standards in the EU”, European Accounting Review, 14(1), 2005a, 181–212. (with A. Tarca) “2005—it’s here, ready or not: a review of the Australian financial reporting framework”, Australian Accounting Review, 15(2), 2005b, 68–78. (with N. Chappel, R. da Silva Rosa, and T. S. Walter) “The reach of the disposition effect: large sample

17

evidence across investor classes”, International Review of Finance, 6(1–2), 2006, 43–78. *[8] (with G. Gallery and O. Goei) “Does market misvaluation help explain share market long-run underperformance following a seasoned equity issue?” Accounting and Finance, 46(2), 2006, 191–219. *[12] (with W. Beekes) “Do better-governed Australian firms make more informative disclosures?” Journal of Business Finance & Accounting, 33(3–4), 2006, 422–450. (with T. Boyd and A. Szimayer) “What determines early exercise of employee stock options in Australia?” Accounting and Finance, 47(2), 2007, 165–185. (with A. Tarca, P. Hancock, D. R. Woodliff, M. E. Bradbury, and T. van Zijl) “The matrix format income statement: a case study about earnings management and reporting financial performance”, Issues in Accounting Education, 22(4), 2007, 607–623. *[17] (with A. Tarca) “Achieving high quality, comparable financial reporting: a review of independent enforcement bodies in Australia and the United Kingdom”, Abacus, 43(4), 2007, 438–473. (with J. Mitchell) “Culture and stock price clustering: evidence from The Peoples’ Republic of China”, Pacific-Basin Finance Journal, 16(1–2), 2008, 95–120. (with A. Tarca, P. Hancock, D. R. Woodliff, M. E. Bradbury, and T. van Zijl) “Identifying decision useful information with the matrix format income statement”,

18

Journal of International Financial Management & Accounting, 19(2), 2008, 184–217. (with B. Al-Shammari and A. Tarca) “An investigation of compliance with international accounting standards by listed companies in the Gulf Co-Operation Council Member States”, The International Journal of Accounting, 43(4), 2008, 425–447. (with A. Ferguson and K. Stone) “Share purchase plans in Australia: issuer characteristics, and valuation implications”, Australian Journal of Management, 33(2), 2008, 307–332. *[19] (with A. Szimayer) “Valuing executive stock options: performance hurdles, early exercise and stochastic volatility”, Accounting and Finance, 48(3), 2008, 363–389. (with J.C.Y. How and P. Verhoeven) “The accuracy of analysts’ dividend forecasts around the world”, Pacific-Basin Finance Journal, 16(4), 2008, 411–435. (with A. Ferguson and A. Jackson) “Pierpont and the capital market”, Abacus, 45(2), 2009, 147–170. (with A. Ferguson and S. Sherry) “Investor behaviour in response to Australia’s capital gains tax”, Accounting and Finance, 50(4), 2010, 783–808. (with W. Beekes and P. Verhoeven) “Corporate governance, accounting and finance: a review”, Accounting & Finance, 51(1), 2011, 96–172.

19

*[15] “International Financial Reporting Standards: what are the benefits?” Accounting and Business Research, 41(3), 2011, 269–285. (with G. W. Dobbie and A. B. Jackson) “Measures of the timeliness of earnings”, Australian Accounting Review, 21(3), 2011, 222–234. IV Miscellaneous Publications (Including Articles in Professional Journals, Book Reviews, and Some Published Conference Proceedings) *[3] “The impact of the annual net profit report on the stock market”, The Australian Accountant, 40(July), 1970, 277–283. “Accounting for changing prices: a comment”, Chartered Accountant in Australia, 42(December), 1971, 24–25. (with T. S. Walter) “Fifty betas”, JASSA, 74(9), 1974, 1–7. (with R. Ball and R. R. Officer) “Risk and return in the share market—part 1: theory and evidence”, The Australian Accountant, 46(2), 1976b, 68–75. (with R. Ball and R. R. Officer) “Risk and return in the share market—part 2: some implications”, The Australian Accountant, 46(3), 1976c, 138–142. *[6] (with R. Ball) “Identifying some issues in C.C.A.”, The Australian Accountant, 47(August), 1977, 413–414 and 419–421. (with R. Ball and F. J. Finn) “Published investment recommendations and share prices: are there free lunches in security analysis?” JASSA, 2(1), 1978, 5–10.

20

(with R. Ball) “The performance of Australian mining equities, 1958–79”, The Investment Analyst, 57(July), 1980b, 11–18. “Rewards to risk-taking—evidence and implications from the Australian share market”, Proceedings of the Tenth National Congress of the Urban Development Institute of Australia, Perth, 1981. (with M. Tippett) “The effect of diversification on Australian portfolios: an analysis”, JASSA, 3(September), 1981, 15–16. “Rewards to risk-taking—evidence and implications from the Australian share market”, Economic Activity, April 1982, 25–33. “Comments on ‘Earnings forecasting research: its implications for capital markets research’ by L. Brown”, International Journal of Forecasting, 9(3), 1993, 331–335. (with R. da Silva Rosa) “Takeovers: who wins?” JASSA, 4(Summer), 1997, 2–5. “Book Review: The value reporting revolution: moving beyond the earnings game”, International Journal of Accounting, 37(2), 2002, 145–148. “Discussion of: Voluntary disclosure of management earnings forecasts in IPO prospectuses”, Journal of Business Finance & Accounting, 30(1–2), 2003, 169–173. “Notes of the University of Sydney Pacioli Society”, Abacus, 40(1), 2004, 132–137.

21

*[5] (with B. Howitt and M. Wee) “Order flow and price effects surrounding an ASX announcement”, 2005 AFAANZ Conference Proceedings, Accounting and Finance Association of Australia and New Zealand (July 2005), CD-ROM. (with R. da Silva Rosa and T. McNaughton) “A portrait of managed fund investors”, BT Quarterly, 2006(2). (with A. Ferguson and K. Stone) “Share purchase plans”, JASSA, 46(1), 2006, 18–23. “Review of: GEOFFREY WHITTINGTON, Profitability, accounting theory and methodology: the selected essays of Geoffrey Whittington”, The Accounting Review, 83(2), 2008, 557–559. “IFRS: How Real Are the Benefits? (in Korean)”, Korean Institute of Certified Public Accountants’ Monthly Journal, 207(8), 2010, 8–14. In addition, from October 1991 to August 2001 I was a part-time member of the Corporations and Markets Advisory Committee (CAMAC; previously known as the Companies and Securities Advisory Committee), which produced 18 reports during my term of office. These reports are available on the CAMAC website http://www.camac.gov.au/.

22

Acknowledgments The authors and publishers would like to thank the following organisations for granting permission to reproduce material in this work: 1. CPA Australia for two papers: “The impact of the annual net profit report on the stock market”, The Australian Accountant, 40(July), 1970, 277–283; and (with R. Ball) “Identifying some issues in C.C.A.”, The Australian Accountant, 47(August), 1977, 413–414 and 419–421. 2. CPA Australia and Wiley-Blackwell for three papers: (with I. Dunlop) “A case of reporting form over substance”, Australian Accounting Review, 1(2), 1991, 40–46; “A note on the inverse (reverse) sum-of-the-years’-digits method and other ways to amortise goodwill”, Australian Accounting Review, 5(1), 1995, 17–21; and (with G. Clinch) “Global harmonisation of accounting standards: what research into capital markets tells us”, Australian Accounting Review, 8(1), 1998, 21–29. 3. The University of Chicago and the University of Chicago Press, and Wiley-Blackwell for: (with R. Ball) “An empirical evaluation of accounting income numbers”, Journal of Accounting Research, 6(2), 1968, 159–178; and “Invited Remarks: Ball and Brown [1968]”, Journal of Accounting Research, 27 (Supplement), 1989, 202–217. 4. The University of Chicago and University of Chicago Press for one paper: (with J. W. Kennelly) “The informational content of quarterly earnings: an

23

extension and some further evidence”, The Journal of Business, 45(3), 1972, 403–415. 5. The Accounting and Finance Association of Australia and New Zealand, and Wiley-Blackwell for three papers: (with B. Howieson) “Capital markets research and accounting standard setting”, Accounting and Finance, 38(1), 1998, 5–28; (with G. Gallery and O. Goei) “Does market misvaluation help explain share market long-run underperformance following a seasoned equity issue?” Accounting and Finance, 46(2), 2006, 191–219; and (with A. Szimayer) “Valuing executive stock options: performance hurdles, early exercise and stochastic volatility”, Accounting and Finance, 48(3), 2008, 363–389. 6. The Accounting Foundation, The University of Sydney and Wiley-Blackwell for four papers: (with H. Y. Izan and A. L. Loh) “Fixed asset revaluations and managerial incentives”, Abacus, 28(1), 1992, 36–57; (with S. L. Taylor and T. S. Walter) “The impact of statutory sanctions on the level and information content of voluntary corporate disclosure”, Abacus, 35(2), 1999, 138–162; (with A. Tarca) “Politics, processes and the future of Australian accounting standards”, Abacus, 37(3), 2001, 267–296; and (with A. Tarca) “Achieving high quality, comparable financial reporting: a review of independent enforcement bodies in Australia and the United Kingdom”, Abacus, 43(4), 2007, 438–473. 7. Wiley-Blackwell for the paper: (with W. Beekes) “Do better-governed Australian firms make more informative disclosures?” Journal of Business Finance & Accounting, 33(3–4), 2006, 422–450.

24

8. Routledge for the paper: “International Financial Reporting Standards: what are the benefits?” Accounting and Business Research, 41(3), 2011, 269–285. 9. SAGE and The University of New South Wales for the paper: (with T. S. Walter) “Sharemarket efficiency and the experts: some Australian findings”, Australian Journal of Management, 7(1), 1982, 19–31. Every effort has been made to contact copyright holders, including coauthors, for their permission to reprint material in this book. The publishers would be grateful to hear from any copyright holder who is not here acknowledged and will undertake to rectify any errors or omissions in future editions of this book. Philip Brown also acknowledges the encouragement and help of many colleagues and friends, including: the editor of the series, Routledge Historical Perspectives in Accounting (Stephen A. Zeff); the journal editors, who agreed to his requests for permission to include material from their journals (Ray Ball, Tyrone Carlin, Graeme Dean, Linda English, Robert Faff, Stewart Jones, Baljit Sidhu, Martin Walker, and Pauline Weetman); his coauthors, all but one of whom (John W. Kennelly) he was able to contact, and who readily agreed to the inclusion of the papers they had written together; and others who helped by answering questions, tracking down material and, occasionally, by suggesting the likely whereabouts of coauthors with whom he had temporarily lost touch (Ray Ball, Bill Beaver, Wendy Beekes, Jon Burrows, Dan Collins, Jeremy Cowcher,

25

Joel Demski, Alexey Feigin, Andrew Ferguson, George Foster, Gerry Gallery, Tony Gleeson, Amanda Grabham, Bryan Howieson, Izan, and Kerrie Ludekens). In addition to the editor of the series, a number of others read the introduction and he thanks them for their advice and for ensuring he did not obviously rewrite history (Ray Ball, Jeff Coulton, David Brown, George Foster, Bryan Howieson, Izan, and Terry Walter).

All rights are reserved to the original publishers of each article.

26

Foreword Philip Brown is one of the most admired and respected accounting academics. He was a pioneer in capital markets research in accounting, and his 1968 article coauthored with Ray Ball, “An Empirical Evaluation of Accounting Income Numbers,” arguably had a greater impact on the course of accounting research, directly and indirectly, than any other article written during the second half of the twentieth century. It was therefore appropriate that the American Accounting Association gave Phil and Ray the inaugural Seminal Contribution to the Accounting Literature Award in 1986. The award is intended to recognize works that have stood the test of time, and it is given no more often than every three years. To be eligible, a work must have been out for at least 15 years. Phil’s other awards are numerous He was a Fulbright Scholar in 1963. In 1979 at the tender age of 38, he became a fellow of the Academy of the Social Sciences in Australia. In 2010, he was inducted as one of the first five members of the University of Melbourne’s Australian Accounting Hall of Fame, and on Australia Day 2012 he was invested with the Order of Australia. Phil has been an academic leader, both in research and academic administration. He established one of Australia’s first MBA programs as the founding director and professor of management at the Australian Graduate School of Management at the University of New South Wales from 1975 to 1979. In 1976, he launched and Ray Ball edited the Australian Journal of Management. Phil

27

then moved to the University of Western Australia, where he was professor of accounting and finance from 1980 until 2001, when he became emeritus. Since then he has been a professor of accounting at the University of New South Wales. He has played an outstanding role in mentoring young researchers at both of these institutions as well as in doctoral colloquia in Australia, the United States, and Europe. In 1970, Phil served as the president of the Australasian Association of University Teachers of Accounting, today the Accounting and Finance Association of Australia and New Zealand (AFAANZ). In 1996, AFAANZ recognized Phil and Ray Chambers with its inaugural Outstanding Contribution to the Accounting Literature Award and made Phil a life member in 2000. Phil has presented his research at numerous conferences and university workshops in Australia, the United States, and Europe. His innovative research has focused on interesting issues that bridge accounting and finance. I will leave it to Phil, in his richly informative introduction, to discuss his development as a scholar and the stories behind his salient research publications. Stephen A. Zeff

28

Introduction Beginnings I had decided to become a primary school teacher. David, my older brother, had graduated from teachers’ college and begun teaching at Smithtown Primary School, near the resort town of South West Rocks in the state of New South Wales, happy in the service. But his joy was short-lived when he was transferred after three weeks to become teacher-in-charge of a one-teacher school in Lower Creek, a tiny, isolated hamlet in rugged terrain halfway between Kempsey and Armidale. David telephoned home soon after arriving at Lower Creek and, with great feeling, said something like “don’t let them do to you what they have done to me”. David survived the experience of teaching children aged 5 to 16 and went on to a fulfilling career in education. So instead of entering teachers’ college, I began work as an office clerk. Shortly after I joined British Leyland (my third job in less than a year), I met another young trainee, Spencer Hird, who persuaded me to study commerce as a part-time student at The New South Wales University of Technology, now The University of New South Wales (UNSW). The campus was a four kilometre walk from Zetland, where I worked as a cost clerk in the Austin engine factory. Classes were held on campus 6–9 P.M. three nights a week. Cold, winter evenings were surely a test of a part-time student’s resolve. I did not find the classes particularly difficult but apart from that, I had little idea how well I was doing. Final 29

examinations were held at the end of each year, and the results were published some weeks later in the Sydney Morning Herald, a daily newspaper, for all to see. I was stunned when David told me it appeared that I had done better than any other student in the course, and his words again changed my direction in life. Instead of studying part time and pursuing a career as an accountant in industry, I decided to become a full-time student and, by the end of my second year, an academic. UNSW offered both a three-year “pass” degree course and a four-year “honours” course and it still does. The honours course at the time involved attending the pass degree classes plus, beginning with the second year, weekly seminars conducted by associate professor W.J.McK. (Bill) Stewart and additional readings plus a thesis in the fourth year. Bill guided me for the rest of my undergraduate days at UNSW, where I was the first accounting student to graduate with first-class honours and a highly coveted University Medal. Although my PhD was in finance and I have worked as an academic in both accounting and finance since the early 1960s, I still regard myself primarily as a financial accountant, which I attribute to the influence of Bill and my initial training in accounting at UNSW. Bill Stewart knew of my interest in an academic career and steered me towards doctoral studies at either the University of Chicago or Cornell University, both of which he had visited. I chose Chicago. Before I left for Chicago in September 1963, I did some tutoring in first year accounting, one of the students being a young chap with reddish hair. His name was Ray Ball.

30

Chicago My first days at Chicago were not easy, for personal reasons. Fortunately I had excellent training in accounting methods and Bill Stewart had schooled me well in the literature of the day. George Sorter agreed to me enrolling in the seminar for accounting doctoral students held in the fall quarter of 1963 and sitting for the accounting PhD comprehensive examination in the winter quarter, a fearsome examination comprising two three-hour papers on a Saturday. Like everyone else I was apprehensive about having to answer one of David Green’s characteristically obscure questions on management accounting. Joel Demski and I became good friends during these early days, even though our academic interests were quite different. In any event we both sat for and passed the accounting comprehensive examination and pressed on with the program. The decision to exempt me from preparatory accounting courses allowed me to immerse myself in the Chicago finance program. Gene Fama and Mert Miller had a profound influence on me with respect to ideas, and Joel Segall when it came to the research process. My fellow doctoral students in finance included Myron Scholes, Victor Niederhoffer, Mike Jensen, Dick Roll, and Marshall Blume. Dick introduced me to the power of Monte Carlo simulations when analytical solutions were less than apparent; and Marshall Blume was gracious enough not to poke too much fun at me when, in my eagerness to try out my newly found Fortran programming skills, I wrote a small program to simulate various characteristics of the term structure of interest

31

rates. Although a lot of the early work on generating random numbers by computer was being done at the University of Chicago, I knew nothing of it, so I extracted some random numbers from a statistical table, diligently punched them onto computer cards, and read them as data into my term structure program. Silly boy! In my defense, just a few years later I wrote a comprehensive Monte Carlo simulation program for one of the Western Australian iron ore producers. That program covered the whole operation: mining and stockpiling the ore, rail operations from the mine to the port, facilities at the port, and shipping. I have now been writing software in various languages for close to 50 years, and those who know me best understand I still enjoy the intellectual challenges it poses. Between 1963 and 1966 I became relatively well schooled in the Chicago research tradition and thoroughly at home in the highly confronting, weekly blood-letting events known as the research workshops in accounting. Bill Beaver, Bill Voss, and Jack Kennelly had been on the doctoral program longer than I, and Joel was a contemporary; all of them influenced my thinking in these formative years as did the accounting faculty, especially Sid Davidson, George Sorter, David Green, and Nick Dopuch. I do have some regrets from those days: while I cannot speak for others who attended the workshops, I am not proud of the way some presenters were treated, including by me. That said, they were great workshops from a research perspective and I know I learnt much about sound argument and the need to test ideas before accepting them.

32

Ray Ball followed me to Chicago, arriving in the fall of 1966. Although I was (I believe) among the first Australians to undertake a PhD in accounting and “modern” finance at a leading U.S. school, it was not long before a steady stream arrived from the Antipodes. Ross Watts and Bob Officer came to Chicago about the same time as Ray, and George Foster went to Stanford in 1971. What started as a trickle became a steady stream, and many have gone on to make their mark in one way or another. It is only fitting that the first paper in this book is “An empirical evaluation of accounting income numbers” (*Ball and Brown 1968; the asterisk indicates the paper is reprinted in this volume), which is the principal product of Ray’s and my time together at Chicago; and that the second paper be “Ball and Brown [1968]” (*Brown 1989), which explains the background to “An empirical evaluation …” and its relationship to the literature that followed. The second paper encapsulates some remarks I made, on Nick Dopuch’s invitation, at the special Journal of Accounting Research (JAR) conference hosted by Washington University in St Louis to celebrate the twentieth anniversary of the publication of not one but two papers in JAR in 1968: Ray’s and my “empirical evaluation” and Bill Beaver’s extraordinary paper, “The information content of annual earnings announcements”. I will not repeat, here, the background to the first and second papers in this collection as it is covered in the second paper. But I must mention Ray’s and my first publication together, Brown and Ball (1967), which was a building block for our 1968 paper. It dealt with the relationship between the earnings of the

33

firm, its industry and the economy. Soon after writing the 1967 paper, I told Ray I thought we should list our names in alphabetical order in any future paper and it is a policy I have tried to maintain, with just a few exceptions. We wrote a third paper around this time, on portfolio theory and its relationship to accounting (Ball and Brown 1969). We wrote this paper because “modern portfolio theory”, or MPT as it was sometimes referred to in the 1960s, was a grey area to many accounting academics. I had great respect for David Green and Joel Segall. But in 1967 they published what to me (and I believe to them) was an odd result: investors were no better placed in predicting this year’s earnings per share when the first quarter’s results were to hand than they were at the end of the previous financial year. The oddity was that Ray Ball and I, and others before us, had observed earnings behaved somewhat like a random walk; there would have to be something substantially different about earnings in the first three months of the financial year to get the Green and Segall result. Victor Niederhoffer thought the same, so we re-examined the Green and Se-gall question and came up with estimates that confirmed our intuition. Our paper, “The predictive content of quarterly earnings”, published in 1968 in the Journal of Business (Brown and Niederhoffer 1968), drew a sharp retort from David and Joel. Victor decided to continue the debate, but I did not join him as I thought we had made our point. I decided to take up a related question instead, to extend the approach Ray and I had used to relate annual earnings numbers to stock returns by applying it to quarterly reports. Jack Kennelly joined

34

me to write the paper. It almost slipped between the cracks when I returned to Australia in 1968 (which, by the way, explains why the author byline in Ball and Brown (1968) shows my affiliation as the University of Western Australia), but was revived when I visited Chicago for three months at the end of 1970. Jack’s and my paper appeared two years later in the Journal of Business under the title “The informational content of quarterly earnings: an extension and some further evidence” (*Brown and Kennelly 1972). UWA In 1963, I accepted the offer of a Fulbright scholarship, which paid my return airfare, to go to Chicago. The terms of the scholarship required me to leave the United States within a few months of completing my PhD. So Edith and I (we married in September 1966) decided we would head back to Australia. I felt some residual obligation to UNSW and enquired about a position there. Much to my disappointment the dean replied I could expect a senior tutorship but probably nothing more (the ranks were tutor, senior tutor, lecturer, senior lecturer, reader/associate professor, and professor). Bill Stewart mentioned a readership had been advertised by the University of Western Australia (UWA); I enquired about that opening. Events moved swiftly, for an Australian university. Andrew Brown, head of UWA’s department of commerce, had been personnel director of Plessey, in the United Kingdom, before taking up his appointment at UWA and was not one for letting opportunities pass by. Andrew soon arranged for Edith and me to fly out to Australia for an interview, and

35

UWA offered me the post before we returned to Chicago. I accepted forthwith. Perth is a wonderful city in which to live, but in 1968 UWA was about as isolated as a university could be. I was raw and inexperienced but helped greatly by the steady hands and minds of two colleagues, Roy Lourens and Les Hollis. Martin Black, from Duke University, was a visitor to UWA and carried, like Roy, Les, and me, a demanding teaching load. Research was a second priority for much of the next six years while we refocused the undergraduate accounting units, strengthened the honours program, introduced a new course in corporate finance (an Australian first), and helped to plan an MBA program and then to raise funds for the university to introduce it. Research infrastructure was minimal. For example there was no Australian equivalent of the CRSP and Compustat databases, so I began to build them and made them freely available to anyone who was interested. My 1970 paper, “The impact of the annual net profit report on the stock market” (*Brown 1970), was an outgrowth of these efforts. It was first presented at Ken Wright’s invitation as an endowed lecture at the University of Adelaide, and was published in a professional accounting journal. I do not know what the professional accountants thought, but it could well have been that I was from another planet. I include the paper in this collection for several reasons: it marked the beginning of empirical capital markets-based academic research in Australia, it confirmed the international character of Ray’s and my principal results for the U.S. market, it confirmed that the post-earnings announcement drift was international in nature, and it

36

contained evidence that price discovery of good news occurred earlier than bad news in Australia. In the early 1970s a young Queenslander, Terry Walter, decided to forsake a burgeoning career in stockbroking to become an academic. Terry joined us at UWA, thereby beginning an association that now extends to about 40 years. Terry enrolled for a PhD and decided to research the Australian market for corporate control. Because monthly returns were too coarse for his liking, and hand collecting daily returns for the whole Australian market was out of the question, Terry built a database of weekly returns on listed stocks. Practitioners had been hearing about beta, which was unquestionably alive at the time, but few beta estimates were available for them to discuss. So we wrote the paper “Fifty betas”, estimated from Terry’s database, and published it in JASSA, a journal for investment professionals (Brown and Walter 1974). Another of our professionally oriented papers (*Brown and Walter 1982) dealt with 157 market professionals’ stock picking ability. They all had attended one-week courses on portfolio management for market professionals held between 1973 and 1979. At the conclusion of each course, my colleagues and I asked the attendees to nominate, individually, two stocks to buy and two to sell. Interestingly, Terry and I found the stocks they recommended be sold did consistently underperform the market over the next year, but their buys did not reveal any worthwhile private information they may have possessed. As my list of publications suggests, I count Terry as one of my closest colleagues. AGSM

37

I had been aware of moves at a political level to establish a national graduate school for managers, which I assumed would be located in Melbourne or Sydney and had little immediate relevance to us, being 3,000 kilometres or more to the west. Richard Cyert, then dean of Carnegie-Mellon’s Graduate School of Industrial Administration and later president of the university, had been appointed by the Australian government to chair a committee to report on whether a specially funded school was needed. The Cyert Committee had recommended in 1970 such a school be established, and in the course of time it was decided the school would be hosted by UNSW. An Interim Board of Management was established and the position of foundation director (dean) was duly advertised. As I understand it, after some months of searching, the Interim Board was not satisfied it had found the right person, so to speak, and asked one of its members, Russell Mathews, to sound me out. Somewhat reluctantly I agreed to fly to Sydney and meet Rupert Myers, vice-chancellor of UNSW, and the other members of the Interim Board. One thing led to another and at the end of 1974 I left UWA, a 34-year-old with a five-year contract to establish the new school, which adopted the name Australian Graduate School of Management in The University of New South Wales, or AGSM for short. I was attracted by the opportunity to create a graduate school from the ground up. I have never been particularly interested in a career as a university administrator, but this was different. In 1972, Reg Gynther had enticed Ray Ball to a chair at the University of Queensland in Brisbane, where he had done some great work with Bob Officer (who had also

38

returned to Australia from Chicago) and some “locals”, including Frank Finn. Ray accepted an offer to join the AGSM as one of a number of foundation professors, and he remained there until 1986. Early in the process the foundation team decided to publish the Australian Journal of Management and I invited Ray to be the first editor. He accepted; he must have enjoyed the experience as he has, since then, held editorial responsibilities with the Journal of Accounting and Economics and the Journal of Accounting Research. The foundation professors were an interesting bunch. Malcolm Fisher was one of them. Not long after I began to work on establishing the school, I wrote to a number of people whom I and others thought might be interested in an appointment at the new school. I then embarked on an international recruitment mission, visiting South Africa, the United Kingdom, and the United States to meet people who had responded to my early enquiries. As Malcolm told me later—he was at Cambridge when I first contacted him—he decided to meet me in London purely because he was curious to know what kind of a person would have the audacity to write to him, a Cambridge Don, with the preposterous idea of him forsaking Cambridge for a risky position at a fledgling school in a former British penal colony. Well, Malcolm took the train from Cambridge to London, we met and got on well together, and it was not all that long before he was on his way to New South Wales as a free man. Others who took the plunge and became part of the new school included John Stringer, who worked in operations research at the Tavistock Institute; Phil Yetton, a behavioural scientist at

39

Manchester University; Di Yerbury, a first assistant secretary in the Australian Public Service in Canberra (Di was an industrial relations lawyer and went on to become vice-chancellor of Macquarie University); and Howard Thomas, then at London Business School and later dean of the business schools at the University of Illinois and Warwick University and now dean of the Lee Kong Chian School of Business at Singapore Management University. The AGSM attracted Peter Wilenski, a former chairman of the Australian Public Service Board in Canberra, partly because of our plans to offer a master’s degree program designed for the needs of the public as well as the private sector. I mention the diversity of backgrounds because it illustrates an important feature of the AGSM: unlike some business or management schools, we valued excellence in management education broadly defined, and the particular lens through which each of my colleagues viewed “management” was less important than the insights they brought and the quality of their work. While my own lens derives in the main from my training in economics, finance, and accounting at UNSW and Chicago, I hope I have always shown proper regard for the views and abilities of colleagues who adopt a different perspective. After Ray Ball arrived at the AGSM we embarked on a number of projects together, although my contribution was less than Ray’s because of the other duties of the director. Frank Finn and Bob Officer joined us on several of them, including the compilation of a book of readings for use in teaching investments classes (Ball, Brown, Finn, and Officer 1980). I have not included in

40

this collection several papers written at this time mostly because they were more obviously in the field of finance (Ball and Brown 1978; Ball, Brown, and Finn 1977; Ball, Brown, and Finn 1978; Brown 1979; Brown, Finn, and Hancock 1977). During my time at the AGSM we extended the share price-price relative database I had started at UWA, and Ray and I collaborated on six papers, including four with Bob Officer, on the ex post relationship between risk and return in the Australian share market: “Asset pricing in the Australian industrial equity market” and “Asset pricing in the Australian industrial equity market: reply”; “Risk and return in the share market”; “Risk and return from equity investments in the Australian mining industry: January 1958–February 1979”; and “The performance of Australian mining equities, 1958–79”. I have a strong belief that accounting and finance academics have a responsibility to work with their professional counterparts and to translate their findings when appropriate; so several of these and other papers were published in professional journals such as The Australian Accountant, Chartered Accountant in Australia, JASSA, and Investment Analyst. There was quite a debate in Australia and elsewhere in the mid- to late 1970s on whether accounting numbers should be adjusted for inflation. One proposal was labeled Current Cost Accounting and it was controversial among Australian academics to say the least. Russell Mathews, Allan Barton, and Reg Gynther were in the thick of the debate and Ray Chambers was “in there swinging” too. In early 1977, an Australian stockbroker, J.M. Bowyer & Co., published a list of 50

41

companies whose 1976 earnings would have been 37 percent less on average under Current Cost Accounting. So Ray Ball and I decided to write a paper for a professional journal to explain why the stockbroker’s analysis was unlikely to have much influence on either inflation or the share market index when it was made public. The title of our paper was “Identifying some issues in C.C.A.”, and it was published in the August 1977 issue of The Australian Accountant (*Ball and Brown 1977). Six years earlier I had written a small piece on inflation accounting for the journal of the other major Australian professional accounting body, The Institute of Chartered Accountants in Australia, which published it under the title “Accounting for changing prices: a comment” (Brown 1971). I tend not to write theoretical pieces in the tradition of other Australian accounting academics such as Chambers, Gynther, Mathews, Barton, Goldberg, or Wright, for example, but there are a couple of exceptions. One is “A note on the inverse (reverse) sum-of-the-years’-digits method and other ways to amortise goodwill” (*Brown 1995). I wrote that paper because, in the mid-1990s, the Australian financial market regulator seemed to think it was impossible to justify a goodwill amortisation method that resulted in an increasing annual charge. Kind of simple to justify, is it not, if we are prepared to make some assumptions? So I spelt them out and it was published in Australian Accounting Review, a journal whose readership comprises mainly “thinking practitioners”, accounting students, and academics. The model was simple and the algebra came easily, because more than 30 years earlier I had written a thesis on depreciation methods, and in the course of writing that 42

thesis I had read a series of papers by Hector Anton, Harold Bierman, Robert Dixon, and a few others on selecting the “most appropriate” depreciation method. I had believed the most appropriate amortisation method was a compound interest method: if initial expectations were realised such that the ex ante and ex post yield on the asset were say r (being a unique, real number), then the “correct” depreciation charge in a given year is the net cash flow for the year less r times the book value of the asset at the end of the previous year. Sound familiar? Harking back, it was really my work on depreciation, as an honours student, that led Bill Stewart to suggest I apply to Chicago and Cornell. David Drake, an assistant professor at Chicago, had written on depreciation as had Hal Bierman at Cornell. I must have disappointed some of my Chicago accounting mentors, as I began to wander off into finance and in the end submitted a thesis on corporate financial policy. However, my thesis did refer to Ray Ball’s and my work on the timeliness of accounting reports, because Miller and Modigliani’s corporate finance theory, which I employed, requires a judgment about which stock price should be used to value the firm. Ray Ball and I have coauthored more than a few papers over the years but, as it has turned out, none since I left the Australian Graduate School of Management in UNSW at the end of my five-year appointment to return to the University of Western Australia. Return to UWA, Via Chicago

43

I was born on 19 April 1940. My health was a concern to my folks and as an infant I had been irradiated to treat an enlarged thymus gland, apparently the result of whooping cough. The longer term effects of radiation therapy caught up with me while I was at the AGSM. Two bouts of thyroid surgery later, Edith and I decided to explore the possibility of returning to Western Australia at the end of my five-year contract. My long-term prognosis was reasonable, but the experience focused our minds on bringing up our two children, Amanda and Matthew, in a more amenable environment. An opportunity opened up at UWA, and with mixed feelings—mainly because the construction of the AGSM building had been delayed due to a funding freeze after a change of government in Canberra—at the end of September 1979 we left Sydney for Perth, taking a six-month “detour” via Chicago. My aim at Chicago was to reinvigorate my research. On arrival I met two top-drawer Australian PhD students, Allan Kleidon and Terry Marsh. We chatted for a while about possibly working on something together and settled on tax-loss selling and the size anomaly that Rolf Banz, who a year earlier had completed his PhD at Chicago, had brought to the fore. We decided to test some of our ideas on Australian data, because Australia’s tax year ends in June and seasonal patterns should be different. Donald Keim, a Chicago PhD student who had linked the size anomaly with the “turn-of-the-year” effect in the United States, joined us on this project. We concluded that the relationship observed in the United States between the end of the tax year and the small firm effect, which appeared to be

44

concentrated in January, could well be correlation rather than causation. We also found that, for the smallest decile of Australian firms, there was a 4 percent share market return premium across all months of the year. Two papers resulted from that visit, both in the same issue of the Journal of Financial Economics (Brown, Keim, Kleidon, and Marsh 1983; Brown, Kleidon, and Marsh 1983). Twenty-five years later I became interested again in ways in which tax loss selling is manifest in financial markets, culminating in two papers. One, “The reach of the disposition effect: large sample evidence across investor classes”, was coauthored with Nick Chappel, Ray da Silva Rosa, and Terry Walter and published in the International Review of Finance (Brown, Chappel, da Silva Rosa, and Walter 2006). Nick, Ray, and Terry had invited me to join them some time after Nick had begun work on his PhD at the University of Sydney, from which the paper was an outgrowth. Ray and Terry, who had been my own PhD students, were Nick’s supervisors. We confirmed an interesting phenomenon. The disposition effect, which is manifest in the tendency of share market investors to sell winners too soon and hang on to losers too long, weakens both over time (perhaps because time heals the pain, implying investors eventually become reconciled to their losses) and especially in the face of opportunities, as at the end of the tax year, for investors to gain some benefit by realizing their losses and offsetting them against taxable capital gains. The second paper was published in Accounting and Finance in 2010 (Brown, Ferguson, and Sherry 2010). This paper was an outgrowth of Sam

45

Sherry’s thesis at UNSW, which Andrew Ferguson and I had co-supervised, and looked at tax loss selling per se. Tax loss selling seems to bob up each June in the Australian financial press and obviously it has been a recurrent theme in my work too. Some questions simply do not go away. I recall about 40 years ago I was asked to read a working paper on corporate dividend policy. The paper’s title indicated the authors believed they were about to have the “last say on dividend policy”, a claim which I thought rather odd at the time and still do: academics thrive on debate and disagreement, and few questions in accounting or finance have been answered to everyone’s satisfaction. That is why I advise PhD students always to ask themselves three questions after they have read an article or working paper (that lends itself to these questions). First, after considering the paper’s findings and the explanations offered by the author, what alternative explanations can I offer? Second, if I had investigated the question, what would I have done differently? Third, what comes next? I have found students who develop this habit are never bereft of ideas. The Brown family arrived back in Perth around the end of March 1980. Roy Lourens was by then professor of accounting and head of the department; Roy had been a lecturer when I first came to UWA in 1968. Not long after my arrival Roy was appointed UWA’s deputy vice-chancellor; and in 1993 he became vice-chancellor of Edith Cowan University, also in Perth. Roy and I had earlier agreed I would have a minimal administrative role at UWA while I re-established my research

46

environment. It took a while. I was fortunate enough to supervise a whole series of superb honours students in this period, some of whom became academics (Patti Dechow and Richard Sloan are among my former honours students) while others have pursued a career elsewhere (Mark Barnaba, Brad Rosser, and Andrew Horin, for example, were outstanding students who chose careers in business). On occasions my former honours students have invited me to join them in writing up papers that built on their work, and others have not: the decision is entirely theirs. My 1986 paper with Andrew Horin, “Assessing competition in the market for corporate control: Australian evidence” is an example of the former (Brown and Horin 1986). Both Patti and Richard went down the second path, and I was very pleased to see papers from their honours theses appear in 1987 in the Australian Journal of Management. Izan is a special colleague. She completed her PhD at Chicago in 1978 and in 1979 joined the AGSM. Two years later she took up a position in the department of accounting and finance at UWA, when her husband, Ken Clements, who holds a PhD in economics from Chicago, was appointed to a chair in economics at UWA. Over the 30 years we have both been at UWA, Izan and I have co-supervised many honours, master’s, and PhD students. Mike Bazley was one such student. Mike wrote a master’s thesis on voluntary disclosure of leases between 1964 and 1980, a time span which saw a twenty-fold increase in lease financing by Australian companies. Building on Mike’s thesis, Izan, Mike, and I reported that the probability of voluntary disclosure by a

47

lessee was related to whether it was a subsidiary company of a foreign parent, its size, and its industry. It was unrelated to the identity of the lessee’s audit firm, the existence of a bonus scheme tied to reported profit and the firm’s risk. Alfred Loh, now at the National University of Singapore, was one of Izan’s and my PhD students. Alfred was interested in understanding asset revaluations by Australian listed companies, a not uncommon event. The literature on accounting policy choice was again a source of the explanatory variables. Subsequently Alfred, Izan, and I coauthored the paper “Fixed asset revaluations and managerial incentives” (*Brown, Izan, and Loh 1992) based on material in his thesis. We concluded that an asset revaluation lowers the probability of a wealth transfer arising from contracting and political costs and conveys a signal to the users of financial statements that the assets are worth more than had been recognized previously. Stanford George Foster came to the AGSM in 1977 and left in 1978. Even though we worked together for only a short while, it was long enough for me to appreciate George as one of the most congenial and generous colleagues an academic could ever have. When I became eligible in 1983 for 12 months of study leave beginning that January, George helped me to arrange a visit to Stanford’s Graduate School of Business. Eric Noreen was also visiting the GSB when I arrived, and Bill Beaver and Joel Demski, friends from my Chicago days, were on the faculty too. Eric, a keen advocate of computerintensive resampling techniques, and I joined

48

George as team leader to write a monograph for the American Accounting Association (Brown, Foster, and Noreen 1985). George, a great writer who has an amazing memory and was a master of the relevant literature, had access to a Wells Fargo database containing internal analysts’ EPS forecasts for up to five years ahead. The monograph canvassed a whole series of research questions on relationships between various properties of analysts’ forecasts and stock prices. An interesting conundrum was the source of the positive correlation between the EPS forecast horizon and how well EPS forecasts “explained” stock prices: the correlation between forecast EPS and stock price was stronger for earnings two years ahead than one year, for three years than two, and so on. The conundrum was whether the increase in explanatory power was due to analysts’ focusing their efforts on near-term earnings and simply using some fraction of price to estimate longer-term earnings, or the result of the analysts recognising that any special knowledge they might have would be short-lived. It is similar to George Sorter’s reductio ad absurdum argument, made at a Chicago workshop many years earlier: the correlation between stock prices and accounting measures could never be the sole criterion for choosing between alternative accounting methods, because, if it were, then accountants need only report market values of equity and debt, set assets equal to their combined value, and define earnings as the change in owners’ equity (with adjustment for dividends and other exchanges between the firm and its equity holders). As far as I am aware the conundrum has not been resolved.

49

Stanford Mark II By the end of 1986 I was eligible for another six months of study leave and again had the opportunity to visit Stanford’s GSB, which I did between January and June. This time George Foster and I teamed up with Greg Clinch, then a PhD student from Melbourne. George had access to high frequency (intraday) stock market trades data and precise earnings announcement times. We were aware of the U-shaped pattern that characterises intraday market activity, which may be loosely described as a flurry of activity at market opening as the backlog of overnight orders is processed, a declining rate of activity followed by a lull when lunches of various lengths are taken (some exchanges, e.g. Tokyo, close for lunch), and a build-up of trading during the afternoon with another peak around market closing. We decided to take a closer look at intraday patterns and their implications for information content studies, which was yet another extension of the original Ball and Brown (1968) study. Since 1968, researchers had taken a progressively closer look at market behaviour: Ray and I had examined monthly stock prices, Bill Beaver had looked at trading volume and volatility over weekly holding periods, and many others had considered daily data; by the 1980s it was possible to observe intraday movements as well. George, Greg and I analysed the data in both calendar and transaction time, using stock returns, volatility, trading frequency, and trading volume as indicators of trading activity. Our study was later published as American Accounting Association Studies in Accounting Research #32 (Brown, Clinch, and Foster 1992).

50

I continued this line of enquiry on my return to Australia. Mike Aitken had been on leave from UNSW and was working with Jim Berry, then head of the Surveillance Division of the Australian Stock Exchange (ASX). Patterns in market behaviour are key to surveillance activities, so Mike was familiar with efforts to identify “abnormal” market behaviour and the difficulties of separating “normal” from “abnormal”. And he had access to ASX data. Mike subsequently took a position at Sydney University, and in 1993 Mike, Terry Walter, who at that time also was at Sydney University, and Izan and I (both at UWA) joined with the ASX to promote market microstructure research in Australia. At first a mirror site was operated at UWA; and I, initially with Paul Vowles and later with Jennifer Cross, developed software to reconstruct the limit order book and to calculate a suite of stock and flow metrics to describe market activity. The stock metrics were designed to capture the state of the limit order book at a given point in time, while the flow metrics tracked the orders and trades that linked successive snapshots of the limit order book. Our first working papers were written in 1993 and 1994. One of the 1994 working papers dealt with two types of announcement: one whose timing is predictable (most earnings announcements are scheduled in advance) and another whose timing is not (on-market takeover bids mostly surprise the market, although it is not unknown for people with inside information to trade in advance). We confirmed what one would expect: reduced activity where the timing was predictable, reflecting the gradual withdrawal of uninformed traders from the market as the

51

odds of them trading with an informed counterparty progressively increased; and, for an unexpected event, no detectable change in activity (“business as usual”) right up to the commencement of the trading halt period, which the ASX imposes when it first learns a takeover bid announcement is imminent. We used resampling methods to demonstrate that the patterns around the time of an announcement were different from intraday patterns observed for the same stocks but on other days. The paper was not published—it contained far too much intuitive material and too many graphs—but it and other working papers served their purpose: to stimulate research in Australia into intraday behaviour of financial markets. More than a decade later Bryan Howitt, Marvin Wee, and I wrote a comprehensive, similar paper based on Bryan’s honours thesis, which was co-supervised by Marvin and me (*Brown, Howitt, and Wee 2005). The ASX classifies company announcements by the nature of the information contained in them. Bryan, Marvin, and I reported that the majority of market sensitive announcement types were associated with abnormal order flow before the release of the announcement document. Similarly, there is abnormal order flow after a market sensitive announcement. The absolute share price change surrounding market sensitive announcements is also significantly larger than at other times. Understandably it was a much more refined version of the earlier working papers and Brown, Howitt, and Wee (2005) is included in this collection to illustrate how that work has evolved. Incidentally, in 1993 Mike, Terry, Izan, and I applied for substantial research funding from our universities, the ASX, and

52

the Commonwealth government. Funds were received over a three-year period beginning in 1994, and the Securities Industry Research Centre of Asia-Pacific, or SIRCA, came into being. Within 15 years SIRCA had grown into a major supplier of high frequency capital market data to the international research community. Corridors We all know how those corridor conversations between academics go; in the extreme, we find something really puzzling and wonder whether we can find order in the chaos. In less extreme situations, we might just wonder why there is a particular gap in the literature, or whether there may be an alternative explanation for results in a paper we have recently read. Like most academics, I have had many of these conversations. Two examples stand out, one in what I describe as the intersection between accounting and finance, which is where I mainly have dwelt, and the other in finance. The first example is a conversation about dividend forecasts. Terry Walter and I had published a paper in the mid-1980s (Brown and Walter 1986) on the dividend-capital gains trade-off, which we inferred from the drop-off ratio, which is the price change (typically a decline) on the ex dividend day expressed as a fraction of the amount of the dividend. Until about that time, broadly speaking, for many individual Australian residents a dividend was taxed as ordinary income and a capital gain was tax exempt. A capital gains tax was introduced by the Australian government on 19 September 1985 and on 1 July 1987, the government introduced a system of dividend imputation whereby, in

53

essence, Australian resident personal taxpayers were credited with their share of federal income tax paid by the company in which they held shares (imputation was extended to superannuation funds a year later). Alex Clarke (now at the Australian National University) revisited the trade-off question in the presence of capital gains tax and imputation tax credits and wrote a master’s degree thesis at UWA on the subject, which I supervised. One way or another I had had a fairly heavy hand in Alex’s work, and we developed a paper from his thesis (Brown and Clarke 1993). That is part of the background. The other part is that Alex and I, and it turned out a UWA colleague, Janice How, had wondered why there were hundreds if not thousands of papers on earnings forecasts, but none that we were aware of on dividend forecasts (we did later come across a 1994 paper by Richard Leftwich and Mark Zmijewski, published in the Journal of Accounting, Auditing and Finance, which had considered relationships between earnings and dividend forecasts). We drew this to the attention of Kadir Lim, a UWA honours student in search of a topic, and he decided to investigate properties of dividend forecasts, drawing parallels between earnings and dividend forecasts since academics in finance have known for 50 years or more that they are connected. The upshot was two papers in the Pacific-Basin Finance Journal (Brown, Clarke, How, and Lim 2000; Brown, Clarke, How, and Lim 2002). In the first paper we reported managers’ dividend forecasts contained in IPO (initial public offering) prospectuses were substantially more accurate and significantly less biased than their

54

earnings counterparts and in the second paper, that the same could be said of analysts’ forecasts. The second example comes from the finance literature on price clustering. Clustering in stock prices results from imprecise beliefs (or “haziness”) about the firm’s intrinsic value. Because traders value immediacy, transaction prices tend to settle more frequently on conventional, salient focal points within regions of haziness. In 1994 Izan and I had co-supervised Christine Buckland’s honours thesis at UWA, in which Christine had found price clustering was pervasive among Australian stocks and largely mirrored the pattern found in U.S. markets. This was a SIRCA-related project and was published as such (Aitken, Brown, Buckland, Izan, and Walter 1996). Three or four years later Jason Mitchell and I were chatting in the corridor about price clustering—Jason had completed his PhD at UWA in 1998 on clustering in Australian dollar exchange rates and had taken a strongly behavioural slant—and realised we had both been wondering, separately, whether the focal points relevant to price clustering might differ by country according to a dominant cultural influence, such as the Chinese preference for 8 and aversion to 4 (Chinese consider 8 to be a “lucky” number, whereas 4 is “unlucky” because it sounds like the Chinese word for “death”). Our first paper, with Amy Kua, an honours student at UWA, reported limited evidence of a “Chinese cultural effect” in the Asian region (Brown, Chua, and Mitchell 2002); but our second attempt, which focused on the Shanghai and Shenzhen exchanges in the PRC, struck pay dirt (Brown and Mitchell 2008). In the second paper we concentrated on the traded prices of A-shares,

55

which were mostly held by Chinese organisations or individuals. For much of our sample period (1994–2002) the last digit of prices on the Shanghai stock exchange was 8 more than twice as often as 4, while the Shenzhen stock exchange also revealed a substantial preference for 8. Interestingly, the preference for 8 on both exchanges had weakened over time. Mathews Committee The Mathews Committee, chaired by the late Russell Mathews, was appointed in 1989 to investigate the state of accounting education in Australia and to make recommendations on its future. Margaret Jackson of RMIT and I were appointed deputy chairs. We had an excellent executive secretary, Lesley Brookes, and research support headed by Gerald Elsworth. It was an onerous task, requiring the committee to visit and report on accounting programs in all Australian universities, among many other things. Australia is approximately the size of the continental United States, and a protracted strike by airline pilots did not help. The Mathews Committee transmitted its three-volume report (Report of the Review of the Accounting Discipline in Higher Education) to the minister for Higher Education and Employment Services on 29 June 1990. The report contained 62 recommendations on matters such as students, courses and programs, funding, staffing, salaries, facilities, teaching, and research. A core recommendation was that undergraduate programs in accounting be extended to a fourth year. The government did not accept this recommendation because (it seemed to us) of the cost

56

involved, since all but a few university-level accounting courses in Australia were publicly funded. We had been concerned about the narrowness of professionally accredited accounting courses, which in many cases were focused too heavily on technical matters and allowed limited scope for students to develop skills and attitudes more suited to lifelong learning. Russell had been ill during the committee’s formative period and it was largely left to Margaret and me to devise the format of our report on each university and get the visiting program under way. Russell, though, was a prodigious writer and his skills came to the fore as the enquiry progressed. Understandably, my own research was largely put on hold during this period. Accounting Standards and Regulation From October 1991 to August 2001 I was a part-time member of the Corporations and Markets Advisory Committee (CAMAC), a national committee, having been nominated as the sole member from Western Australia. Previously known as the Companies and Securities Advisory Committee (CASAC), the committee’s role was to advise the federal minister largely on matters to do with corporate law and regulation of financial markets. The committee for the most part comprised senior partners of professional accounting and law firms, and industry leaders, with one or two others like me. The chairperson of the main corporate regulatory body, the Australian Securities and Investments Commission (ASIC), is a member ex officio. The committee was busy, and during my almost 11 years as a member it produced 18 reports. It was a

57

leavening experience; and firsthand knowledge of “how things work” kindled my interest in the politics and processes of regulation and accounting standard setting generally, and International Financial Reporting Standards (IFRS) in particular. Australia embarked on a process of harmonization of Australian and international accounting standards in 1996, following the issue of a policy statement by the Australian Accounting Standards Board. The Australian Stock Exchange supported harmonization, possibly because it saw benefits flowing to Australian companies and to itself as market provider. Greg Clinch and I had views on the soundness of some of the debate in favour of harmonization and wrote a paper on the subject for the Australian Accounting Review (*Brown and Clinch 1998). We concluded that harmonization would affect markets and their participants, but research to that point was silent on exactly what the consequences would be. Harmonization might even weaken the ASX’s position by reducing barriers to international competition that were provided by domestic accounting standards; but without additional research, this possibility was conjectural. We also believed that, because there had been little research on the effects of harmonization, we could not be confident about what it would mean for companies’ cost of capital. I make no secret of the fact that I was among many who were chary of any proposal to adopt international accounting standards in lieu of Australian domestic standards, as I could see that both costs and benefits

58

would be involved. I was given the opportunity to present my ideas in the form of endowed lectures at the University of Tasmania in 1998 and the University of Melbourne in 2000. Ann Tarca joined me in 2000 to flesh out and extend these ideas, resulting in our 2001 paper in Abacus (*Brown and Tarca 2001). Ann and I had met in 1996, when she enrolled in the Advanced Financial Accounting Class I taught on Friday nights. Ann had lived in Karratha, in the Pilbara, for the previous eight years, teaching in the local technical college and “country contracting” for another Perth-based university, conducting first year classes in accounting, management, and law. Ann wished to pursue a teaching career rather than return to public practice, where she had worked before moving to Karratha. She completed a master’s degree in 1998 and a PhD, on international convergence of accounting practice, in 2003. Since 2001 Ann and I have collaborated on various questions to do with international accounting standards and regulation of their usage. For example, we have looked at the efficacy of different enforcement models (Brown and Tarca 2005a; *Brown and Tarca 2007) and examined the degree of compliance in the Gulf region with international accounting standards (Al-Shammari, Brown, and Tarca 2008). The last paper was developed from material collected by Bader Al-Shammari, a PhD student from Kuwait who was co-supervised by Ann and me. Early Exodus In 2001 UWA, where I had been a full-time professor for about 25 years, committed to an organisational

59

restructuring. I was offered an early retirement package and accepted it, with effect from 31 December. Ken Peasnell, a former editor of Accounting and Business Research with whom I had worked as a member of the editorial board, indicated that Lancaster University would be happy to have me as a half-time visitor in their department of accounting and finance. What a good move that proved to be. Between January and December 2002 I taught the PhD seminar in financial accounting and co-taught, with James Huang, a graduate course in investments. Paul Taylor and I graded the 60 or so accounting master’s theses, and I was asked to mentor James Huang and another lecturer by the name of Wendy Beekes. James is a finance theorist who tends to work alone, whereas Wendy is like me in that she prefers to work with others. Wendy was interested in corporate governance. While I had not researched corporate governance questions to that point, I was willing to try to get up to speed. So we agreed to collaborate. We chatted about possibilities, and decided to investigate whether the professional pundits were correct when they asserted that “better”-governed Australian firms are more forthcoming in their disclosures and evenhanded in their disclosure of good and bad news. The Australian market is unique among public equity markets. Traditionally the law has given backing to the Australian Securities Exchange (ASX; formerly the Australian Stock Exchange), which from 1987 to 2011was the sole, major stock exchange in Australia. Companies listed on the ASX are required by law to keep the market informed about market sensitive events

60

(*Brown, Taylor, and Walter 1999); and market sensitive announcements must be provided first to the ASX, which then releases them to subscribers, the market, and the public at large. Jim Psaros (University of Newcastle) generously made available to Wendy and me his ratings (published in association with the Horwath accounting firm) of the 2001 governance practices of the largest 250 Australian companies. The corporate governance ratings were supplemented with analysts’ EPS forecasts from I/B/E/S and daily stock returns from SIRCA. We confirmed that the pundits were correct: better-governed firms, as rated by Jim and his colleague Michael Seamer, were indeed associated with more frequent disclosures, analysts’ forecasts were more accurate for those firms, and price discovery was both faster and more evenly balanced between good and bad news. To demonstrate the last result, we harked back to the timeliness ideas in Ball and Brown (1968, pp.174–176). Wendy presented our results at a conference sponsored by the Journal of Business Finance and Accounting in the summer of 2005, and the paper was published in the conference issue (*Beekes and Brown 2006). That work has grown into an international comparative study of corporate governance outcomes in 24 countries and is funded by grants from the Australian Research Council and the Leverhulme Trust. In 2010 the Accounting and Finance Association of Australia and New Zealand (AFAANZ) had its fiftieth anniversary, and a special edition of Accounting and Finance was published in 2011 to celebrate the occasion. I was asked to write a survey article and corporate

61

governance was mentioned as a possibility. I knew there were far too many “works” on corporate governance for one person to survey, so I enlisted the aid of Wendy Beekes and Peter Verhoeven. Both are well-organised, hard working, and knowledgeable on governance matters. (Peter is an econometrician, with interests in finance, who worked with me at UWA for a while.) The answer to the question, “How many works are there on corporate governance?” is at least 50,000, as we soon found out when we trolled electronically through the literature. We found the number of published articles and working papers is still increasing at a rate well in excess of 500 a year. We also confirmed some academics have no idea how to write a useful abstract (seeming to think abstracts should arouse the reader’s curiosity but not say what was found); “corporate governance” can mean substantially different things in different academic settings; and there is a major disconnect between the issues which many researchers consider important and matters the business and financial community have had in mind when complaining about “poor quality” corporate governance. We did not cite the 50,000 works in our 77-page review (Brown, Beekes, and Verhoeven 2011), but we did get up to about 400! Return to UNSW Edith and I may well have stayed longer in the United Kingdom, but looking for a job there was not on the agenda. We both wished to return to Australia to be closer to our son, Matthew, who lived in Western

62

Australia at the time, and our daughter, Amanda, in Sydney. So we returned to Perth. My former colleagues wanted me to rejoin UWA, but the terms of my 2001 separation agreement precluded me from doing so until January 2004. Terry Walter, who had expressed an interest in me joining his department at some stage, had moved to UNSW in 2002, where he was a professor and head of the School of Banking and Finance. I was offered and accepted a half-time appointment at UNSW, dividing my time equally between Terry’s school and the School of Accounting, where Wai-Fong Chua was head. Over the last nine years I have had the good fortune to have worked on papers with a number of colleagues at UNSW, including Gerry Gallery and “the two Andrews” (Andrew Ferguson and Andrew Jackson) in the School of Accounting, and Sian Owen and Terry Walter in the School of Banking and Finance. In late 2004 Sian, Terry and I received a sizeable grant over three years (2005–2007) from the Australian Research Council to study the equity issue anomaly, whereby firms that grow by issuing equity seem to under-perform their peers over the next three to five years. Our collective interest in that particular anomaly has led to a number of published papers; others are still in progress. For example, Kate Stone, an honours student in the School of Accounting, researched share purchase plans (SPPs), which have become a popular form of raising additional equity especially from small shareholders. She found that firms that engaged in SPPs also substantially underperformed the market after the

63

issue. Andrew Ferguson, now at UTS, and I co-supervised Kate’s thesis and put together two papers from it, one designed for finance professionals (Brown, Ferguson, and Stone 2006) and the other for an academic audience (Brown, Ferguson, and Stone 2008). In a related paper, Gerry Gallery (formerly in the School of Accounting at UNSW and now at Queensland University of Technology) and I supervised an honours thesis by Olivia Goei, who investigated whether firms that grew via SEOs appeared to have issued the additional shares at a time when their intrinsic value, as indicated by an Ohlson-type model fitted to analysts’ earnings forecasts, was substantially less than the stock’s market price. If so, then perhaps subsequent corrections to their stock price to more closely approximate intrinsic value could account for at least some of the subsequent underperformance. This too led to an academic paper (*Brown, Gallery, and Goei 2006) in which we confirmed that mispricing, in the form of a higher ratio of market price to intrinsic value, could explain a significant part of the subsequent underperformance of SEO firms. For almost 40 years, readers of the (Australian) Financial Review were entertained by a regular column written by a shrewd observer and critic of questionable corporate behaviour, especially by the directors of exploration companies traded in the Australian share market. Trevor Sykes, the columnist, wrote under the pseudonym Pierpont, and was soon to retire. It seemed a shame that such a popular columnist should be allowed to go into retirement without any remark by the

64

academic community. So I joined “the two Andrews” to document Pierpont’s track record, using methods that paralleled George Foster’s two studies of Abraham Briloff published in the Journal of Accounting Research in 1979 and the Journal of Accounting, Auditing and Finance in 1987. Our award-winning paper (Brown, Ferguson, and Jackson 2009) established that Pierpont’s columns contained superior investment analysis as well as being “a great read”. Options Although I had taught investments at undergraduate and graduate levels since the early 1970s, I had not considered all that closely the issues which options of various kinds posed for accountants. Reading about a special options deal done between two of the then giants of the Australian corporate world, Robert Holmes à Court and John Spalvins, was a wake-up call. Ian Dunlop, a UWA colleague, was equally impressed by the way Holmes à Court and Spalvins had kept one of Australia’s largest industrial companies in the dark about Holmes à Court’s growing stake in the company. The two businessmen had entered into simultaneous put and call contracts that had the effect of delivering 99.25 million BHP shares supposedly controlled by Spalvins into the hands of Holmes à Court, whose ambition seemed to be to acquire control of the industrial giant. There were two pairs of matching put and call options, described as the 1985 put and call and the 1986 put and call agreements, but only USD2,000 changed hands and in both directions. We argued (*Brown and Dunlop 1991) that the “economic substance” of the option

65

agreements indicated Australian accounting standards were deficient if the treatment was indeed permitted under those standards. Each transaction separately was worth of the order of AUD50 million. The USD2,000 consideration made no commercial sense unless it was seen simply as a contra between two pairs of offsetting transactions and was used to establish the legality of the contracts and to minimise any transaction costs. Had the contracts been combined and recognized in the accounts as a sale, the financial performance and leverage ratios of Holmes à Court’s Bell Group and Spalvins’ Adelaide Steamship group would have been very different. In the 1990s executive share options became common in Australia, as in other countries, the argument of convenience being that they aligned the interests of managers and shareholders. I was impressed by the fact that few accounting practitioners and directors I knew had much of a clue about how to value them even with a simple procedure such as the Black-Scholes formula. That meant many self-interested executives must have considered shareholders were “ripe for the picking”. So I again teamed up with Ian Dunlop and wrote a small paper containing four case studies, to illustrate how the Black-Scholes formula could be used to value the ESOs, and published it in the Australian Accounting Review (Brown and Dunlop 1994). That same issue contained a second paper that concentrated on accounting issues raised by employee stock options (Brown and Howieson 1994). Eight years later I returned to the topic of ESOs, writing a paper with a UWA honours student, Elissa Yew (Brown and Yew 2002), whose thesis I had

66

supervised before leaving UWA at the end of 2001. We concluded that the Australian standard setter’s proposal at the time, whereby options’ fair value would be recognized and immediately expensed on the vesting date, was inconsistent with the financial market’s perception that many ESOs have significant value on their vesting date. That was not quite the end of the options matter for me. Share-based compensation became a hot issue for accounting standard setters in various countries, including the United Kingdom. While I was visiting Lancaster in 2002, I helped Peter Pope plan a symposium on ESO accounting and valuation issues associated with the IASB’s ED2 (Share-based Payment), which was held in London on 16–17 January 2003. Later, I instigated a symposium on valuing employee share options, held by FIRN (Financial Integrity Research Network) in Sydney on 13 February 2006. The year before the FIRN symposium, Alex Szimayer, a statistician with a good sense of humour and a keen interest in valuing derivatives, and I had co-supervised three gifted UWA honours students, each of whom had researched a different aspect of the valuation of ESOs. Their work was presented at the FIRN symposium, and one publication, on the determinants of early exercise of ESOs in Australia, was the outcome (Boyd, Brown, and Szimayer 2007). Valuing ESOs can be a lot more complicated than many realise. Executives have a strong personal interest in understanding them and in the opportunities available to them, sometimes because of the ignorance of others, to

67

increase their wealth through seemingly innocuous variations in the issue terms. An example is the use of performance hurdles, whereby options do not vest unless some prespecified conditions are met. For instance it might be the case that a particular tranche of ESOs granted to the CEO does not vest unless the firm outperforms a comparator group of companies either in accounting or market terms. There are many possibilities, and by carefully crafting a performance condition it can be rendered of next to no effect. In other cases, the performance condition might have real bite, perhaps halving the options’ value relative to that indicated by the Black-Scholes model. The problem of estimating an option’s value as perceived by knowledgeable investment analysts (who presumably are price setters) should trouble those who research accounting issues from a capital markets perspective if their sample contains a significant number of ESOs subject to performance conditions. The values placed on those options may be biased, perhaps seriously, making any conclusions suspect at best. Alex Szimayer and I looked into this possibility and two other matters affecting options’ fair values (early exercise and stochastic volatility). The results were published, along with two high-profile Australian case studies where we applied the valuation methods. Alex now has a chair at Bonn University, hopefully gained at least in part as a result of our paper (*Brown and Szimayer 2008), which won the Peter Brownell Prize for the best paper published in Accounting and Finance that year. Reflections

68

From time to time, we academics get opportunities to reflect on the state of knowledge generally. My first real opportunity came at the Journal of Accounting Research conference held to celebrate the twentieth anniversary of the Ball and Brown and Beaver 1968 papers, which I mentioned earlier. A second opportunity was the Mathews Committee in 1989, also mentioned earlier; a third was in 1991, when I was appointed Distinguished International Visiting Lecturer by the American Accounting Association; and a fourth was when, in 1991/92, I was the inaugural Coopers & Lybrand-Accounting Association of Australia and New Zealand Visiting Research Professor in Australasia. As C&L-AAANZ Professor I was expected to write a monograph on research methods in financial accounting (Brown 1994), which has since been translated into Chinese, Japanese, and Korean. Plenary addresses are another good time for reflection. On one such occasion Bryan Howieson and I gave a joint address to the 1997 annual conference held by AAANZ (*Brown and Howieson 1998). In that paper, Bryan and I set out a research agenda for those with an interest in connecting capital markets-based research with the needs of standards setters. On another occasion I focused on connections between financial markets and financial accounting, which was published in China Accounting and Finance Review (Brown 1999). In that paper I spelt out my view of the role of accounting standards in mitigating agency costs, and have reiterated that view on more than one occasion since then (e.g. *Brown and Tarca 2001).

69

Incidentally, I first met Bryan Howieson as an undergraduate student in an investments class I taught at UWA. I did not spare the students “the Greeks” (characters used in finance), despite the fact that many among the approximately 200 attending my lectures had not progressed beyond first year undergraduate mathematics. Bryan, who had some difficulty deciphering my squiggles, was the student with the opera glasses. I never did find out just how close he got to those symbols, figuratively speaking, although I suspect it was not all that close because Bryan is “more a poet than an engineer” in MBA speak. I have always greatly enjoyed Bryan’s company and we remain very good friends, despite the shaky beginning. I was quite taken by his remarks about me when I was inducted into the Australian Accounting Hall of Fame in 2010. My story would not be complete if I did not mention Steve Zeff. I met Steve in the summer quarter of 1966, when he was a visiting associate professor teaching a financial accounting course at the University of Chicago. Steve was gathering data on Henry Rand Hatfield’s years at the University of Chicago from 1892 to 1904, when he left for Berkeley. To cut a long story short, Steve beat me hands down at paddle ball but I did not hold that against him as on other occasions we just kicked around a soccer ball on Stagg Field and had a good chat. When Steve was president of the American Accounting Association he introduced the Seminal Contribution to the Accounting Literature Award, to recognize works, published 15 or more years previously, that had stood the test of time. I was humbled, as was Ray, to be the inaugural recipients of the award, in 1986, for our 1968

70

paper. Much later, Steve recommended that I be invited to synthesise the literature on the benefits of IFRS for presentation to the Information For Better Markets conference sponsored by the Institute of Chartered Accountants in England and Wales and held in London in December 2010, a task which I greatly enjoyed. The published version (*Brown 2011) acknowledges that the literature is voluminous but it is still far from complete. My final reflection, insofar as this book is concerned, is on my own career as an accounting and finance academic. Making a mark on the research literature at a young age has allowed me the luxury of pursuing diverse questions simply because I found them interesting. At the same time, I have tried not to lose sight of questions important to others. It has been a privilege to work with many wonderful students and great colleagues, and I owe every one of them a big “thank you” for the opportunities I have been given. Looking back over more than 50 years I admit it has been a long journey, with many sometimes unexpected twists and turns. It reminds me of a saying attributed to Yogi Berra, which is the title of his book (ISBN 0-7868-6775-2): “When you come to a fork in the road, take it!” I feel as if I have done just that: on many occasions I have chosen a path without knowing quite where it leads. Undoubtedly serendipity has played a big part in my career, but at the same time I can see my life has not been a random walk. Well, not entirely. Philip Brown October, 2011

71

Part I Returns and Earnings

72

An Empirical Evaluation of Accounting Income Numbers Ray Ball* and Philip Brown† Accounting theorists have generally evaluated the usefulness of accounting practices by the extent of their agreement with a particular analytic model. The model may consist of only a few assertions or it may be a rigorously developed argument. In each case, the method of evaluation has been to compare existing practices with the more preferable practices implied by the model or with some standard which the model implies all practices should possess. The shortcoming of this method is that it ignores a significant source of knowledge of the world, namely, the extent to which the predictions of the model conform to observed behavior. It is not enough to defend an analytical inquiry on the basis that its assumptions are empirically supportable, for how is one to know that a theory embraces all of the relevant supportable assumptions? And how does one explain the predictive powers of propositions which are based on unverifiable assumptions such as the maximization of utility functions? Further, how is one to resolve differences between propositions which arise from considering different aspects of the world? The limitations of a completely analytical approach to usefulness are illustrated by the argument that income numbers cannot be defined substantively, that they lack “meaning” and are therefore of doubtful utility.1 The argument stems in part from the patchwork development of accounting 73

practices to meet new situations as they arise. Accountants have had to deal with consolidations, leases, mergers, research and development, price-level changes, and taxation charges, to name just a few problem areas. Because accounting lacks an all-embracing theoretical framework, dissimilarities in practices have evolved. As a consequence, net income is an aggregate of components which are not homogeneous. It is thus alleged to be a “meaningless” figure, not unlike the difference between twenty-seven tables and eight chairs. Under this view, net income can be defined only as the result of the application of a set of procedures {X1, X2, … } to a set of events {Y1, Y2, … } with no other definitive substantive meaning at all. Canning observes: What is set out as a measure of net income can never be supposed to be a fact in any sense at all except that it is the figure that results when the accountant has finished applying the procedures which he adopts.2 The value of analytical attempts to develop measurements capable of definitive interpretation is not at issue. What is at issue is the fact that an analytical model does not itself assess the significance of departures from its implied measurements. Hence it is dangerous to conclude, in the absence of further empirical testing, that a lack of substantive meaning implies a lack of utility. An empirical evaluation of accounting income numbers requires agreement as to what real-world outcome constitutes an appropriate test of usefulness. Because net

74

income is a number of particular interest to investors, the outcome we use as a predictive criterion is the investment decision as it is reflected in security prices.3 Both the content and the timing of existing annual net income numbers will be evaluated since usefulness could be impaired by deficiencies in either. An Empirical Test Recent developments in capital theory provide justification for selecting the behavior of security prices as an operational test of usefulness. An impressive body of theory supports the proposition that capital markets are both efficient and unbiased in that if information is useful in forming capital asset prices, then the market will adjust asset prices to that information quickly and without leaving any opportunity for further abnormal gain.4 If, as the evidence indicates, security prices do in fact adjust rapidly to new information as it becomes available, then changes in security prices will refleet the flow of information to the market.5 An observed revision of stock prices associated with the release of the income report would thus provide evidence that the information reflected in income numbers is useful. Our method of relating accounting income to stock prices builds on this theory and evidence by focusing on the information which is unique to a particular firm.6 Specifically, we construct two alternative models of what the market expects income to be and then investigate the market’s reactions when its expectations prove false. Expected and Unexpected Income Changes

75

Historically, the incomes of firms have tended to move together. One study found that about half of the variability in the level of an average firm’s earnings per share (EPS) could be associated with economy-wide effects.7 In fight of this evidence, at least part of the change in a firm’s income from one year to the next is to be expected. If, in prior years, the income of a firm has been related to the incomes of other firms in a particular way, then knowledge of that past relation, together with a knowledge of the incomes of those other firms for the present year, yields a conditional expectation for the present income of the firm. Thus, apart from confirmation effects, the amount of new information conveyed by the present income number can be approximated by the difference between the actual change in income and its conditional expectation. But not all of this difference is necessarily new information. Some changes in income result from financing and other policy decisions made by the firm. We assume that, to a first approximation, such changes are reflected in the average change in income through time. Since the impacts of these two components of change—economy-wide and policy effects—are felt simultaneously, the relationship must be estimated jointly. The statistical specification we adopt is first to estimate, by Ordinary Least Squares (OLS), the coefficients (a1jt, a2jt) from the linear regression of the change in firm j’s income (∆Ij,t−τ) on the change in the average income of all firms (other than firm j) in the

76

market (∆Mj,t−τ)8 using data up to the end of the previous year (τ = 1,2, … , t − 1) :

where the hats denote estimates. The expected income change for firm j in year t is then given by the regression prediction using the change in the average income for the market in year t:

The unexpected income change, or forecast error (ûjt), is the actual income change minus expected:

It is this forecast error which we assume to be the new information conveyed by the present income number The Market’s Reaction It has also been demonstrated that stock prices, and therefore rates of return from holding stocks, tend to move together. In one study,9 it was estimated that about 30 to 40 per cent of the variability in a stock’s monthly rate of return over the period March, 1944 through December, 1960 could be associated with market-wide effects. Market-wide variations in stock returns are triggered by the release of information which concerns all firms. Since we are evaluating the income report as it relates to the individual firm, its contents and timing should be assessed relative to changes in the rate of return on the firm’s stocks net of market-wide effects.

77

The impact of market-wide information on the monthly rate of return from investing one dollar in the stock of firm j may be estimated by its predicted value from the linear regression of the monthly price relatives of firm j’s common stock10 on a market index of returns:11

where PRjm is the monthly price relative for firm j and month m, L is the link relative of Fisher’s “Combination Investment Performance Index” [Fisher (1966)], and vjm is the stock return residual for firm j in month m. The value of [Lm − 1] is an estimate of the market’s monthly rate of return. The m-subscript in our sample assumes values for all months since January, 1946 for which data are available. The residual from the OLS regression represented in equation (3) measures the extent to which the realized return differs from the expected return conditional upon the estimated regression parameters (b1j, b2j) and the market index [Lm − 1]. Thus, since the market has been found to adjust quickly and efficiently to new information, the residual must represent the impact of new information, about firm j alone, on the return from holding common stock in firm j. Some Econometric Issues One assumption of the OLS income regression model12 is that Mj and uj are uncorrelated. Correlation between them can take at least two forms, namely the inclusion of firm j in the market index of income (Mj), and the presence of industry effects. The first has been eliminated by construction (denoted by the j–subscript 78

on M), but no adjustment has been made for the presence of industry effects. It has been estimated that industry effects probably account for only about 10 per cent of the variability in the level of a firm’s income.13 For this reason equation (1) has been adopted as the appropriate specification in the belief that any bias in the estimates a1jt and a2jt will not be significant. However, as a check on the statistical efficiency of the model, we also present results for an alternative, naive model which predicts that income will be the same for this year as for last. Its forecast error is simply the change in income since the previous year. As is the case with the income regression model, the stock return model, as presented, contains several obvious violations of the assumptions of the OLS regression model. First, the market index of returns is correlated with the residual because the market index contains the return on firm j, and because of industry effects. Neither violation is serious, because Fisher’s index is calculated over all stocks listed on the New York Stock Exchange (hence the return on security j is only a small part of the index), and because industry effects account for at most 10 per cent of the variability in the rate of return on the average stock.14 A second violation results from our prediction that, for certain months around the report dates, the expected values of the vj’s are nonzero. Again, any bias should have little effect on the results, inasmuch as there is a low, observed autocorrelation in the j’s,15 and in no case was the stock return regression fitted over less than 100 observations.16

79

Summary We assume that in the unlikely absence of useful information about a. particular firm over a period, its rate of return over that period would reflect only the presence of market-wide information which pertains to all firms. By abstracting from market effects [equation (3)] we identify the effect of information pertaining to individual firms. Then, to determine if part of this effect can be associated with information contained in the firm’s accounting income number, we segregate the expected and unexpected elements of income change. If the income forecast error is negative (that is, if the actual change in income is less than its conditional expectation), we define it as bad news and predict that if there is some association between accounting income numbers and stock prices, then release of the income number would result in the return on that firm’s securities being less than would otherwise have been expected.17 Such a result (û < 0) would be evidenced by negative behavior in the stock return residuals ( < 0) around the annual report announcement date. The converse should hold for a positive forecast error. Two basic income expectations models have been defined, a regression model and a naive model. We report in detail on two measures of income [net income and EPS, variables (1) and (2)] for the regression model, and one measure [EPS, variable (3)] for the naive model. Data

80

Three classes of data are of interest: the contents of income reports; the dates of the report announcements; and the movements of security prices around the announcement dates. Income Numbers Income numbers for 1946 through 1966 were obtained from Standard and Poor’s Compustat tapes.18 The distributions of the squared coefficients of correlation19 between the changes in the incomes of the individual firms and the changes in the market’s income index20 are summarized in Table 1. For the present sample, about one-fourth of the variability in the changes in the median firm’s income can be associated with changes in the market index. Table 1 Deciles of the Distributions of Squared Coefficients of Correlation, Changes in Firm and Market Income*

* Estimated over the 21 years, 1946–1966. The association between the levels of the earnings of firms was examined in the forerunner article [Ball and Brown (1967)]. At that time, we referred to the existence of autocorrelation in the disturbances when the levels of net income and EPS were regressed on the appropriate

81

indexes. In this paper, the specification has been changed from levels to first differences because our method of analyzing the stock market’s reaction to income numbers presupposes the income forecast errors to be unpredictable at a minimum of 12 months prior to the announcement dates. This supposition is inappropriate when the errors are autocorrelated. We tested the extent of autocorrelation in the residuals from the income regression model after the variables had been changed from levels to first differences. The results are presented in Table 2. They indicate that the supposition is not now unwarranted. Table 2 Deciles, of the Distributions of the Coefficients of First-Order Autocorrelation in the Income Regression Residuals*

* Estimated over the 21 years, 1946–1966. Annual Report Announcement Dates The Wall Street Journal publishes three kinds of annual report announcements: forecasts of the year’s income, as made, for example, by corporation executives shortly after the year end; preliminary reports; and the complete annual report. While forecasts are often imprecise, the preliminary report is typically a condensed preview of 82

the annual report. Because the preliminary report usually contains the same numbers for net income and EPS as are given later with the final report, the announcement date (or, effectively, the date on which the annual income number became generally available) was assumed to be the date on which the preliminary report appeared in the Wall Street Journal. Table 3 reveals that the time lag between the end of the fiscal year and the release of the annual report has been declining steadily throughout the sample period. Table 3 Time Distribution of Announcement Dates

a Indicates that 25 per cent of the income reports for the fiscal year ended 12/31/1957 had been announced by 2/07/1958. Stock Prices Stock price relatives were obtained from the tapes constructed by the Center for Research in Security Prices (CRSP) at the University of Chicago.21 The data used are monthly closing prices on the New York Stock Exchange, adjusted for dividends and capital changes, for the period January, 1946 through June, 1966. Table 4 presents the deciles of the distributions of the squared coefficient of correlation for the stock return 83

regression [equation (3)], and of the coefficient of first-order autocorrelation in the stock residuals. Table 4 Deciles of the Distributions of the Squared Coefficient of Correlation for the Stock Return Regression, and of the Coefficient of First-Order Autocorrelation in the Stock Return Residuals*

* Estimated over the 246 months, January, 1946 through June, 1966 Inclusion Criteria Firms included in the study met the following criteria: earnings data available on the Compustat tapes for each of the years 1946–1966; fiscal year ending December 31; price data available on the CRSP tapes for at least 100 months; and Wall Street Journal announcement dates available.22 Our analysis was limited to the nine fiscal years 1957–1965. By beginning the analysis with 1957, we were assured of at least 10 observations when

84

estimating the income regression equations. The upper limit (the fiscal year 1965, the results of which are announced in 1966) is imposed because the CRSP file terminated in June, 1966. Our selection criteria may reduce the generality of the results. The subpopulation does not include young firms, those which have failed, those which do not report on December 31, and those which are not represented on Compustat, the CRSP tapes, and the Wall Street Journal. As a result, it may not be representative of all firms. However, note that (1) the 261 remaining firms23 are significant in their own right, and (2) a replication of our study on a different sample produced results which conform closely to those reported below.24 Results Define month 0 as the month of the annual report announcement, and APIM, the Abnormal Performance Index at month M, as:

Then API traces out the value of one dollar invested (in equal amounts) in all securities n (n = 1,2, … , N) at the end of month −12 (that is, 12 months prior to the month of the annual report) and held to the end of some arbitrary holding period (M = −11, −10, …, T) after abstracting from market affects. An equivalent interpretation is as follows. Suppose two individuals A and B agree on the following proposition. B is to construct a portfolio consisting of one dollar invested in

85

equal amounts in N securities. The securities are to be purchased at the end of month −12 and held until the end of month T. For some price, B contracts with A to take (or make up), at the end of each month M, only the normal gains (or losses) and to return to A, at the end of month T, one dollar plus or minus any abnormal gains or losses. Then APIM is the value of A’s equity in the mutual portfolio at the end of each month M.25 Numerical results are presented in two forms. Figure 1 plots APIM first for three portfolios constructed from all firms and years in which the income forecast errors, according to each of the three variables, were positive (the top half); second, for three portfolios of firms and years in which the income forecast errors were negative (the bottom half); and third, for a single portfolio consisting of all firms and years in the sample (the line which wanders just below the line dividing the two halves). Table 5 includes the numbers on which Figure 1 is based. Table 5 Summary Statistics by Month Relative to Annual Report Announcement Date

86

a

Column headings:

(1) Abnormal Performance Index—firms and years in which the income forecast error was positive. (2) Abnormal Performance Index—firms and years in which the income forecast error was negative. (3) Chi-square statistic for two-by-two classification by sign of income forecast error (for the fiscal year) and sign of stock return residual (for the indicated month). Note: Probability (chi-square ≥ 3.84 | χ2 = 0) = .05, for 1 degree of freedom. Probability (chi-square ≥ 6.64 | χ2 = 0) = .01, for 1 degree of freedom.

87

Fig. 1 Abnormal Performance Indexes for Various Portfolios Since the first set of results may be sensitive to the distributions of the stock return disturbances,26 a second set of results is presented. The third column under each variable heading in Table 5 gives the chi-square statistic 88

for a two-by-two classification of firms by the sign of the income forecast error, and the sign of the stock return residual for that month. Overview As one would expect from a large sample, both sets of results convey essentially the same picture. They demonstrate that the information contained in the annual income number is useful in that if actual income differs from expected income, the market typically has reacted in the same direction. This contention is supported both by Figure 1 which reveals a marked, positive association between the sign of the error in forecasting income and the Abnormal Performance Index, and by the chi-square statistic (Table 5). The latter shows it is most unlikely that there is no relationship between the sign of the income forecast error and the sign of the rate of return residual in most of the months up to that of the annual report announcement. However, most of the information contained in reported income is anticipated by the market before the annual report is released. In fact, anticipation is so accurate that the actual income number does not appear to cause any unusual jumps in the Abnormal Performance Index in the announcement month. To illustrate, the drifts upward and downward begin at least 12 months before the report is released (when the portfolios are first constructed) and continue for approximately one month after. The persistence of the drifts, as indicated by the constant signs of the indexes and by their almost monotonie increases in absolute value (Figure 1), suggests not only that the market begins to anticipate

89

forecast errors early in the 12 months preceding the report, but also that it continues to do so with increasing success throughout the year.27 Specific Results 1. There appears to be little difference between the results for the two regression model variables. Table 6, which classifies the sign of one variable’s forecast error contingent upon the signs of the errors of the other two variables, reveals the reason. For example, on the 1231 occasions on which the income forecast error was positive for variable (1), it was also positive on 1148 occasions (out of a possible 1231) for variable (2). Similarly, on the 1109 occasions on which the income forecast error was negative for variable (1), it was also negative on 1026 occasions for variable (2). The fact that the results for variable (2) strictly dominate those for variable (1) suggests, however, that when the two variables disagreed on the sign of an income forecast error, variable (2) was more often correct. While there is little to choose between variables (1) and (2), variable (3) (the naive model) is clearly best for the portfolio made up of firms with negative forecast errors. A contributing factor is the following. The naive model gives the same forecast error as the regression model would give if (a) the change in market income were zero, and (b) there were no drift in the income of the firm. But historically there has been an increase in the market’s income, particularly during the latter part of the sample period, due to general increase in prices and the strong influence of the protracted expansion since 1961. Thus, the naive

90

model [variable (3)] typically identifies as firms with negative forecast errors those relatively few firms which showed a decrease in EPS when most firms showed an increase. Of the three variables, one would be most confident that the incomes of those which showed negative forecast errors for variable (3) have in fact lost ground relative to the market. This observation has interesting implications. For example, it points to a relationship between the magnitudes of the income forecast errors and the magnitudes of the abnormal stock price adjustments. This conclusion is reinforced by Figure 1 which shows that the results for positive forecast errors are weaker for variable (3) than for the other two. 2. The drift downward in the Abnormal Performance Index computed over all firms and years in the sample reflects a computational bias.28 The bias arises because

where E denotes the expected value. It can readily be seen that the bias over K months is at least of order (K − 1) times the covariance between υm and, υm−129 Since this covariance is typically negative,30 the bias is also negative. While the bias does not affect the tenor of our results in any way, it should be kept in mind when interpreting the values of the various API’s. It helps explain, for example, why the absolute changes in the indexes in the bottom panel of Figure 1 tend to be greater than those in the top panel; why the indexes in the top panel tend to 91

turn down shortly after month 0; and finally, why the drifts in the indexes in the bottom panel tend to persist beyond the month of the report announcement. 3. We also computed results for the regression model using the additional definitions of income: a. cash flow, as approximated by operating income,31 and b. net income before nonrecurring items. Neither variable was as successful in predicting the signs of the stock return residuals as net income and EPS. For example, by month 0, the Abnormal Performance Indexes for forecast errors which were positive were 1.068 (net income, including nonrecurring items) and 1.070 (operating income). These numbers compare with 1.071 for net income [Table 5, variable (1)]. The respective numbers for firms and years with negative forecast errors were 0.911, 0.917, and 0.907. 4. Both the API’s and the chi-square test in Table 5 suggest that, at least for variable (3), the relationship between the sign of the income forecast error and that of the stock return residual may have persisted for as long as two months beyond the month of the announcement of the annual report. One explanation might be that the market’s index of income was not known for sure until after several firms had announced their income numbers. The elimination of uncertainty about the market’s income subsequent to some firms’ announcements might tend, when averaged over all firms in the sample, to be reflected in a persistence in the drifts in the API’s 92

beyond the announcement month. This explanation can probably be ruled out, however, since when those firms which made their announcements in January of any one year were excluded from the sample for that year, there were no changes in the patterns of the overall API’s as presented in Figure 1, although generally there were reductions in the χ2 statistics.32 Table 6 Contingency Table of the Signs of the Income Forecast Errors—by Variable

A second explanation could be random errors in the announcement dates. Drifts in the API’s would persist beyond the announcement month if errors resulted in our treating some firms as if they had announced their income numbers earlier than in fact was the case. But this explanation can also probably be ruled out, since all announcement dates taken from the Wall Street Journal Index were verified against the Wall Street Journal. A third explanation could be that preliminary reports are not perceived by the market as being final. Unfortunately 93

this issue cannot be resolved independently of an alternative hypothesis, namely that the market does take more time to adjust to information if the value of that information is less than the transactions costs that would be incurred by an investor who wished to take advantage of the opportunity for abnormal gain. That is, even if the relationship tended to persist beyond the announcement month, it is clear that unless transactions costs were within about one per cent,33 there was no opportunity for abnormal profit once the income information had become generally available. Our results are thus consistent with other evidence that the market tends to react to data without bias, at least to within transactions costs. The Value of Annual Net Income Relative to Other Sources of Information34 The results demonstrate that the information contained in the annual income number is useful in that it is related to stock prices. But annual accounting reports are only one of the many sources of information available to investors. The aim of this section is to assess the relative importance of information contained in net income, and at the same time to provide some insight into the timeliness of the income report. It was suggested earlier that the impact of new information about an individual stock could be measured by the stock’s return residual. For example, a negative residual would indicate that the actual return is less than what would have been expected had there been no bad information. Equivalently, if an investor is able to take advantage of the information either by selling or by

94

taking a short position in advance of the market adjustment, then the residual will represent, ignoring transactions costs, the extent to which his return is greater than would normally be expected. If the difference between the realized and expected return is accepted as also indicating the value of new information, then it is clear that the value of new, monthly information, good or bad, about an individual stock is given by the absolute value of that stock’s return residual for the given month. It follows that the value of all monthly information concerning the average firm, received in the 12 months preceding the report, is given by:

where TI denotes total information.35 For our sample, averaged over all firms and years, this sum was 0.731. For any one particular stock, some of the information between months will be offsetting.36 The value of net information (received in the 12 months preceding the report) about the average stock is given by:

where NI denotes net information. This sum was 0.165. The impact of the annual income number is also a net number in that net income is the result of both income-increasing and income-decreasing events. If one

95

accepts the forecast error model,37 then the value of information contained in the annual income number may be estimated by the average of the value increments from month −11 to month 0, where the increments are averaged over the two portfolios constructed from (buying or selling short) all firms and years as classified by the signs of the income forecast errors. That is,

where II denotes income information, and N1 and N2 the number of occasions on which the income forecast error was positive and negative respectively. This number was 0.081 for variable (1), 0.083 for variable (2), and 0.077 for variable (3). From the above numbers we conclude: (1) about 75 per cent [(.731 − .165)/.731] of the value of all information appears to be offsetting, which in turn implies that about 25 per cent persists; and (2) of the 25 per cent which persists, about half [49 %, 50 %, and 47 %—calculated as .081/.165, .083/.165, and .077/.165—for variables (1)-(3)] can be associated with the information contained in reported income. Two further conclusions, not directly evident, are: (3) of the value of information contained in reported income, no more than about 10 to 15 per cent (12 %, 11 %, and 13 %) has not been anticipated by the month of the report;38 and (4) 96

the value of information conveyed by the income number at the time of its release constitutes, on average, only 20 per cent (19 %, 18 %, and 19 %) of the value of all information coming to the market in that month.39 The second conclusion indicates that accounting income numbers capture about half of the net effect of all information available throughout the 12 months preceding their release; yet the fourth conclusion suggests that net income contributes only about 20 per cent of the value of all information in the month of its release. The apparent paradox is presumably due to the fact that: (a) many other bits of information are usually released in the same month as reported income (for example, via dividend announcements, or perhaps other items in the financial reports); (b) 85 to 90 per cent of the net effect of information about annual income is already reflected in security prices by the month of its announcement; and (c) the period of the annual report is already one-and-one-half months into history. Ours is perhaps the first attempt to assess empirically the relative importance of the annual income number, but it does have limitations. For example, our results are systematically biased against findings in favor of accounting reports due to: 1. the assumption that stock prices are from transactions which have taken place simultaneously at the end of the month; 2. the assumption that there are no errors in the data; 3. the discrete nature of stock price quotations; 4. the presumed validity of the “errors in forecast” model; and

97

5. the regression estimates of the income forecast errors being random variables, which implies that some misclassifications of the “true” earnings forecast errors are inevitable Concluding Remarks The initial objective was to assess the usefulness of existing accounting income numbers by examining their information content and timeliness. The mode of analysis permitted some definite conclusions which we shall briefly restate. Of all the information about an individual firm which becomes available during a year, one-half or more is captured in that year’s income number. Its content is therefore considerable. However, the annual income report does not rate highly as a timely medium, since most of its content (about 85 to 90 per cent) is captured by more prompt media which perhaps include interim reports. Since the efficiency of the capital market is largely determined by the adequacy of its data sources, we do not find it disconcerting that the market has turned to other sources which can be acted upon more promptly than annual net income. This study raises several issues for further investigation. For example, there remains the task of identifying the media by which the market is able to anticipate net income: of what help are interim reports and dividend announcements? For accountants, there is the problem of assessing the cost of preparing annual income reports relative to that of the more timely interim reports.

98

The relationship between the magnitude (and not merely the sign) of the unexpected income change and the associated stock price adjustment could also be investigated.40 This would offer a different way of measuring the value of information about income changes, and might, in addition, furnish insight into the statistical nature of the income process, a process, little understood but of considerable interest to accounting researchers. Finally, a mechanism has been provided for an empirical approach to a restricted class of the controversial choices in external reporting. References BALL, RAY AND PHILIP BROWN (1967). “Some Preliminary Findings on the Association between the Earnings of a Firm, Its Industry and the Economy,” Empirical Research in Accounting: Selected Studies, 1967, Supplement to Volume 5 of the Journal of Accounting Research, pp. 55–77. BEAVER, WILLIAM H. (1968). “The Information Content of Annual Earnings Announcements,” forthcoming in Empirical Research in Accounting: Selected Studies 1968, Supplement to Volume 6 of the Journal of Accounting Research. BLUME, MARSHALL E. (1968). “The Assessment of Portfolio Performance” (unpublished Ph.D. dissertation, University of Chicago). BREALEY, RICHARD A. (1968). “The Influence of the Economy on the Earnings of the Firm” (unpublished

99

paper presented at the Sloane School of Finance Seminar, Massachusetts Institute of Technology, May, 1968). BROWN, PHILIP AND VICTOR NIEDERHOFFER (1968). “The Predictive Content of Quarterly Earnings,” Journal of Business. CANNING, JOHN B. (1929). The Economics Accountancy (New York: The Ronald Press Co.).

of

CHAMBERS, RAYMOND J. (1964). “Measurement and Objectivity in Accounting,” The Accounting Review, XXXIX (April, 1964), 264–274. CHAMBERS, RAYMOND J. (1966). Accounting, Evaluation, and Economic Behavior (Englewood Cliffs, N.J.: Prentice-Hall). CHAMBERS, RAYMOND J. (1967). “Continuously Contemporary Accounting—Additivity and Action,” The Accounting Review, XLII (October, 1967), 751–757. COOTNER, PAUL H., ed. (1964). The Random, Character of Stock Market Prices (Cambridge, Mass.: The M.I.T. Press). EDWABBS, EDGAR O. AND PHILIP W. BELL (1961). The Theory and Measurement of Business Income (Berkeley, Cal.: The University of California Press). FAMA, EUGENE F. (1965). “The Behavior of Stock Market Prices,” Journal of Business, XXXVIII (January, 1965), 34–105. FAMA, EUGENE F., AND MARSHALL E. BLUME (1966). “Filter Rules and Stock Market Trading,” Journal of 100

Business, 226–241.

XXXIX

(Supplement,

January,

1966),

FAMA, EUGENE F., LAWRENCE FISHER, MICHAEL C. JENSEN, AND RICHARD ROLL (1967). “The Adjustment of Stock Prices to New Information,” Report No. 6715 (University of Chicago: Center for Mathematical Studies in Business and Economics; forthcoming in the International Economic Review). FISHER, LAWRENCE (1966). “Some New Stock Market Indices,” Journal of Business, XXXIX (Supplement, January, 1966), 191–225. GILMAN, STEPHAN (1939). Accounting Concepts of Profit (New York: The Ronald Press Co.). IJIRI, YUJI (1967). The Foundations of Accounting Measurement (Englewood Cliffs, N.J.: Prentice-Hall). JENSEN, MICHAEL C. (1968). “Risk, the Pricing of Capital Assets, and the Evaluation of Investment Portfolios” (unpublished Ph.D. dissertation, University of Chicago). KING, BENJAMIN F. (1966). “Market and Industry Factors in Stock Price behavior,” Journal of Business, XXXIX (Supplement, January, 1966), 139–190. LIM, RONALD S. (1966). “The Mathematical Propriety of Accounting Measurements and Calculations,” The Accounting Review, XLI (October, 1966), 642–651. PATON, W. A., AND A. C. LITTLETON (1940). An Introduction to Corporate Accounting Standards (American Accounting Association Monograph No. 3).

101

SAMUELSON, PAUL A. (1965). “Proof That Properly Anticipated Prices Fluctuate Randomly,” Industrial Management Review, 7 (Spring, 1965), 41–49. SCHOLES, MYRON J. (1967). “The Effect of Secondary Distributions on Price” (unpublished paper presented at the Seminar on the Analysis of Security Prices, University of Chicago). SHARPE, WILLIAM F. (1964). “Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk,” Journal of Finance, XIX (September, 1964), 425–442. STERLING, ROBERT R. (1967). “Elements of Pure Accounting Theory,” The Accounting Review, XLII (January, 1967), 62–73. VATTER, WILLIAM J. (1947). The Fund Theory of Accounting (Chicago: The University of Chicago Press).

* University of Chicago. † University of Western Australia. The authors are indebted to the participants in the Workshop in Accounting Research at the University of Chicago, Professor Myron Scholes, and Messrs. Owen Hewett and Ian Watts. 1

Versions of this particular argument appear in Canning (1929); Gilman (1939); Paton and Littleton (1940); Vatter (1947), Ch. 2; Edwards and Bell (1961), Ch. 1; Chambers (1964), pp. 267–68; Chambers (1966), pp. 4 and 102; Lim (1966), esp. pp. 645 and 649; Chambers 102

(1967), pp. 745–55; Ijiri (1967), Ch. 6, esp. pp. 120–31; and Sterling (1967), p. 65. 2

Canning (1929), p. 98.

3

Another approach pursued by Beaver (1968) is to use the investment decision, as it is reflected in transactions volume, for a predictive criterion. 4

For example, Samuelson (1965) demonstrated that a market without bias in its evaluation of information will give rise to randomly fluctuating time series of prices. See also Cootner (ed.) (1964); Fama (1965); Fama and Blume (1966); Fama, et al. (1967); and Jensen (1968). 5

One well documented characteristic of the security market is that useful sources of information are acted upon and useless sources are ignored. This is hardly surprising since the market consists of a large number of competing actors who can gain from acting upon better interpretations of the future than those of their rivals. See, for example, Scholes (1967); and footnote 4 above. This evaluation of the security market differs sharply from that of Chambers (1966, pp. 272–73). 6

More precisely, we focus on information not common to all firms, since some industry effects are not considered in this paper. 7

Alternatively, 35 to 40 per cent could be associated with effects common to all firms when income was defined as tax-adjusted Return on Capital Employed. [Source: Ball and Brown (1967), Table 4.]

103

8

We call M a “market index” of income because it is constructed only from firms traded on the New York Stock Exchange. 9

King (1966).

10

The monthly price relative of security j for month m is defined as dividends (djm) + closing price (pi, m + 1), divided by opening price (pjm):

A monthly price relative is thus equal to the discrete monthly rate of return plus unity; its natural logarithm is the monthly rate of return compounded continuously. In this paper, we assume discrete compounding since the results are easier to interpret in that form 11

Fama, et al. (1967) conclude that “regressions of security on market returns over time are a satisfactory method for abstracting from the effects of general market conditions on the monthly rates of return on individual securities.” In arriving at their conclusion, they found that “scatter diagrams for the [returns on] individual securities [vis-à-vis the market return] support very well the regression assumptions of linearity, homoscedasticity, and serial independence.” Fama, et al. studied the natural logarithmic transforms of the price relatives, as did King (1966). However, Blume (1968) worked with equation (3). We also performed tests on the alternative specification:

104

where lne denotes the natural logarithmic function. The results correspond closely with those reported below. 12

That is, an assumption necessary for OLS to be the minimum-variance, linear, unbiased estimator. 13

The magnitude assigned to industry effects depends upon how broadly an industry is defined, which in turn depends upon the particular empirical application being considered. The estimate of 10 per cent is based on a two–digit classification scheme. There is some evidence that industry effects might account for more than 10 per cent when the association is estimated in first differences [Brealey (1968)]. 14

The estimate of 10 per cent is due to King (1966). Blume (1968) has recently questioned the magnitude of industry effects, suggesting that they could be somewhat less than 10 per cent. His contention is based on the observation that the significance attached to industry effects depends on the assumptions made about the parameters of the distributions underlying stock rates of return. 15

See Table 4, below.

16

Fama, et al. (1967) faced a similar situation. The expected values of the stock return residuals were nonzero for some of the months in their study. Stock return regressions were calculated separately for both exclusion and inclusion of the months for which the stock return residuals were thought to be nonzero. They report that both sets of results support the same conclusions.

105

An alternative to constraining the mean vj to be zero is to employ the Sharpe Capital Asset Pricing Model [Sharpe (1964)] to estimate (3b):

where RF is the risk-free ex ante rate of return for holding period m. Results from estimating (3b) (using U.S. Government Bills to measure RF and defining the abnormal return for firm, j in month m now as ) are essentially the same as the results from (3). Equation (3b) is still not entirely satisfactory, however, since the mean impact of new information is estimated over the whole history of the stock, which covers at least 100 months. If (3b) were fitted using monthly data, a vector of dummy variables could be introduced to identify the fiscal year covered by the annual report, thus permitting the mean residual to vary between fiscal years. The impact of unusual information received in month m of year t would then be estimated by the sum of the constant, the dummy for year t, and the calculated residual for month m and year t. Unfortunately, the efficiency of estimating the stock return equation in this particular form has not been investigated satisfactorily, hence our report will be confined to the results from estimating (3). 17

We later divide the total return into two parts: a “normal return,” defined by the return which would have been expected given the normal relationship between a stock and the market index; and an “abnormal return,” the difference between the actual return and the normal

106

return. Formally, the two parts are given by: b1i, + b2i [Lm − 1]; and υim. 18

Tapes used are dated 9/28/1965 and 7/07/1967.

19

All correlation coefficients in this paper are product-moment correlation coefficients. 20

The market net income index was computed as the sample mean for each year. The market EPS index was computed as a weighted average over the sample members, the number of stocks outstanding (adjusted for stock splits and stock dividends) providing the weights. Note that when estimating the association between the income of a particular firm and the market, the income of that firm was excluded from the market index. 21

The Center for Research in Security Prices at the University of Chicago is sponsored by Merrill Lynch, Pierce, Fenner and Smith Incorporated. 22

Announcement dates were taken initially from the Wall Street Journal Index, then verified against the Wall Street Journal. 23

Due to known errors in the data, not all firms could be included in all years. The fiscal year most affected was 1964, when three firms were excluded. 24

The replication investigated 75 firms with fiscal years ending on dates other than December 31, using the naive income-forecasting model, over the longer period 1947–65 25

That is, the value expected at the end of month Τ in the absence of further abnormal gains and losses.

107

26

The empirical distributions of the stock return residuals appear to be described well by symmetric, stable distributions that are characterized by tails longer than those of the normal distribution [Fama (1965); Fama, et al. (1967)]. 27

Note that Figure 1 contains averages over many firms and years and is not indicative of the behavior of the securities of any particular firm in any one year. While there may be, on average, a persistent and gradual anticipation of the contents of the report throughout the year, evidence on the extent of autocorrelation in the stock return residuals would suggest that the market’s reaction to information about a particular firm tends to occur rapidly. 28

The expected, value of the bias is of order minus one-half to minus one-quarter of one per cent per annum. The difference between the observed value of the API computed over the total sample and its expectation is a property of the particular sample (see footnote 26). 29

In particular, the approximation neglects all permutations of the product vs.vt, s = 1, 2,…, K− 2, t = s + 2, …, K, as being of a second order of smallness. 30

See Table 4.

31

All variable definitions are specified in Standard and Poor’s Compustat Manual [see also Ball and Brown (1967), Appendix А]. 32

The general reduction in the χ2 statistic is due largely to the reduction in sample size.

108

33

This result is obtained as follows. The ratio APIm/APIm−1 is equal to the marginal return in month m plus unity:

Similarly,

and, in general,

Thus, the marginal return for the two months after the announcement date on the portfolio consisting of firms for which EPS decrease would have been 0.878/0.887 − 1 ≅ −.010; similarly, the marginal return on the portfolio of firms for which EPS increased would have been 1.059/1.056 − 1 ≅ .003. After allowing for the computational bias, it would appear that transactions costs must have been within one per cent for opportunities to have existed for abnormal profit from applying some mechanical trading rule. 34

This analysis does not consider the marginal contribution of information contained in the annual income number. It would be interesting to analyze dividends in a way similar to that we have used for income announcements. We expect there would be some overlap. To the extent that there is an overlap, we

109

attribute the information to the income number and consider the dividend announcement to be the medium by which the market learns about income. This assumption is highly artificial in that historical income numbers and dividend payments might both simply be reflections of the same, more fundamental informational determinants of stock prices. 35

Note that the information is reflected in a value increment; thus, the original $1.00 is deducted from the terminal value. 36

This assertion is supported by the observed low autocorrelation in the stock return residuals. 37

Note that since we are interested in the “average firm,” an investment strategy must be adopted on every sample member. Because there are only two relevant strategies involved, it is sufficient to know whether one is better off to buy or to sell short. Note also that the analysis assumes the strategy is first adopted 12 months prior to the announcement date. 38

The average monthly yield from a policy of buying a portfolio consisting of all firms with positive forecast errors and adopting a short position on the rest would have resulted in an average monthly abnormal rate of return, from −11 to −1, of 0.63%, 0.66%, and 0.60% for variables (1), (2), and (3) respectively. The marginal rate of return in month 0 for that same strategy would have been 0.92%, 0.89%, and 0.94% respectively. However, relatively much more information is conveyed in the month of the report announcement than in either of the two months immediately preceding the announcement

110

month or in the two months immediately following it. This result is consistent with those obtained by Beaver (1968). 39

An optimum policy (that is, one which takes advantage of all information) would have yielded an abnormal rate of return of 4.9% in month 0. 40

There are some difficult econometric problems associated with this relationship, including specifying the appropriate functional form, the expected statistical distributions of the underlying parameters, the expected behavior of the regression residuals, and the extent and effects of measurement errors in both dependent and independent variables. (The functional form need not necessarily be linear, if only because income numbers convey information about the covariability of the income process.)

© 1968 by Institute of Professional Accounting, The University of Chicago

111

Invited Remarks Ball and Brown [1968] Philip Brown* 1. Introduction It is a great honor for me to be at this conference and speaking to you today. Nick Dopuch was not all that explicit when it came to a topic for this last session. Nick said, “Talk about anything you like.” I interpreted that to mean I should talk in general terms about my early work with Ray Ball and its influence on the accounting literature; and in particular, about a paper we wrote about 20 years ago, which is now commonly referred to simply as “Ball and Brown [1968].” I begin by covering the paper itself: the antecedent conditions, why the paper was written, what we attempted to do in the paper, and what was novel as far as we were concerned. Next, I give my view of the main strands in the accounting literature, since 1968, that are related to our work. I conclude by offering a few thoughts on directions that the capital-markets-based accounting literature might take in the years ahead. 2. Background 2.1 Antecedent Conditions Ray Ball and I are accounting honors graduates of the University of New South Wales (UNSW). Bill Stewart

112

was my mentor, and it was he who was responsible for me deciding to go to the United States to pursue a career in accounting education. Bill had been a visiting professor at the University of Chicago’s Graduate School of Business, where he had met and been greatly impressed by Sid Davidson and his colleagues. I believe I was the first of what was to become a steady procession of Australians and New Zealanders who went to the States in search of strong accounting doctoral programs, where, incidentally, many of them have excelled. I went to Chicago in 1963; but just before I left UNSW I met Ray, who was a student in a first-year accounting tutorial group I taught. At Chicago, my basic discipline was economics; and my two fields of concentration were accounting and finance. Honors students at UNSW studied the accounting classics. We had been exposed to Canning, Gilman, Hatfield, Paton and Littleton, Edwards and Bell, and Chambers, to mention a few. We were also well trained in accounting procedures. So on my arrival at Chicago I was exempted from all accounting courses other than the doctoral seminar which, in any event, was confined to ground I had already traversed. I was, however, programmed into a full complement of courses in Chicago-style economics and finance. With some notable exceptions, much of Chicago’s finance research program was built around the Center for Research into Security Prices (CRSP). Jim Lorie had arranged the funding of CRSP’s New York Stock Exchange monthly data history, which was the core of CRSP; Larry Fisher had the computer expertise to make

113

the CRSP data base a reality; Mert Miller and Gene Fama provided inspiration and real scholarly leadership; and a whole group of amazingly able students had been attracted to the finance program. Dick Roll, Mike Jensen, Marshall Blume, and Myron Scholes are a few of the better-known graduate students from those heady days. It did not take long for me to be completely seduced by the sheer vitality of the Chicago finance group which, at that time, was rapidly developing lines of research fundamentally at odds with much of the accounting literature to which I had been exposed. Chicago’s research workshops in accounting and finance were lively battlegrounds, to say the least. The second part to this “formative” story is the role of Chicago’s Economics Department. I and many of my doctoral program classmates chose Economics as our basic discipline (the alternatives were Mathematics or Behavioral Science). We then trotted off to the Economics Department where we inevitably were schooled in applied microeconomics and given a heavy dose of so-called positive economics, often taught by Milton Friedman himself. The empirical mindset was so dominant in the 1960s that it influenced almost all of the doctoral students’ choices of research topics for a generation or more. My doctoral dissertation was on Miller and Modigliani’s theoretical and empirical work on valuation. That is when I learned about Ordinary Least Squares, instrumental variables, two-and three-stage least squares, Zellner’s λ-estimators, reverse regressions, and the like.

114

Forgive me for taking this opportunity to quash a rumor that has lived too long and too dangerously. I did in fact submit my dissertation. Its title is “Some Aspects of Valuation in the Railroad Industry.” Moreover, I graduated on the strength of that dissertation, not on other papers which I wrote, sometimes with Ray, at the same time as I worked on my doctorate—papers which, admittedly, I found more interesting than a dissertation. They addressed such topics as the comparative predictive abilities of alternative income measures, market and industry effects in accounting income series, portfolio theory and accounting, and, of course, an empirical evaluation of accounting income numbers. But I am jumping ahead. Ray Ball came to Chicago in the autumn of 1966 against the advice of Bill Stewart, who, according to Ray, was not particularly pleased with the direction my own interests had taken by then. I am glad Ray came, because he completed the three C’s, which contributed to our success: Chicago, CRSP, and colleagues. Incidentally, Ross Watts joined the Chicago accounting doctoral program the same year as Ray. 2.2 The Aim of Ball and Brown [1968] King [1966] had pointed to clustering of the off-diagonal commonalities in the stock return variance–covariance matrix, due to what he identified as industry effects. Ray and I conjectured that these same commonalities would be present in earnings, being reflected in the “top down” approach to security analysis, for example. We thus chose industry and market commonalities in earnings as our first major project. We quickly realized, however,

115

that it was leading us to the far more important question of the connection between earnings and stock returns. When we presented our first paper at the 1967 JAR Empirical Research Conference, so sure were we that it was merely the precursor to a much more important second paper that we were taken aback when a discussant, Thomas H. Williams, did not obviously share our expectations. Despite its grandiose title, “An Empirical Evaluation of Accounting Income Numbers,” Ball and Brown [1968] had a quite modest aim: to test the null hypothesis that accounting income numbers are “not useful” to share-market investors, against the specific alternative hypothesis that they are. In simple terms, we had had an earful, first in our honors course at UNSW and then in our early Chicago accounting doctoral seminar, about why accounting practice was little more than mere ritual. How could accounting reports have survived for so long, we reasoned, if they cost real dollars to prepare yet had no bearing on the wealth positions of the shareholders to whom they were addressed? The fact that reports did not match exactly what this or that accounting theorist prescribed was, and still is, insufficient grounds for rejecting reports prepared under Generally Accepted Accounting Principles. Once we rejected the null, we then went on to assess what we called “usefulness,” in terms of the “relevance” and “timing” of preliminary net income reports. We concluded as follows: “Of all the information about an individual firm which becomes available during a year,

116

one-half or more is captured in that year’s income number. Its content is therefore considerable. However, the annual income report does not rate mgniy as a timely medium, since most of its content (about 85 to 90 per cent) is captured by more prompt media which perhaps include interim reports.” 2.3 The Paper’s Impact Why did our paper have such an enormous impact? I can offer you only my personal view. I have seven reasons. 1. It was cast in the mold of a traditional experiment: question, hypothesis, data collection, hypothesis testing, conclusion. It was strong on other methodological grounds, e.g., in demonstrating robustness to alternative models and variable definitions. If you read the footnotes you will see one of the first empirical applications of Sharpe’s Capital Asset Pricing Model, which was used to demonstrate that our results were robust to our choice of the market model to measure abnormal returns. (Interestingly, we reported that a proxy for cash flow did not outperform accrual income.) 2. It expressed a view than ran counter to the critics of accounting, in that it rejected the null hypothesis in a convincing way. 3. It emphasised the use of data to test a belief. As Bill Voss, a fellow doctoral student, often said when one of us took to pontificating, “That’s an interesting hypothesis—why don’t you test it?” 4. It reflected the then-thinking in finance: the Efficient Market Hypothesis was a maintained hypothesis. 117

5. It documented an association between earnings and prices that has proved robust over time and different markets. 6. It introduced a focus on income expectations and on announcements which revised those expectations (in contrast to say the Fama, Fisher, Jensen, and Roll classic study on stock splits, which focused on a split’s effective date, which is some time after the split’s announcement). 7. It developed an approach to measuring the shareholder wealth effects of information of various types. 2.4 Weaknesses Ball and Brown [1968] must also have had its weaknesses, although I am hardly the person to ask to point them out. It was rejected by the editor of The Accounting Review on the grounds that it was not suited to an accounting journal. The editor did offer to reconsider the manuscript if Ray and I wished to cut down the empirical stuff and expand the “bridge” we had tried to build between our paper and the accounting literature, when we argued that our null hypothesis was consistent with a common belief expressed in that literature. While I am on the subject of weaknesses, Ray and I might well have started the habit, which others continued, of drawing attention to securities market anomalies and then proceeding to ignore them. This habit

118

culminateci in Ray’s paper (Ball [1978]), where earnings-related anomalies were brought out into the open once again in a forceful way. That will have to do for criticisms from me. Let me turn now to the literature since 1968 and how it is connected to our paper. 3. The Literature Since Ball and Brown [1968] I am no historian, so there is no way I can hope to do complete justice to the literature which has referenced Ball and Brown since 1968. My problem is exacerbated by the fact that Ball and Brown [1968] is one of the most frequently cited articles in the accounting literature since at least as far back as 1956.1 So if you will allow me to make a highly personalized, selective classification, I shall group the empirical literature into eight main areas: 1. shareholder wealth effects of accounting income numbers 2. shareholder wealth effects of other accounting numbers 3. time-series behavior of accounting numbers 4. preannouncement drifts in share prices 5. earnings-related anomalies in share rates of return 6. earnings response coefficients 7. information transfers 8. multiperiod earnings forecasts and share prices It is a selective classification and I shall take them one at a time. 3.1 Shareholder Wealth Effects of Accounting Income Numbers 119

An early extension of our work was the research into the “cosmetic” effects of accounting changes, an example of which is Ray’s paper, “Changes in Accounting Techniques and Stock Prices” (Ball [1972]). Ray argued that, in the absence of cash flow effects, changes in accounting methods that changed the accounting representations of economic events, but not the events themselves, would be “seen through” by security analysts, because they had almost costless access to alternative information sources. This literature was relatively short-lived, because it begged the question of why a firm would change its accounting methods if change is costly and there are no apparent economic benefits. The more interesting question that has been asked many times since is “What are the economic incentives that lead managers to choose a particular set of accounting policies?” Accounting changes can have cash flow consequences because of their implications for income tax payments, borrowing costs, political costs, and management compensation agreements (although the connection between management compensation and accounting methods is less rigid in Australia than in the United States). This literature, which often is identified with the Rochester school, has had a major impact on the accounting literature since the mid-1970s. Ball and Brown compared the predictive powers of operating income and net income both before and after extraordinary items. We concluded that the mean abnormal return from a perfect foresight model was

120

greatest for net income inclusive of extraordinary items. Since then, there have been many studies of the “incremental” share price effects of alternative accounting constructs—e.g., net cash flows, lease capitalization, GPLA and replacement cost disclosures, and LIFO adoptions, to name just a few. Indeed, it is now hard to imagine any proposal to change U.S. accounting reporting requirements that will not have to run the gauntlet of a market-based test as part of the routine evaluation of its economic consequences.2 Ball and Brown [1968] had measured simply the average share price movement for all firms. Others have questioned whether the nature of the adjustment of prices to information is homogeneous. They have related the adjustment to attributes such as a firm’s size, its industry, the time delay from the end of the fiscal year until the profit report is released, or, as in Skinner [1989], whether there are exchange-traded options on the underlying shares. In general, the answer seems to be that there are important and to some extent predictable differences in the adjustment process across firms. 3.2 Shareholder Wealth Effects of Other Accounting Numbers Ball and Brown focused tightly on the bottom line, yet there are many other numbers in the financial statements. The question of their relevance was put (but not answered) by Brown [1970]: “Of what importance are other items contained in the annual report?” In that paper, I proposed that a multivariate discriminatory technique be used to classify reports into “good” and

121

“bad” news categories, based on financial statement data. Over the years, several attempts were made to answer this question, one conclusion being that there may not be a whole lot more to be said, once earnings had been accounted for. For example, in his [1974] paper, Nick Gonedes looked at six ratios, including working capital/ total assets, financial leverage, the asset turnover rate, cash flow to debt plus preferred stock, two measures of the accounting rate of return, and EPS. Nick concluded that the numbers, jointly, were uninformative and that, given EPS, the other six explained little. However, more recently Ou and Penman [1989] have been pushing this line of enquiry considerably further. Using 68 initial accounting variables,3 they have fitted a multivariate logit model first to predict future earnings and then to predict future stock market returns, with a success rate that has startled many believers in market efficiency! 3.3 Time-Series Behavior of Accounting Numbers The time-series behavior of earnings was perhaps the first major outgrowth of Ball and Brown [1968]. We had referred to the issue as “the statistical nature of the income process, a process little understood but of considerable interest to accounting researchers” [1968, p. 177]. As Brown [1970] explains more directly, “in the absence of explicit knowledge of investors’ EPS forecasts they must be simulated by mechanical rules.” Ball and Brown had used the classical naive model, which predicts that EPS this year will be no different 122

from EPS last year, to identify good and bad EPS news; and an adaptation of the “market model” popularized by Fama et al. [1969], which has embedded in it a submartingale assumption.4 Although Little [1962] had concluded that earnings in the United Kingdom seemed to approximate a random walk, it was important that the efficiency of earnings forecast methods be addressed directly. For if the forecast was inefficient, then the earnings forecast error—in essence, the “surprise” component of the earnings signal—would be measured with error, and the association between earnings and share returns would be understated. Ray and Ross Watts enrolled in a corporate finance class I taught in the summer of 1967 in Chicago’s evening M.B.A. program. I required a term paper from them and admit to steering them in the direction of the time-series properties of EPS; whence came Ball and Watts [1972], one of the ten most-cited articles in the finance journals of the 1970s. Since then, people have gone on to examine and reexamine the question, mostly arriving at the same result: that Ball and Brown’s choice was perhaps fortuitous, but nevertheless a sound one. It went back to a second-year undergraduate statistics course I took from Dr. Sheila Rowley at UNSW, who impressed upon her students the value of genuinely parsimonious models in empirical applications. It is a lesson I have never forgotten. Interest in the efficiency of time-series forecasts expanded to encompass the comparative efficiencies of

123

security analysts’ and managers’ forecasts. In many applications that are modeled on Ball and Brown’s, time-series extrapolations have been replaced by security analysts’ forecasts, since they are more efficient.5 The time-series literature too has developed a life of its own, far beyond that Ray and I ever contemplated. To illustrate, George Foster, in his book Financial Statement Analysis [1986, p. 212], lists six applications of time-series forecasting: A. Performance evaluation of management where a key concern is what percentage of the earnings change is due to non-firm-oriented factors; B. Examining allegations that management is “manipulating” earnings to (say) avoid violating restrictive covenants in a bank loan agreement; C. Designing a “profit-sharing” component of an executive compensation plan where a central concern is risk sharing between management and other parties associated with the firm; D. Management decisions on alternative accounting methods where an important factor is the time-series variability in the reported earnings series; E. Litigation where allegations of excess profits have been made and the concern is to explain the sources of a reported earnings series; F. Litigation where business operations have been disrupted by a fire or a strike and estimation must be made of the earnings that would “normally” have occurred.

124

3.4 Preearnings Announcement Drifts in Share Prices Ball and Brown [1968] noted the preannouncement drift and concluded that, on average, 85% of the earnings-related change in share prices6 could be attributed to events that predate the annual EPS announcement month. They then asked [1968, p. 177] “of what help are interim (i.e., quarterly earnings) reports and dividend announcements” in explaining the anticipation? So it was not long before quarterly or interim profit reports7 and dividend announcements8 were studied in the United States and in Australia.9 Clearly, a variety of signals can result in the contents of an annual report being anticipated. It is only to be expected that share prices in a competitive market will reflect the contents of many reports well before their release. Since 1968 this line of research has spread to New Zealand, Canada, the United Kingdom, Europe, and Singapore. I have little doubt that, somewhere in the Soviet Union, there is a researcher hoping that peres–troika will extend to the introduction of a capital market and that one day he will be able to relate share prices to published accounting reports in the U.S.S.R. … One thing we have learned from all the replications and extensions is that the original Ball and Brown experiment is extremely robust. 3.5 Earnings-Related Anomalies in Share Rates of Return

125

Ball and Brown observed a postannouncement drift which, although not particularly significant from a statistical viewpoint, nevertheless was troublesome because of the maintained hypothesis of market efficiency. The drift was observed on average. It need not be true of any one individual company, as the averaging across firms of jumps which occur at random intervals for individual firms can induce an apparent drift, whether it be downward or upward. This “anomalous” result was one of the first in what is now a whole series of anomalies. Unusual market behavior has been documented for all of the following: a. Time of Day. On the NYSE, for example, the mean return at the end of the day is consistently large and positive, which means you should sell, but not buy, at the end of the day. b. Day of the Week. Australian stockmarket returns are typically lower on Mondays, Tuesdays, and Wednesdays and higher on Thursdays and Fridays; hence you should buy on a Wednesday afternoon but sell on a Friday afternoon. c. Day of the Month. Australian market returns are below the grand mean up to the last two days of the month, when they become positive; thus, you should buy toward the end of the month, but sell halfway through it. d. Month of the Year. In the United States you should buy at the end of December, but not sell before the end of January, because of the turn–of–the–year anomaly.

126

e. Around Holidays. In Australia, according to Easton [1989], you should sell after holidays but buy before them, because returns are anomalously high around holiday periods. f. Shares with High Versus Low Earnings Yields. Abnormally high returns are available if only you can figure out which stocks command low P/E ratios…. g. Shares with High Versus Low Dividend Yields. Or, if you find it easier to calculate the dividend yield, just stick to the high-yielding shares…. h. Small Versus Large Firms. Finally, as everyone now knows, all you have to do is figure out which stocks have low market capitalizations and, if you live in the United States, buy them for Christmas! I’ll stop there, but the list is seemingly endless. This circumstance, recurring phenomena which are present but not understood, has led to a number of papers that explore research design questions, beginning with Ball’s [1978] “Anomalies” paper and extending to Bernard and Thomas’s [1989] Conference paper, “Post-Earnings Announcement Drift: Delayed Price Response or Risk Premium?” It may be claiming too much to attribute them to questions raised in Ball and Brown [1968], but in a sense they are a natural extension of the postan–nouncement drift we found so puzzling. 3.6 Earnings Response Coefficients There is a burgeoning literature on what has been called “earnings response coefficients”; but it has a long

127

history. That history dates back at least to Miller and Modigliani’s [1966] work on estimating the cost of capital, and it illustrates why R2 is not a helpful way to summarize the association between earnings and prices. It is easy to find a high correlation between earnings and prices. Simply choose a relatively homogeneous industry with a wide range of firm size, and correlate the capitalized market value of the firm with Miller and Modigliani’s variable, X(1—τ), which is just net income plus after-tax interest payments. However, as Miller and Modigliani noted, the resultant estimate of the “earnings response coefficient”—they referred to its inverse as “the cost of capital”—may be nonsensical. Standard statistical procedures, such as deflated and instrumental variable regressions, may yield more efficient estimates of the coefficients and at the same time reduce the apparent explanatory power. Ball and Brown [1968, p. 177] referred to this class of question (i.e., earnings response coefficients), as follows: “The relationship between the magnitude (and not merely the sign) … could … be investigated.” And “be investigated” it was! You might well wonder why Ray and I did not investigate it ourselves. There were two reasons. First, we did not need to, because we were interested in the mean abnormal return from taking a long or short position in every share, and to do this, we needed to know only the sign of the earnings forecast error. Second, we foresaw “some difficult econometric problems associated with this relationship,” some of which we spelled out in a footnote.

128

3.7 Information Transfers Ball and Brown, and Beaver [1968], had very limited views of the financial world: we simply did not contemplate a major role for information transfers.10 For example, there was no industry index in the Ball and Brown earnings regression model, and Bill Beaver steadfastly refused to throw out any observations other than those for which the firm itself made an announcement. So it has been left for others to go down the information transfer track, one that has been trodden well by Australians, including Richard Morris [1980], George Foster [1981], and Greg Clinch and Norm Sinclair [1984]. George Foster [1981], for example, reported that earnings information transfers contributed to positive covariances between the returns on shares of firms in the same industry. To quote Foster again: “Those earnings releases that were associated with the largest increase in security return variability for the announcing firm also were associated with the largest increase in security return variability for the other firms in its industry. Further analysis revealed that earnings releases that were associated with positive/negative price changes for the announcing firm … were also associated with positive/negative price changes for the other nonannouncing firms in the same industry…. Clinch and Sinclair [1984] report similar results for a sample of 328 announcements by 47 Australian firms in ten industries in the 1977–81 period” [1986, pp. 388–89]. What we seem to have learned so far is that EPS-determined information transfers are a second-order

129

phenomenon, a result which George Foster, Greg Clinch, and I are finding holds up at the transaction level as well. 3.8 Multiperiod Earnings Forecasts and Share Prices The last strand I wish to mention is multiperiod modeling of the relationship between earnings forecasts and prices. This literature is small and, I believe, undeservedly so. I am referring to the Brown, Foster, and Noreen [1985] AAA Monograph no. 21 and to a paper by Brad Cornell and Wayne Landsman [1989], who explore issues raised in the Brown, Foster, and Noreen monograph. Brown, Foster, and Noreen asked the question, “Is the market myopic?” Managers have been criticized for acting as if short-run market considerations were paramount. Managers act this way, so the argument goes, because portfolio managers have developed a short-term focus and pressure managers into acting against their long-term interests. The monograph tried to answer the question by seeing if changes in short-or long-term EPS forecasts were more closely associated with stock price movements. Far from finding evidence of market myopia, we found strong evidence of a dominant long-term focus. One possible explanation for this result is that shorter-term forecasts are recognized by analysts as being influenced relatively more by transitory components, whereas longer-term forecasts are more allied in concept to “permanent” earnings and have a closer association with share prices.

130

Academics are interested in sorting out transitory versus sustainable earnings, to borrow an analysts’ phrase. So too are the analysts, and their short-and long-term forecasts are available. 4. Prospect Let me now turn from my admittedly selective interpretation of the post-Ball-and-Brown [1968] literature to give what is an even more selective view of some likely directions that capital-market-based accounting research will take over the next few years. I have seven of them. 4.1 Derivative Securities Patell and Wolfson [1984] used option prices to infer that most of the adjustment of prices to information releases occurs within the first two trading hours, although detectable traces “linger into the following day” [1984, p. 240]. Skinner [1989], in a different tack again, reports that profit announcements have less effect on share prices if options are traded on those shares than if they are not. Derivative securities, however, have a much wider interest to accountants. To illustrate, they underlie many executive compensation schemes, yet the time value of the option is ignored in accounting for management costs; they affect the balance sheet classification of equities, yet convertible securities may be treated as pure debt; they have been used via put options to manipulate profits from certain well-publicized Australian property transactions; and they are relevant to such contracts as

131

interest-rate swaps and currency hedges, which do affect credit ratings but need not be capitalized. Derivative securities raise so many accounting issues and the research agenda is so rich that my mind positively boggles. 4.2 Market Microstructure Over the years we have had access to market data for ever–Shorter trading intervals: yearly, monthly, weekly, daily, transaction by transaction, and now bid and ask prices in real time. There are some major challenges in the use of the latest data sets, including modeling market equilibrium conditions as well as designing efficient computer-processing methods. Market microstructure is concerned with details of the trading process, such as the generation and spreading of information, patterns in order arrivals, the distribution or price changes, the variance–covariance structure of security returns, the impact of regulation or deregulation, the institutional rules which translate orders into trades, price pressures exerted by block trades, and so forth. Although interest in these issues dates back at least a quarter of a century,11 there is an accelerating awareness of the potential implications of market microstructure for making valid inferences about market behavior, including the adjustment of prices to information as it enters the public domain. Share-markets, such as the New York Stock Exchange (NYSE), are often held out as everyday examples of perfectly competitive markets. While in many cases it may be reasonable to assume away from the NYSE the 132

frictions that characterize most markets, in some cases, particularly when we get down to the transaction level, we simply cannot ignore them. Market microstructure is thus becoming a key issue as violations of the assumptions of competitive markets become more crucial in experimental design. I am sure we shall see an explosion of interest in micro-structure considerations in security market research, as transaction data become more widely available. 4.3 Information Transfers Information transfers have not been exploited fully, by any means. At issue is explaining the variance–covariance matrix of abnormal security returns. I expect to see much more work done on accounting data at a less aggregated level than net income. For example, dollar sales are often modeled in corporate plans as the net result of market size, market share, product life cycles, and a pricing policy. Clearly, one firm’s actions and performance can impinge on other firms along each of these sales dimensions. A potentially richer way to model information transfers would be to disaggregate the earnings signal into key components. 4.4 Shareholder Wealth Effects It is unlikely that any new major financial reporting method or mandate will be introduced without it being subject, at some stage, to a market-based study of its wealth effects. In the United States, a hot research agenda item is accounting for the cost of, and liability for, employees’ postretirement benefits other than

133

pensions.12 According to some commentators,13 between 400 and 1,000 billion U.S. dollars of unfunded, off-balance sheet liabilities are not being recognized because such costs are being accounted for on a pay-as-you-go basis. I have no doubt that we will read, and some of us maybe will also write, much about this phenomenon in the next few years. 4.5 “Quality” of Earnings Baruch Lev [1989] has called for more research into the “quality” of earnings, and he has a point. We know that analysts make ad hoc adjustments to P/E ratios for various matters, including the “quality” of earnings.14 So far we have done little in security market research applied to accounting questions to look behind the accounting numbers, preferring rather to adopt the view that we are studying the numbers as reported. If I may quote Baruch: “Research on the quality of earnings shifts the focus to an explicit consideration of accounting issues by calling for a systematic examination of the extent to which the specific principles underlying accounting measurements and valuations, as well as managerial manipulations, detract from the usefulness of earnings and other financial variables. Such research has the potential both to further our understanding of the role of financial information in asset valuation and to contribute meaningfully to accounting policymaking” (Lev [1989, p. 176]). 4.6 Globalization of Capital Markets15

134

International capital markets are now highly integrated. The same securities are often listed on exchanges in several countries and professionally managed portfolios are diversified internationally. And there is much interest in establishing electronic linkages to create a global capital market. These trends must affect our research agenda. A prime candidate for research is explaining differences in the P/E ratios of companies listed in different countries. An article in the July 17, 1989 issue of Business Week (p. 141) tabulated P/E ratios in ten countries, ranging from 100 in Japan (reported to have the largest market capitalization) down to 10 in Australia (which had the smallest capitalization of the ten countries). The P/E ratio in the U.S. was reported to be 21, while the market capitalization of the U.S. was about two-thirds that of Japan. Are these differences in P/E ratios explained by differences in growth rates, in the quality of earnings, in leverage, in inflationary expectations, in tax regimes, in real returns, or in relative risk? There is a long list of possible explanations which can be probed, now that the necessary data are becoming available. An upsurge in research related to global trading could lead to more substantive interaction between accounting researchers in different countries. 4.7 Market Efficiency Then there is that old chestnut, “Was the security market efficient with respect to …?”—and you can fill in the blanks. There are so many “anomalies” around

135

nowadays that I sometimes wonder if there are more anomalies than instances of efficiency. However, there is enough disagreement about how prices should behave in an efficient market that I fear the efficiency question will be around for many a year yet. The argument will presumably be sealed for me when proponents of inefficiency, such as my good friend Steve Penman at Berkeley, give up their family station wagons and take to traveling in private jets. Until then, I am afraid my Chicago training has left me too skeptical to believe that competitive capital markets could remain so obviously inefficient for so long. 5. Concluding Remarks It is now 22 years since Ray and I wrote the first draft of Ball and Brown [1968], and it has been a hard act, for me at least, to follow. Nevertheless we both plug on, each in the hope that he still has something to discover, something to say. References BALL, R. “Changes in Accounting Techniques and Stock Prices.” Journal of Accounting Research (Supplement 1972): 1–38. BALL, R. “Anomalies in Relationships Between Securities’ Yields and Yield Surrogates.” Journal of Financial Economics (June/September 1978): 103–26. BALL, R., AND P. BROWN. “An Empirical Evaluation of Accounting Income Numbers.” Journal of Accounting Research (Autumn 1968): 159–78.

136

BALL, R., AND R. WATTS. “Some Time Series Properties of Accounting Income.” Journal of Finance (June 1972): 663–81. BEAVER, W. H. “The Information Content of Annual Earnings Announcements.” Journal of Accounting Research (Supplement 1968): 67–92. BEAVER, W. H., AND W. R. LANDSMAN. Incremental Information Content of Statement 33 Disclosures. Stamford, Conn.: FASB, 1983. BEAVER, W. H., AND D. MORSE. “What Determines Price-Earnings Ratios?” Financial Analysts Journal (July/August 1978): 65–76. BERNARD, V. L., AND J. K. THOMAS. “Post-Earnings-Announcement Drift: Delayed Price Response or Risk Premium?” Journal of Accounting Research (Supplement 1989): 1–36. BROWN, P. “The Impact of the Annual Net Profit Report on the Stock Market.” Australian Accountant (July 1970): 277–82. BROWN, P., Bulletin No. 13: Those Half–Yearly Reports. Melbourne: Australian Society of Accountants, 1972. BROWN, P., AND J. W. KENNELLY. “The Information Content of Quarterly Earnings: A Clarification and an Extension.” Journal of Business (1972): 403–15. BROWN, P., G. FOSTER, AND E. NOREEN. Security Analyst Multi–Year Earnings Forecasts and the Capital Market. Sarasota, Fl: American Accounting Assn., 1985.

137

CLINCH, G. J., AND N. A. SINCLAIR. “Intra–Industry Information Releases: A Recursive Systems Approach.” Journal of Accounting and Economics (April 1984): 89–106. CORNELL, B., AND W. LANDSMAN. “Security Price Response to Quarterly Earnings Announcements and Analyst Forecast Revisions.” The Accounting Review (October 1989): 680–92. EASTON, S. “Returns to Equity Before and After Holidays: Australian Evidence and Tests of Plausible Hypotheses.” Paper presented to the AAANZ Conference, University of Melbourne and Royal Melbourne Institute of Technology, July 1989. FAMA, E. F., L. FISHER, M. С. JENSEN, AND R. ROLL. “The Adjustment of Stock Prices to New Information.” International Economic Review (1969): 1–21. FOSTER, G. “Intra-Industry Information Transfers Associated with Earnings Releases.” Journal of Accounting and Economics (March 1981): 201–32. FOSTER, G. Financial Statement Analysis. 2d ed. Englewood Cliffs, N.J.: Prentice-Hall, 1986. GONEDES, N. J. “Capital Market Equilibrium and Annual Accounting Numbers: Empirical Evidence.” Journal of Accounting Research (Autumn 1974): 26–62. KING, B. F. “Market and Industry Factors in Stock Price Behavior.” Journal of Business (Supplement 1966): 139–90.

138

LEV, B. “On the Usefulness of Earnings and Earnings Research: Lessons and Directions from Two Decades of Empirical Research.” Journal of Accounting Research (Supplement 1989): 153–92. LITTLE, I. M. D. “Higgledy Piggledy Growth.” Bulletin of the Oxford Institute of Economics and Statistics (November 1962): 387–412. MILLER, M. H., AND F. MODIGLIANI. “Some Estimates of the Cost of Capital to the Electric Utility Industry.” American Economic Review (June 1966): 333–91. MORRIS, R. “The Effect of Firm’s Earnings Announcements on the Share Price of the Other Firms in the Same Industry.” Working paper, University of New South Wales, 1980. OU, J. Α., AND S. H. PENMAN. “Accounting Measurement, Price-Earnings Ratios, and the Information Content of Security Prices.” Journal of Accounting Research (Supplement 1989): 111–44. PATELL, J. M., AND M. A. WOLFSON. “The Intraday Speed of Adjustment of Stock Prices to Earnings and Dividend Announcements.” Journal of Financial Economics (June 1984): 223–52. PETTIT, R. R. “Dividend Announcements, Security Performance and Capital Market Efficiency.” Journal of Finance (December 1972): 993–1007. PETTIT, R. R. “The Impact of Dividend and Earnings Announcements: A Reconciliation.” Journal of Business (1974): 86–96.

139

SKINNER, D. J. “Options Markets and the Information Content of Accounting Earnings Releases.” Paper presented to the Workshop in Accounting and Financial Management, Macquarie University, May 30, 1989. STIGLER, G. J. “Public Regulation of the Securities Markets.” Journal of Business (1964): 117–42. WATTS, R. “The Information Content of Dividends.” Journal of Business (1973): 191–211. WATTS, R. “Comments on the Impact of Dividend and Earnings Announcements: A Reconciliation.” Journal of Business (1974): 97–106.

* University of Western Australia. This paper, which is based on my plenary address at the 1989 Accounting Association of Australia and New Zealand Conference, is an elaboration of my presentation at the 1989 JAR Conference. This version has benefited from comments by George Foster. 1

The citation Ray and I received in 1986 for the American Accounting Association’s inaugural Seminal Contribution to the Accounting Literature award reads as follows: “No other paper has been cited as often or has played so important a role in the development of accounting research during the past thirty years.” 2

A key development in the process was Beaver and Landsman [1983], which won the 1985 Wildman Medal.

140

3

Ou and Penman have reduced the 68 initial variables to 28 plus an intercept, thereby giving new meaning to the term “parsimonious model.” 4

That is, when the market effect is zero, predicted EPS is last year’s plus the average change in EPS since the start of the time series. 5

For example, Brown, Foster, and Noreen [1985] document that analysts’ consensus forecasts are on average 50% more accurate than the Ball and Brown naive model when the forecast is made 12 months prior to the announcement. 6

Net of market effects.

7

For example, Brown and Kennelly [1972].

8

For example, Pettit [1972] and Watts [1973].

9

For example, Brown [1972].

10

Information transfers” refer to the process whereby information about one firm has value implications for other firms in the same industry. 11

For example, Stigler [1964].

12

Current and future pensioners’ medical bills are a prime source of the liability. 13

Australian Financial Review (April 17, 1989), p. 36.

14

See, e.g., Beaver and Morse [1978].

15

This discussion was added at George Foster’s suggestion.

141

© 1990 by Institute of Professional Accounting, The University of Chicago

142

By Philip Brown, B.Com., M.B.A., Ph.D., A.A.S.A. Professor in Accounting, University of Western Australia The Impact of the Annual Net Profit Report on the Stock Market* I. INTRODUCTION

T

about the fundamental determinants of share prices have existed for many years. One of the more impressive is that put forward by Miller and Modigliani1, which relates the total market value of the firm to, among other things, its future earnings. But, like most others, the Miller-Modigliani (MM) theory is silent on how and when the investor learns about future earnings. This is not intended as a criticism of MM, because when testing their theory they were able to by-pass these questions … HEORIES

Nevertheless the questions remain; and are interesting, for if the phenomena which trigger price movements on the stock market were identified, then security analysts could focus their efforts to gather and process information and hence perform their tasks more efficiently. Some early work, based on data for firms listed on the New York Stock Exchange (NYSE), has been done. For example, Fama et al2 investigated “… the process by which common stock (ordinary share) prices adjust to the information … that is implicit in a stock (share) split”; Scholes3 examined the stock market’s reaction to large block sales; Ball and Brown4 assessed the relevance to investors of the annual report of net

143

profit; and Brown and Kennelly5 extended the Ball and Brown results to interim reports. The Australian Society of Accountants’ 1969 Endowed Lecture in the University of Adelaide This paper will be concerned mostly with efforts to further the research of Ball and Brown, and Brown and Kennelly. The paper reports tentative findings on the impact of the annual net profit report on the prices of shares listed on stock exchanges in Australia. The results from a more comprehensive survey will not be available until late 1970. II. EXPERIMENTAL DESIGN The experimental design is similar to that described in detail in Ball and Brown6. It is (1) assumed that future Earnings (Net Profit) Per Share is relevant to investors, who (2) forecast future Earnings Per Share (EPS) from known data, but (3) react to errors in their forecast, as errors are revealed, by adjusting the price at which their shares are traded. The ability of investors to make advance predictions determines the amount of information contained in an EPS report at the time of its release. Thus, if investors could predict exactly the contents of an annual report say, six months before its release, then, apart from confirming their prediction, the release of the report is of no consequence. Conversely, the greater the discrepancy between actual EPS and the EPS as forecast by investors, the greater the amount of information contained in the EPS report; and presumably the greater the impact of the

144

report’s announcement on share prices. Basically, the impact of the annual net profit report is assessed by relating this discrepancy, or EPS forecast error, to the adjustment in share prices associated with its announcement. There are two parts to the experimental design. First, in the absence of explicit knowledge of investors’ EPS forecasts they must be simulated by mechanical rules. Second, investors’ reactions (via share price adjustments) must be measured and related to the errors in “their” forecasts as those errors are revealed. Forecasting Future EPS It is difficult to develop efficient mechanical rules for forecasting EPS without some knowledge of the behaviour of earnings through time. Ball and Watts7 concluded, after a detailed study of several hundred firms listed on the NYSE, that EPS appears to behave as if it is generated by a process which is either a random walk or at least “very close” to it. Unfortunately, to my knowledge, no similar study has been conducted in Australia, although I am hopeful some results will be known soon. In the meantime I have fallen back on the Classical Naive Model, which assumes simply that the world does not change8. In other words, it predicts that tomorrow will be the same as today which, in the present context, reduces to the forecast that this year’s EPS will be the same as last year’s. For example, suppose investors forecast, on the basis of News Ltd.’s EPS of 20 cents for the year ended June 30, 1967, that EPS for 1967/68 will also be 20 cents. If

145

actual EPS proves later to be 25 cents, then there will be a forecast error of −5 cents. Moreover, when the actual EPS is announced and investors realise they have under-estimated EPS (that is, they have made a positive forecast error), they will react by adjusting the price of News Ltd. upwards. On the other hand, if actual EPS proves to be only 15 cents, then there will be a forecast error of +5 cents, suggesting that the price of News Ltd., will decline—other things being equal. Investors’ Reactions The difficulty when measuring investors’ reactions is that those “other things” are not always equal. It is a well known fact, for example, that share prices, and therefore rates of return from investing in shares, tend to go up and down together, behaving as if they are all subject to common expectational influences. Some factors which contribute to this related movement in Australia are changes in monetary and fiscal policy (the Federal Budget), in foreign investment (“London Support”), in interest rates (at home and abroad) and in the rate of overall economic activity. Therefore, when measuring investors’ reactions to a particular firm’s profit report, an adjustment should be made for that part of the movement in the price of the firm’s share which is due to fluctuations and drifts in the stock market in general. The adjustment procedure is the same as that used elsewhere (see notes 2, 3, 4, 5). For each particular share it consists of three steps. First, estimate, using past price data, the normal relationship9 between the rate of return10 on this share and the rate of return on a market index. Second, for any particular month which is of interest,

146

predict, from this normal relationship, how much the rate of return on this share should have been, given whatever change happened to take place in the market index. Third, compare the actual rate of return on this share with the rate of return predicted under step two. The result, the extent to which the actual rate of return is greater or less than the predicted rate of return, reflects the impact of new information about this particular share. Continuing the News Ltd. example will help clarify the procedure. Suppose, when there is no change in the market index, that the price of a share in News Ltd. on average, increases by one-half of one per cent per month; whereas if the market index increases by say 1 per cent, the price of that same share increases by a further per cent. Now take any particular month, for example August 1963 when the market index increased by about 5 per cent. Then other things being equal the price of News Ltd. shares should have increased by per cent. If, in fact, the shares rose by only 4 per cent, then they did not rise as much as they normally should have risen, implying that some adverse information concerning that particular firm, News Ltd., must have been received by investors during August. The argument applies equally to months when the market index decreases, as, for example, it did in November 1960. If the market declined by 10 per cent, then we would have expected News Ltd. shares normally to change by per cent. If News Ltd. shares declined by only 10 per cent, then shareholders in effect have ended up some per cent

147

better off than they normally would have expected. This per cent, obtained by subtracting the predicted (or “normal”) return from the actual return, is defined as the “abnormal rate of return”. And it is clear that the sign of the abnormal rate of return determines whether the sum of information received about News Ltd., in any one month has been deemed, on balance, to be “good” or “bad” news: if positive, then it was “good”, if negative, then it was “bad”. Summary The two parts can now be assembled in the following way. Because of the nature of the forecasting rule, investors will have forecast this year’s EPS as soon as last year’s EPS was known; which for convenience is assumed to be exactly twelve months prior to this year’s announcement11. In the intervening time (that is, between last year’s and this year’s announcements) the shareholder would have received many pieces of information of various types—for example, an interim dividend, the half-yearly profit report, routine newspaper items—some of which, in particular the half-yearly profit report, would have improved his ability to forecast this year’s EPS. An improved ability to forecast the final outcome would result in an early realisation of any forecast error, such that investors would adjust share prices some time before the annual report’s announcement. For this reason abnormal rates of return are tracked from a starting point twelve months prior to the announcement date. Further, by studying the behaviour of the abnormal rates of return over the full twelve months, we can measure the

148

relevance for investors of both the content of the annual report and the timing of its release. At this stage we are more interested in what happens on average rather than what happens as a result of any one particular annual report or in any one particular year. Thus data for more than 100 firms over a ten-year period are the base for this report. III. DATA Three types of data are required: annual EPS; EPS announcement dates; and share prices. Annual EPS data were extracted from the 1968 and 1969 editions of Ian Potter & Co’s Australian Company Reviews. The 1968 edition lists EPS since 1958 for 138 separate companies. Eight mining companies were excluded12 because they will be the subject of a special study in the near future. EPS announcement dates were obtained for the ten years 1959-68 from two sources: the Official Record of the Melbourne Stock Exchange and records maintained by the Research and Statistical Bureau at the Sydney Stock Exchange. These dates correspond to the month in which each preliminary profit statement was released to the press, subject to the proviso that, if the press release was made within the last five calendar days of the month, the profit statement was assumed to have become available to the general public in the following month13. Share price data, corresponding to the last recorded sale for each month since January 1958, were extracted from publications of the Adelaide, Melbourne and Sydney

149

stock exchanges. Data were also collected for all dividends and rights to issues (whether bonus or cash issues of deferred, ordinary or preference shares, issues of debentures and convertible unsecured notes, etc.) The price, dividend and rights14 data were then expressed in the form of monthly price relatives, with each relative defined as the ratio of the end-of-month price cum dividends and rights to the begin-ning-of-month price (that is, the end-of-month price from the previous month). A share price relative so defined is just the (adjusted) proportion by which the price changed in the particular month, plus unity. For example, in October 1960 News Ltd. issued approximately three million 5/-ordinary shares to existing shareholders on the basis of 1-for-1 at a 3/-premium. The last recorded share sales were 30/6 (September, cum rights) and 18/3 (October, ex rights). Based on the October ex rights price of 18/3, the theoretical value of the rights attached to one of the old shares was 10/3. The shares had already been quoted ex the final dividend for 1959/60 in September, hence the price relative for October was 28/6 (=18/3 + 10/3) divided by 30/6, or 0.934. The rate of return was –0.066, or a loss of 6.6% for the month. For statistical reasons it is essential that each share be actively traded for a number of years15. I arbitrarily selected five years, or 60 monthly price relatives, as a minimum16 which excluded a further twelve firms, leaving a total of 118. Some properties of the 118 are interesting. Table 1 gives the month-by-month results from investing $100 at the

150

end of January 1958 (spread equally over all firms for which prices were quoted)17. It shows that the Potter survey must include a number of firms which, historically, have been extremely successful. Table 2 gives the deciles of the distributions of: (1) the square of the co-efficient of simple correlation between the monthly rate of return on each security and the monthly rate of return on the market index18 and (2) the coefficient of first-order auto-correlation in the estimated abnormal rates of return, that is, the residuals from the OLS regression implied in (1). The median r2 value (.5 decile, in Table 2) indicates that about 20 per cent of the variance in the monthly rate of return on the “typical” share can be associated with widespread movements in share prices19. The distribution of the co-efficients of first-order auto-correlation suggests that there is very little relationship, if any, between abnormal rates of return in consecutive months20. Table 1 Investment Performance Index—118 Firms Surveyed by Ian Potter & Co. January 1958 = 100

Table 2 Deciles of the Distributions of die Squared Co-efficient of Correlation for the Rate of Return

151

Regression, and of i Co-efficient of First-Order Auto-correlation in the Estimated Abnormal Rates of Return.

IV. RESULTS The easiest way to reduce the obvious complexity of the results is to treat the 12 months leading up to the announcement of the annual report as the basic unit of time. The results can then be summarised, by month relative to the annual report announcement, in the form of (a) Abnormal Performance Indexes21 and (b) chi-square statistics calculated from a two-by-two classification by the sign of the EPS forecast error and the sign of the abnormal rate of return. Abnormal Performance Indexes (API’s) are computed separately for three portfolios, the first consisting of all EPS reports which contain “good” news (that is, when EPS increased over the previous year), the second consisting of all EPS reports which contain “bad” news, and the third consisting of all EPS reports regardless of what they contain. This last portfolio provides a control. Suppose, for example, that News Ltd.’s EPS for the year to June 30 1964 was ten cents, and for 1964/65 fifteen cents. Thus according to our forecast rule the 1964/65 Net Profit report, announced in September 1965,

152

contained “good” news. Now assume that the relevant monthly abnormal rates of return for News Ltd., were:

Then the sequence of the API for News Ltd.’s 1964/65 annual report would be calculated as follows: Month Relative to AnnualAnnouncement Performance Report Abnormal Index

−11

1.001

−10

1.001 × 1.002

− 9

1.001 × 1.002 × .997

− 8

1.001 × × 1.008

1.002

×

.997

Finally, entries to the two-by-two contingency tables would be: Entries to Contingency Tables

Table 3 presents the three API’s (the first two arc adjusted)22 and chi-square statistics calculated for the 153

118 firms and those years for which data were available23. Figure 1 plots the three API’s. Several results emerge. Table 3 Summary Statistics by Month Relative to Ana nal Report Announcement Date

Note: Probability (chi-square ≥ 3.84 | x2 = 0) = .05, for 1 degree of freedom. Probability (chi-square ≥ 6.64 | x2 = 0) = .01, for 1 degree of freedom.

154

Fig. 1 Abnormal Performance Indices for Various Portfolios 1. Net profit reports contain information relevant to investors. To illustrate, suppose an investor were told annual EPS twelve months in advance of the 155

release of each EPS report and that his investment decisions had negligible effect on stock prices. Then by buying those shares which later report an increase in EPS he could have made on average 5.4 per cent p.a. more than he normally would have expected24; and, at the same time, he could have avoided making investments which would have returned him on average 9.0 per cent less than he normally would have expected25. 2. The larger values of the chi-square statistics and the larger proportionate changes in the first two API’s appear to cluster in two distinct time intervals, namely months −7 and −6, and months −1 and 0. These correspond to the announcements of the half yearly and yearly profit reports respectively. Although the greatest monthly adjustment in share prices takes place in the annual report’s announcement month, the results in Table 3 indicate that it is only about 20–25 per cent of the total adjustment for the year. Presumably investors anticipate some three-quarters of the content of the annual profit report before the month of its release. Unfortunately, due to data deficiencies26, these estimates might improve inaccurate27. Future work will identify more precisely the date the annual report became generally available, which will improve our assessment of the roles the reports play in informing investors about EPS. 3. If the half-yearly and yearly profit reports are themselves the media by which the stock market becomes informed about EPS, then it appears that the market reacts speedily to adjust share prices to that information. Both the API’s and the chi-square 156

statistics (Table 3) suggest once the annual report has become generally available there are no further opportunities for abnormal profits. At that point the individual investor can earn further abnormal profits only by forecasting the next annual report more accurately or in advance of all other investors28. 4. It is quite possible that investors look further than twelve months into the future when predicting EPS. Moreover, EPS is certainly not the only datum on which investors base their decisions. It is of some interest, therefore, to measure the extent to which share price adjustments can be associated with (a) EPS reports, and (b) information from all other sources. The estimation procedure is intricate so I will provide only a heuristic interpretation29. Suppose investor X were told every single piece of information about a firm which would be announced to the public at any time during the next twelve months, after which time the firm would announce EPS for its most recent fiscal year. Suppose further that: (a) X must either buy, or adopt a short position on, a share in the firm and hold it until the annual report is released; and (b) X’s decision has no effect on share prices. Then X would have earned on average 14.6 per cent more than he normally could have expected30. On the other hand suppose investor Y is also clairvoyant, but only with respect to annual EPS. That is, he can foresee an EPS report twelve

157

months in advance of its announcement, but he can foresee no other news. If Y is identical to X in all other respects, (that is, he is permitted to decide on an investment strategy only once and his decisions do not influence share prices), then Y who acts only on EPS information, would have earned on average 6.9 per cent more than he too normally would have expected31. I conclude that approximately half the annual adjustments in share prices can be related to EPS reports and, by implication, that approximately half are related to information from other sources. 5. The above results for the Australian stock market are remarkably consistent with results obtained for the New York Stock Exchange. From their study of the NYSE, Ball and Brown (p. 176) conclude: “Of all the information about an individual firm which becomes available during a year, one-half or more is captured in that year’s income number.” Our preliminary estimate is the same. Ball and Brown continue: “However, the annual income report does not rate highly as a timely medium, since most of its content (about 85 to 90 per cent) is captured by more prompt media which perhaps include interim reports.” Again, the estimates are entirely consistent. The annual report probably has greater impact in the month of its announcement in Australia because 158

profit reports are made typically only once every six months here, compared with once every three months in the U.S. Finally, the tendency of the proportionately larger movements in the API’s and the relatively larger values of the chi-square statistics to cluster in the neighbourhood of the release of the half-yearly report indicates that results, such as those recently obtained by Brown and Kennelly, will likewise be obtained in Australia. V. CONCLUDING REMARKS The objective of this paper was to report on some preliminary attempts to measure the impact of the annual net profit report on the Australian stock market. We conclude: although net profit reports contain information relevant to investors, in the sense that at least half the adjustments in share prices over a 12-month period can be related to EPS reports, only about 20-25 per cent of those adjustments take place in the month in which the annual report is announced32. Moreover, by the time the annual report is generally available, the market has fully-adjusted share prices such that no further opportunity for abnormal profit remains. While these findings are important in their own right, there are many other questions to be investigated. A few of them are listed below. 1. What effect did the adoption of the Uniform Companies Acts have on the relevance of the annual profit report?

159

2. Does the amount of information contained in annual reports differ between industries—for example, mining versus manufacturing? 3. Some have argued net profit should be adjusted for “unusual items”. What is the evidence? 4. How relevant is the half-yearly profit report? 5. How do EPS series behave through time33? 6. Of what importance are other items contained in the annual report? This last question is particularly intriguing. It could be answered by employing discriminant analysis techniques to differentiate those firms and years in which the API was greater than unity from those in which it was less. Unfortunately this exercise requires far more data than are at my disposal. The implications of our findings for the investing public are as follows. For the Fundamentalist, there is little doubt that profit reports are important. Generally, however, the market is able to make advance forecasts of the contents of an annual report, these forecasts having greatest impact on share prices during the two months immediately prior to a report’s release. Thus the investor who can develop a superior ability to forecast EPS, even if it is only from a point two months prior to the release of the report, will be in a position to make substantial gains over and above what he normally could expect—provided he acts quickly. There appears to be no point to waiting until the audited details of profits are published, for example, because by then all price adjustments have been made and it is too late. This does not mean that those glossy annual reports are worthless,

160

for they contain information in addition to profits; although the value of that other information is at present unknown. Finally, an unfortunate implication for the Chartist. Little is to be gained from studying month-by-month share price adjustments for clues to future months, since, to all intents and purposes, monthly adjustments are serially independent. This means that the stock market reacts both speedily and fully to new information, such that the rewards belong to those who look to the future, and are brave enough to act quickly on what they see.

* The author wishes to acknowledge the helpful co-operation of: Mr. D. Patterson, formerly of the Sydney Slock Exchange; Mr D. Smalley, formerly of Messrs. Ian Potter & Co.; Mr. I Steele, of the Stock Exchange of Perth; Mr. R. Watters, of the Sydney Stock Exchange; and Mr. F. Wray, of the Stock Exchange of Adelaide. 1. Miller, Merton H., and Modigliani, Franco. “Some Estimates of the Cost of Capital to the Electric Utility Industry, 1954-1957.” American Economic Review, LVI (June, 1966) 333-391. 2. Fama, Eugene F.; Fisher, Lawrence; Jensen, Michael C.; and Roll, Richard. “The Adjustment of Stock Prices to New Information.” International Economic Review, X (February, 1969) 1-26.

161

3. Scholes, Myron J. “The Effect of Secondary Distributions on Price.” Unpublished Ph.D. dissertation, University of Chicago, 1969. 4. Ball, Ray, and Brown, Philip. “An Empirical Evaluation of Ac counting Income Numbers.” JoutmqI of Accounting Research, VI (Autumn, 1968) 159-78. 5. Brown, Philip, and Kennelly, John W. “The Information Content of Quarterly Earnings: A Clarification and an Extension.” Available 6. Ball and Brown, op. cit., pp. 160-165. 7. Ball, Ray, and Watts, Ross. “The Behavior of Earnings through Time.” Unpublished paper presented to the Workshop in Accounting Research, University of Chicago, April 23, 1968. 8. It should be pointed out that the Classical Naive Model is, naive only in the statistical sense. Several economic time series appear to be described well by this model. It has been used extensively in applications similar to the present one. 9. Estimated by ordinary least squares regression. 10. That is, the proportionate change in price, adjusted for dividends and rights to new issues. 11. This assumption is not an accurate description of what in fact takes place. However, for a rather subtle reason which I will not argue here, it happens to make little difference as lens as the two announcements are at least about nine months apart.

162

12. They were: BH South. Consol Goldfields, CRA, Hamersley, Mt Isa, New BH, North BH and WMC. However, the remaining 130 firms include several firms which are classified as industrial but for which mining activities play an important part in overall operations: for example, BHP, CSR and EZ Industries. 13. The reason for this adjustment is that the price quotations are taken as the last sale for the month, which must be post-announcement for the procedure to be accurate. The effect of the assumption can be investigated in several ways, but I have not yet had the necessary time to do so. 14. The theoretical value of the rights was used in the case o(issues of ordinary shares, and the actual value (that is, quoted price) of the rights in all others. 15. The reason is that the distribution of the disturbance terms in the OLS regression between the investor’s rate of return on an in dividual share and the rate of return on the market as a whole i characterised by fat tails. 16. The minimum number of monthly price relatives for firms which met the cut-off criterion turned out to be 75. 17. Expressed in the form of an Investment Performance Index, which “… assumes the same dollar amount is invested in each share at the beginning of the month, all shares are then held to the end of the month, when the whole portfolio (including rights) is sold out. The proceeds are then re-invested and the process repeated.” See Philip Brown, “The Stockmarket in Retrospect,” Economic Activity, 12 (July. 1969) 7. 163

18. For each firm, the market index is estimated by the arithmetic mean of the share price relatives for the remaining 117 firms. 19. Because the Potter survey contains firms which historically have been more successful than the average Australian firm, the estimate of 20 per cent is possibly biased downwards. 20. Some negative serial correlation Is to be expected because of small data violations of the assumptions of the model we are using. However, the estimates do not vary much from those reported by Ball and Brown (footnote 4). 21. Ball and Brown [1, p. 168] define and illustrate the Abnormal Performance Index as follows: Define month O as the month of the annual report announcement, and APIM, the Abnormal Performance Index at month M, as:

(where vnm is the abnormal rate of return on security n in month m. Then APIM traces out the value of $1.00 invested (in equal amounts) in all securities n (n = 1, 2, …, Ν) at the end of month −12 (that is, twelve months prior to the month of the annual report) and held to the end of some arbitrary holding period (M = −11, −10, …, T) after abstracting from market effects. An equivalent interpretation is as follows. Suppose two individuals, A and B, agree on the following proposition. Β is to construct a portfolio consisting 164

of $1.00 invested in equal amounts in Κ securities. The securities are to be purchased at the end of month −12 and held until the end of month T. For some price, Β contracts with A to take (or make up), at the end of each month M, only the normal gains (or losses) and to return to A, at the end of month T, one dollar plus or minus any abnormal gains or losses. Then APIM is the value of A’s equity in the mutual portfolio at the end of each month M. 22. The adjustment (for each month) was made by subtracting the third API from unity, and adding this difference to the other two. 23. Nine hundred and seven EPS reports are included. 24. Table 3, row for month 0, column for EPS increases:

25. Similalry, –9.0 per cent = 26. Recall the arbitrary rule by announcement month was determined.

which

the

27. For example, I suspect 20-25 per cent is too low. 28. In this connection it should be noted that the persistence in the drifts in the API’s from month −11 to month 0 (see Fig. 1) is the result of averaging over many firms and years. No such persistence is to be found in the A P I for the “typical” firm and the “typical” EPS report. If it were to be found, then there would be opportunities to profit by applying a mechanical trading rule (which is

165

most unlikely on a priori ground’s). Further, those opportunities would be evidenced by positive auto-correlation in the abnormal rates of return. However, Table 2 shows that autocorrelation, if anything, is slightly negative. 29. See Ball and Brown (footnote 4) p 176 for a detailed description 30. That is, based on the firms and years surveyed in this report. 31. See above, result 1 32. Both calculations are biased against findings in favour of EPS reports; see Ball and Brown (footnote 4) p 176. 33. See above.

© 1970 by CPA Australia

166

Philip Brown* John W. Kennelly† The Informational Content of Quarterly Earnings: An Extension and Some Further Evidence Introduction The usefulness of quarterly earnings per share (EPS) data is subject to some debate.1 Our intent in this paper is to extend empirical knowledge of current reporting in one sense. While we adopt a different approach from the previous studies, our results have implications regarding them, and for future research.2 Previous empirical research addresses the contribution of quarterly EPS to predictions of (1) annual EPS3 or (2) ex post rates of return on common stocks for one- to twelve-year horizons.4 The results appear anomalous.5 We chose a different empirical design, based upon a different though still restricted, view of informational content. Experimental Design Our experimental design follows closely that of Ball and Brown.6 Basically, the method is to use various forecasting rules to predict EPS. Actual EPS for the predicted period is then compared with predicted EPS, and, on the basis of the comparison, actual EPS reports are classified into three categories: those which provide “good” news; those which provide “bad” news; and those which provide “no” news at all. Next, an investment portfolio is constructed (one portfolio for each forecasting rule) by (a) buying those stocks which 167

will report “good” news, (b) selling short those which will report “bad” news, and (c) taking no action on the rest7 as if the contents of the earnings report were known at some arbitrary time in advance of the actual announcement.8 Abnormal monthly rates of return9 are then computed for each portfolio and expressed in the form of an index. Finally, the forecasting rules are evaluated by their ability to maximize this index. This experimental design permits a direct comparison between alternative forecasting rules. Also, one can assess the impact of disaggregating annual EPS into its quarterly components. Both contentions perhaps require explanation. The first contention is based on the hypothesis (and evidence) that the stock market is “both efficient and unbiased in that if information is useful in forming capital asset prices, then the market will adjust asset prices to that information quickly and without leaving any opportunity for further abnormal gain.”10 There is, then, a presumption that the consensus of the market reflects, at any point in time, an estimate of future EPS which is the best possible from generally available data. Since the abnormal rate of return measures the extent to which the market has reacted to errors in its previous expectations, the abnormal rate of return can be used to assess the predictive accuracy of any device which attempts to forecast a number that is relevant to investors. To our knowledge, Ball and Brown were the first to make such use of this fact.11 The second contention can be illustrated by the following heuristic argument. Suppose investor A were

168

told annual EPS, for a set of 100 stocks, twelve months in advance of the rest of the market and that his investment decisions had negligible effect on stock prices. Presumably, A could classify all 100 stocks into three groups, composed of those whose reports he thought were “good” news, those which he thought were “bad” news, and those to which he was “indifferent.” He then could construct a portfolio and hold it for twelve months until the annual reports were released. As a result, A would earn an average rate of return over the twelve months of, say, X percent above what he would “normally expect.”12 Now suppose investor B were told not only annual EPS (twelve months in advance) for the same set of 100 stocks, but also EPS by quarter. Then his investment strategy would probably require him to undertake some additional trading during the twelve months preceding the announcement of the annual report. For example, he may choose to buy Continental Can (assuming it is included in the 100) and hold until the first quarter’s report is released; then sell it short, maintaining a short position for the rest of the second quarter; switch positions at the start of the third quarter by buying after the second quarter’s report is released; and hold for the remainder of the year. Averaged over the 100, B’s portfolio will yield over the full twelve months an abnormal rate of return of Y percent. Ignoring transactions costs (and the cost of providing the additional reports), if the quarterly EPS data allow improvement over reliance solely upon annual EPS data, we would expect Y to be greater than X.

169

This conclusion implies two qualifications of our methodology. First, ignoring costs tends to overstate any “value” assessment of interim reports. Second, a special definition of informational content, which operates against positive results, is employed here. We assess the impact of EPS “information” by the reduction of bias in predictions and ignore potential improvements in predictive efficiency which might arise from sources other than bias reduction. Even negative findings here would leave open the question of the presence of this latter kind of information. Forecasting Models Clearly, the validity of the experiment depends largely on how well earnings reports are classified into “good,” “bad,” and “indifferent.” Errors in classification can result from weaknesses in the data themselves, which is the issue about which we are concerned. If the disaggregation of annual EPS into quarterly EPS adds information, it should reduce errors in classification for shorter periods. That is, by using quarterly data, and trading more frequently, advantage can be taken of the short-run intra-year price movements. On the other hand, errors in classification can result from inefficiencies in the forecast models employed. If the annual forecast model chosen is less efficient than the quarterly models, then models employing the quarterly data may seem “better” independent of differences in the data. Unfortunately, there is no clear way to eliminate this difficulty.13

170

Following Ball and Brown, we define the informational content of an EPS number by the difference between the number and some conditional expectation, or forecast, of that number. For example, suppose we forecast, on the basis of last year’s EPS of $4.00, that EPS for this year will be $4.00. If actual EPS were $5.00, then the EPS report would be deemed “good” news; if actual EPS were $3.00, then it would be “bad” news; and if actual EPS were $4.00, then actual would equal forecast and, apart from any second-order effects,14 the informational content is zero and we are indifferent to reported EPS. The key difficulty is to construct a forecast. This note relies exclusively on previously published forecasting models, which fall into two classes: naïve models and relatively less naïve regression models.15 a) Naïve Forecasting Models Only two naïve models are used here to predict annual EPS, denoted N1 and N2. Model N1 is the classical naïve model which predicts that this year’s EPS will be the same as last year’s.16 Model N2, which predicts that this year’s EPS will be equal to last year’s plus the average change over the available history of the data (that is, since 1951 for the present group of firms),17 is included because it is just N1 corrected for one type of systematic bias.18 Two naïve models are used to predict quarterly EPS, being counterparts to N1 and N2. Thus, the first naïve quarterly model, QN1, predicts that EPS for a specified quarter of this year will be the same as EPS for the same

171

quarter of last year;19 the second one, QN2, predicts that it will be equal to EPS for the same quarter of last year plus the average change in that quarter EPS which has occurred over available history. b) Regression Models Regression models malee use of the fact that, historically, EPSs of firms have tended to move together.20 Thus, a decrease in the EPS of a particular firm by $1.00 would not indicate “bad” news if, given what happened to other firms, its EPS were expected to have decreased by $2.00. Models based on the association between the EPS of a firm and the EPS of some market average (or average of other firms) are labeled “regression models” simply because ordinary least squares is used to estimate that relationship from previous history. Knowledge of the historical relationship, together with a knowledge of the EPS of those other firms for the present year, yields a conditional expectation for the firm’s present EPS. The specification adopted here is identical to that employed by Ball and Brown except that market EPS is approximated by Standard and Poor’s Industrials Average EPS.21 A major difficulty with estimating the regression relationships is the limited number of degrees of freedom. The naïve model adjusted for drift and the regression model are closely related: the regression model degenerates to the naïve model when the market term is suppressed. It turns out, in the present case, that the loss of another degree of freedom, when the market term is added, is sufficient to offset any benefits that

172

derive from the improved specification.22 Thus, the regression model results are reported more to show specification and estimation weaknesses than for any other purpose. c) Summary of Forecasting Models Two types of models are used to forecast EPS, naïve models and regression models. The forecast is then subtracted from actual EPS to define a forecast error. If that forecast error is positive, then EPS for that particular period and according to that particular forecaster is deemed to convey “good” news; if it is negative, then EPS is deemed to convey “bad” news; and finally, if it is zero, then EPS is deemed to convey “indifferent” news. Data Data are for ninety-four of the 100 Compustat firms referred to by Brown and Niederhoffer. The 100 were systematically selected as being of interest to a particular group of security analysts. They are typical of most Compustat firms in that they are larger and tend to be less risky23 than the average firm. The ninety-four selected here also close their books on December 31. Strictly speaking, any conclusions should be confined to the group of firms and the earinings period (195–167) included in the study. Although there is no obvious reason to believe that the ninety-four are not typical of a wide class of firms, the task has only begun, as part of a larger study, of collecting additional data to validate the results.24 The EPS data, adjusted for stock splits and dividends, are as supplied by the analysis. Monthly stock price 173

relatives from January 1946 through March 1966 are obtained from tapes constructed by the Center for Research in Security Prices (CRSP) at the University of Chicago.25 These CRSP data are augmented by price relatives calculated from price and other data reported in Standard and Poor’s ISL Daily Stock Price Index of the New York Stock Exchange. Quarterly and annual report announcement dates are taken from the Wall Street Journal Index and verified in the Wall Street Journal.26 As in Ball and Brown, the preliminary report date is assumed to be the date on which reported earnings were considered to be generally available. To maintain a standard of comparison (with the Ball and Brown study), announcement dates are collected beginning with the first quarter of 1958. Results Six different forecasting rules have been defined, three of which identify the informational content of annual EPS and three more, counterparts to the others, which identify the informational content of each quarter’s announcement. If we think of the annual forecasting rules as indicating that the same investment strategy should be adopted for each of the four quarters in any one fiscal year, then we have for each firm effectively forty periods (10 years × 4 quarters), with six different predictions about the appropriate strategy for each period. To reduce the complexity of the results, we chose to treat the fiscal year as the basic unit of time and pool all years together. The results are then summarized, by month

174

relative to the annual EPS announcement, in the form of (a) abnormal performance indexes27 and (b) χ2 statistics calculated from a two-by-two classification of each report by the sign of the EPS forecast error and the sign of the abnormal rate of return. The abnormal performance indexes give an indication of the average magnitude of the association between the informational content of the half-yearly report and the percentage change in a share’s price, while the χ2 test is a test of the statistical significance of the association, for it could well be the case that the association is only the product of chance. The following example illustrates the procedure. Suppose the four 1968 quarterly reports for Continental Can were released in April, July, October (1968), and February (1969). Suppose further that, according to the annual EPS regression model, 1968 was a “good” year, although according to the quarterly EPS regression model, the first, third, and fourth quarters were “good” while the second was “bad.” Then the investment strategies in table 1 would be assumed in constructing the 1968 APIs for Continental Can. Now assume that the relevant monthly abnormal rates of return for Continental Can were as given below. Table 1 Investment Strategy Date February 29, 1968 ……………………

Annual Model Buy

175

Quarterly Model Buy

April 30, 1968 ……………………… July 31, 1968 ………………………. October 31, 1968 …………………….

1968:

(Hold)

Sell short

(Hold)

Buy

(Hold)

(Hold)

March, −.02 April, +.03

July, −.00 August, +.01 September, −.03 May, +.01 1969: June, −.05 October, +.10

November, −.02 December, −.01 January, +.05 February, +.02

Then the sequence of the two APIs (for the first four months of Continental Can’s fiscal year 1968) would be calculated as in table 2. Note that, since both models predict a buy policy for the first quarter, the APIs for the first two months are identical. At the start of the second fiscal quarter, however, the quarterly regression model predicts a switch from a buy to adopting a short position. Thus, May’s abnormal rate of return, a gain of 1 percent, would be converted to a loss of 1 percent, while June’s loss of 5 percent would become a gain of 5 percent. Table 2 Abnormal Performance Index Month −11 ………………. −10 ………………. − 9 ………………. − 8 ……………….

Annual Model

Quarterly Model

.98 .98 .98 × 1.03 .98 × 1.03 .98 × 1.03 × 1.01 .98 × 1.03 × 99 .98 × 1.03 × 1.01 .98 × 1.03 × .99 × .95 × 1.05

176

Finally, entries to the two-by-two contingency tables, for the quarterly model only, would be as given in table 3. Table 3 Entries to Contingency Tables—Quarterly Model

Tables 4 and 5 present APIs and χ2 statistics for the three annual and the three quarterly forecasting models, respectively, calculated over all firms and years in this study. Figure 1 plots four of the six APIs,28 plus the API that results from buying and holding every firm in the group. This last API serves as a control. Table 4 Summary Statistics by Month Relative to Annual Report Announcement Date—Annual Forecasting Models

177

NOTE.—P (χ2≥ 3.84| χ2 = 0) = .05 for 1 df; P (X2 ≥ 6.64] = χ2 = 0) = .01 for 1 df. Table 5 Summary Statistics by Month Relative to Annual Report Announcement Date—Quarterly Forecasting Models

178

Fig. 1.—Abnormal performance index for selected portfolios Four aspects of our results, and their limitations, should be noted. 1. The systematic inferiority of the annual models to their quarterly counterparts is striking. For example, the API on N1 (table 4) reaches 1.058 by month 0, while the API for its counterpart (QN1 in Table 5) reaches 1.076; the corresponding numbers are for N2 and QN2, 1.063 and 1.089, and for the regression models, 1.064 and 1.086. Thus, interim reports increase, albeit only by some 30–40 percent, the predictive content of annual EPS.29 179

Why only 30–40 percent and not 300 percent? We believe the answer lies in three different directions: (a) a relationship between the magnitude of the difference between actual and expected EPS and the magnitude of the stock market’s reaction,30 (b) shortcomings in our quarterly predictive model’s specification (for example, see n. 19 above); and (c) the special difficulties accountants face when measuring EPS over time periods shorter than one year. Future research will be pursued in each of these directions. 2. The regression models are not obviously better or worse than the naïve models corrected for drift.31 Presumably, this casts doubt on estimates of the association between earnings of firms which are based on only a few years’ data. Alternatively, it may imply only that the process generating income data over time is well approximated by the N2 model. In any event, it is difficult to find regression models which outperform the naïve ones adjusted for drift. Ball and Brown (who employed N1 to check on the efficiency of their estimates of the earnings association) found the regression model better than N1, a finding supported by table 4. 3. Other results of Ball and Brown are also substantiated.32 In particular, most of the information contained in reported annual EPS has been incorporated in stock prices well in advance of the announcement date. Furthermore, the market’s anticipation of actual EPS is sufficiently accurate that its release does not appear to

180

cause any unusual jumps in the API in the announcement month. 4. The clustering of the larger values of the χ2 statistics around what appear to be the quarterly report announcement dates indicates that the mechanism whereby the market is able to anticipate annual EPS bears closer examination. Tables 6 and 7 employ the same data as in tables 4 and 5, but give the χ2 statistics calculated by month relative to the quarterly report announcement month. If quarterly reports were the sole source of information about annual EPS, then the χ2 statistics in table 7 should be significant in the four quarterly announcement months but insignificant in all others. Table 6 Summary Statistics by Month Relative to Quarterly Report Announcement Date—Annual Forecasting Models

181

Table 7 Summary Statistics by Month Relative to Quarterly Report Announcement Date—Quarterly Forecasting Models

182

In all but the last quarter, the χ2 statistics in table 7 are higher in month 0 than in any other month. This suggests that the market is limited in its anticipation of the contents of a quarterly report except possibly during the announcement month itself. Unfortunately, at this stage the experimental design imposes limitations. The fact that price relatives are observed only at the end of each month prevents a closer examination of the announcement mechanism. Presumably, weekly or even daily prices would be of greater help. Conclusions This inquiry has deliberately avoided selecting any one particular quarter’s report for special examination. Accordingly, there are just two major conclusions: (1) The information contained in quarterly EPS reports is

183

useful in that it can be used to predict aggregate abnormal rates of return on the securities to which the EPS numbers relate. (2) Disaggregation of annual EPS into its quarterly components improves the predictive ability of the EPS series by at least 30 to 40 percent. In addition to these major conclusions, the results have interesting implications relating to previous research. First, as Green and Segall and Brown and Niederhoffer recognized,33 prediction of the next annual EPS may not be the only objective of quarterly reports. Given the validity of the alternative purpose specified here, our results suggest that currently prepared reports are useful independent of any annual prediction difficulty.34 Further, the fact that second and third quarter reports allow additional improvement of API predictions suggests that first quarter results are less than perfect predictors of the remaining nine months’ results. Our results also suggest that one- to twelve-year holding periods may be inappropriate return bases for evaluating the usefulness of quarterly data.35 Such a holding policy implies a decision interval of such great length that the information derived by disaggregating the annual series would probably be overwhelmed by residual noise. We hope that this note has presented some new insights and some new evidence which will be useful to others exploring the topic.

* University of Western Australia.

184

† University of Iowa. We are indebted to Ray Ball and Nicholas Gonedes (University of Chicago), William Beaver (Stanford University), and the members of the Seminar in Accounting Research at the University of Iowa for helpful comments. 1. See David Green, Jr., and Joel Segall, “The Predictive Power of First-Quarter Earnings Reports,” Journal of Business 40 (January 1967): 44–55; “Brickbats and Straw Men: A Reply to Brown and Niederhoffer,” ibid. 41 (October 1968): 498–502; “Return of Strawman,” ibid. 43 (January 1970): 63–65; Philip Brown and Victor Niederhoffer, “The Predictive Content of Quarterly Earnings,” Journal of Business 41 (October 1968): 488–97; Victor Niederhoffer, “The Predictive Content of First-Quarter Earnings Reports,” Journal of Business 43 (January 1970): 60–62; Thomas L. Holton, “Discussion of the Predictive Power of First-Quarter Earnings Reports: A Replication,” Journal of Accounting Research 4, suppl. (1966): 37–39; George J. Staubus, “Earnings Periods for Common Share Analysis,” Journal of Business 41 (October 1968): 472–76; and Glenn A. Welsch, “Discussion of the Predictive Power of First-Quarter Earnings Reports: A Replication,” Journal of Accounting Research 4, suppl. (1966): 40–43. 2. For example, John W. Kennelly, “The Utility of Interim Reporting,” unpublished manuscript (1969), extends the analysis to more firms and to some other possible reporting techniques. 3. Green and Segall, “Predictive Power” and “Brickbats and Straw Men”; and Brown and Niederhoffer.

185

4. Staubus. 5. Green and Segall, “Brickbats and Straw Men”; and Staubus. 6. Ray Ball and Philip Brown, “An Empirical Evaluation of Accounting Income Numbers,” Journal of Accounting Research 6 (Autumn 1968): 159–78. 7. The number of stocks which fall into each class depends upon the forecasting rule adopted. Forecasting rules used in this note resulted in all EPS reports being classified as either (a) or (b). 8. One reason for constructing the portfolio in advance of the actual announcement is that the contents of the report are anticipated by the market in advance of the announcement date (see Ball and Brown). 9. An abnormal monthly rate of return is defined as the one-month ex post rate of return net of market effects. The estimation process was established previously (see Eugene F. Fama et al., “The Adjustment of Stock Prices to New Information,” International Economic Review 10 [February 19691: 1–26). Market effects are removed by taking the residuals about the ordinary least-squares regression of the monthly price relatives of security j on a market index. The market index is constructed by averaging the price relatives of all securities (other than security j) in the set of firms included in the study. 10. Ball and Brown. For a survey of the theoretical and empirical evidence, see Eugene F. Fama, “Efficient Capital Markets: A Review of Theory and Empirical Work,” Journal of Finance 25 (May 1970): 383–417.

186

11. Earlier, George J. Benston, “Published Corporate Accounting Data and Stock Prices,” Journal of Accounting Research 5, suppl. (1967): 1–14, 25–54, used rates of change in security prices “adjusted for the average effect of changes an general market conditions.” However, he presumed the changes in such net of market prices to be linearly related to changes in accounting data. If reported accounting changes convey information about expected risk as well as return (see William H. Beaver et al., “The Association between Market Determined and Accounting Determined Risk Measures,” Accounting Review 45 [October 1970]: 654–82), a linear relation might not be expected. Hence, we, as Ball and Brown (see n. 6 above), look only at directions of change. 12. “Normally expect” is defined in relation to the rate earned by the market. That is, in the absence of information uniquely applicable to an individual firm, we would expect its return responses to reflect only market-wide adjustments in expectations. These are abstracted from the returns of firms by regression on a market index. The residuals are the “unique” changes in expectations (see Ball and Brown, and William H. Beaver, “The Information Content of Annual Earnings Announcements,” Journal of Accounting Research 6, suppl. [1968]: 67–92). Earlier, Benjamin F. King, “Market and Industry Factors in Stock Price Behavior,” Journal of Business 39, pt. 2 (January 1966): 139–90, estimated the extent to which market, industry, and unique factors “explain” price movements. Ball and Brown did the same for earnings data; but see our regression results below.

187

13. An attempt to deal with this problem is part of the larger study of quarterly reporting in Kennelly. 14. Information regarding risk, such as that reported by Beaver et al., is not evaluated in our design. 15. While regression models are intuitively more appealing, their empirical superiority over the “naïve” models is yet to be conclusively demonstrated. Compare, for example, the relative predictive abilities of variables (2) and (3) in Ball and Brown, and see below. Actually, all of these models are members of a more broadly defined “naïve” class (Victor Zarnowitz, An Appraisal of ShortTerm Economic Forecasts, Occasional Paper 104 [New York: National Bureau of Economic Research, distributed by Columbia University Press, 1967], p. 83). Also, Ray Ball and Ross Watts, “The Behavior of Earnings through Time” (unpublished paper presented at the Workshop in Accounting Research, University of Chicago, April 23, 1968), provide empirical evidence on the behavior of income times series which implies that results such as those reported in Ball and Brown might be expected. 16. We chose this purposely for its relationship to the previous empirical studies. 17. However, as Ball and Watts point out, N2 is a near-optimal rule for the extensive set of alternative rules they considered. 18. See Ball and Watts. 19. QN 1 results from comparing moving annual sums of quarterly EPS. It has its origin, in the present context,

188

in Green and Segalľs Interim 4 (it is a well-known naïve correction for seasonalities). Its major shortcoming lies in the fact that any information contributed in the intervening three quarters is automatically ignored. 20. See, for example. Philip Brown and Ray Ball, “Some Preliminary Findings on the Association between the Earnings of a Firm, Its Industry and the Economy,” Journal of Accounting Research 5, suppl. (1967): 55–77: Ball and Brown; and Richard A. Brealey, “The Influence of the Economy on the Earnings of the Firm” (unpublished paper presented at the Sloane School of Finance Seminar, Massachusetts Institute of Technology, May 1968). 21. That is, defining M as Standard and Poor’s Industrial Average EPS and Ij as the earnings of firm j for a period, we employed ordinary least squares to estimate, over available history to, but not including, the period forecasted:

The coefficients were used, with the S and P index, to generate a conditional expectation of the change in earnings, defined:

Then, the error.

189

is the information conveyed. (Recall that we are interested only in the sign of ûjį.) See Ball and Brown, pp. 161–62. 22. One obvious avenue for investigation lies in the pooling of quarterly data, perhaps by using dummy variables to remove seasonalities. This procedure may tend to overcome the shortcoming mentioned above in n. 17. 23. In the Sharpe “systematic risk” sense. 24. Kennelly. 25. The Center for Research in Security Prices at the University of Chicago is sponsored by Merrill Lynch, Pierce, Fenner, and Smith, Incorporated 26. We are indebted to Messrs. James Nelson and Eustace Nedd at the University of Iowa for their assistance in verifying portions of these data. 27. Ball and Brown, p. 168, define and illustrate the abnormal performance index as follows: Define month 0 as the month of the annual report announcement and APIM, the abnormal performance index at month M, as:

where vnm is the abnormal rate of return on security n in month m. Then API M traces out the value of $1.00 invested (in equal amounts) in all securities n (n = 1, 2, …, N) at the end of month −12 (that is, twelve months prior to the month of the annual report) and held to the 190

end of some arbitrary holding period (M = −11, −10, …, T) after abstracting from market effects. An equivalent interpretation is as follows. Suppose two individuals A and B agree on the following proposition: B is to construct a portfolio consisting of $1.00 invested in equal amounts in N securities. The securities are to be purchased at the end of month −12 and held until the end of month T. For some price, B contracts with A to take (or make up), at the end of each month M, only the normal gains (or losses) and to return to A, at the end of month T, $1,00 plus or minus any abnormal gains or losses. Then APl M is the value of A’s equity in the mutual portfolio at the end of each month M. 28. Only the four naive models are plotted to avoid cluttering the picture. 29. These estimates depend on the forecasting rules adopted (see below for further discussion of this important constraint). 30. However, one possible interpretation of the “added information” of the quarterly reports is that the number of, for example, positive forecast errors detected by the quarterly model is a crude measure of the magnitude of the annual error. 31. At month 0 the APIs were: for N2 and the annual regression model. 1.063 and 1.064; and for QN2 and the quarterly regression model, 1.089 and 1.086. The sign test statistic z from a comparison of the APIs (for each firm and each year) at month 0 were: N2 versus annual regression model, z = 0.00; and QN2 versus quarterly regression model, z = 0.26. In large samples, z, is

191

approximately normally distributed, with probability (|z|) ≥ 0.26 | Z = 0) = 0.8. The more powerful Wilcoxon signed-rank test yields the same conclusion: that any differences presumably were due to chance. 32. One reason is that seventy-one of the ninety-four were also included in the 261 studied by Ball and Brown. 33. Green and Segall, “Predictive Power,” p. 44, and Brown and Niederhoffer, p. 489. 34. However, our results in no way suggest that Green was incorrect in suggesting that seasonal adjustment of interim data would improve these reports. Rather, we provide a benchmark against which such improvement can be compared. It is interesting to observe that Brown and Niederhoffer’s worst interim predictors (as well as Green and Segall’s worst) are those most affected by the seasonal aberrations noted in Green’s earlier paper. 35. See Staubus. In fact, we wonder, as well, about holding periods greater than one year for annual data evaluation.

© 1972 by The University of Chicago. All rights reserved.

192

Order Flow and Price Effects Surrounding an ASX Announcement Philip Brown*, Bryan Howitt** and Marvin Wee** * Schools of Accounting and Banking & Finance, UNSW; and UWA Business School ** UWA Business School Abstract Brown, Forbes and Wee (2003) examine order flows surrounding four types of announcement on the Australian Stock Exchange (ASX). They conclude that ASX “largely gets it right” when classifying some announcements as market sensitive and others as not. We extend their study in two important ways: (1) by examining trade- to-trade price changes and time elapsed between successive trades as well as order flow; and (2) by testing a substantially more comprehensive set of announcement types. We find the majority of market sensitive announcement types display abnormal order flow behaviour before the release of the announcement document. Similarly, order flow subsequent to a market sensitive announcement exhibits abnormal behaviour. Absolute share price change surrounding market sensitive announcements is also significantly larger than at other times. The added time delay between trades is harder to interpret because of the interposition of the trading halt itself. Keywords: Market sensitive announcements; Order Flow; Price Impact; Event Studies; Trading Behaviour

193

JEL Classification: G12, G14, G18 1. Introduction Our primary aim is to investigate order flow and share price changes when companies on the Australian Stock Exchange (ASX) announce “market sensitive” news. Specifically, we address the question, “Do market sensitive announcements contain information that causes significant changes in order flow and price?” Our contribution is to extend recent papers by Brown, Ng and Wee (2003) and Brown, Forbes and Wee (2003). Brown, Forbes and Wee (2003) find a higher level of order flow around announcements assessed by the ASX to be market sensitive but not for non–market sensitive announcements. Brown, Ng and Wee (2003) investigate the order reaction to companies announcing takeovers and earnings news via the ASX information service. They find the release of the headline of an earnings announcement is associated with significant increases in order imbalance and order revisions. However, investors do not react immediately to the release of market sensitive news but are influenced by the trading process. We extend Brown, Ng and Wee (2003) and Brown, Forbes and Wee (2003) in two major respects. First, neither study considers price change or time between trades, whereas we do. Second, due to constraints, these two previous studies review only a few event types. We expand the number of event types substantially. The remainder of our chapter is structured as follows. Section 2 provides a review of literature relevant to our study. Section 3 discusses the institutional background of 194

the ASX. Section 4 presents the hypotheses, and Section 5 details the various data sources and the research method. Section 6 presents the results, and our conclusions are summarised in Section 7. 2. Related Literature In this section, we discuss relevant literature on investor behaviour, order flow and price movements around news releases. As news releases are often associated with a trading halt, we discuss briefly the literature on trading halts. 2.1. Models of Investor Behaviour Around a Market Sensitive Release Two key articles that discuss the linkage between information asymmetry and forthcoming announcements are Demski and Feltham (1994) and McNichols and Trueman (1994). Both studies consider information asymmetry before a price-sensitive announcement. Demski and Feltham (1994) demonstrate that, subject to certain conditions, a fraction of traders are motivated to engage in costly private information acquisition in order to speculate on the forthcoming public announcement. Similarly, McNichols and Trueman (1994) predict that a forthcoming public announcement will motivate traders to privately acquire costly information before the announcement. Other studies address information asymmetry after a price-sensitive disclosure. Diamond (1985) shows that the public disclosure (announcement) of company information reduces the incentives for private information acquisition because it makes traders’ beliefs 195

more homogeneous and decreases the magnitude of speculative positions that informed traders take. On the other hand, Lundholm (1991) and Kim and Verrecchia (1994) find that information asymmetry can increase after a public company announcement due to (1) private information being concentrated in the hands of very few informed traders or (2) differences in traders’ abilities to process the information contained in the announcement. 2.2. Empirical Studies of Order Flow Around a Market Sensitive Release On ASX, market sensitive releases prompt a trading halt, which restricts investor actions as they are unable to trade until the halt is terminated. This leads to interesting behaviour by market participants before, during and after the halt. Earlier analyses of order flow around a market sensitive release by ASX include Aitken, Brown and Walter (1994). They examine order flow and the state of the limit order book around two types of market sensitive announcements—preliminary final statements and takeovers—using an event window of ±5 trading days, which they partitioned into 30-minute intervals. Their source of release times was ASX’s Signal G, which typically lags the “official” ASX release.1 They report that before an announcement there is a decrease in the number of new orders, a widening of the bid ask spread, less depth on the limit order book and fewer trades. Subsequent to the announcement there is an increase in order flow activity that persists for several days. Brown, Ng and Wee (2003) and Brown, Forbes and Wee (2003) examine order flow activity around a market

196

sensitive release using data that more accurately identifies the timing of the release. Brown, Forbes and Wee (2003) find a higher level of order flow around the release of information. Also, order flow around non–market sensitive announcements does not exhibit the same degree of abnormal behaviour. Similarly, Brown, Ng and Wee (2003) find the release of the headline of an earnings announcement is accompanied by significant increases in order imbalance and order revisions. They also find an increase in order imbalance and order revisions for takeover announcements. An important result is that investors do not react immediately to the release of market sensitive news. These effects are not confined to Australia. In the U.S., Corwin and Lipson (2000) document that the number of orders placed during a trading halt on the New York Stock Exchange (NYSE) is abnormally high. However, they find, in general, no consistent abnormal activity for new orders entered, which may imply that traders are unsure about the price implications of the announcement and adopt a wait and see approach. This suggests that investors need time to digest the information contained in market sensitive releases. Similarly, they find trading volume increases after trading resumes. 2.3. Price Changes and Trading Halts Hopewell and Schwartz (1978) conducted one of the earliest analyses of price adjustments that occur during the trading halt period. They find the magnitude of price adjustments around temporary trading suspensions on the NYSE is consistent with large and permanent changes in equilibrium prices. Similarly, Ackert, Church

197

and Jayaraman (1999) compare investor behaviour across three different market phases: market closure, temporary halt and no interruption. The absolute deviation of price from the expected price is not significantly different across the three. They also observe that order flow activity is significantly affected by a trading halt. Other studies examine price volatility after a trading halt. Lee, Ready and Seguin (1994) find trading volume and price volatility are significantly higher for less than a day before a halt and for over two days after it. They conclude that, despite the significant amount of news released during a halt, there appears to be greater uncertainty and trading costs around the halt. Aitken, Frino and Winn (1996) replicate Lee et al. (1994) and find that in Australia the price discovery process after a market sensitive announcement has similar properties to the U.S. market. In particular, trading halts are accompanied by elevated stock price volatility and abnormally high trading volume.2 2.4. Trading Halts and Information The costs and benefits of trading halts are widely debated. Proponents such as Stein (1987) and Greenwald and Stein (1991) argue that halts allow investors time to react to material news events and market makers to search for a new price level. However, critics such as Slezak (1994) and Subrahmanyam (1994) argue that halts are an unnecessary barrier to trading, which is supported by empirical studies that find unusually high levels of volatility following halts.

198

3. Institutional Background Most trading on ASX is conducted via the Stock Exchange Automated Trading System (SEATS), which is a network of terminals located in brokers’ offices. Through each terminal, member firms perform a variety of tasks, such as submitting new orders, viewing the limit order book and receiving company announcements. When an entity listed on ASX decides to make an announcement, it is required by ASX Listing Rule 15.7 to transmit the announcement to the ASX Company Announcements Office (CAO) before it is released to external parties.3 When an announcement is received by the CAO, it is first examined on-screen by specialist staff, who assess the announcement in several steps. First, the announcement is analysed to decide whether it is for public consumption. If so, it is classified as either “market sensitive” or “not market sensitive.”4 If an announcement is classified as market sensitive and is received during normal trading hours (from around 10:00 a.m. to about 4:00 p.m.),5 the CAO requests a halt to trading in securities for that particular announcing firm (and any other listed firms directly affected by the announcement). Trading is then halted by SEATS Market Control (SMC), which tags a “Notice Received” flag against the stock and changes the trading phase of the stock from “Normal” to “Preopening.” During the preopening phase, no trading in the particular stock takes place. However, brokers can submit new orders and amend or cancel existing orders. Typically within a minute after trading in a stock is halted, SMC enters a brief alert header message into the

199

SEATS system, which appears on all SEATS trading terminals.6 This header message contains a number of fields, including the summary headline of the announcement, the announcement’s classification,7 the issuer code, an indicator of whether the announcement is market sensitive and the date and time at which the header is released by the CAO. Almost immediately8 after the header’s release, the CAO releases the full PDF file of the announcement document to ASX’s dissemination server, for distribution to licensees.9 At the time the announcement is released by ASX, an automatically generated acknowledgement is both e-mailed and faxed by the CAO to the listed entity. Shortly after the announcement is released to the dissemination server, a summary of the announcement is broadcast over ASX’s Voiceline public address system.10 These summary releases can be made in several parts depending on the nature of the announcement. About 10 minutes11 after the transmission of the header message, SMC changes the market status from Preopening to Opening and then Opening to Normal. These status changes occur within an instant of each other and mark the end of the trading halt. After the release of the full announcement document to the dissemination server, the original announcement document is edited into a text message and transmitted to subscribers via an electronic signal known as Signal G. The Signal G records contain both the header message and the essence of the announcement. Similar to the Voiceline summary, during busy periods the Signal G records are often

200

released in separate chunks. The timing of the Signal G releases is, however, not fixed. While the Signal G records could be transmitted before trading resumes, some are sent afterward. Figure 1 shows a timeline of the different stages of the ASX announcement process.

Figure 1 Timeline of steps in the ASX announcements process for the release of market sensitive announcements There are nine steps in total, five of which are used in our study. Note that the time duration of 10 minutes between Step 3 and Step 6 does not apply to takeover announcements. The duration of a trading halt for a takeover announcement is at least 50 minutes. Signal G transmissions usually begin between Steps 4 and 6 although they are shown after Step 7 on this diagram. The timeline is not to scale. 4. Hypotheses This section discusses our hypotheses concerning market behaviour and price changes across the ASX announcements process. They relate to order flow leading up to and subsequent to an announcement, order flow during a trading halt and the price change and time

201

lag between the last trade before and the first trade after a market sensitive release. 4.1. Before the Announcement Theoretical models posit that information asymmetry between informed and uninformed traders increases before a timely announcement, due to traders’ incentives to acquire private information (Demski and Feltham, 1994; Kim and Verrecchia, 1991a).12 Models by Admati and Pfliederer (1988) and Foster and Viswanathan (1990) show that discretionary liquidity traders (uninformed traders) have an incentive to delay their transactions when they believe informed traders are present in the marketplace. An alternative strategy is for uninformed traders to revise their standing limit orders to reduce the probability of trading with an informed counterparty. H1: Prior to a market sensitive announcement, uninformed investors protect themselves by moving away from the market or withdrawing from it. While it is hypothesised that uninformed investors will price protect themselves, these options may only be available when an announcement is predictable and timely, because uninformed traders need time to implement their price protection strategies. Kim and Verrechia (1991b) demonstrate that informed traders prefer to trade around information events because their information advantage is maximized at this time. In a related study, Foster and Viswanathan (1990) posit that information held by informed traders has a limited life span. They imply that informed traders will enter the 202

market before the announcement to maximise their potential profit.13 However, we do not expect to observe abnormal order flow activity around non–market sensitive announcements, because according to ASX they should not contain information not already incorporated into the share price. H2: Before a market sensitive announcement, informed traders attempt to maximise their profit by placing orders, which is reflected in greater order imbalance. 4.2. After the Announcement An uninformed trader’s risk of loss by trading with an informed counterparty is lowered by a public announcement, which substantially reduces the value of private information (Verrecchia, 1982). As uninformed investors respond to the information in the announcement, there will be an abnormally high level of order flow and trading activity. H3: After a market sensitive announcement, there is an increase in order flow as uninformed traders re-enter the market. 4.3. During the Trading Halt As previously discussed, theoretical models such as Verrechia (1982) and Diamond (1985) indicate that the public disclosure of private information reduces information asymmetry because uninformed traders receive the information. While trades cannot occur during a trading halt, uninformed traders can react to the new information in the form of new and revised orders. This was shown in Corwin and Lipson (2000), who

203

found increases in order flow during temporary trading halts on the New York Stock Exchange. H4: There is an increase in new orders, revised orders and order imbalance in the release to resume phase. Not all traders possess the same ability to process information. Thus, we predict that traders with superior processing ability will conceal their orders until the final minutes before the resumption of trading to maximise profit (Kim and Verrecchia, 1994). Uninformed traders will not be able to gain any advantage from the signalling of order flow prior to the resumption of trading, as modelled by Glosten and Milgrom (1985). H5: There is abnormal order flow in the final minutes of the release to resume phase as informed traders begin to reveal their private information. 4.4. Price Change and Time between Surrounding a Market Sensitive Announcement

Trades

In efficient capital markets, as defined by Fama (1976), “prices reflect all available information promptly and correctly, including the information comprised in announcements.” Announcements classified by ASX as market sensitive are presumed likely to contain information that will affect the company’s share price. The Efficient Markets Hypothesis predicts that this information will be quickly impounded in the share price. We also expect traders to respond quickly to market sensitive news to adjust their portfolios to their new preferred positions.

204

H6: There is an abnormal share price change surrounding a market sensitive announcement. H7: The time between the last trade before and the first trade after a market sensitive announcement is less than on other days when there is no announcement. 5. Data This section details the sources of data and our sample design. We first describe the RB records used to identify the timing of the news release. We then discuss the SEATS transactions database (which contains information relevant to our order flow metrics), our sample selection criteria and sample design. 5.1. RB Records The RB record is a company announcement alert header. It briefly describes the nature of the detailed announcement that will follow. The header has several data fields: • Issuer Code: A three-letter ASX code that identifies the company to which the announcement relates. • Report Code: A five-digit code that corresponds to the CAP report type. • Industry Subgroup: The subgroup to which the company belongs. • Market Sensitivity Flag: A flag indicating whether an announcement is market sensitive (Y) or non–market sensitive (N), as judged by ASX. • Entered Date: The date on which the announcement was entered into SEATS.

205

• Entered Time: The time at which the announcement was entered into SEATS. We use the report code to identify the type of announcement. 5.2. SEATS Intraday Database All data are originally sourced from ASX and obtained from the Securities Industry Research Centre of Asia-Pacific (SIRCA). Specifically, we use Stock Exchange Automated Trading System transaction records and ASX information releases. The SEATS archive held by SIRCA contains control records designating trading phase changes and records of order flow and trades. This intraday database captures all computer transactions processed by SEATS within a trading day and includes information on orders and trades. Each transaction has several fields attached that are important to our research. In our study we utilise the transaction type (e.g., a new bid order entered on SEATS, an amendment to an existing ask limit order or a trade), ASX issuer code (denotes the security to which the transaction relates), the date and time in which the transaction took place and the price and volume of the transaction. Our order flow and price metrics were derived from these transactions.14 5.3. Sample Design By arrangement with ASX, the sample includes announcements released by firms during the period 1 July 1998 to 30 June 2003. To observe trading behaviour before and after each announcement, we impose a minimum trading time of 30 minutes before and after the

206

release. This reduces but does not eliminate the confounding effects problem associated with start of day and end of day empirical regularities.15 The following sections outline the selection process implemented for the three subsamples used in this study. 5.3.1. Comparison of Market and Non–market Sensitive News Releases—Sample A Sample A comprises announcements used to compare order flow between market sensitive (MS) and non–market sensitive (NMS) announcements. This sample contains four announcement types: (1) disclosure documents, (2) asset acquisitions, (3) asset disposals and (4) responses to ASX queries. We exclude announcements that were not unique in the sense that they contain more than one announcement type. Announcements are also excluded if the halt starting date, halt ending date and announcement release date are not the same. To avoid contaminating cases with market movements possibly related to other announcements by the same firm, we also exclude from the sample each case where a separate announcement that ASX judged to be market sensitive was released on the same day or on either of the two trading days immediately preceding it. Table 1 summarizes the final sample. Table 1 Final sample size by announcement type and market sensitivity. The table shows the number of cases, by announcement type, after all filters have been applied for non–market sensitive announcements (NMS), market

207

sensitive announcements anchored on the release time (MS-RT)

5.3.2. Order Flow during a Trading Halt—Sample B Sample B is used for the analysis of timely (predictable) market sensitive announcements from the release to resume market phase. We initially identified 5,097 earnings announcements known as preliminary final statements (PFSs).16 We exclude from Sample B all such announcements classified as non–market sensitive and announcements with a missing halt or resumption date or time, which reduces the sample to 1,360 cases. We only include announcements that have halt times after the normal market opening time of around 10:00 a.m. and resumption times before the normal market closing time of 4:00 p.m. Missing control records, mostly those designating the start of a trading halt, limit the initial sample set. The final sample of PFSs comprises 1,221 announcements. 5.3.3. Order Flow, Price Change and Time between Trades—Sample C First, announcement types that are deemed by ASX never to be market sensitive are excluded. The remaining 63 announcement types are classified by ASX as “always sensitive” or “may be sensitive.” A filter is applied to eliminate releases deemed non–market sensitive by ASX. Of the 63 remaining announcement

208

types, we exclude 22 announcement types with fewer than 40 individual market sensitive releases.17 This process results in a final potential sample of 41, from which 19 different announcement types are selected.18 A number of restrictions are placed on the data used to estimate the price change and time between trades. First, we process a company’s announcement (to calculate experimental and control metrics) only if there are at least 50 useable event days within the window of ± 125 trading days. We also exclude announcements if there is no price recorded either before the announcement (within the specified window of 1 hour) or after the announcement (within the specified window of 1 hour).19 The final sample is shown in Table 2. Table 2 Announcement types used. Column three presents the number of releases for each announcement type made between 10:30 a.m. and 3:30 p.m. for the period 1 July 1998 to 30 June 2003. Columns four and five present the number of market sensitive and non–market sensitive announcements, respectively. The percentage of market sensitive announcements is shown in column six. The seventh column presents the number of announcements tested in the price change program and time between trades analysis.

209

6. Method 6.1. Metrics Four aspects of market activity are measured: (1) order imbalance, (2) new orders placed, (3) orders revised and (4) trades. We normalise the metrics so that when portfolios are formed for announcements of a particular type, large firms do not dominate the sample. Normalisation is done in the following manner. First, for a given intraday time interval, we measure the raw metric for that time interval for each trading day within the event time window ±125 trading days and compute the metric’s mean and standard deviation. For each day’s metric we then 210

subtract the mean and divide the result by the standard deviation. Activity is measured in dollar volume and transaction frequency (intensity), which is the number of occurrences within the set time period. Frequency is calculated because of the potential for dollar volume to be influenced by one or two very large trades. In particular, institutions that trade in large amounts will disproportionately influence the metric. However, a problem with transaction frequency is that all traders, whether informed or not, have the same weighting.20 6.1.1. Order Placements and Order Revisions To measure order activity, four metrics are examined: (1) ENTBID—new bid orders submitted; (2) ENTASK—new ask orders submitted; (3) REVBID—the aggregation of amended bids, cancelled bids and ticks of bid orders that move the order one price tick closer to the market; and (4) REVASK—the aggregation of amended asks, cancelled asks and ticks of ask orders. 6.1.2. Trading Activity We use the metric TRADES to capture on-market trades that result from executing a market order against a standing limit order. 6.1.3. Order Imbalance Two measures of order imbalance are adopted: (1) order imbalance of new orders (ORDIMB) and (2) order imbalance of revised orders (ROR-DIMB). The

211

deviation of order imbalance from 0.5 is used as the benchmark of zero imbalance.21 • If the number of bids + the number of asks = 0, order imbalance = 0. • If the number of bids ≥ asks, order imbalance = bids/ (bids + asks) −0.5. • If the number of asks > bids, order imbalance = asks/ (bids + asks) − 0.5. 6.2. Price Change and Time between Trades The most commonly used statistic to measure price change is the standard deviation of return, in part because it is a standard measure of risk in theories of portfolio selection and asset pricing. However, this measure cannot be used in our study, because we do not have sufficient return data. Two alternative metrics are the price change and its absolute value. The price change metric is expressed as a signed relative price change. It is calculated as the natural log of the ratio of the first price after the release to the last price before the release. The signed price change can indicate whether market participants view the announcement positively, negatively or otherwise. If the aggregation of all the price changes around announcements of type X is greater than zero, then it can be inferred that type X announcements, on average, contain market sensitive information that is viewed positively by the market. If the value is zero, then it can be inferred that they are not biased toward either good or bad news; a type X announcement contains negative or positive information depending on the specific release.22 212

The second metric utilised is the absolute value of the (relative) price change. A significant absolute price change suggests that the announcement altered investors’ prior beliefs and thus contained important information. We expect price change metrics to be significant around only market sensitive releases. Additionally, we predict that the price change is greater around announcements that are unexpected by the market. The third metric is the time from the last trade before an announcement to the first trade after it. The time between trades reflects investor reaction to information and how quickly market sensitive information is disseminated and impounded into a trade price. It also reflects the importance of the information released. A total of 21 metrics are examined in our study. They are summarised below: TRADE (TRADE$)

Total frequency (value) of on-market trades resulting from a crossing of a bid and ask order.

ENTBID (ENTBID$$)

Total frequency (value) of new bid orders entered

ENTASK (ENTASK$)

Total frequency (value) of new ask orders entered

REVBID (REVBID$)

Total frequency (value) of amended bids, cancelled bids and amendments to bids one price step closer to the market

213

REVASΚ (REVASK$)

Total frequency (value) of amended asks, cancelled bids and amendments to asks one price step closer to the market

IMBNEW (IMBNEW$)

Imbalance of new orders, calculated as the ratio of the total frequency (value) of new bid orders (or the total frequency (value) of new ask orders) to the sum of the two values/frequencies

IMBREV (IMBREV$)

Imbalance of revised orders, calculated as the ratio of the total frequency (value) of revised bid orders and the total frequency (value) of revised ask orders, to the sum of both revised bids and asks values/frequencies

ABIMBNEW Absolute imbalance of new orders, (ABIMBNEW$) calculated as the ratio of the absolute frequency (value) of new bid orders and the absolute frequency (value) of new ask orders to the sum of the absolute values/frequencies of both revised bids and asks ABIMBREV Absolute imbalance of revised orders, (ABIMBREV$) calculated as the ratio of the absolute value/frequency of revised bid orders and the absolute value/frequency of new ask orders, to the sum of the absolute values/frequencies of both revised bids and asks

214

PRICE CHANGE

Signed price change, calculated as the natural log of the ratio of the first price after the release to the last price before the release

ABSOLUTE PRICE CHANGE

Absolute (value of the) price change

TIME BETWEEN TRADES

Time between the last trade before to the first trade after an announcement

It must be noted that these metrics may be significantly influenced by market microstructure factors, which can induce patterns in trading attributes orders of magnitude larger than the effect associated with many information releases (Brown, Clinch, and Foster, 1992). 6.3. Significance Testing Statistical significance is established in the same way for all metrics, using a resampling procedure. 6.3.1. Parameters The width of the event window is defined as the amount of time prior to and after an announcement that is examined. Several issues need to be considered in deciding the event window’s width. Widening the window addresses problems such as infrequent trading, which affects many firms on ASX, but introduces noise into the experiment. Because the research literature provides relatively little guidance on this aspect of experimental design where high-frequency equity market 215

data are employed, we examine the 60 minutes before and the 60 minutes immediately after each announcement (120 minutes in total), in subintervals of 1 to 15 minutes. The control periods are formed using up to 125 days before and 125 days after an announcement. The control metrics are resampled 500 times from within the event window ±125 days, excluding the 3 days centred on the announcement date itself. 6.3.2. Determining Significance To test our hypotheses we need to establish whether a particular metric in a given time interval is abnormal relative to a benchmark. That is, to know what is abnormal requires a benchmark of normal behaviour. We use a resampling technique similar to Aitken, Brown and Walter (1994) and other studies to compare the experimental metric with the corresponding metric for 500 randomly constituted control portfolios. We test for statistical significance by generating the empirical distribution of each normalised metric under the null hypothesis of no association between the market’s behaviour at a given time of day and whether a particular event occurred that day. This compares a time with no information (pseudo event)23 to the real event time, which contains information. Therefore, by comparing an event to non-event times we are able to gauge whether there is abnormal performance. For example, suppose our portfolio of releases has just one case, a PFS released by BHP at 1:59 p.m. on the announcement day. We first identify the 12:59

216

p.m.−2:59 p.m. window as the one of interest, it being divided into, say, 120 one-minute intervals.24 Within the ±125 day window, each raw metric is observed within each one minute interval, from 1259 p.m.−2:59 p.m. Next, the raw metric is normalised by subtracting the mean (which is calculated for that metric and that one-minute time interval, using each day within the ±125 day window) and dividing the result by the standard deviation. The experimental metric (one for each minute) is BHP’s normalised metric on the announcement day. We next form 500 control portfolios by choosing a day randomly from within the control period and observe the normalised metric for each of the 120 one-minute intervals on that day.25 Statistical significance is indicated by comparing the experimental portfolio metric to the distribution of control portfolios’ metric. Specifically, for each one-minute time interval, we calculate the relative frequency with which the experimental metric exceeds a control portfolio’s metric. In this way we are able to determine whether the experimental portfolio’s metric is significantly different from the metric that is observed by chance. 7. Results The results are divided into five sections: (1) a comparison of market sensitive with non–market sensitive announcements (Sample A), (2) order flow during a trading halt when a PFS is released (Sample B), (3) order flow around 19 market sensitive release types (Sample C), (4) price change and time between successive trades before and after market sensitive releases (Sample C) and (5) an important robustness

217

check, whereby we include cases where there was no recorded trade price before or after a release by assuming had a trade occurred the price would have been unchanged. The robustness test is important because, although it comes at the cost of invoking a sometimes inaccurate assumption, it does substantially increase the power of the tests. 7.1. Comparison of Market and Non–Market Sensitive News Releases—Sample A This section contains results for order flow around four announcement types that contain some announcements the ASX classified as market sensitive and others non–market sensitive. The four types are (1) disclosure document, (2) asset acquisition, (3) asset disposal and (4) response to ASX query. Table 3 summarises the results for each announcement type (detailed results are available from the authors). Panel A relates to order flow from 30 minutes before to 60 minutes after a non–market sensitive announcement is released to the dissemination server. Panel B relates to announcements judged to be market sensitive. It differs from Panel A in that it summarises abnormal order flow relative to the time trading was halted. Table 3 Mean standardised experimental period metrics for six 15-minute intervals around the announcement release. The metric is averaged across the four different announcement types in Sample A. The first number in the brackets shows the number of announcement types with experimental metrics significantly greater than the benchmark portfolios at the 10% level. The second

218

number in the brackets shows the number of announcement types that have metrics that are significantly less.

219

Generally there is no significant abnormal order flow around a non–market sensitive announcement. An exception is a response to an ASX query, which exhibits some evidence of abnormal order flow, but only in some intervals. Taken as a whole, the results demonstrate that market participants do not change their orders either before or in response to an announcement that ASX classifies as non–market sensitive. We find little evidence of increases in new order placements before a market sensitive announcement, except for responses to ASX queries. However, outstanding orders are revised significantly more often before the release of responses to ASX queries, asset acquisitions and disposals. It is possible that these announcements are widely anticipated and that traders adjust their orders prior to the announcement in line with their expectations. Also, announcements such as a response to an ASX query are released as a result of previous unusual market activity, so it is no surprise to observe higher abnormal activity prior to the release. In contrast, order flow around a disclosure document does not behave abnormally either before or after the announcement. In sum, the evidence here is consistent with hypothesis H1—that uninformed traders who know of an impending announcement price protect themselves by revising their orders prior to a market sensitive announcement. It is also consistent with informed traders who in the past did not demand immediacy (i.e., they placed limit orders) progressively modifying their positions before their private information is made public.26

220

Asset acquisitions, asset disposals and responses to ASX queries exhibit abnormally high order flow behaviour for all metrics after the announcement. Obviously these announcements contain important information that causes investors to adjust their orders significantly. 7.2. Order Flow During a Trading Halt—Sample B Table 4 summarises order flow around the release of a PFS. When first released, these announcements are market sensitive and are associated with a trading halt. The table contains statistics for the time period from the release of the document to the resumption of trading. Panel A contains the results for announcements with release to resume durations of 7 to 8 minutes, Panel B for durations of 8 to 9 minutes and Panel C for durations of 9 to 10 minutes. Table 4 Order flow during release to resume phase for earnings announcements (Sample B). The table shows the mean standardised experimental period metrics for 1 minute intervals after an announcement release. Panel A is for announcements with release to resume duration of 7 to 8 minutes, Panel B for a duration of 8 to 9 minutes and Panel C for a duration of 9 to 10 minutes.

221

*** Significant at 1%. ** Significant at 5%. * Significant at 10%.

222

Earnings announcements with a release to resume phase of 7 to 8 minutes have abnormally high (at 1%) levels of revised orders in the first minute after the document’s release. This may reflect uninformed traders removing themselves from the market. If so, it supports Kim and Verrechia’s (1994) proposition that information asymmetry is greater after a release due to investors’ differential information-processing abilities. Perhaps for the same reason, revised orders are significantly higher (at 1%) for earnings announcements with release to resume times of 8 to 9 and 9 to 10 minutes. Conversely, new order placements are significantly less than on non-event days or not significant at all. This is generally true for all three earnings release to resume times across the entire event window,27 which contrasts with Corwin and Lipson (2000), who report that the number of order submissions is abnormally high during NYSE trading halts.28 This suggests that investors refrain from trading until they comprehend the information contained in the release. Alternatively, because no on-market trades occur during a trading halt, market participants have an incentive to delay their order submission while they process the information.29 Hence, the existence of a trading halt around the release may cause traders to delay submitting new orders and slow the process of price discovery. Absolute order imbalance based on new orders is mostly lower than the benchmark for all three subsamples of announcements, but these numbers are noisy because of the relative absence of new orders. Absolute order imbalance based on revised orders is significant (at

223

1%–5%) across all three subsamples. This indicates that traders react to earnings information by revising their preexisting limit orders. We find in general no consistent abnormal activity for new orders entered, which may imply that traders are unsure about the price implications of the announcement and take a wait and see approach. This seems logical, as investors need time to acquire and process the information in the PFS before placing a new order. Alternatively, the abnormal levels of revised orders suggest that, in the short-term, investors respond to earnings information by revising their preexisting limit orders. Therefore, our hypothesis (H4)—that there is a significant increase in new orders, revised orders and order imbalance in the release to resume phase—is only partially supported. Hypothesis (H5)—that traders with superior information processing abilities will conceal their new orders until the final minute to avoid the premature signalling of their private information—is supported. New orders and imbalance of new orders are significant (typically at the 1%–5% level) in the final minute before trading resumes. Informed traders are apparently active, as the weighting of orders is predominantly on one side of the market. 7.3. Order Flow Around the News Releases—Sample C Table 5 summarises the results30 for order flow, from 1 hour before to 1 hour after the release in 15-minute intervals, for the 19 market sensitive announcement types in Sample C.

224

Table 5 Order flow around ASX news releases (Sample C). The table shows the mean standardised experimental period metrics for eight 15-minute intervals around the announcement release. The metric is averaged across the 19 different announcement types. The first number in the brackets shows the number of announcement types that have metrics that are larger than the benchmark period at the 10% significance level. The second number in the brackets shows the number of announcement types that have metrics that are smaller than the benchmark period at the 10% significance level.

New orders and imbalance of new orders are significant (typically at 1%–5%) in the final minute before trading resumes. As before, this is possibly due to traders with superior information-processing ability concealing their knowledge until the last minute. For the majority of market sensitive announcements, order flow before the release is significantly higher than for the control portfolios in intervals such as [−30 to −15] and [−15 to 0]. Announcements that have no consistent abnormal activity prior to their release are:

225

first-, third- and fourth-quarter activities reports, renounceable and nonrenounceable issues, on-market buybacks and directors’ appointments and resignations. After the release, most announcements exhibit some form of abnormal order flow. The exception is third-quarter activities reports. This confirms for a wider range of announcements that releases classified by ASX as market sensitive do contain information important to investors. An interesting result is that, for all 19 announcement types, there are significant revisions to bids and asks for at least 15 minutes before and after the release. This suggests that a common investor reaction to a market sensitive announcement is to revise their standing limit orders. Probably they revise their orders away from the market leading up to an announcement due to the information asymmetry and then revise their orders closer to the market after the public information is released. 7.4. Price Change and Time Between Trades Around the News Releases—Sample C Table 6 contains results for price change and time between trades for Sample C. Specifically, the three metrics we examine are price change, absolute price change and time between trades. Table 6 Price change and time between trades (Sample C). The table shows the mean standardised experimental period metrics around an announcement release. Panel A presents the results for signed relative price change, Panel B presents the results for absolute relative price

226

change metric and Panel C reports results for time between the first trade after a release and the last trade before a release. The first column shows the announcement type while the second and third columns give the experimental mean and control mean for a particular metric. The fourth column shows the relative frequency with which the experimental mean exceeds the control means. The fifth column shows the number of control metrics (C) that exceed the experimental metric (E). The sixth column shows how many control metrics (C) were exceeded by the experimental metric (E).

227

228

*** Significant at 1%. ** Significant at 5%. * Significant at 10%. In general we find significant price changes surrounding a market sensitive announcement (H6). The announcements that had no significant absolute price change were the third- and fourth-quarter activities reports, renounceable issues and disclosure documents. The fact that there is no significant price change for disclosure documents is surprising because they exhibit abnormal order flow both before and after their release. Third-quarter activities reports are the only announcements to have no abnormal order flow either 229

before or after the release and no significant absolute price change. Thus, information contained in these releases is of limited use to investors.31 We observe that an abnormal level of order flow, measured by frequency and dollar volume, surrounding a market sensitive announcement is generally followed by a significant absolute price change. This result provides some support to the literature that suggests higher volume and more frequent trading are associated with a larger absolute price change. However, this correlation is not formally tested here. Contrary to H7, we find that the elapsed time between the last trade before and the first trade after a market sensitive announcement is significantly longer compared to other days when there is no similar announcement. Superficially, it appears that investors are slow to respond to market sensitive information. The difficulty with this interpretation is that the interposition of the enforced trading halt in the experimental portfolio’s metric affects this result. 7.5. Robustness Check When testing the significance of the results in the accompanying tables, we exclude from the control portfolios days when there is no trade before the release or after it. As a robustness test, we do not discard these cases. Instead, if an announcement has no recorded price before or after the release (within the specified window), we assume the price change is zero.32 This reinterpretation substantially increases the sample size, but our conclusions are unchanged.

230

8. Conclusion We examine order flow and price effects surrounding company announcements on the Australian Stock Exchange. We find that, leading up to a market sensitive announcement, traders withdraw from the market or amend their live limit orders by decreasing their volume traded or moving the price limit away from the market. The flow of revised orders is abnormally high while the flow of new orders is abnormally low before a market sensitive announcement (especially where its timing is widely anticipated). Correspondingly, completed trades are lower. Order imbalance is abnormally higher. We find mixed evidence on order imbalance after a market sensitive release, as it depends largely on the information content of the particular event. In contrast, measures of order flow and trading are abnormally high as investors process and react to the market sensitive event. Order flow activity around non–market sensitive announcements does not behave unusually. In general we find the absolute share price change is significantly larger surrounding a market sensitive announcement, while the time elapsed from the last trade before to the first trade after a market sensitive announcement is generally significantly longer than on other days. Interpreting this last result is complicated by an enforced delay to trading because of the trading halt applied by ASX to all market sensitive announcements made during trading hours. Appendix Description of Announcement Types

231

This section describes the 19 announcement types examined in this study. Specifically, we describe the information contained in the announcement and regulatory requirements under the ASX listing rules. 1. Intention to Make Takeover Bid (01001) As the title suggests, when a company intends to make a takeover bid for another company it must first release an intention to make a takeover bid announcement to the market. 2. Preliminary Final Report (03003) All companies listed on the ASX must prepare a preliminary final report. Following the end of the financial year of an entity (except a mining exploration entity) ending on or after 30 June 2003, the entity must give ASX an Appendix 4B (Preliminary Final Report). 3. Half-Yearly Report (03004) The Corporations Act requires Australian listed entities to prepare half-yearly and annual reports and accounts. This report does not include all the notes of the type normally included in an annual financial report. The half-yearly report is regulated by the Listing Rules 4.1 and 4.2A. 4. Quarterly Reports (04001–04004) Mining exploration entities are required to provide quarterly reports on their cash flows in the form of an Appendix 5B document. A mining entity as classified by ASX listing rules is a mining exploration entity or a

232

mining producing entity. A mining exploration entity is an entity whose activities or the principal part of whose activities consist of the extraction of a mineral, oil or gas.33 5. Renounceable Issue (06001) The ASX defines a renounceable issue as a rights offer that may be taken up, forfeited or traded on the market. A rights offer is presented to shareholders as an invitation to buy new shares in the same company, usually below the prevailing market price. 6. Placement (06003) Placement is defined in the Corporations Act 2001 as the process of selling new issues of securities through a stock broker or other intermediary to existing shareholders and the public. To process a placement issue, the company in question must fill out an Appendix 3B document and submit it to the ASX. In addition, the company must report what its proposed use of the capital will be. 7. Nonrenounceable Issue (06008) The main difference between a nonrenounceable issue and a renounceable issue is that the former cannot be traded on the market. Therefore, it is a rights issue that may only be taken up or forfeited when it is presented to shareholders. 8. Disclosure Document (06010)

233

A disclosure document is a document or prospectus that outlines the details of a new issue of securities to ensure that investors are fully informed. Another type of disclosure document that is similar to a prospectus is an offer information statement. 9. On-Market Buyback (06011) Companies are required to announce to the market their intention to buy back shares by completing the Appendix 3C form. This document includes information on the type of buyback, class of shares which is the subject of the buyback, voting rights, whether fully paid or partly paid, number of shares in the class on issue and reason for buyback. 10. Asset Acquisition (07001) Companies have to report assets acquired if they are substantial in value. An asset is substantial (and thus needs to be reported) if its value, or the value of the consideration for it is, or, in ASX’s opinion, is 5% or more of the equity interests of the entity as set out in the latest accounts given to ASX under the listing rules. 11. Asset Disposal (07002) Similar to an asset acquisition, companies must report any asset disposed that is substantial in value. An asset is substantial if its value, or the value of the consideration for it is, or, in ASX’s opinion, is 5% or more of the equity interests of the entity as set out in the latest accounts given to ASX under the listing rules. 12. Dividend Rate (10003) 234

A dividend is a payment, in the form of cash or stock, by a company to its shareholders. The dividend rate is usually released in the half-yearly report or preliminary final report. However, in some instances, it is released as a separate announcement. 13. Progress Report (11001) Progress reports are released by mining/exploration companies to inform the market of developments relating to specific projects. 14. Director Appointment/Resignation (12001) When a new director is appointed or a director resigns at a company, he or she must disclose this to the market through a director appointment/resignation announcement. 15. Open Briefing (14006) An open briefing is used by listed companies to brief the market. It involves a record of an interview between subscribing company executives and analysts. It is used to explain company announcements, results or other corporate issues of interest to the market. Access to company briefings is available through the ASX and Corporate File Pty Ltd. 16. Response to ASX Query (17003) The ASX will often submit a query to a company to which the company must respond. Listing Rule 18.7 states that an entity must give ASX any information, document or explanation that ASX asks for to enable it to be satisfied that the entity is, and has been, complying

235

with the listing rules. The entity must do so within the time specified by ASX. ASX may submit or require the entity to submit any information given to ASX to the scrutiny of an expert selected by ASX. The entity must pay for the expert. The ASX will usually query a company when there has been unexplained abnormal volatility in the company’s share price. Notes 1. See below for a summary of the announcement process on ASX. 2. Uncertainty is measured by the bid-ask spread. 3. Listing Rule 15.7 states: “An entity must not release information that is for release to the market to any person until it has given the information to ASX and has received an acknowledgement that ASX has released the information to the market.” 4. A market sensitive announcement is one which is likely to cause a price impact on the market; a non–market sensitive announcement is any other announcement. 5. We ignore the after-market (i.e., the special single price auction held separately sometime after 4:00 p.m.). 6. See Section 5.1 for a description of the headline record. 7. See the appendix for a list of announcement (report) types used in our study. The header record allows for up to 10 types to be associated with a single document. 8. 236

Discussions with ASX staff indicate that the time between the release of the header and the release of the full announcement is typically less than a minute. 9. The dissemination server stores the following information: (1) a header message, (2) a summary of the announcement and (3) a PDF file of the announcement document. 10. The message is broadcast to participating organisations and other subscribers to the “Voiceline” service, also known as the “squawk box”. 11. Takeover announcements are an exception. They trigger trading halts of around an hour. 12. “Timely” means the release is regular and expected, as is often the case with an earnings release. It does not mean the content of the release is known. 13. Their entry is measured by new orders entered into SEATS, imbalance of new orders and trades processed. A significant new order imbalance means new orders are biased to one side of the market and is thus a sign of informed traders trading on private information. 14. See Section 6 for a description of the metrics. 15. Also, previous literature such as Lee, Mucklow and Ready (1991) state that information releases made outside trading hours might have attributes different from releases made during trading hours. 16. We have taken some license in describing PFSs as earnings announcements, since they include abbreviated

237

P&L statements, balance sheets, statements of cash flows, and so on. 17. This number is chosen to allow the more commonly observed/utilised announcement types to be included while maintaining a reasonably large sample size. 18. Nineteen announcement types were chosen due to time pressures. 19. These restrictions were relaxed subsequently. See Section 7.5. 20. Uninformed traders add noise to the experiment as we are principally interested in the effect of information releases on trading behaviour. 21. The deviation from 0.5 was purely an aid to interpreting the raw metric (i.e., before normalisation). 22. If the absolute price change for the same announcement type is zero, then on average the announcement type does not contain market sensitive information. 23. When resampling, no regard is paid as to whether the day randomly chosen is one on which the firm made some other market sensitive announcement. 24. The first (semi-open) interval is (12:59−1:00], the second is (1:00−1:01] and the last is (2:58−2:59]. 25. There are 500 control metrics and 1 experimental metric in this example of a portfolio comprising a single announcement.

238

26. In future research we will attempt to distinguish between these two explanations. 27. An exception is found in announcements with release to resume durations of 9 to 10 minutes. In the minute immediately before the resumption of trading, entered (new) bids are significantly higher (at 1%). This behaviour may suggest that traders (informed) hide their judgment of the fundamental value of the asset until the final minute, perhaps to discourage front running by uninformed traders or simply because it takes time to process the information. We intend to take a closer look at these possibilities in future work. 28. Different market structures and trading mechanisms may explain part of this result. 29. However, earlier orders have higher priority when trading resumes. 30. Detailed results are available from the authors. 31. Nonetheless it would seem odd if ASX were to classify third-quarter reports as non–market sensitive when other quarterly reports are market sensitive. 32. Time between trades is not assigned a zero value. It is calculated from the first time after the release to the last time before. This biases the time between trades upward. 33. Mining exploration entities must provide quarterly cash flow reports under listing Rule 5.3 for as long as they are mining exploration entities. The obligation ceases if and when they become mining producing entities. 239

References Ackert, L., B. Church and N. Jayaraman, 1999, An experimental study of circuit breakers: The effects of mandated market closures and temporary halts on market behavior, Working paper, Federal Reserve Bank of Atlanta. Admati, A. and P. Pfleiderer, 1988, A theory of intraday patterns: Volume and price variability, The Review of Financial Studies 1: pp. 3–40. Aitken, M., P. Brown and T. Walter. 1994. The behaviour of the bid/ask schedule around the time of a price sensitive announcement. Paper presented at the Asia-Pacific Finance Association Conference, Sydney. Aitken, M., A. Frino and R. Winn, 1996, Consensus analysts’ earnings forecasts and security returns, Asia Pacific Journal of Management 13, no. 2: pp. 101–110. Brown, P., C. Forbes and M. Wee, 2003, Liquidity differences surrounding announcements classified by ASX according to their market sensitivity, Working paper, University of Western Australia. Brown, P., C. Ng and M. Wee, 2003, An examination of the ASX announcement process and information pricing policy, Working paper, University of Western Australia. Brown, P. R., G. Clinch and G. Foster, 1992, Market microstructure and capital market information content research, Studies in Accounting Research 32. Sarasota, FL: American Accounting Association.

240

Corwin, S. A., and M. L. Lipson, 2000, Order flow and liquidity around NYSE trading halts, The Journal of Finance 55, no. 4: pp. 1771–1802. Demski, J. S., and G. A. Feltham, 1994, Market response to financial reports, Journal of Accounting and Economics 17, no. 1–2: pp. 3–34. Diamond, D. W., 1985, Optimal release of information by firms, The Journal of Finance 40, no. 4: pp. 1071–1094. Fama, E., 1976, Efficient capital markets reply, The Journal of Finance 31, no. 1: pp. 143–145. Foster, F. D., and S. Viswanathan, 1990, A theory of the interday variations in volume, variance, and trading costs in securities markets, The Review of Financial Studies 3, no. 4: pp. 593–624. Glosten, L. R., and P. Milgrom, 1985, Bid, ask and transaction prices in a specialist market with heterogeneously informed traders, Journal of Financial Economics 14, no. 1: pp. 71–100. Greenwald, B. C., and J. C. Stein, 1991, Transactional risk, market crashes, and the role of circuit breakers, The Journal of Business 64, no. 4: pp. 443–463. Hopewell, M. H., and A.L.J. Schwartz, 1976, Stock price movement associated with temporary trading suspensions: Bear market versus bull market, Journal of Financial and Quantitative Analysis 11, pp. 577.

241

Kim, O. and R. E. Verrecchia, 1991a, Market reaction to anticipated announcements, Journal of Financial Economics 30, no. 2: pp. 273–309. Kim, O. and R. E. Verrecchia, 1991b, Trading volume and price reactions to public announcements, Journal of Accounting Research 29, no. 2: pp. 302–321. Kim, O. and R. E. Verrecchia, 1994, Market liquidity and volume around earnings announcements, Journal of Accounting and Economics 17, no. 1–2: pp. 41–67. Lee, C., M. J. Ready and P. J. Seguin, 1994, Volume, volatility, and New York stock exchange trading halts, The Journal of Finance 49, no. 1: pp. 183–215. Lundholm, R. J., 1991, Public signals and the equilibrium allocation of private information, Journal of Accounting Research 29, no. 2: pp. 322–349. McNichols, M. and B. Trueman, 1994, Public disclosures, private information collection, and short-term trading, Journal of Accounting and Economics 17, no. 1–2: pp. 69–94. Slezak, S. L., 1994, A theory of the dynamics of security returns around market closures, The Journal of Finance 49, no. 4: pp. 1163–1212. Stein, J., 1987, Informational externalities and welfare-reducing speculation, Journal of Political Economy 95: pp. 112–145. Subrahmanyam, A., 1994, Circuit breakers and market volatility: A theoretical perspective, The Journal of Finance 49, no. 1: pp. 237–255.

242

Verrecchia, R. E., 1982, The use of mathematical models in financial accounting, Journal of Accounting Research 20: pp. 1–42.

© 2005 by the Authors

243

Part II Miscellaneous Issues

244

Identifying Some Issues in C.C.A. Ray Ball and Philip Brown

Ray Ball is Professor of Management in the Australian Graduate School of Management. His major areas of interest are accounting, corporate finance and the securities industry. He is a senior associate member of the Australian Society of Accountants and a member of the Securities Institute of Australia. He has degrees from the University of New South Wales and the University of Chicago.

245

Philip Brown is Professor of Management and Director of the Australian Graduate School of Management. His current research interests are centred on the impact of accounting and other information on the Australian sharemarket. He is a senior associate member of the Australian Society of Accountants and a member of the Securities Institute of Australia. He is a graduate of the University of New South Wales and the University of Chicago. Postponement of Current Cost Accounting (C.C.A.) for a year could be seen as evidence of a vacillating accounting profession. Alternatively, it could be evidence that the accounting profession must respond to the political and market pressures inherent in the nature of its policy choices and that it is the various political and market forces which are uncertain. A third explanation is that the practical issues in C.C.A. implementation are greater than the profession believed. Whichever way it is viewed, the postponement does provide an opportunity to canvass the issues more widely, and to that extent it is to be welcomed. We enter the debate, not because we support the introduction of C.C.A. in the manner proposed by the Statement of Provisional Accounting Standards of October 1976, but because we believe that two allegedly major consequences of the introduction of C.C.A. are relatively unimportant. In particular, we argue that the introduction of C.C.A. will not significantly affect inflation, nor will it significantly affect the sharemarket averages.

246

There are several other significant issues, however, which do warrant careful debate. We discuss them in the context of the costs and benefits of introducing C.C.A., and the difficulties that policy-makers face when assessing whether C.C.A.’s benefits are likelv to exceed its costs. C.C.A. and Inflation There is a fear that the introduction of C.C.A. will exacerbate inflation. A version of the argument runs as follows: Businesses base their pricing decisions on buyer demand and product replacement cost, such that the higher the product replacement cost, the higher the price. If businesses do not know replacement costs, they base their pricing decisions on the available measure, namely historical costs (book values) of the products. When C.C.A. is introduced, many businesses will know for the first time the replacement costs of the products they sell. Since “inflation” implies costs on average are rising through time and since businesses typically hold stocks for some time before they are sold, products’ replacement costs at the time of sale, on average, exceed their historical costs (book values). On the basis of their new knowledge – that product costs are higher – many businesses will raise their product prices, thus exacerbating inflation. Key assumptions in this argument are that many businesses are unaware of replacement costs so that the introduction of C.C.A. will bring a significant number of surprises, and that the net effect of those surprises will 247

be biased in a particular way – namely, towards increased costs. But how many businesses do not already know, approximately, their product replacement costs? Do the guesses which are now being made tend to over or under estimate replacement costs? If there is a widespread commercial incentive to provide internal information to managers on product replacement costs, is it not more reasonable to assume, as a working hypothesis, that replacement costs are already accessible by management, whether or not they are supplied in a routine fashion by an accounting system? We argue, then, that even if there were an inflationary impact due to businesses becoming aware of the excess of replacement over historical cost, that impact would be felt long before the time of C.C.A.’s introduction. To argue otherwise is to argue that the average business is underestimating inflation. We submit that most businesses have a rough idea of replacement costs, even if those costs are not routinely reported in the accounts. Further, we would expect the “rough guesses” to be unbiased – to be over estimates about half the time and under estimates for the other half.1 Suppose the introduction of C.C.A. is linked to its introduction for income tax purposes such that lower C.C.A. profit figures are associated with reduced taxable incomes. Then one possible equilibrium solution is for lower rather than higher product prices, the crucial parameters being how, and to what extent, the Government acts to maintain its total revenue.2

248

In arguing that C.C.A. is inflationary, some seem to assume the implied target rate of return under C.C.A. is much the same as under conventional (historical cost) reporting. It is unclear to us, however, that decision-makers who use C.C.A.’s figures will choose to exclude from profit the holding gains on stock and equipment, as the profession’s Provisional Standard proposes. Even if they do so choose, why should their target rates of return bear any close resemblance to their target rates under conventional reporting? It may well be the case, for example, that the current target rate of return under C.C.A. is low or even negative for many firms, to be consistent with the rational depletion of capital stock in a situation of continuing excess of supply of in vestment assets. In our opinion, there are good grounds for believing C.C.A.’s introduction will have little effect on inflation. C.C.A. and the Sharemarket It is well known that the introduction of C.C.A. will result in lower disclosed profits (and lower earnings per share) for most companies. But can the sharemarket averages be expected to fall as a result? We believe the answer is, by and large, “No”. Our reasoning is, once again, based on our belief that the market-makers have made unbiased guesses about the effects of inflation, in this case on share values, and hence C.C.A. cannot be expected to affect the sharemarket averages in any particular direction. There seems as much reason to expect the market to overestimate the effect of inflation on profits as there is to expect it to underestimate the

249

effect. Which way share prices will be affected by C.C.A. reports seems an even-money bet. In addition, we can’t see much reason for the market’s guess to be far wrong in either direction. It is not uninformed. Appreciation of C.C.A.’s effects on disclosed profits is widespread. Many accountants in commerce and industry (in sharemarket terms, “insiders”) have looked closely at the question of how much difference C.C.A. will make to the profits of companies with which they are associated closely. Inevitably, their conclusions will be reflected, if only indirectly, in the overall structure of share prices. The important point is that the sharemarket is presently making guesses, however well-informed they may be on an individual company basis, of the effect of inflation on the levels of company profits. Numerous companies have disclosed C.C.A.’s effects on their profit figures including, for example, BP, C.S.R., Ford, Direct Acceptance, John Fairfax, Philips, and B. Seppelt & Sons (only some of these are listed on Australian Stock Exchanges). Finally, security analysts’ and other sharemarket “outsiders” have not ignored the situation. For example, J. M. Bowyer & Co. have published widely-quoted estimates for “Australia’s 50 largest companies, by market capitalization, (excluding Woodside and Pancontinental)”, which showed that reported profits would have averaged 37% less if C.C.A. had been adopted by each company for its financial year ending in 1976.

250

Thus, the sharemarket is hardly ill-informed on the matter. The Bowyer estimates emphasize this point – they were derived from publicly-available data and from estimates of asset purchase patterns and depreciation rates, without recourse to companies’ internal records. Belief in a general awareness of the average impact of C.C.A. on corporate profits, coupled with the observation that market-makers act quickly on useful information (and have an enviable tendency to ignore data that ex ante are useless), leads us to predict that the introduction of C.C.A. will not affect the sharemarket averages by any discernible amount, other things being equal. If C.C.A. information is useful, some share prices will undoubtedly rise and others will fall, as the individual company profit results either exceed or fall 251

short of the analysts’ predictions. Thus, even if the sharemarket finds C.C.A. information to be useful in valuing individual shares, the share averages should not be affected significantly by C.C.A. reports. In an attempt to gain some limited appreciation of the effect of C.C.A. profit reports on share prices, we computed the price movements which occurred when the Bowyer estimates were announced. Prices were observed two weeks before and two weeks after the day upon which the estimates were released. Appropriate allowance was made for dividends and other changes in the basis of quotation for each share. The accompanying table and graph reveal what happened. The general picture is that nothing happened. On average, prices rose moderately by about 1%, compared with the average profit reduction under C.C.A. of 37%. The sharemarket as a whole rose by about the same percentage (the Statex-Actuaries Accumulation Index rose by one twentieth of one per cent less than the average of the 50 Bowyer & Co. companies). Based upon the average price behaviour of the 50 shares, the sharemarket does not seem to have been perturbed by the C.C.A. estimates. At the individual share level, there was, if anything, some tendency for price movements to be inversely related to the change in reported profit: the greater the reduction in profit from historical cost accounting to C.C.A., the better the average price performance.3 This tendency is shown by the downward-sloping line on the graph, which is the best estimate of the relation which occurred between share prices and the C.C.A. estimates.

252

However, the tendency to move in the opposite direction (than that which most observers would expect) is slight. On average, a 10% reduction in profit under C.C.A. was associated with only a four-tenths of 1% increase in share price. It would be close enough to say that individual share prices were unaffected by the Bowyer & Co. estimates. There seems little evidence in these data to support the view that investors will be penalized in the sharemarket if C.C.A. is introduced. Either the Bowyer & Co. estimates are not perfect, or the sharemarket disregards C.C.A. figures, or the sharemarket anticipates C.C.A. figures with considerable accuracy, or (as seems likely) some combination of these effects occurred. Other Issues We wonder whether some of the opposition to C.C.A. can be explained by the fact that most of the alleged benefits of replacement cost accounting are already obtainable at a less formal level, whereas the costs of formally implementing C.C.A. are yet to be met. Incidentally, the potential benefits of C.C.A. include the promise of taxation relief, which some company executives might now consider is obtainable regardless of methods used in accounting to investors. As pointed out earlier, much depends upon the extent to which the Government acts to maintain its revenue. There is growing awareness of C.C.A.’s costs. Some costs are “internal”; for example the costs of staff training, accounting system development, replacement price data acquisition, and increased auditor

253

involvement; others will be imposed externally, such as the increased costs of policing by regulatory agencies. There are also costs associated with re-negotiating agreements, such as borrowing contracts (e.g. adjustments to debenture trust deeds), executive compensation plans and other incentive schemes, and contractual profit-sharing arrangements such as partnership agreements. The costs are many, and they are not insignificant to users. The accounting policy-makers would do well to study them closely. Some potential benefits of C.C.A. defy measurement. Suppose we accept, for argument’s sake, that C.C.A. would increase economic efficiency. How are we to satisfy ourselves that it is achieved? While there may be something to be said for the disclosure provisions of the C.C.A. proposal, are the empirical research techniques sufficiently advanced to allow a follow-up study which can draw valid conclusions? There is also the question of equity. The tax reform proposals, which bear some resemblance to the C.C.A. proposal, reduce the general level of company tax. However, some companies (notably those with substantial investments in “monetary items” such as cash, accounts receivable, loans, and fixed-interest securities) will not experience noticeable tax reductions. This can be seen from the Bowyer & Co. estimates in the attached table – the smallest reductions in profit under C.C.A. are for the banks and finance companies. If the company tax rate is increased to maintain the Government’s revenue, then these companies will pay more tax under the tax reform proposals. Since C.C.A.

254

and the tax reform proposals are not divorced, banks and finance companies could be excused for believing that C.C.A. is inequitable. Why should their profits appear to be higher? Finally, there is the dubious relevance of replacement costs in an economy which is not replacing its capital. The conventional argument for replacement costs goes along the lines that non-replacement of capital is “bad” and that replacement-cost accounting is needed to alert managers and investors to this fact. This argument falls down if non-replacement of the capital stock is a rational occurrence. If investors assess investments as now being riskier than when initially undertaken, as we believe they have in recent experience, then non-replacement of the capital stock is rational. We suggest this as an explanation of the current depression of investment activity. And if it is not worth replacing capital (i.e., if the total capital stock is to be run down), then on average the market worth of capital must be less than its replacement cost. In such times, how relevant is replacement cost information to investors? Conclusion The suggestions that C.C.A. will reduce share prices and increase inflation seem wildly impractical. Managers, investors and other users of accounting reports are making implicit or explicit estimates of the effects of inflation, whatever they may be, at this very moment. It is feasible that those estimates are too high – that C.C.A. income will be greater than currently expected; but it is

255

also feasible that they are too low. Either way, we do not expect C.C.A.’s introduction will have much impact on either share prices or inflation. The impact issues lie elsewhere. Perhaps the accounting profession has underestimated the direct and indirect costs of C.C.A. We believe that it has also overestimated the relevance of replacement costs in the present economic climate. Share Price Reaction to J. M. Bowyer & Co. Estimates of C.C.A. Pre-Tax Profits

256

* Adjusted for dividends and rights (i.e., the only changes in the basis of share capitalization).

1. An alternative version of the argument is that businesses will act to restore their profit levels, however calculated, and that they will therefore raise product

257

prices if C.C.A. profits are reported. This version seems to assume that prices can be raised at will. 2. It could also be argued that the introduction of C.C.A. might slightly reduce inflation through its effect on risk. If C.C.A. provides information superior to that now available to investors and managers, then its introduction will eliminate some investor and management guesses – that is to say, it will reduce uncertainty. Now, ignoring for the moment the costs involved in C.C.A., a greater certainty consequent upon C.C.A.’s introduction, other things being equal, will lead risk-averse suppliers of capital to accept a lower expected return on investment, and this lower cost of capital will be passed on in the form of reduced product prices. 3. For the statistically-minded, the Spearman rank-order correlation coefficient was −0.32, which (on the basis of a two-tailed t-test) is significant at the .05 level. We note that the t-statistic probably is overestimated, due to dependence among shares’ rates of return. The product-moment coefficient was −0.21; and the r-square from the regression plotted on the graph was 4%.

© 1977 by CPA Australia

258

Sharemarket Efficiency And The Experts: Some Australian Findings by Philip Brown Terry Walter* Abstract: This paper analyses the investment performance of 625 investment recommendations made by practitioners attending five Australian courses on portfolio management held between 1973 and 1979. It was found that course participants consistently out-performed the market, on a risk-adjusted basis. Their success was not due to picking winners to buy, but to nominating losers to sell. Keywords: EFFICIENT MARKET HYPOTHESIS; INVESTMENT PERFORMANCE; SHARE GAME; STRONG FORM EFFICIENCY Introduction From 1973 to 1979, five one-week courses ol modern portfolio theory were held at Macquarie University and the Australian Graduate School of Management. This paper shows that the participants in those courses consistently could out-perform the market. Their success was not due to picking winners to buy, but to nominating losers to sell.

259

Course participants included security analysts, investment advisers, portfolio managers, actuaries, share brokers and personal investors. They varied widely in experience and seniority. Some professed to be fundamental analysts and others admitted to more than just a passing interest in share charting. Prior knowledge that the results would be published, a course organisers’ guarantee of confidentiality, the general impression gained by the instructors and the results all point to the conclusion that the share game, as it was called, was taken seriously.1 Because the game was taken seriously and because the participants were professionals and keen students of the Australian market place,2 we have a direct test of the ability of Fama’s (1970) strong form of the Efficient Market Hypothesis3 to explain Australian share price behaviour. As foreshadowed in our opening remarks, our evidence is inconsistent with the EMH in its strong form.4 The Share Game Each participant was invited to nominate two shares to buy and two shares to sell, to be executed at the last sale up to the close of trading for the week in which the course was held. Altogether 157 participants made 625 recommendations (314 buy and 311 sell) involving 415 separate securities chosen from the total of 6192 securities listed on the Australian Associated Stock Exchanges when the courses were held. Table 1 sets out some features of the game. Table 1

260

Summary features of the share game

Notes: (1) One participant nominated an unlisted share (2)

One participant did not nominate any securities to sell

A degree of concordance underlies Table 1 in that more than one participant recommended the same share be bought or sold. Interestingly, in every course there was closer agreement on sell than buy recommendations, a point which we take up later. Table 2 explores this concordance by cross-tabulating a security according to the number of times it was recommended to be bought or sold. The null hypothesis, of independence between buy and sell recommendations, is rejected at the α = .01 confidence level for all but one of the five courses.5 Table 2 Cross-tabulation of buy/sell recommendations

261

Note: Table 2 is a summary (aggregation) of five similar tables one for each course. Table 2 indicates that, in the five separate tables, there was a total of 20 entries where one person recommended a security be bought and one other person (only) recommended that same security be sold. Similarly there were 8 occasions on which one person recommended a security be bought and two others recommended it be sold. Closer analysis showed that participants in all courses had a significant preference (α = .01) for securities issued by companies in the top 100 by market capitalisation, and in all but Course 4, for securities in the top 25. There was no significant difference in the proportions of securities in the top 100 companies that were bought and sold (α = .05). These results are as one might expect if the game was taken seriously and the participants made their recommendations from the shares they would have studied the most closely as professionals.6 Measuring Investment Performance Investment performance was measured in three ways:

262

i. did each investment result in a profit or a loss? ii. how did its rate of return compare with the market average, before allowing for relative (beta) risk? iii. how did its rate of return compare on a risk-adjusted basis? Measure (ii) and particularly measure (iii) are more in keeping with the spirit of the portfolio management courses.7 Consequently most of our discussion focusses on them. Weekly prices were collected for every security nominated by the participants in a course, for the 105-week period beginning 53 weeks prior to the Friday of the event week, i.e., the execution date. There were some “problem” securities, as one might expect in a game of this type where single investment recommendations are analysed without penalty for catastrophies. Table 3 lists them and their stories. The final price was assumed to be zero for Stella options and the five that were liquidated. The final price for the five others, all taken over, was the last recorded sale.8 Adjustments were made to the prices for any dividends, bonus and rights issues etc., using the theoretical value of each change in the basis of quotation. Table 3 Eleven “problem” securities and their stories

263

The Statex–Actuaries Accumulation Index (SAAI) was used as the market index, which we denote by Rm. An index computed by one of the authors and reported in Walter (1980), was used for the period prior to February 1973, when the SAAI became available on a daily basis. The 90-day Treasury Bill rate, supplied by the Reserve Bank of Australia, was used as the risk-free rate. This risk-free rate, Rf, was subtracted from the portfolio (R ) and market rates of return. Thus we worked with an ex post analogue of Sharpe s (1964) excess return model.9 Results Initially we divided the 105-week period into two: weeks –52 to week 0 (i.e., the 53 weeks ended on the Friday of the week a course was held); and week +1 to week +52. For simplicity we refer to them as the pre- and post-event periods. Although the share game was stated to conclude within 14 to 29 weeks of the end of the course week (refer Table 1), we concentrated our tests on a full year, essentially for statistical reasons. We report later on a validity check which confirmed our initial

264

belief that the conclusions would be unaffected by extending the game ending dates. Results are presented for portfolios and for individual securities. Portfolio price relatives were formed by weighting the price relative of each security included in a portfolio, by the number of times it was recommended for inclusion. Thus 90 price relatives from 73 separate securities are averaged in the “buy” portfolio for Course 1; 60 securities appeared once, 10 securities appeared twice, two appeared three times, and one four times, for a total of 90 recommendations. Rates of return are continuously compounded for the portfolios and discrete for the individual securities, because the latter involved some securities with total losses. Tables 4 and 5 contain the overall results for the Buy and Sell recommendations respectively. From column (2) it is evident there was a positive risk premium (i.e., Rm > Rf) in the post-event period for the last three games whereas it was negative for the first two. Column (3) gives the means of Sharpe’s excess return (Rp – Rf) measure for the portfolios. The buy and sell recommendations yielded identical excess return rankings across courses (Courses 5, 3, 4, 1 and 2 ranked from highest to lowest), which implies that they were diversified sufficiently for the market risk premium to reveal itself.10 Before risk adjustments, buy recommendations outperformed the market for four of the five courses, whereas portfolios constructed on the basis of sell recommendations still returned more than the market for three of the five courses. However, based on a t-test, the average weekly portfolio excess returns

265

were not significantly different from the market excess return (α = ·05), apart from the sell recommendations for Course 5, for which the t-statistic for the difference in the means was 2.60. The sign of the difference was in the predicted direction implicit in the recommendation to sell. Risk-adjusted results are contained in Columns (4) to (7). They are based on Scholes-Williams’ (1977) estimation procedure, which is designed to overcome bias in alpha and beta when they are estimated by Ordinary Least Squares (OLS) and when the securities are either inactively or very actively traded. The alpha value is an estimate of the abnormal portfolio rate of return. We see from Table 4, for example, that the portfolio constructed according to the first course participants’ buy recommendations, yielded an average weekly abnormal rate of return of 0.92 percent over the period week −53 to week 0. Note that the buy portfolio (Table 4) recommended by each course always outperformed the sell portfolio (Table 5) in the pre-event period. The maximum average weekly margin was 1.10% (Course 4) and the minimum was 0.20% (Course 5). The differences for Courses 1, 2 and 4 were significant (α = .01), whereas the others were not. Table 4 Buy recommendations--overall results

266

Table 5 Buy recommendations--overall results

The key to the share game, however, is the post-event period. On three of five courses the mean excess return on the buy portfolio did exceed that on the sell, though the difference between them was never significant (α = .05). Only six of the ten post-event alphas (3 buys, 3 sells) were in the correct direction, which is hardly a strong result in favour of rejecting the null hypothesis11 of no revealed investment expertise. Thus it is difficult to gain a clear impression from the portfolio results. The results for the individual securities are much more revealing, as we shall see. Scholes-Williams portfolio beta estimates are contained in columns (6) and (7). Eighteen of the 20 pre- and post-event betas were less than the market average beta 267

of unity.12 Columns (8) and (9) show that in every case the correlation between the portfolio and the market excess returns increased from the pre- to the post-event period. Intuitively, we would expect this if the participants’ attention were attracted by large, abnormal individual security returns in the pre-event period. The absolute values of those large, abnormal security returns probably would have attenuated in the post-event period and thus have become a less important component of the total return (on a diversified portfolio). Individual security results are contained in columns (10) and (11). For example, of the 73 separate securities recommended for purchase by participants in the first course, 34 had negative post-event alphas. Over all courses, only 123 out of 248 securities, or very slightly less than half, had post-event alphas in the expected direction. The inescapable conclusion: no revealed expertise in picking winners. However there is evidence in Table 5 that participants could identify securities which were more likely to be losers. There is less than a 1% probability that we would have observed a total of 124 negative post-event alphas were the null hypothesis true.13 This result holds despite the anomalous result on Course 3. On an individual course basis, only Course 5’s recommendations were significant at the α = .05 level. An interesting feature of this result is that, since participants agreed more on the likely losers than winners, we expected the basis of those views to have been widely known and reflected in the pre-event prices.14 Obviously we were wrong.15

268

Suppose we pooled the results in such a way that we excluded from the “buy” portfolio any security for which there was at least one sell recommendation; and vice versa for the portfolio of securities to be sold. Would the results change? The answer, evident in Table 6, is that they would change very little. There is only one change in the rankings of Sharpe’s post-event excess return measure: an exchange of rankings between the portfolios of buy recommendations of Courses 1 and 4. Based on mean portfolio excess returns, buy recommendations still outperformed sell-recommendations from only three of the five courses. Exactly half the 202 buy recommendations were winners and half were losers. In contrast, sell recommendations were significantly more accurate predictors of future abnormal losses, than we would expect by chance (α = .05). Table 6 Consensus recommendations--summary results

Validity Checks Because studies such as this one can be affected by data errors,16 we screened all data and verified questionable 269

rates of return. We examined the effect of switching from continuous to discrete compounding in the portfolio rates of return and found it made negligible difference. It also made negligible difference whether OLS or the Scholes-Williams procedure was used to estimate the post-event alphas. When we restricted our analysis to the post-event period stated in the game handouts, we reached conclusions similar to those discussed in detail above. For example, individual security post-event alphas were positive for 52% of the buy recommendations, which is insignificantly different from 50% (α = .05). For sell recommendations, 57% of the post-event alphas were negative, revealing significant expertise (α = .05) in picking securities to sell. Post-event risk-adjusted portfolio results are also consistent with the extended period results. Conclusions There is little doubt that there were groups of professionals operating in the Australian sharemarket in the 1970’s, who were able to outperform the sharemarket on a risk-adjusted basis. Their superior performance, which was measured without accounting for transactions costs,17 was due to their systematic ability to nominate future losers to sell. These findings have interesting parallels with Finn’s (1981) study of 361 recommendations made by security analysts working for a major Australian institutional investor. Finn concluded that the recommendations, if acted upon, would have yielded significant abnormal

270

returns, particularly recommendations to sell. He also observed very few sell recommendations were acted upon. .

(Date of receipt of final typescript: February 1982). References Brown, S.J. and J.B. Warner, 1980, “Measuring Security Price Performance,” Journal of Financial Economics, 8, 205-258. Fama, E.F., 1970, “Efficient Capital Markets: A Review of Theory and Empirical Work,” Journal of Finance, 25, 383-417. Finn, F.J., 1981, “Internal Performance Evaluation of Managed Investment Funds,” Unpublished Ph.D. thesis, University of Queensland. Gonedes, N., 1976, “The Capital Market, The Market for Information, and External Accounting,” Journal of Finance, 31, 611-30. Scholes, M. and J. Williams, 1977, “Estimating Betas from Nonsynchronous Data,” Journal of Financial Economics, 5, 309-27. Sharpe, W.F., 1964, “Capital Asset Prices: A Theory of Market Equilibrium Under Conditions of Risk,” Journal of Finance, 19, 425-44. Shevlin, T.J., 1981, “Measuring Abnormal Performance on the Australian Securities Market,” Australian Journal of Management, 6, 67-107.

271

Walter, T.S., 1980, “Australian Takeovers: Capital Market Efficiency and Shareholder Risk and Return,” Unpublished Ph.D. thesis, University of Western Australia.

* University of Western Australia. This study would not have been possible without the co-operation of the respective course organisers. Eatna Sengupta collected the rate of return data. Her skill and care are acknowledged, gratefully. 1

One participant recommended the same shares be bought and sold. His recommendation was included, since it was a valid strategy under the rules of the game. 2

Participants from outside Australia were not excluded from the game although they tended not to take part. 3

The strong form of the EMH is “concerned with whether given investors or groups have monopolistic access to any information relevant for price formation” [Fama (1970), p.383; Cf Gonedes (1976)]. 4

see Finn (1981), for similar contradictory to the strong form.

micro-evidence

5

The expected frequencies under the null hypothesis were generated by assuming a Poisson process. 6

Australian institutions often have policies that tend to restrict most of their investments to shares issued by relatively large companies. For example, they often prefer: companies with proven skills in “remote” management (i.e., through formal planning and control 272

systems); actively traded securities, which offer greater liquidity for larger parcels; and investments in companies that are sufficiently large, so that a “substantial” dollar holding is not a significant proportion (say, more than 10%) of issued capital. Advertiser Newspapers Ltd. was ranked the 100th largest A.A.S.E.–listed company, by market capitalisation, on January 8, 1982. Its capitalisation then was $101.6m (source: Australian Stock Exchange Journal, January 1982, p.28). 7

For example, given the exactly equal numbers of requested buys and sells, there is little incentive for participants to forecast the market average. 8

This price represents the last available market consensus of the worth of a share in the acquired firm, given the terras of the offer. 9

The equation fitted was:

where the “hats” denote estimates and upt is the disturbance term. 10

Note that the ranking across courses for the market risk premium is identical to that of the buy/sell portfolios. 11

The probability of observing these results (under the hypothesis) is 0.38. 12

The sharply-increased post-event estimated beta for course 4’s buy portfolio ( = 1.57) has no obvious explanation. Eighteen of the 30 individual security

273

estimated betas increased from the pre- to the post-event period. Possibly the apparent instability is due to sampling error: note the portfolio r2 values, particularly in the pre-event period. 13

The assumed null hypothesis is that positive and negative post-event alphas were equally likely. 14

It is possible that some participants had decided not to act, in their professional capacities, on their sell recommendations (prior to attending the courses), because a sale would establish an income tax liability, or because the marginal transactions costs (of realisation and re-investment) would have exceeded their avoidable losses. 15

A referee suggests the apparent asymmetry in the results (success at nominating “sells” but not “buys”) could mean we are also wrong in assuming participants accepted the organisers’ guarantee of confidentiality. He reasons that maybe the participants did not trust the course organisers with their “buys”, believing they could act on them; whereas they could be trusted with the “sells”, in the belief that it was unlikely teachers (i.e., non-practitioners) would own “sells” or engage in short-selling. 16

See also Brown and Warner (1980) and Shevlin (1981).

17

It is often a moot point whether transactions costs are relevant, and this study is no exception.

274

© 1982 by The University of New South Wales

275

Does Market Misvaluation Help Explain Share Market Long-Run Underperformance Following a Seasoned Equity Issue? Philip Browna, Gerry Galleryb, Olivia Goeic a

Schools of Accounting and Banking & Finance, University of New South Wales, Sydney, 2052, Australia and UWABusiness School, University of Western Australia, Crawley, 6009, Australia b

School of Accounting, Queensland University of Technology, Brisbane, 4000, Australia c

PT. Cargo International Logistics, Jakarta, Indonesia

Abstract We examine the relation between pre-seasoned equity offering (SEO) announcement date misvaluation and long-run post-SEO performance for a large sample of Australian SEOs made between 1993 and 2001. Our study is motivated by inconsistent findings across countries with respect to the SEO long-run underperformance anomaly first documented in the USA, inconclusive findings with respect to the hypothesis that managers exploit market misvaluation when timing equity issues, and a recent Australian Stock Exchange proposal to loosen SEO regulation. We find SEO firms underperform common share market benchmarks for up to 5 years after the announcement. Using a residual income valuation method, we show that this underperformance is related to pre-announcement date misvaluation. An unexpected result is that underperformance and misvaluation are more severe for 276

private placements than rights issues. Institutional factors unique to the Australian setting, particularly the large number of smaller loss-making firms among private placement issuers, appear to explain the poorer performance of placement firms. Our results are robust to various measurement methods and assumptions, and demonstrate the importance of researching SEO performance in alternative institutional settings. Key words: Residual income valuation; Seasoned equity offerings; Long-run underperformance JEL classification: G10, G14 doi: 10.1111/j.1467-629X.2006.00172.x 1. Introduction We study three aspects of the investment performance of Australian firms that make seasoned equity offerings (SEOs). First, we examine whether SEO firms underperform market benchmarks for holding periods up to 5 years following the SEO announcement date. Second, we investigate whether subsequent performance can be explained by market misvaluation before the SEO announcement date. Third, we analyse long-run performance and misvaluation across SEO type (rights issues and private placements). Our study is motivated by three factors. First, the SEO underperformance anomaly found in the USA (Loughran and Ritter, 1995; Spiess and Affleck-Graves, 1995) has not been fully explored in Australia. Normally that might not be of great interest because the share markets are

277

similar; however, the anomaly we study is not robust across countries (Abhyankar and Ho, 2002). Second, various theories have been proposed to explain SEO firms’ long-run underperformance, such as the ‘windows of opportunity’ hypothesis; namely, that managers of issuing firms take advantage of market misvaluation that occurs from time to time. Hitherto, tests of the misvaluation hypothesis have used the level of issue overpricing to measure misvaluation (Ritter, 1991; Allen and Soucik, 1999a), with inconsistent results. We use a different measure; one based on the residual income value. Third, from a practitioner’s perspective, there is some merit in empirical studies that offer insights into how capital markets operate under changing circumstances. When one considers issues such as the level of equity issues fluctuating substantially over time, the structural changes to the new issues market brought about by the 1994 amendments to the Corporations Law,1 and recent attempts to loosen capital raising rules (ASX, 2003), further Australian research seems warranted. According to information and signalling theories, long-run underperformance is predicted to occur in the Australian SEO market, and to be related to pre-announcement day misvaluation. Long-run underperformance and misvaluation are expected to be less severe for private placement issuers than for rights issuers because investors in placement issues are likely to have better access to private information about the SEO firm. We test our hypotheses for a large sample of SEOs made between 1993 and 2001 by applying a residual income model to measure misvaluation. As

278

predicted, SEO firms do underperform common benchmarks for up to 5years after the announcement date and underperformance is related to announcement day misvaluation. However, contrary to expectation, misvaluation and underperfonnance are more severe for private placement issuers. Institutional factors unique to the Australian setting, particularly the large number of small loss-making firms among private placement issuers, appear to explain the poorer performance of placement firms. The remainder of the paper is organised as follows. Section 2 provides an overview of the SEO underperformance literature. Section 3 develops the hypotheses. Section 4 explains the data, sample and research design. Section 5 presents results from univariate and multivariate tests. Section 6 outlines results of robustness tests and the paper concludes in Section7 with a discussion and summary of findings. 2. Prior Literature 2.1. Evidence on SEO Long-Run Underperformance Anomalous returns to equity offerings were noted in the 1960s (Friend and Longstreet, 1967) and have been studied closely since the 1980s (Asquith and Mullins, 1986; Masulis and Korwar, 1986). Early studies focused on relatively short-term share price performance, but in the 1990s attention shifted to longer periods. Loughran and Ritter (1995) and Spiess and Affleck-Graves (1995) were among the first to show returns to US firms following SEOs were significantly lower than their non-issuing counterparts for up to 5 years. Studies in

279

other countries have often reported similar findings.2 Some evidence of the anomaly is also evident in Australian SEOs. Based on a sample of 102 Australian SEOs between 1983 and 1994, Allen and Soucik (1999a,b) find that these firms, over the 5 year period after the issue, underperform (by 124 per cent) benchmark firms matched on size and industry. 2.2. Explanations for SEO Long-Run Underperformance Reasons given for the long-run underperformance anomaly include delayed investor reactions to signalling, managers exploiting windows of opportunity, earnings management, and outright rejection of the anomaly on the grounds of experimental error. From a signalling perspective, an SEO financing decision is likely to convey a negative signal because if proceeds are used to retire debt, the decision will signal an increase in the firm’s underlying risk (Masulis, 1983), or if proceeds are used to fund capital expenditures, the decision will signal that existing assets generate insufficient funds to finance ongoing investment (Myers and Majluf, 1984). Although the negative signal conveyed by an SEO announcement can help explain a short-term price decline (Asquith and Mullins, 1986; Masulis and Korwar, 1986), it is not clear how, in efficient markets, it can explain long-run underperformance. The windows of opportunity rationale is a behavioural explanation. A ‘window of opportunity’ arises when there is sufficient information asymmetry between management and outside investors regarding the firm’s

280

true value. During windows in which the market has overpriced the firm’s shares, managers can exploit their information advantage by issuing shares at inflated prices (Loughran and Ritter, 1995; Brav et al., 2000). Over certain periods, windows of opportunity might be so pervasive that the market as a whole is overly optimistic in its expectation of the future performance of equity-issuing firms. This over-optimism can obscure an SEO’s negative signal to such an extent that price does not decline in the short run (Spiess and Affleck-Graves, 1995). Bayless and Chaplinsky (1996), Loughran and Ritter (1997) and Jindra (2000) provide empirical evidence in support of the windows of opportunity hypothesis. The earnings management perspective suggests that managers actively deceive the market by managing earnings upwards before an SEO. The market is subsequently surprised when future performance reflects unmanaged earnings (Teoh et al., 1997; Rangan, 1998). Teoh et al. (1997) show discretionary accruals, which proxy for earnings management, in the year before the offering year are directly related to the issuer’s subsequent stock performance over 4 years following the offering.3 However, in a more recent study, Shivakumar (2000) argues that earnings management is not designed to mislead investors but reflects the issuer’s rational response to anticipated market behaviour when the offer is announced. In contrast to prior studies, Shivakumar finds that earnings management per se does not explain the anomaly.

281

Experimental error in the measurement of risk and return has been suggested as another explanation. Lee (1997) argues that if managers of SEO firms have an information advantage they would rationally trade their shares in anticipation of underperformance. As he finds the volume of insider trading is unrelated to the long-run stock return of SEO firms, he concludes that long-run underperformance is not induced by information asymmetry, but results from risk mismeasurement. Others claim inconsistent findings are a result of differences in abnormal return measurement (Eckbo et al., 2000; Shivakumar, 2000). Mitchell and Stafford (2000) show the SEO anomaly disappears when the calendar-time abnormal return (CTAR) method is used to measure returns. Brav et al. (2000) find US firms making SEOs between 1972 and 1992 underperform a matched-firm benchmark, but not a benchmark measured according to a more broadly based (four-factor) asset pricing model. Similarly, Cheng (2003) and Li and Zhao (2003) use a propensity score matching procedure and find no evidence of long-run underperformance. 2.3. The Association Between Overvaluation and Long-Run Underperformance Research design issues aside, most explanations for SEO long-run underperformance stem from a belief that investors overvalue an SEO firm’s equity before the issue. Despite this belief, only a few studies (e.g. Ritter, 1991; Schaeffer, 2003) have directly tested for any link between overvaluation on the offer date and subsequent long-run underperformance. Typically these studies

282

measure mispricing by the relative difference between the closing stock price on the offer date and the offer price itself. We are unaware of any published study that has used a fundamental valuation method to explain the SEO long-run underperformance anomaly. Since Ball and Brown (1968), the information content literature shows that accounting numbers used in fundamental valuations explain a significant part of contemporaneous and future stock return. The residual income model (RIM) has become popular as a fundamental valuation method, largely as a result of Ohlson’s theoretical work (Feltham and Ohlson, 1995; Ohlson, 1995). RIM has been shown to predict stock returns better than price-earnings multiples, book-to-price ratios, or discounted cash flow methods (Frankel and Lee, 1998; Penman and Sougiannis, 1998; Dechow et al, 1999; Lee et al, 1999; Ali et al, 2003; Curtis and Fargher, 2003). RIM’s superior predictive ability is often attributed to its lower sensitivity to input measurement error and assumptions inherent in other valuation models (Penman, 2001). The comparative advantages of RIM have stimulated interest in using RIM to measure misvaluation. D’Mello and Shroff (2000) use RIM to show that managers repurchase stock when their assessment of the firm’s economic value exceeds market value. Dong et al. (2003) report the residual-income-value-to-price (RIV/P) ratio helps explain long-run underperformance experienced by bidder firms in takeover situations better than the book-to-price ratio. In the SEO context, Jindra (2000) examines the relation between overvaluation,

283

measured by the difference between the current market price and an earnings-based value, and the probability that a firm will issue new equity. Using a residual income value and a dynamic earnings approach to measure earnings-based value, he demonstrates both approaches are powerful in explaining announcement period returns. However, Jindra does not examine the relation between market misvaluation and the SEO firms’ long-run performance. 3. Hypotheses 3.1. Long-Run Underperformance of Australian SEO Firms Although there are exceptions, most international studies report an underperformance anomaly. Institutional differences can cause differences to be observed across countries. In the UK, a pre-emptive right given to existing shareholders leads to rights issues being more common than in the USA. This difference might partly explain the lower level of reported negative abnormal returns following a UK SEO (Levis, 1995; Abhyankar and Ho, 2002). Although Australian rules on pre-emptive rights are similar to the UK, Australia has a much smaller capital market and relatively smaller firms. Because Loughran and Ritter (1995) show negative long-run underperformance is more pronounced among smaller firms, we would expect to observe significant long-run underperformance in the Australian context. The early Australian evidence of Allen and Soucik (1999a,b) appears to support these findings. Therefore, our first hypothesis is:

284

H1a: Australian firms that announce an SEO subsequently experience share price underperformance in the long-run. As previously indicated, differences in the nature and regulation of equity issues might also impact on the extent of underperformance. Since the Australian Corporations Law amendments of 1994, a prospectus has not been required for a private placement but one is still required for a (public) rights issue. Private placements have become an attractive method of raising capital because they allow firms to obtain capital relatively cheaply and quickly.4 In addition, a private placement could be preferred to a rights issue because of lower proprietary costs. A firm making a rights issue might have to make specific public disclosures about its future growth plans or its intellectual property to attract enough investors to subscribe to the issue. In contrast, a private placement issuer can avoid detailed public disclosures. Cronqvist and Nilsson (2003) argue private placements signal undervaluation, as the willingness of private investors to take up a sizeable stake in the firm is likely to credibly convey undervaluation to the stock market. Uninformed investors assume that private investors are exploiting an information advantage about the prospects of the SEO firm.5 Therefore, assuming other issuer characteristics remain constant across issue type, the level of information asymmetry between managers and investors should be lower for private placements, leading to relatively superior performance. Although we are unaware of any Australian evidence on the relative long-run performance of the two types of SEO, Dehnert

285

(1992; 1994) shows that placement firms tend to outperform rights issuing firms in a narrow window surrounding the SEO announcement date.6 Assuming these findings can be generalised to long-term performance, we hypothesise the following: H1b: Private placements are followed by a lower level of long-run underperformance than rights issues. 3.2. The relation between long-run underperformance and offer date misvaluation As discussed previously, the long-run underperformance anomaly might stem from market misvaluation during the offer period. Penman (2000) proposes that in a less than perfectly efficient market, stock price can stray from fundamental (or ‘true’) value in the short term but would be expected to move toward it in the longer term. According to this mispricing view, if RIV is a more accurate reflection of the true asset value than stock price before the SEO announcement date, and if the costs of arbitrage are sufficiently large to prevent immediate convergence (Lee et al., 1999), a positive correlation should be observed between an SEO firm’s RIV/P ratio and subsequent performance. We therefore hypothesise: H2: The pre-announcement RIV/P ratio of SEO firms is positively correlated with post-announcement long-run underperformance. 4. Data and Method 4.1. Data The SEO sample is sourced from the SDC Platinum database and includes SEOs made between 1993 and 286

2001 by Australian companies listed on the Australian Stock Exchange (ASX). We exclude issues relating to dividend re-investment plans, bonus plans, other non-cash-related share issues, and any by SEO firms that subsequently change their name or principal activity in the study period. To ensure completeness, the SDC sample was checked against a list of SEOs from Connect 4.7 Table 1 summarizes the sampling procedure. Table 1 Sample selection procedure Number of cases

Selecting criteria

Number of Australian SEOs from the SDC4636 platinum database, 1993–2001, after excluding non-SEO issues or SEO firms changing names or principal activities Less: SEOs by trusts, including real estate(87) investment trusts SEOs with missing announcement dates

(257)

SEOs where the rights or placement status(4) cannot be clearly established Observations with missing share prices (in(343) CRIF share price database) Observations with missing financial(210) information (in the Aspect/Huntley DatAnalysis database) 287

Observations with negative book value of(81) equity Observations with RIV/P outliers

(4)

Final sample

3650

Rights issues

664

Private placements

2986

SEO, seasoned equity offering; CRIF; Centre for Research in Finance; RIV/P, residual-income-value-to-price. Share prices used to calculate returns are sourced from the Centre for Research in Finance (CRIF) share price-price relative and Securities Industry Research Centre of Asia Pacific (SIRCA) Core Research Data (CRD) databases.8 Accounting input variables are sourced from the Aspect/Huntley DatAnalysis database and analysts’ earnings forecasts are obtained from the Institutional Brokers Estimate System (I/B/E/S). Observations with missing announcement dates, information about SEO type (rights or placements), share prices, or accounting data were deleted following unsuccessful manual searches. Firms with negative book value of equity were also deleted from the main analysis.9 Our final sample comprises 3650 SEOs (664 rights issues and 2986 private placements). Further analysis (not shown in tables) reveals the number of SEOs peaked in 1999 (593), reflecting an active issue period following soon after a low in 1998 (377). The modal 288

number of SEOs made by any one firm during the sample period is one and the maximum 34. Approximately half the SEOs were issued by resource firms (1804), which include production as well as exploration companies. Other industrials constitute the second largest group (627), followed by technology (434), financial services (263), manufacturing (181) and utilities (35). 4.2. Measuring Long-Run Performance Following Ritter (1991) and Spiess and Affleck-Graves (1995), SEO firms’ returns are calculated for holding periods up to 5 years after the issue. Firms that were suspended or delisted are excluded from the relevant 2, 3 or 5 year return observations.10 Share prices are collected from the offer date until the earliest of the firm’s suspension or name change date, the offering’s fifth anniversary, or 31 December 2002. Constraints on the availability of reliable daily share price data prevent their use as the primary source of returns. Instead, holding period returns for both issuers and the benchmark are calculated as the product of their monthly price relatives. Long-run performance is measured by computing the return from purchasing the shares at the closing price on the day before the SEO announcement date.11 Because CRIF provides monthly data, it is not possible to calculate the return over a period including the announcement date when it does not fall on the first trading day of the month. In such cases, SIRCA’s CRD is used to patch the CRIF monthly data to include the return from the beginning date to the last trading day of the announcement month. In the few cases where the

289

SIRCA daily prices are not available for a particular SEO firm, the closing price of the month before the announcement month, obtained from CRIF, is used instead. Since the 2000 Corporations Law reform, ASX listing rules allow a company to request a trading halt up to 48 h before announcing a share placement. Where this occurs, the last available price before the announcement date is used as the beginning price. As in previous studies, we use both raw returns and benchmark-adjusted (abnormal) returns to measure long-run performance. Abnormal return is computed two ways: (i) as the natural logarithm of the ratio of the cumulative price relative on the stock and on its benchmark; and (ii) as the difference between these two cumulative price relatives. Our main results are presented for the second measure.12 The ASX All Ordinaries Accumulation Index (XAO) as used by Allen and Soucik (1999a) is one possible benchmark. However, because XAO is value-weighted and less appropriate as a benchmark for small firms, we use a different ‘market’ index, namely the average return for all firms covered by the CRIF monthly database. This benchmark is more consistent with that used by Brav etal. (2000) and Schultz (2003). Because CRIF provides only monthly values of its market index, we interpolate in part months by assuming the market benchmark return is earned evenly over the month. A market-wide benchmark suffers benchmark contamination because the SEO firms are themselves included in the benchmark (Loughran and Ritter, 2000; Ang and Zhang, 2002). The returns of matched firms’

290

securities could be used to overcome this bias. However, the Australian equity market has a relatively small number of listed companies with a large number of SEOs. As a result, the sample size drops substantially because of the difficulty in finding suitable matches when the matched-firm method is adopted. The outcome is lower statistical power and possible sample selection bias. Hence, although we also calculate results for a matched-firm benchmark, they are reported later in robustness tests. 4.3. Calculating The RIV/P Ratio RIV/P is the ratio of residual income value to share price. RIM expresses the intrinsic value of the firm’s equity as the current book value of equity plus the present value of an infinite series of expected residual incomes (Ohlson, 1995). In practical applications, we work with a finite series. Frankel and Lee (1998) show a naïve model that uses current earnings (assumed to be earned in perpetuity) performs as well as, or is superior to, models using three or more forecasting periods when explaining future stock prices.13 Models using analysts’ consensus forecasts perform marginally better, but few of our sample firms are on the I/B/E/S database and even fewer have an analyst following sufficient to generate reliable consensus forecasts. Therefore to estimate RIV/ P, we adopt the following naïve version of the model (i.e. with a two-period expansion and assuming no growth):14

291

Future return-on-equity (FROE) is calculated as net income divided by average book value (B). The firm’s required rate of return (r) is estimated according to the capital asset pricing model (Sharpe, 1964). The risk-free rate is proxied by the 11.00 hours cash rate, which is the Australian equivalent of the US short-term treasury bill rate.15 The firm’s beta is sourced from CRIF and is the estimate for the quarter before the SEO’s announcement. If beta is unavailable (e.g. because the firm is a recent listing) then the average beta of all firms in the same industry is substituted. We assume a constant market risk premium of 6 per cent per annum, which Officer (1994) estimates is the long-term premium in Australia. RIV is divided by the last available closing price before the announcement date, to obtain the RIV/P ratio.16 For profitable firms, Frankel and Lee (1998) show RIV/P outperforms B/P in explaining future stock prices but they do not examine loss-making firms. As 71 per cent of our SEO firms are loss firms (i.e. current period ROE is negative), the application of RIM is problematic. To address this concern, we assume residual income is zero for loss firms and for the small number of profit firms with negative residual income. As a consequence, the RIV/P ratio collapses to the book-to-price (B/P) ratio. This assumption is supported by Collins et al. (1999), who find that negative earnings have little predictive ability, whereas book value is a value-relevant proxy for future residual income (and for the abandonment option) for loss firms.17 We also control for differences in the previous loss sequence across SEO firms because, as Joos and Plesko (2004)

292

show, the market prices firms with transitory losses differently from those with persistent losses. 4.4. Hypothesis Testing To test for evidence of underperformance (Hypothesis la), we test the significance of buy-and-hold abnormal returns (BHAR) using a parametric one-sample t-test for BHAR means and a more robust, non-parametric sign test for BHAR medians. An independent sample t-test and a non-parametric, Mann-Whitney U-test are used to test whether the BHAR of private placements and rights issues follow different return distributions (Hypothesis lb). Two procedures are used to test for evidence of an association between the RIV/P and BHAR (Hypothesis 2). The first involves dividing sample firms into RIV/P quintiles and comparing BHAR across quintiles. Mann–Whitney U-tests and Wilcoxon signed ranks tests are then used to test return differences between quintile 1 (lowest RIV/P) and quintile 5 (highest RIV/P). As a more powerful procedure, we estimate an ordinary least squares (OLS) regression model to test the long-run performance explanatory power of RIV/P, after controlling for other potential determinants of long-run returns. The regression model is specified as follows:

In equation (2) α and ε are the constant and error terms, respectively; BHAR is the firm’s long-run abnormal performance; SEOTYPE has a value of one if the SEO is 293

a rights issue and zero if it is a private placement (a negative coefficient is expected in accordance with Hypothesis lb); LRIV/P is the natural logarithm of the RIV-to-price ratio as previously defined (a positive coefficient is expected in accordance with Hypothesis 2); LOSS is an indicator variable for loss firms, equal to one if previous period net income is negative and zero if positive; and LOSS_SEQ is the number of sequential (net income) losses over the 3 years before the pre-announcement reporting period (i.e. t – 1 to t – 3). The remaining variables are control variables and are explained below. The variables DISCOUNT, LISSIZE and ISSUEVOL are included to control for differences across issues. DISCOUNT reflects the SEO’s discount (or premium) and is measured as the SEO offer price divided by the market price prevailing on the day before the SEO announcement. As in previous studies, it is expected to provide a strong signal of the extent of overvaluation. LISSIZE is the natural logarithm of the ratio of SEO shares offered to ordinary shares outstanding pre-announcement and controls for the relative size of the offering. Bayless and Chaplinski (1996) find an offer of a larger percentage of total equity is a stronger negative signal. ISSUEVOL is the total proceeds of all SEO issues in the same year in which the issue occurred and is included to control for ‘hot’ and ‘cold’ issue periods. Consistent with the windows of opportunity hypothesis, Bayless and Chaplinski (1996) report a delayed and more severe market reaction following hot (high-volume) issue markets.

294

The variables AGE, LMCAP, CASH, LEVERAGE and BETA control for differences in firm-specific characteristics across SEO firms. AGE is the number of years from the firm’s listing date to the SEO announcement date. Spiess and Affleck-Graves (1995) report that post-offering underperformance is usually more severe for younger, newly listed, firms. LMCAP proxies for firm size and is measured as the natural logarithm of the firm’s market capitalisation on the last trading day of the month before the announcement month. CASH is cash divided by total assets. Firms raising capital to fund cash flow deficits are less likely to be investing funds in long-term, value-creating projects. LEVERAGE is the firm’s ratio of total interest-bearing debt to total assets. Bayless and Chaplinski (1996) show that leverage moderates the level of market reaction to SEO announcements. BETA is either the individual firm OLS beta, or if unavailable, the industry beta, calculated at the last quarter pre-announcement and is included to control for a firm’s market-related risk. Allen and Soucik (1999b) find that beta might partly explain long-run performance. The last two variables are included to control for industry differences. RESDUM is equal to one if the SEO firm is in a resources industry, defined as ASX primary industry classification codes 1 to 4, and zero otherwise. TECHDUM is equal to one if the firm is in a technology industry, defined as ASX primary industry code 18 (telecommunications) or code 21 (healthcare and biotechnology), and zero otherwise. Spiess and Affleck-Graves (1995) report differences in the post-offering performance of SEO firms across

295

industries, and in our sample, the resources and the technology industries are two distinctly different and dominant industry groupings.18 5. Results 5.1. Sample Characteristics and Univariate Tests of Long-Run Performance Descriptive statistics for all firms in the SEO sample are presented in Table 2. Panel A shows the issue size is relatively small with a mean (median) 12.3 per cent (6.8 per cent) of issued capital, which when combined with the mean (median) discount of 4 per cent (8.3 per cent) implies SEOs have a limited dilution effect. The mean (median) age of an SEO firm is 9.9 (8) years, which implies that the majority of SEO firms are not start-up companies. They are relatively small (median market capitalisation $10.8m) and riskier than the market in general (median beta 1.2).19 Table 2 Descriptive statistics of SEOs issued during 1993–2001 and univariate tests of long-run abnormal return calculated using the CRIF market index benchmark

296

*, ** characteristics are significantly different at the 0.05, and 0.01 levels, respectively (Mann-Whitney U test)./)-values in parenthesis are the probabilities that the return measure differs from zero based on ѓ-statistics for means and sign-tests for medians; ISSUE SIZE is the ratio of seasoned equity offering (SEO) shares offered to ordinary shares outstanding pre-announcement DISCOUNT is the SEO offer price divided by the market price prevailing on the day before the announcement date; ISSUEVOL is the total capital raised via SEOs by all firms in the sample during the year of the SEO; AGE is years since listing; MCAP is the pre-announcement market capitalisation; ASSETS is pre-announcement total assets; CASH is pre-announcement cash divided by ASSETS; ROA is pre-announcement net income (before abnormal items net of tax) divided by ASSETS’, LOSS_SEQ is the number of sequential (net income) losses over the 3 years before the pre-announcement reporting period (i.e. t—1 to t—3); LEVERAGE is pre-announcement total interest bearing debt divided by ASSETS’, BETA is either the individual firm ordinary least squares beta or the industry beta calculated at the last quarter pre-announcement; BIP (book-to-price ratio) is pre-announcement book value of equity to market capitalisation ratio; RIV/P is pre-announcement residual income value to price ratio; Raw return is the buy-and-hold return from investing in an SEO firm for 297

the indicated period; BHAR is the abnormal return, calculated as the difference in buy-and-hold return on the SEO firm and the Centre for Research in Finance (CRIF) market portfolio, (1–5 indicates the holding period in years subsequent to SEO announcement date). As is further evident in Table 2, Panel A are the financial characteristics of SEO firms. Profitability is poor with a mean (median) ROA of −26.5 per cent (−7.7 per cent). In addition, the large number of loss-making firms (71 percent of the sample made losses) and firms with persistent losses (median loss sequence is two for the full sample and three for the loss firm subsample) suggest SEO firms are frequently financially stressed. SEO firms have limited borrowings (median debt is 3 per cent of assets), which is not surprising given their poor profitability and low level of total assets (median $8.4m) to offer as security. The low median B/P ratio (0.54) implies the market expects future growth from most SEO firms. In other words, the market seems to be optimistic that SEO firms will be able to overcome their current financial difficulties and achieve better results in the future. However, the low median RIV/P ratio (0.58) suggests this optimism is not reflected in fundamental value, as hypothesised in our overvaluation hypothesis (Hypothesis 2). Statistics presented in Table 2, Panel B provide strong support for Hypothesis la. A portfolio that consists of all SEO stocks earns a median raw return of −16.4 per cent over the first year following the equity offering.20 The median raw returns are −30.5, −34.2 and −47.8 per cent 298

when accumulated over 2, 3 and 5 years following the equity issue, respectively. Consistent with the raw returns, both the mean and median long-run abnormal returns (BHAR) are negative and are magnified as the holding period lengthens. For example, in the first year following the equity offer, half the SEO firms underperform the CRIF market benchmark by at least 25.8 per cent and by at least 51.6 per cent, 68.6 percent and 112.4 per cent over 2, 3 and 5 years after the issue.21 The returns earned by Australian SEO stocks are much lower than those reported in US studies. For example, Loughran and Ritter’s (1995) sample of SEO firms has a mean raw return of 45.9 per cent over the 5 years following the SEO issue. Various institutional factors (which we discuss later in the paper) might explain the difference. In contrast to Hypothesis la, H1b is not supported. Table 2, Panel B shows the raw returns and BHAR characteristics separately for rights and placement firms. The latter generate 8.3, 16.4, 9.7 percent, and 6.7 per cent lower median raw returns than rights issue firms over 1, 2, 3 and 5 years after the issue and this poor performance is reflected in benchmark-adjusted (BHAR) returns. The return differences are statistically significant over all holding periods except the last 5 year period. Characteristics of subsamples reported in Table 2, Panel A reveal that the lower (abnormal) return earned by placement firms does not seem to be driven by a small firm effect. The median differences show placement firms are similar in market capitalisation to rights issuers. However, they differ in most other

299

characteristics. Placements are smaller issues,22 offer lower discounts, and the larger ISSUEVOL shows they are made in more active markets. Furthermore, placement firms are less profitable, are more highly levered and tend to have higher levels of market-related risk. A lower B/P ratio indicates that market participants attach higher growth expectations to private placement firms. However, a low B/P ratio can indicate a stock is overpriced and this seems to be reflected in the lower median RIV/P ratio for placement firms. If this is the case, greater underperformance is expected of placement firms than of rights firms. We investigate this issue later, when we consider how the RIV/P ratio is related to future performance. Although the signalling hypothesis and proprietary cost theory that underpin Hypothesis lb are not supported, the results are not necessarily inconsistent with previous studies when institutional factors are considered. For example, Abhyankar and Ho (2002) find calendar-time, value-weighted UK rights issue portfolios earn + 110 basis points per month, whereas value-weighted share placement portfolios earn +75 basis points per month. Australian rights issues tend to be larger than private placements, and are considered more costly and time consuming (Dunn, 2003; ASX, 2004).23 As rights issues typically require 40 business days between the announcement or proposal date and the issue date, issuers must anticipate and incorporate any downward price movement during that period into the pricing of the issue. A higher discount for rights issues helps ensure that the issue is more fully subscribed, In contrast, a private placement that is less than 15 per cent of issued

300

capital requires no shareholder approval or prospectus and can be arranged at short notice (often within 48 h) with a lower discount. Hence the size and urgency of capital raising are important determinants of SEO type in Australia. A private placement is likely to be preferred when a firm wishes to raise a small amount of new funds quickly at minimal cost. A rights issue is likely to be preferred when a firm wishes to raise a large amount of new funds without diluting existing equity interests because all current shareholders are entitled to subscribe to the new shares in proportion to their present holdings. In addition, the substantial fixed costs of a rights issue are more justifiable when the amount raised is large. 5.2. Univariate Tests of RIV/P’s Predictive Ability Our first test of Hypothesis 2 is conducted by comparing the abnormal return characteristics of quintile portfolios formed according to the RIV/P ratio. The results are presented in Table 3. In Panel A, RIV/P quintiles are ranked from the lowest (Ql) to the highest (Q5). A degree of market misvaluation, as captured by the RIV/P portfolios, is clearly evident. Mean and median 1 year abnormal returns (BHAR1) increase monotonically down the portfolios and the two extreme quintiles are significantly different, with the means (medians) differing by 44 per cent (36 per cent). This pattern is found for the 2, 3 and 5 year holding periods. Table 3

301

Mean and median return characteristics of RIV/P and B/P quintile portfolios of firms issuing SEOs between 1993 and 2001

*, ** Characteristics of quintile 1 and quintile 5 portfolios are significantly different at the 0.05 and 0.01 levels based on t-statistics for means and Mann–Whitney U-tests shown under medians (two-tailed tests). B/P (book-to-price ratio) is pre-announcement book value of equity to market capitalisation ratio; RIV/P is pre-announcement residual income value to price ratio; BHAR is the abnormal return, calculated as the

302

difference in buy-and-hold return on the seasoned equity offering (SEO) firm and the Centre for Research in Finance market portfolio, (1–5 indicates the holding period in years subsequent to SEO announcement date). Table 3, Panel B shows similar results for the profit-firm subsample to those reported in Panel A for the full sample. However, when the portfolios are sorted according to the B/P ratio (Panel C), the pattern is not as evident and the difference between the extreme portfolios is less significant. These results demonstrate the RIV/P ratio is a superior valuation measure to the B/P ratio for profitable firms. In contrast, for loss firms (where the RIV/P ratio collapses to the B/P ratio, following our assumption of zero future residual income), the B/P ratio performs as well as the RIV/P ratio does for profit firms. Panel D shows that abnormal returns for the loss firms increase monotonically down the portfolios and the extreme quintile difference is significant across all holding periods. Therefore, these results provide support for Hypothesis 2; SEO firms with lower RIV/P (greater overvaluation) have significantly lower abnormal returns than their higher RIV/P counterparts. 5.3. Multivariate Analysis Table 4 reports the regression analysis of factors expected to explain the level of abnormal return for 1, 2, 3 and 5 year holding periods. Contrary to Hypothesis lb, the result for the 1-year holding period shows a significantly positive SEOTYPE coefficient of 0.099 (t = 2.44, p < 0.01). Similar results with varying levels of significance are observed across the other holding 303

periods. Consistent with our earlier univariate findings, private placement issuers rather than rights issuers have lower abnormal returns after controlling for other SEO and firm-specific factors. In economic terms, the coefficients signify that, on average, private placement firms earn 10 to 20 cents (depending on the holding period) less than other firms for every dollar of return. Further analysis (not shown in tables) reveals larger and significant SEOTYPE coefficients for loss-making firms than for profitable firms, implying the lower long-run performance by placement firms is primarily driven by loss firms. Hypothesis 2 proposes a positive association between the RIV-to-price ratio and long-run abnormal return (i.e. more overvaluation leads to lower returns). As hypothesised, a significant positive LRIV/P coefficient of 0.08 (t = 6.98, p < 0.01) is observed in Table 4 for the 1 year holding period and similar results are evident over longer holding periods. When the model is estimated for the issue type subsamples, the results (not reported in tables) clearly show the more dominant role of placement issuers relative to rights issuers. The LRIV/P coefficients for placement firms are significant across all holding periods, whereas for rights firms, they are generally smaller and significant only in the 1 and 2 year holding periods. These findings reflect the greater overvaluation (lower average RIV/P ratio) of placement firms relative to rights firms reported in the univariate analysis. Table 4

304

Regression analysis of valuation and other factors explaining post-announcement abnormal returns over 12 3 and 5 year holding periods for firms making SEOs between 1993 and 2001

ˆ, *, ** Significant at the 0.1, 0.05, and 0.01 levels (one-tailed test when coefficient sign is predicted, two-tailed when coefficient sign is not predicted), respectively. Abnormal return (BHAR) is calculated by

305

subtracting the Centre for Research in Finance benchmark return from the seasoned equity offering (SEO) firm’s buy-and hold return for the respective period. SEOTYPE is equal to 1 for a rights issue or 0 for a private placement; LRIVIP is the pre-announcement natural log of the residual income value to price ratio; LOSS is equal to 1 if pre-announcement net income is negative and 0 if positive; LOSS_SEQ is the number of sequential (net income) losses over the 3 years before the pie-announcement reporting period (i.e. t−1 to t−3); LISSIZE is the natural log of the ratio of SEO shares offered to ordinary shares outstanding pie-announcement; DISCOUNT is the SEO offer price divided by the market price prevailing on the day before the announcement date; ISSUEVOL is the total capital raised via SEOs by all firms in the sample during the year of the SEO; AGE is years since listing; LMCAP is the natural log of pre-announcement market capitalisation; CASH is pie-announcement cash divided by total assets; LEVERAGE is pie-announcement total interest bearing debt divided by total assets; BETA is either the individual firm ordinary least squares beta or the industry beta calculated at the last quarter pre-announcement; RES_DUM is equal to 1 if the firm is in the resources industries and 0 otherwise; TECH_DUM is equal to 1 if the firm in the technology industries and 0 otherwise. Also evident in Table 4 for the 1 year holding period is the significantly negative LOSS coefficient of −0.154 (t = −3.54, p < 0.01). Firms with current period losses significantly underperform profitable firms. This result,

306

which persists across all holding periods, highlights the importance of controlling for the direction of prior earnings performance in combination with valuation proxies when explaining long-run performance.24 Consistent with Collins et al. (1999), current earnings and book value appear to be important proxies for expected future abnormal earnings and returns for profit firms, whereas for loss firms, negative current period earnings are largely irrelevant and book value proxies for expected normal earnings (and the abandonment option).25 Following Joos and Plesko (2004), LOSS_SEQ is included in the model to capture differences in the expected performance of loss firms based on the previous history of losses. Firms with persistent losses are more likely to continue to perform poorly relative to firms with more transitory losses. However, our findings show only weak support for this argument among SEO loss firms. The LOSS_SEQ coefficient is significantly negative only in the 3 year holding period (t = −2.27, p < 0.05) and is not significant over other holding periods. The SEO and firm-specific control variables included in the model have varying explanatory power. Among the SEO-related variables, DISCOUNT is the most dominant with significant positive coefficients across all holding periods, indicating, as expected, that greater discounts are associated with lower returns. Also, as expected, larger issuers experience greater underperformance as indicated by the negative LISSIZE coefficient for the 2 year and longer holding periods. In contrast, the coefficient on the issue volume year variable

307

(ISSUEVOL) is significantly negative only for 2 year holding periods (t = −2.36, p < 0.01). Over other holding periods, the coefficient is insignificant or positive. Therefore, we find little evidence that abnormal returns are impacted by the nature of the prevailing market (i.e. hot or cold issue periods).26 Our study period is possibly too short to capture sufficient variation in market activity. Among the firm-specific control variables, the listing age (AGE) and firm size (LMCAP) variables exhibit persistent explanatory power. As in previous research (e.g. Spiess and Affleck-Graves, 1995), the significantly positive AGE coefficient for holding periods up to 3 years demonstrates that more mature firms do not perform as poorly as younger firms subsequent to the issue. The significantly negative LMCAP coefficient across holding periods is consistent with Fama and French (1992). A negative relation is expected if larger firms tend to issue equity when internally generated funds are insufficient for short-term requirements. Smaller firms, if they manage to survive in the long run, are more likely to generate extreme positive returns as they are riskier. Leverage is significant in explaining abnormal returns only for the 1 year holding period. Neither the level of cash holding before the issue (CASH) nor systematic risk (BETA) is significantly related to abnormal return. As others have found, beta is unrelated to abnormal return when size and B/P are included in the regression (Fama and French, 1992). Furthermore, beta is a noisier risk measure for SEO issuers, as most are small growth firms.

308

Noticeable differences are evident in the results for the two dominant industry groups. Technology firms tend to outperform other SEO firms in earlier holding periods. In contrast, resource sector firms tend to underperform other SEO firms in later holding periods. Untabulated results reveal technology SEO firms are larger in size (as measured by market capitalisation), have larger cash holdings, better previous earnings performance and lower systematic risk than resource firms. Technology firms are therefore more likely to be using SEO proceeds to fund growth opportunities, whereas resource firms are more likely to be funding existing, less profitable projects.27 Across all holdings periods, F-statistics show that the independent variables jointly explain a significant amount of abnormal returns. Not surprisingly, the explanatory power of the model, which is typically low for this type of returns model, is greater for shorter holding periods (adjusted R2 drop from 7.5 per cent in the 1 and 2 year holding periods to 4.2 per cent and 4.4 per cent in the 3 and 5 year holding periods). In summary, the multivariate results show consistent support for Hypothesis la and Hypothesis 2 but not Hypothesis lb. The unexpected result for Hypothesis lb implies that institutional characteristics could be important for several reasons. First, ASX listing rules allow funds to be raised within a relatively short period and on a more frequent basis via a private placement. In addition, the abolition of a prospectus requirement for share placements makes them a low-cost alternative for raising capital (Dunn, 2003); therefore, they are more

309

likely to be used by financially distressed firms that need to raise capital quickly. Because they are likely to have poor operating performance, subsequent stock returns are likely to be lower than those for rights issues.28 Second, placements are typically for smaller dollar amounts than rights issues. As the proceeds are smaller, placements often will be used to fund shorter-term objectives, such as reducing short-term borrowings, bridging temporary operating deficits, or acquiring short-lived assets. Rights issues, however, are more likely to be used to finance a major investment.29 Accordingly, rights issuers should perform better because they have superior growth potential. Third, in Australia most rights issues are underwritten by financial institutions, but most private placements are not. Underwriting is likely to signal higher quality and reduce the risk to less well-informed investors.30 6. Robustness Tests 6.1. Offer Frequency and Survivorship Long-run performance measures might be sensitive to the treatment of multiple issuing firms and firms not surviving the holding periods tested. To assess the sensitivity of our results to the inclusion of multiple issuers, we re-estimate the regression model for the subsample with only one issue over the 5 year holding period, or with an issue number variable included as an additional model regressor using the full sample. Both approaches do not change our main findings. To address the survivorship issue, all tests are repeated for the subsamples of SEO firms that (i) survived to the end of each holding period, (ii) survived

310

to the beginning of each holding period and (iii) survived the full 5 year study period. For all three subsamples the results are consistent with those previously reported. 6.2. Non-Overlapping Holding Periods As our overlapping holding period approach to reporting returns tends to obscure each subperiod’s contribution to total return, we repeat our tests for each non-overlapping period. For firms that survived to the end of each respective period, the (untabulated) median 1, 2, 3 and 4 to 5 year abnormal returns continue to be significantly negative (−26.2, −27.4, −20.3 and −41.8 percent, respectively).31 When the non-overlapping holding period returns are regressed on the valuation and other control variables (according to equation (2)) a noticeable decline in the model’s explanatory power is evident. As shown in Table 5, most of the explanatory power of the valuation proxies (LRIV/P and LOSS) and issue-specific variables is confined to the first 2 years immediately after the issue. Table 5 Regression analysis of valuation and other factors explaining post-announcement abnormal returns (BHAR) for firms making SEOs between 1993 and 2001 for each non-overlapping holding period.

311

^, *, ** Significant at the 0.1, 0.05, and 0.01 levels (one-tailed test when coefficient sign is predicted, two-tailed when coefficient sign is not predicted), respectively. Abnormal return (BHAR) is calculated by subtracting the Centre for Research in Finance benchmark return from the seasoned equity offering (SEO) firm’s buy-and hold return for each holding period (1, 2, 3 and 4–5 years) after the announcement month. SEOTYPE is equal to 1 for a rights issue or 0 for 312

a private placement; LRIVIP is the pre-announcement natural log of the residual income value to price ratio; LOSS is equal to 1 if pre-announcement net income is negative and 0 if positive; LOSS_SEQ is the number of sequential (net income) losses over the 3 years before the pre-announcement reporting period (i.e. t−1 to t −3); LISSIZE is the natural log of the ratio of SEO shares offered to ordinary shares outstanding pre-announcement; DISCOUNT is the SEO offer price divided by the market price prevailing on the day before the announcement date; ISSUEVOL is the total capital raised via SEOs by all firms in the sample during the year of the SEO; AGE is years since listing; LMCAP is the natural log of pre-announcement market capitalisation; CASH is pre-announcement cash divided by total assets; LEVERAGE is pre-announcement total interest bearing debt divided by total assets; BETA is either the individual firm ordinary least squars beta or the industry beta calculated at the last quarter pre-announcement; RES_DUM is equal to 1 if the firm is in the resources industries and 0 otherwise; TECH_DUM is equal to 1 if the firm in the technology industries and 0 otherwise. 6.3. Abnormal Return Measurement Method A characteristics-matched firm is often claimed to offer a better benchmark than a market index (Barber and Lyon, 1997; Ang and Zhang, 2002). To test the robustness of our results to the application of such a benchmark, we first identify all non-SEO firms within ±30 per cent of the market value of the event firm and within the same

313

industry. Then, the firm with a B/P ratio closest to that of the event firm is selected as the benchmark. The matching procedure leads to a substantial reduction (of approximately 44 per cent) in the number of observations because of the unavailability of many suitable matches. Subsequent analysis shows that the deleted, unmatched SEO subsample is significantly larger in pre-announcement date market capitalisation and cash holdings, and has significantly smaller B/P and RIV/P ratios (i.e. greater overvaluation) than the matched subsample. Despite this sample selection bias, the matched SEO subsample still exhibits long-run underperformance.32 For example, the median 1, 2, 3 and 4 to 5 year abnormal returns are −12.1 per cent, −20.7 per cent, −27.1 per cent, −11.4 per cent, respectively, and all are significant at the 5 per cent level. Therefore, Hypothesis la is robust to this alternative abnormal return benchmark. Untabulated results confirm that our prior findings for Hypothesis lb and H2 are also robust to the use of this alternative benchmark. 6.4. Residual Income Model (RIM) Assumptions Although the residual income model is less sensitive to variation in assumptions than alternative valuation models, errors in input measures can nevertheless reduce its accuracy (Lundholm and O'Keefe, 2001). To demonstrate the robustness of the RIM approach, we re-estimate the regression model using RIV/P calculated according to alternative inputs. First, each firm’s required rate of return is re-computed using a 4 per cent, 5 per cent or 7 per cent market risk premium.

314

Second, we use a simple naïve one-period expansion model where current earnings are assumed to be the best predictor of future earnings.33 Third, two alternative earnings metrics are used to estimate the future ROE: net operating profit after abnormal items (net of income tax), and median analysts’ consensus forecasts, provided by I/B/E/S. Finally, the B/P ratio is used instead of the RIV/P ratio as an alternative valuation measure. That is, we assume all SEO firms (both profit and loss) have zero future residual income. Apart from the B/P and analysts’ forecast-based regressions, the results from each set of tests are similar to those we have reported in detail. In the B/P regressions, the B/P coefficient is noticeably smaller and less significant than in the main results for RIV/P, and is insignificant in the 5 year holding period results. These findings are consistent with those shown in Table 4 and those of Dong et al. (2003), who report the RIV/P ratio has incremental explanatory power over the B/P ratio. When RIV/P is measured using analysts’ forecasts, we find the variable is no longer positively related to long-run abnormal return and it is negative in some versions of the model. We also re-estimated the main regression model inclusive of a dummy variable for I/B/ E/S-covered firms. The related coefficient is insignificant in all our tests, which further suggests that analyst-covered firms are not significantly different in explaining abnormal return. The reduced power of the model resulting from a much smaller sample size (less than 10 per cent of our sample firms have I/B/E/S coverage) is a possible explanation. Analysts’ earnings

315

forecasts might also be overly optimistic before the SEO, which would add noise to our RIV/P measure. However, in a related paper, Brown and Wong (2005) find no evidence of pre-SEO announcement date bias after controlling for factors known to be associated with bias in analyst forecasts. We therefore leave resolution of this issue to further research. 7. Discussion and Conclusion Overall, empirical support is found for the primary hypotheses (1a and 2) but not for Hypothesis lb. In accordance with Hypothesis 1a, both univariate and multivariate tests confirm the existence of a long-run underperformance anomaly following Australian SEOs. These results are consistent with most prior US and non-US research. In addition, the larger magnitude of underperformance reported here, relative to US studies, is likely to be a result of the dominance of small, loss-making firms in our sample. Although we do not attempt to distinguish between explanations for long-run underperformance, our results are consistent with the windows of opportunity hypothesis, whereby managers take advantage of temporary overvaluation in their firm’s stock to issue equity. With respect to Hypothesis lb, contrary to expectations, firms making private placements exhibit larger, negative abnormal returns than firms making rights issues. The Australian institutional environment, in particular the costly nature of rights issues, means larger, more profitable firms are more likely to use this method to raise capital. Therefore, in the case of Australian SEOs, a self-selection rationale appears to dominate explanations

316

based on the proprietary cost or signalling theories that would otherwise suggest less misvaluation and superior performance by placement issuers rather than rights issuers. Consistent with Hypothesis 2, more overvalued (undervalued) SEO firms perform worse (better) than their benchmark in the post-issue period. Our study is the first known to document such a link between SEO misvaluation (using a residual income-based valuation model) and SEO long-run underperformance. The Australian evidence also supports similar findings in other contexts (Frankel and Lee, 1998; D’Mello and Shroff, 2000; Ali et al., 2003; Dong et al., 2003). That is, RIV/P has explanatory power in predicting future abnormal stock returns. That said, we should also acknowledge that the predictive power of RIV/P declines rapidly after the first 2years following the SEO announcement, and is mainly confined to loss films and placement firms over holdings periods beyond 2years. Even where the market apparently misvalues SEO firms, it would be difficult for a would-be arbitrageur to ‘get set’ because most of them are small and thinly traded. Our results are robust to alternative methods and assumptions used to measure returns and fundamental value, and to different subperiods and issuer characteristics. Our misvaluation findings highlight an important implication for researchers using a matched-firm benchmark in SEO long-run performance studies; matching on characteristics correlated with residual income value, such as the book-to-price ratio,

317

might inadvertently introduce matching or sample selection bias leading to spurious inferences about long-run performance. Furthermore, our findings demonstrate the importance of researching SEO performance in alternative institutional settings. Finally, we offer several insights into the behaviour of the Australian SEO market that could interest investors. References Abhyankar, A., and K. Y. Ho, 2002, Long-horizon abnormal performance after SEOs revisited: an asset allocation perspective [online]. Available: http://ssm.com/abstract=338140 Ali, A., L. S. Hwang, and M. A. Trombley, 2003, Residual income based valuation predicts future stock returns: evidence on mispricing vs. risk explanations, The Accounting Review 78, 377–396. Allen, D. E., and V. Soucik, 1999a, Long run underperformance of seasoned equity offerings: fact or an illusion? [online]. Available: http://ssrn.com/ abstract=172633 Allen, D. E., and V. Soucik, 1999b, Performance of seasoned equity offerings in a risk adjusted environment [online]. Available: http://ssm.com/abstract=172648 Ang, J. S., and S. Zhang, 2002, Choosing benchmarks and test statistics for long horizon event study [online]. Available: http://ssm.com/abstract=303803 Asquith, P., and D. W. Mullins, 1986, Equity issues and offering dilution, Journal of Financial Economics 15, 61–89. 318

Australian Stock Exchange (ASX), 2003, Capital raising mechanisms in a disclosure-based market—ASX proposal for informed choice, Exposure Draft, October. Ball, R., and P. Brown, 1968, An empirical valuation of accounting income numbers, Journal of Accounting Research 6, 159–178. Barber, B. M., and J. D. Lyon, 1997, Detecting long-run abnormal stock returns: the empirical power and specification of test statistics, Journal of Financial Economics 43, 341–372. Bayless, M., and S. Chaplinsky, 1996, Is there a window of opportunity for seasoned equity issuance? Journal of Finance 51, 253–278. Brav, A., C. Geczy, and P. A. Gompers, 2000, Is the abnormal return following equity issuances anomalous? Journal of Financial Economics 56, 209–249. Brown, P., and D. Wong, 2005, Properties of Australian security analysts’ forecasts around the time of a seasoned equity offering, working paper (University of New South Wales and University of Western Australia). Cai, J., and T. Loughran, 1998, The performance of Japanese seasoned equity offerings, 1971–1992, Pacific-Basin Finance Journal 6, 395–425. Cheng, Y., 2003, Propensity score matching and the new issues puzzle [online]. Available: http://ssrn.com/ abstract=421006

319

Collins, D. W., M. Pincus, and H. Xie, 1999, Equity valuation and negative earnings: the role of book value of equity, Journal of Accounting Research 74, 29–61. Cronqvist, H., and M. Nilsson, 2003, The choice between rights offerings and private equity placements EFA 2001 Barcelona Meetings [online]. Available: http://ssrn.com/abstract=275846 Curtis, A., and N. Fargher, 2003, A comparison of residual income and comparable firm valuation of initial public offerings [online]. Available: http://ssm.com/ abstract=408320 Dechow, P. M., A. P. Hutton, and R. G. Sloan, 1999, An empirical assessment of the residual income valuation model, Journal of Accounting and Economics 26, 1–34. Dehnert, J., 1992, A study of rights issues of equity: a theoretical and empirical analysis, unpublished PhD thesis, Australian Graduate School of Management, University of New South Wales. Dehnert, J., 1994, A study of private placements in Australia, Miscellaneous Working Paper No. 352 (CS First Boston, Australia). D’Mello, R., and P. K. Shroff,2000, Equity undervaluation and decisions related to repurchase tender offers: an empirical investigation, Journal of Finance 55, 2399–2424. Dong, M., D. Hirshleifer, S. Richardson, and S. H. Teoh, 2003, Does investor misvaluation drive the takeover market? EFA 2003 Annual Conference Paper No. 652;

320

Dice Center Working Paper No. 2003–7 [online]. Available: http://ssm.com/abstract=393021 Dubois, M., and P. Jeanneret, 2000, The long-run performance of seasoned equity offerings with rights: evidence from the Swiss market [online]. Available: http://ssrn.com/abstract=239873 Dunn, J., 2003, Easy placements leave rights behind, The Australian, Nov. 3, Wealth Supplement, 4. Eckbo, B. E., R. W. Masulis, and O. Norli, 2000, Seasoned public offerings: resolution of the ‘new issue puzzle’, Journal of Financial Economic 56, 251–291. Elder, R. J., and J. Zhou, 2003, Audit quality and earnings management by seasoned equity offering firms [online]. Available: http://papers.ssm.com/ abstract=321042 Fama, E., and K. R. French, 1992, The cross-section of expected stock returns, Journal of Finance 47,427–465. Feltham, G. A., and J. A. Ohlson, 1995, Valuation and clean surplus accounting for operating and financial activities, Contemporary Accounting Research 11, 689–731. Frankel, R., and C. M. C. Lee, 1998, Accounting valuation, market expectation, and cross-sectional stock returns, Journal of Accounting and Economics 25, 283–319. Friend, I., and J. R. Longstreet, 1967, Price experience and return on new stock issues, Investment banking and

321

the new issues market (World Publishing, Cleveland, Ohio). Frith, C., 2003, Wee fees and a rights rebound: underwriting fees are falling and rights issues are suddenly back in fashion, The Australian, 28 October, p.18. Jeanneret, P., 2000, Use of the proceeds and long-term performance of French SEO firms [online]. Available: http://ssm.com/abstract=239872 Jindra, J., 2000, Seasoned equity overvaluation, and timing [online]. http://ssrn.com/abstract=216010

offerings, Available:

Joos, P., and G. A. Plesko, 2004, Valuing loss firms, working paper (Massachusetts Institute of Technology, Massachusetts). Kang, J. K., Y. C. Kim, and R. M. Stulz, 1999, The underreaction hypothesis and the new issue puzzle: evidence from Japan, Review of Financial Studies 12, 519–534. Lee, C. M. C., J. Myers, and B. Swaminathan, 1999, What is the intrinsic value of the Dow? Journal of Finance 54, 1693–1741. Lee, I., 1997, Do managers sell overvalued equity? Journal of Finance 52, 1439–1466. Levis, M., 1995, Seasoned equity offerings and the short and long-term performance of initial public offerings in the U.K., European Financial Management 1, 125–146.

322

Li, X., and X. Zhao, 2003, Is there an SEO puzzle? AFA 2004 San Diego Meetings [online]. Available: http://ssm.com/abstract=451740 Llorca, M. J. P., and J. C. G. Sala, 2003, Is the market overoptimistic about the prospects of firms that issue equity? Evidence from the Spanish market, working paper (University of Alicante, Alicante). Loughran, T., and J. R. Ritter, 1995, The new issue puzzle, Journal of Finance 50, 23–51. Loughran, T., and J. R. Ritter, 1997, The operating performance of firms conducting seasoned equity offerings, Journal of Finance 52, 1823–1850. Loughran, T., and J. R. Ritter, 2000, Uniformly least powerful tests of market efficiency, Journal of Financial Economics 55, 361–389. Lundholm, R. J., and T. B. O’Keefe,2001, On comparing residual income and discounted cash flow models of equity valuation: a response to Penman 2001, Contemporary Accounting Research 18, 693–696. Masulis, R., 1983, The impact of capital structure change of firm value: some estimates, Journal of Finance 38, 107–126. Masulis, R. W., and A. N. Korwar, 1986, Seasoned equity offerings: an empirical investigation, Journal of Financial Economics 15, 91–118. Mathew, P. G., 2002, Long-horizon seasoned equity offerings performance in Pacific Rim markets, Review of Financial Economics 11, 317–333.

323

Miller, M. H., and F. Modigliani, 1966, Some estimates of the cost of electric utility industry, The American Economic Review 56, 333–391. Mitchell, M. L., and E. Stafford, 2000, Managerial decisions and long-run stock price performance, Journal of Business 73, 287–320. Myers, S. C., and N. Majluf, 1984, Corporate financing and investment decisions when firms have information that investors do not have, Journal of Financial Economics 13, 187–221. Officer, R. R., 1994, Cost of capital of a company under an imputation tax system, Accounting and Finance 34, 1–17. Ohlson, J. A., 1995, Earnings, book values, and dividends in equity valuation, Contemporary Accounting Research 11, 661–687. Pastor, M. J., and J. F. Martin, 2001, Efectos a largo plazo de las ampliaciones de capital en el Mercado espanol, Instituto Valenciano de Investigacions Economicas, WP-EC 2001–2026. Penman, S., 2000, Financial statement analysis & security valuation (McGraw-Hill, New York, NY). Penman, S., 2001, On comparing cash flow and accrual accounting models for use in equity valuation: a response to Lundholm and O’Keefe, Contemporary Accounting Research 18, 681–692. Penman, S., and T. Sougiarmis, 1998, A comparison of dividend, cash flow, and earnings approaches to equity

324

valuation, Contemporary Accounting Research 15, 343–383. Rangan, S., 1998, Earnings management and the performance of seasoned equity offerings, Journal of Financial Economics 50, 101–122. Ritter, J. R., 1991, The long-run performance of initial public offerings, Journal of Finance 46, 3–27. Schaeffer, J., 2003, The missing link: establishing the connection between IPO underpricing and underperformance, working paper (University of Chicago, Chicago). Schultz, P., 2003, Pseudo market timing and the long-run underperformance of IPOs, Journal of Finance 58, 483–517. Sharpe, W. F., 1964, Capital asset prices: a theory of market equilibrium under conditions of risk, Journal of Finance 19, 425–442. Shivakumar, L., 2000, Do firms mislead investors by overstating earnings before seasoned equity offerings? Journal of Accounting and Economics 29, 339–371. Spiess, K. D., and J. Affleck-Graves, 1995, Underperformance in long-run stock returns following seasoned equity offerings, Journal of Financial Economics 38, 243–267. Stehle, R., O. Ehrhardt, and R. Przyborowsky, 2000, Long-run stock performance of German initial public offerings and seasoned equity issues, European Financial Management 6, 173–196.

325

Suzuki, K., 2000, Seasoned equity offerings in the UK, usage of funds, method of issue and share price reaction of issuers, working paper (London Business School, London). Teoh, S. H., I. Welch, andT. J. Wong, 1997, Earnings management and the underperformance of seasoned equity offerings, Journal of Financial Economics 50, 63–99.

The authors thank Asher Curtis, Jesse Brooke, Patricia Dechow, Richard Sloan, Natalie Gallery and seminar participants at the University of Queensland and Queensland University of Technology, participants at the 2004 Symposium on Market-based Accounting Research at the University of Melbourne, and delegates at the 2004 Accounting and Finance Association of Australia and New Zealand Conference for their helpful comments. We also thank the Deputy Editor, Professor Ian Zimmer and an anonymous referee for their useful suggestions. This paper is based on Olivia Goei’s honours thesis at the University of New South Wales (UNSW), supervised by Philip Brown and Gerry Gallery. Olivia gratefully acknowledges the financial assistance provided by a Faculty of Commerce and Economics, UNSW honours scholarship. Received 14 March 2005; accepted 6 June 2005 by Ian Zimmer (Deputy Editor).

326

1

Most Australian firms now opt to raise new equity through private placements instead of rights or public issues. 2

For example, long-rununderperformance is observed following SEOs issued in the UK (Levis, 1995; Suzuki, 2000), Japan (Cai and Loughran, 1998; Lang et al., 1999; Mathew, 2002), Hong Kong (Mathew, 2002), Germany (Stehle et al, 2000), France (Jeanneret, 2000), Spain (Pastor and Martin, 2001; Llorca and Sala, 2003) and Switzerland (Dubois and Jeanneret, 2000). 3

Elder and Zhou (2003) observe discretionary accruals from 3 years before to 3 years after the SEOs are lower where the auditor is a Big 5 firm or an industry specialist. 4

In 1993, 47 per cent of SEOs in our sample were placements but after 1994 the percentage increased to more than 80 per cent of total issues. 5

Although Australian continuous disclosure rules prohibit the selective release of inside information, carve-out provisions and enforcement difficulties can result in private information being exchanged ahead of a public release. 6

Dehnert (1992) finds a significant negative abnormal return averaging −1.2 per cent over the 3 day announcement window for a sample of 174 rights issues. However, for the sample of 84 placements, Dehnert (1994) observes no significant negative abnormal returns over the 5 day window surrounding the announcement date and some evidence of positive abnormal returns in the 7 days leading up to the announcement. 327

7

The SDC Platinum database for Australian SEOs starts in 1993 and the relevant Connect 4 database begins in 1999. 8

The CRIF share price-price relative database is available from the CRIF at the Australian Graduate School of Management. The SIRCA CRD are available from SIRCA. 9

These financially distressed firms were excluded to avoid biasing our findings in favour of underperforrnance, but subsequent analysis shows our main findings are not sensitive to their inclusion. 10

If these events occur in the second year, the firm’s 1 year return is calculated accordingly, but the 2, 3 year and 5 year returns are assigned a return of −100 per cent (see Ritter, 1991). As a robustness check, the sample was confined to surviving firms only, with similar results. 11

Spiess and Affleck-Graves (1995) use the closing price on the day of the offering as the starting point for return calculations. We use the closing price the day before the announcement date as a result of Australian evidence that the announcement day return already reflects information relayed by the announcement (see Dehnert, 1992,1994). 12

Untabulated results show that the abnormal returns are qualitatively similar for the two measures. 13

Lee et al. (1999) and Ali et al. (2003) also show the model’s valuations are generally not sensitive to forecast horizons beyond 3 years.

328

14

The zero growth assumption is not unreasonable given our expectation of underperformance. See Curtis and Fargher (2003) who apply a similar model in an initial public offering context. 15

Lee et al. (1999) demonstrate the short-term t-bill rate is a better measure of the risk-free rate than the long-term t-bond rate when used in RIV calculations. The Australian Government no longer issues short-term treasury bills so the cash rate is used as a substitute. 16

We report sensitivity tests of our RIM assumptions later in the paper. 17

Miller and Modigliani (1966) had earlier argued losses (negative earnings) are a poor proxy for the earning power of a firm’s assets. 18

Our results are not sensitive to the inclusion of controls for the other minor industry groupings. 19

The higher betas are not surprising given the dominance of resources and technology firms in the sample. 20

Although we describe the return as following the offer, recall that it includes the return on announcement day. 21

Using the continuously compounded returns method, SEO firms underperform the CRIF market portfolio by 26, 54, 59 and 90 per cent over 1, 2, 3 and 5 years following the SEO announcement. 22

ASX listing rule 7.1 limits the percentage of securities that can be issued via a private placement without

329

shareholder approval to 15 per cent of total shares outstanding; it was 10 per cent until November 1996. 23

Most rights issues are underwritten whereas most placements are not. The underwriter’s fee is estimated to be between 2 and 3 per cent for major capital raisings (Frith, 2003). 24

In sensitivity analysis we also include a slope dummy (LOSS multiplied by LRIV/P) in the model. A significant positive coefficient is observed in the 1 and 2 year but not the 3 and 5 year holding period regressions. All other results remain substantially the same. 25

Recall that for loss firms, RIV/P equals B/P because we assume zero residual income. We test this assumption in sensitivity analysis by including current period net income (as recorded in the period prior to the SEO) deflated by total assets as a separated independent variable in our model for loss firms. The coefficient remains insignificant across all holdings period, confirming the irrelevance of current period earnings for predicting future performance. 26

In sensitivity tests we continue to find inconclusive results using alternative proxies for hot/cold issue years including raw number of issues per year, ranked number of issues per year, ranked value of issues per year, and issue year dummy variables. 27

Estimating the regression model inclusive of additional dummy variables to cover other major industry groups produces results qualitatively similar to those reported with little evidence of other industry effects. 330

28

Obviously we are assuming the market is unaware of or unable to respond quickly to the overvaluations. 29

Although the ASX requires firms to state the objective of a capital raising, it is not always possible to identify the various purposes for the new capital because firms often provide only general obscure statements. 30

In our sample, 75 per cent of rights issues and 22 per cent of placements are underwritten. Untabulated results show the average long-run performance of underwritten SEOs is superior to non-underwritten SEOs. 31

Year 1 excludes the announcement month return.

32

A possible counteracting bias is introduced through matching on pre-announcement B/P. As our earlier results show, misvaluation is likely to lead to inflated B/P ratios for many of our SEO firms. Assuming the matched (non-SEO) firm B/P ratios are correct, the matching procedure will introduce a benchmark bias in favour of finding long-run underperformance. 33

This is equivalent to model 1 in Frankel and Lee (1998). © 2006 The Authors; Journal compilation © 2006 AFAANZ

331

PHILIP BROWN AND IAN DUNLOP A CASE OF REPORTING FORM OVER SUBSTANCE

T

he Australian Accounting Research Foundation’s (AARF) recent discussion paper1 on accounting for financial instruments is a timely reminder that the growth in Australia’s financial services may have outstripped the capacity of current accounting standards to keep financial statement users adequately informed. The Bell Group/Adelaide Steamship one-on-one option agreements are an example of how an unusual transaction can side-step conventional reporting practices and keep large amounts “hidden” from the companies’ public records. These agreements constituted possibly the largest single share-option deal in Australia’s history, yet their economic substance was not reported in either company’s accounts. Had they been reported in the way we suggest, the Bell Group’s financial profile would have been substantially different, and Adelaide Steamship most likely would have reported a major short position in BHP shares. The option agreements came out of Robert Holmes à Court’s epic bid for control of BHP, which was launched in August 1983 with the unlikely offering of two Wigmores shares for one BHP share. Throughout the bid, Holmes à Court opted to hold his cards close to his chest. On earlier occasions2, he had capitalised on his “corporate-raider” reputation by announcing his presence well before acquiring the 10% which would have obliged him to file a substantial-shareholder notice.3 It was in this context that Holmes à Court set in 332

motion the BHP share option play, a play which gave him effective control over 40 million BHP shares without disclosing his position. The deal was done in New York; it was documented in An alleged difficulty in regulating the reporting of novel financial instruments is that their legal form precludes their treatment as assets or liabilities; eg, the contingent, executory nature of option contracts. We argue that, at bottom, all financial instruments consist of a rebundling of basic, well understood economic transactions and so yield to much the same type of analysis. The Gordian Knot of accounting for financial instruments can be cut by having firms account for the economic substance of their transactions. The Bell Group/Adelaide Steamship one-on-one option agreements of 1985 and 1986 illustrate our point. Bermuda by Titmuss, Sainer and Webb of London; yet the consideration, which changed hands between the parties for the signing of each contract, in essence netted out to zero. Four option agreements, all dated 27 April 1985, spelled out the terms of the deal: the “1985 put option”, the “1985 call option”, the “1986 put option”, and the “1986 call option”. The 1985 options covered 29.25 million BHP shares, were exercisable at $181.935 million ($6.22 per share) and expired in July 1985. The Bell Group (“Bell”) had the right to put and Adelaide Steamship (“Adsteam”) the right to call the shares. The 1986 options covered 70 million BHP shares, were exercisable at $518.7 million ($7.41 per share) and expired in September/October 1986. This time, Adsteam had the 333

right to put and Bell4 the right to call. The consideration for granting each option was $US 1,000, so that $US2,000 was payable by Bell to Adsteam and the identical amount by Adsteam to Bell. Secrecy provisions were written into the agreements to keep their existence and terms confidential. We ask and answer the following questions: • Was $US 1,000 “fair” consideration for each option agreement? • What was the effect of the overall deal on Adsteam’s and Bell’s holdings in BHP? • How was the deal reported by the companies involved? • How ought the deal have been recorded in the Adsteam and Bell accounts, if the aim were to reflect its economic substance? For brevity we analyse only the 1986 option agreements. Much the same analysis and conclusions apply to the 1985 options. The appendix contains a chronology of events.

SECRECY

PROVISIONS

WERE

WRITTEN

INTO

THE

AGREEMENTS TO KEEP THEIR EXISTENCE AND TERMS CONFIDENTIAL.

Fair Consideration Both 1986 options were protected against capitalisation changes and against dividends paid by BHP that

334

exceeded 30 cents over the options’ lives. In the event, BHP paid dividends of 20 cents in October 1985 and 24 cents in June 1986 (the 24 cents is calculated by adjusting for a l-for-5 bonus issue). The 1986 put option was exercisable between 1 August and 26 September, the call between 29 September and 1 October. For both contracts the exercise price was $7.41 per share ($518.7m/70m shares). On 27 April 1985 BHP’s ordinary shares closed at $5.92. Using last-sale prices of BHP’s at-the-money exchange-traded options, we calculated the implied standard deviation on BHP’s ordinary shares was 32.6% per annum. Commonwealth government securities of the appropriate maturity were then yielding 13.66%. The unprotected parts of the two dividends, when added together, were worth 27 cents on 27 April (their value when discounted at 13.66%), so that a BHP share, without the two dividend entitlements, was worth $5.65 ($5.92 less $0.27). Given these data, it is a short step to apply the Black-Scholes options model, which places a valuation on the 1986 call option of 70 cents a share, or a total of $49 million on the contract. The 1986 put option can be valued by simple application of the Put-Call Parity Theorem5: that contract was worth, according to our calculations, $86 million. Was $US 1,000 “fair” consideration for each option contract? If the contracts were independent, the answer

335

is a resounding No! Their interdependence is dealt with next. The Overall Deal and its Effect on the Companies’ bhp Holdings According to BHP’s David Adam (general manager corporate affairs), “between April and October 1985 … every public document and every document in our share register led us to believe that Holmes à Court was selling out and Spalvins was buying”.6 Gideon Haigh wrote: “While Holmes à Court’s stake appeared to be dwindling, he actually had his foot on 70 million shares.”7 Seventy million? Not quite, because Holmes à Court had also signed the 1985 options agreements, in which Bell effectively pre-sold 29.25 million BHP shares to Adsteam at $6.22. Leaving aside the question of the 1985 options, Haigh obviously interpreted the 1986 contracts as going “hand-in-hand”. They were tantamount to a forward sale because there was only the remotest possibility that one or other of the parties would not wish to exercise its option. As long as BHP’s share price was not exactly $7.41 at the end of September 1986 (ic, almost 18 months after the contracts were signed), then either Adsteam would wish to put the shares or Bell to call them.8 As far as Adsteam shareholders were concerned, their company had effectively bought 29.25 million BHP shares at $6.22 per share for delivery within three months (the 1985 options), and simultaneously sold 70

336

million at $7.41 for delivery late the next year (the 1986 options). It is worth noting that, at the time of the deal, Adsteam was thought to hold around 30 million BHP shares9, or about 10 million less than the net amount of 40.75 million shares contracted for forward delivery. The situation for Bell Group was complicated by the interposition of its indirectly controlled subsidiary10, Weeks Petroleum Limited, in one of the 1986 agreements. From the viewpoint of a Bell Group shareholder, the consolidated position was fundamentally the reflection of Adsteam’s. Reporting the Deal As mentioned, confidentiality clauses prohibited the disclosure of the existence and terms of the agreements. Exceptions were to be made if disclosure was required “to comply with any applicable law or the requirements of any regulatory body (including any relevant stock exchange)”, or “to advance or protect [either company’s] own commercial interest”. It was further provided that “each party shall so far as reasonably practical notify the other party prior to making any such disclosure”. It was against this contractual background that the terms, if not the existence, of the deal were “one of Australia’s best-kept corporate secrets”.11 Rumours that Adsteam’s BHP holding had been “bought by none other (than) Robert Holmes à Court”12 were met with guarded comments:13 Holmes à Court said “he did not think anything that had been written regarding his involvement had represented the true picture”,14 and he wasn’t saying what that picture was. The rumours persisted; for

337

example, on 20 September 1985 The Sydney Morning Herald (p. 17) carried a story that “Mr Robert Holmes à Court is believed to hold 9.92 per cent of the Broken Hill Proprietary Co Ltd and to be on the brink of some form of takeover offer”15. When, on 2 October, Bell acquired a further 10 million shares as well as call options, it was time to disclose. As Haigh put it: “BHP must have thought it had seen a ghost: Holmes à Court, whom they believed had almost completely disappeared from their register, suddenly had 11 per cent.”16 On 7 October it was revealed that Bell Resources (as Wigmores by now was called) had a relevant interest in 113 million (10.95%) BHP shares: 19.6 million were held beneficially; and another 83 million shares were under option, including the 70 million that had been optioned by Adsteam.17 According to early press reports, the 1985 options had lapsed.18 However when, following legal advice, Adsteam disclosed just a few days later its own BHP holdings, it became clear that Adsteam had acquired the 29.25 million shares after all.19 The annual accounts of Adsteam and relevant members of the Bell Group were not released until after the substantial-shareholder notices had been filed and the options’ existence had become public knowledge. In its 1985 accounts, Adsteam disclosed (note 7) that Weeks held a call option over 70 million BHP shares; Adsteam’s put option was undisclosed, as far as we know. Adsteam’s 1986 accounts hinted at trouble, disclosing (note 7) that the call option would not be completed until 15 June 1987 and that the put option had

338

been extended until three days later. The 1987 accounts again suggested trouble, stating that the option agreement was “duly completed” on 16 June 1987 “strictly in accordance with the obligations of Adsteam … as advised by Adsteam’s solicitors … and confirmed by Senior Counsel” (note 7). The June 1985 annual accounts of Bell Group, released in November, disclosed all four option contracts (note 22). They also disclosed that the 1985 put option was exercised20 by Bell. Similar disclosures were made by Bell Resources, which reported that “a subsidiary company” held the 1986 call option. Bell Resources’ December 1986 annual accounts disclosed that the 1986 call option had been exercised that September by Weeks, with settlement deferred until June 1987. Weeks’s December 1986 annual accounts reported the exercise of its option on 26 September 1986 to acquire 90,177,636 BHP shares for a contract price of $557,011,662. Weeks’s 1987 annual accounts close the story as follows: “In June 1987, the Adelaide Steamship Company Limited was due to deliver … 110 million BHP shares and 40 million BHP Gold Mines Limited (BHP Gold) shares in terms of a deferred settlement agreement. At settlement, Adsteam only delivered 73 million BHP and 20 million BHP Gold shares. As compensation for the short delivery, Adsteam has paid to Weeks a total of $US70 million in full and final settlement.” We are not told how the 70 million BHP shares had grown to 110 million. We treat it as another story.21

339

Throughout the proceedings, the forgotten party seems to have been the Australian Stock Exchange which was, according to it, uninformed. Malcolm Maiden’s comment22 that “Bell Group’s substantial shareholding has highlighted the fact that the substantial shareholding provisions of the Companies Code are incapable of keeping the market fully informed about trading in Australia’s largest company” was also applicable to the AASE’s Listing Requirements as they then stood.23 How Ought Such a Deal be Reported? To answer this question, we reconsider the first question: what was the deal worth? Earlier we concluded that, given the data, as independent transactions the call option was worth $49 million and the put $86 million. Customarily, we think of options markets as places where agents trade risk. If we buy a call option from you, we must think the stock is riskier than you do, other things being equal. Put another way, if the price of the underlying share is accepted as fairly reflecting its worth, then the only stimulus for options trading is different beliefs about future risk. That proposition is fundamental to equilibrium models for the valuation of options. However, in the case of the 1986 options contracts, it is clear that neither Robert Holmes à Court nor John Spalvins believed BHP was fairly priced when it sold for $5.92 on 27 April 1985. All else being equal, their “break-even” price was $6.45, which is readily seen by once again applying the Put-Call Parity Theorem.24 From the terms of the option agreements, Spalvins’s present value estimate of BHP did not top $6.45, while Holmes à Court was probably more optimistic.25

340

An alternative view of the deal can be gleaned by examining the economic substance of the transactions. Since there was a negligible probability that one or other of the options would not be exercised, it was only a matter of time before either party would be able to instigate legal action to complete the relevant contract.26 If Adsteam was “sufficiently” certain that the Bell Group would pay for the 70 million shares, it could have employed the realisation principle and recorded the options as a forward sale contract. As for Bell, the only question was whether Adsteam would be able to deliver. In our opinion, Bell could fairly have relied on Adsteam’s commercial standing in 1985, and have recorded the purchase at that time.

THERE WAS ONE, PERHAPS BELL MIGHT HAVE WISHED

COMPELLING, REASON WHY TO AVOID RECORDING THE

OPTIONS AGREEMENTS AS A FORWARD PURCHASE.

There was one, perhaps compelling, reason why Bell might have wished to avoid recording the options agreements as a forward purchase. To do so would have meant introducing an asset (the shares) and a liability (the amount payable to Adsteam) of approximately the same dollar amount. Because the dollar amounts were very large in relation to Bell’s balance sheet, such inclusions would have caused a substantial deterioration in Bell’s financial profile. Had the options agreement been recorded as a purchase in Bell Group’s June 1985 annual report, Bell’s total assets

341

would have increased from $983.4 million to $1,416 million and its debt/assets ratio would have jumped from 65% to 76%. As for Adsteam, we wonder if it was reluctant to recognise income from what appears to have been a very substantial short position in the sharemarket. Should not this have been disclosed? To complete our analysis, we solved for the discount rate implicit in equating the $5.92 last sale price for BHP on 27 April 1985, with the present value of the sum of three amounts: the two dividend payments which were retained by Adsteam,27 and the $7.41 exercise price. The interest rate turns out to be 21%, which is expensive money when we consider that either Bell or Adsteam would probably have paid about 15% for one-year money in April 1985. On this reasoning, Adsteam financed Bell into 70 million BHP shares at a premium interest rate.28 Robert Holmes à Court the investor and John Spalvins the banker is the way we figure it. But that’s not the way it was reported. It is unreasonable not to expect players in the sharemarket to further their economic self-interest to the fullest extent but, by the same token, accounting standards are justified, in part, by the need to maintain an informed market. As the Bell/Adstcam options agreements illustrate, innovations in financial instruments threaten to reduce to farce the recognition and disclosure requirements of accounting standards based on historical cost accounting. Phillip Hancock’s AARF discussion paper is a timely reminder that all is not well in accounting for financial

342

instruments. We hope it promotes full consideration of the issues. Chronology of Events 15 August 1983: Robert Holmes à Court launches bid for BHP, using Wigmores as the bidding vehicle. 27 April 1985: Date of writing 1985 Put and Call Options covering 29.25 million BHP shares. Bell Group had the right to put and Adelaide Steamship the right to call. The options were to expire in July 1985. Date of writing 1986 put and call options covering 70 million BHP shares. Adelaide Steamship had the right to put and Bell Group through Weeks Petroleum the right to call. The put was exercisable between 1 August and 26 September 1986 and the call between 29 September and 1 October of the same year. May 1985: Rumours that Bell was buying out Adelaide Steamship holdings of BHP. May 1985: Holmes à Court denies rumours of buyout. 14 May 1985: The Australian Financial Review reports that Holmes à Court is understood to have negotiated an option deal with Adelaide Steamship. 20 September 1985: The Sydney Morning Herald reports that Holmes à Court is believed to hold 9.92 per cent of BHP. 2 October 1985: Bell acquires 10 million BHP shares as well as call options. 7 October 1985: Revealed that Bell Resources had a relevant interest in 10.95% of BHP’s ordinary shares. 343

13 October 1985: Adelaide Steamship 1985 annual report released. 30 October 1985: Bell Group 1985 annual report released. 21 January 1986: Bell Resources 1985 annual report released. 23 March 1986: Weeks Petroleum 1985 annual report released. 10 October 1986: Adelaide Steamship 1986 annual report released. 21 October 1986: Bell Group 1986 annual report released. 21 January 1987: Weeks Petroleum annual report released. 24 March 1987: Bell Resources annual report released.

INNOVATIONS

THREATEN TO REDUCE TO FARCE THE

RECOGNITION

&

ACCOUNTING

STANDARDS

DISCLOSURE

REQUIRE

BASED

ON

MENTS

OI

HISTORICAL

ACCOUNTING.

Philip Brown is KPMG Peat Marwick Professor of Accounting in the Department of Accounting and Finance, University of Western Australia. Ian Dunlop is a lecturer in the same department. The authors recognise the helpful comments of Phillip Hancock and colleagues at the University of Western Australia,

344

particularly Raymond da Silva Rosa, and the research assistance of Richard. Maltman and Paul Leong. The views in this paper are those of the authors and do not necessarily represent the viws of KPMG Peat Marwick. Notes 1 Phillip Hancock, Financial Reporting for Financial Institutions and Accounting for Financial Instruments, Australian Accounting Research Foundation, Discussion Paper No. 14, Melbourne 1991. 2 Business Review Weekly 19 September 1986, p. 19. 3 Holmes à Court freely admitted his strategy: “We are active traders in BHP – I think that is on the record – but to do this successfully we must avoid any signalling of our intentions” (1 May 1985, p. 27). There is little doubt that he threw up an effective smokescreen. Bryan Frith had this to say on Holmes à Court’s trades in BHP scrip: “The market has been totally confused by his activities. There have been varying rumours that he was buying, selling or that his holding was static at around 45 million shares (The Australian, 8 October 1985, p. 16). 4 The 1986 call option was held by Bell’s indirect subsidiary, Weeks Petroleum Limited. See note 11. 5 The Put-Call Parity Theorem states that the worth of a put option for the same underlying share, exercisable at the same price and with the same maturity date, is calculated by the price of the call, plus the present value of the exercise price, less the dividend-adjusted price. If we ignore the trivial difference in the expiry dates of the

345

1986 options, the put option was worth $1.23, calculated as $0.70 + $6.18 – $5.65. 6 As quoted by Gideon Haigh, The Battle for BHP, Information Australia, 1987, p. 17. John Spalvins was the chief executive of Adsteam. 7 Ibid. 8 We assume no market frictions. 9 The Australian Financial Review, 30 April 1985, p. 68. 10 Bell Resources held 93.6% of Weeks but Bell Group held only 48.5% of the voting rights in Bell Resources. In both 1985 and 1986 annual reports, the directors of Weeks Petroleum stated that they believed Bell Resources Ltd was the ultimate holding company. Bell Group ended its financial year on 31 December. There is, however, little doubt that Bell Group exercised effective control over both Bell Resources and Weeks Petroleum. 11 Haigh, op. cit., p. 16. 12 The Australian Financial Review,1 May 1985, p. 72. The Age reported that the sharemarket was “awash with rumours this week of Bell buying out the Adelaide Steamship group’s holding in BHP”, 1 May 1985, p. 27. 13The Australian Financial Review, 2 May 1985, reported speculation “that Mr Holmes à Court, now in London, had clinched an option deal with Mr Spalvins. When asked if he had placed a call option for Adelaide Steamship’s BHP stake at $7 a share, he replied, ‘that was a new one’.” Despite the denials, the rumours

346

continued. For example, on 14 May 1985, p. 82, The Australian Financial Review reported that “earlier this month Mr Holmes à Court is understood to have negotiated an option deal with Adsteam’s Mr John Spalvins to buy at a strike price of $7 a share”. 14 The Sydney Morning Herald, 1 May 1985, p. 17. 15 Holmes à Court proffered no comment on the story. 16 Op. cit., p. 13. 17 The Australian, 5 October 1985, p. 29 18 The Australian Financial Review, 8 October 1985, pp. 1,80. 19 eg, The Australian Financial Review, 10 October 1985, p.58; The Australian, 10 October 1985, p. 1. 20 According to Bryan Frith in 10 October 1985, p. 17: “Technically neither option was exercised, but Adsteam on July 1 bought 29.25 million BHP shares from Bell at $6.22, which had the same effect as if Bell had exercised its put.” However, note 22 (ii) states that the option “was exercised jointly by the Bell Group Ltd and its assignee” subsequent to 30 June 1985. 21 Part of the explanation may be the BHP 1-for-5 bonus issue of 3 February 1986. 22 The Australian Financial Review, 9 October 1985, p. 63. 23 The situation today is different, in that similar agreements must be disclosed at the outset.

347

24 Under the theorem, the stock price may be expressed as the difference between the call and put prices (ie, $0)’ plus the present value of the exercise price ($6.18), plus the present value of the div idend entitlement ($0.27) – which comes to $6.45. See also note 27. 25 The option agreements were termed a “$500 million corporate wager” in, inter alia, The Australian, 8 October 1985. It is not obvious how the logic of this deal differed from the thousands of share transactions or “wagers” that take place each day. 26 “Mr Holmes à Court agreed last night that: the effect of the agreement was for Adsteam to bankroll a large part of Bell’s BHP stake for about 15 months” (The Age, 8 October 1985, p. 39). 27 Recall that the options were protected for dividends exceeding 30 cents. Thus the dividends that we discounted were the 20 cents paid in October 1985, and 10 of the 24 cents paid in April 1986. 28 The premium of 7.34% over the risk-free rate possibly incorporates three components: (i) lending risk; (ii) the premium Adsteam demanded to compensate for losing the anticipated increase in BHP’s share price (assuming Adsteam believed that BHP’s shares were currently undervalued and that a “market correction” would occur by the exercise dates); and (iii) a small premium for expected control of BHP. COMMENTARY PHILLIP HANCOCK

348

T

he article on the Bell Group/Adelaide Steamship one-on-one options agreements raises some interesting questions about the appropriate accounting treatment for financial instruments. As noted by the authors, the appropriate accounting treatment for all financial instruments, including option Contracts, is considered in Discussion Paper 14 (DP 14) which was recently issued by the Australian Accounting Research Foundation. The authors state in their paper: “An alleged difficulty in regulating the reporting of novel financial instruments is that their legal form precludes their treatment as assets or liabilities; eg, the contingent, executory nature of option contracts.” These types of contracts are discussed in DP 14 and in the Proposed Concepts Statements on assets and liabilities. The view put forward in all these publications is that all contracts which are equally proportionately unperformed (sometimes referred to as executory contracts) should result in the recognition of assets and liabilities when it is probable that such contracts will be executed. The conclusion in DP 14 is that the rights and obligations embodied in contracts which are equally proportionately unperformed meet the definition of assets and liabilities. Therefore, if the recognition criteria as stated in the Proposed Concepts Statements are satisfied, such assets and liabilities should be admitted to the balance sheet. The first recognition criterion requires for assets that it is probable that the service potential or future economic benefits will eventuate and for liabilities that it is probable that settlement of the

349

liability will be required. The second recognition criterion requires that the value or cost of the asset or amount required to settle the liability can be reliably measured. In chapter 6 of DP 14, some specific financial instruments are examined in relation to the application of the definitions and criteria for recognition of assets and liabilities as contained in the Proposed Concepts Statements. For option contracts, the recommendation is that the rights and obligations embodied in such contracts meet the definitions of assets and liabilities. Therefore, when the recognition criteria are satisfied, such assets and liabilities should be admitted to the balance sheet. The Bell Group/Adelaide Steamship one-on-one option agreements would clearly satisfy the first recognition criterion because, as the authors state: “Since there was a negligible probability that one or other of the options would not be exercised, it was only a matter or time before either party would be able to instigate legal action to complete the relevant contract.” Therefore, for both the 1985 and 1986 option contracts, the first recognition criterion was satisfied as it was almost certain one of the two parties would exercise its option. The authors have argued, in the case of the 1986 options, that the measurement of the liability (asset) for Bell Group (Adsteam) would be the present value of the exercise price under the option contract. While the measurement phase of the conceptual framework is yet to be completed, it is common practice with many

350

financial institutions to use present values, as suggested by the authors, for financial instruments. The valuation of the shares to be acquired by Bell Group (asset) and to be given up by Adsteam (liability) could be at the market price of BHP shares at balance date. This would result in a gain or loss being recorded by Adsteam and Bell Group at balance date. Therefore, the conclusion by the authors that the one-on-one 1986 option contracts should give rise to the recognition of assets and liabilities in the accounts of both Adsteam and Bell Group is consistent with the recommendation in DP 14 in relation to option contracts. Further support for this treatment will be provided if the principles contained in the Proposed Concepts Statements on assets and liabilities in relation to contracts equally proportionately unperformed are retained in the final Statement of Accounting Concept on the elements of financial statements to be released later this year. In commenting about the lack of disclosure of information about the option contracts, the authors state: “Throughout the proceedings, the forgotten party seems to have been the Australian Stock Exchange which was, according to it, uninformed.” It could also be argued that users of general-purpose financial reports were denied relevant information which would have been useful for making and evaluating investment decisions in either Bell Group or Adsteam as required by the application of Statement of Accounting Concept 2 “Objective of General Purpose Financial Reporting” and Statement of

351

Accounting Concept 3 “Qualitative Characteristics of Financial Information”. It is also stated in paragraph 24 of SAC 3: “If financial information is to be both relevant and reliable it is necessary that the substance rather than the form of transactions or events be reported.” This, as the authors have argued, has not been applied to the Bell Group/ Adsteam option contracts. While these concept statements did not apply in 1985 or 1986, it is questionable whether they would have resulted in any different treatment of the option contracts in the accounts of either Bell Group or Adsteam. It is the author’s opinion that until the requirements contained in the Proposed Concepts Statements on assets and liabilities, particularly in relation to contracts equally proportionately unperformed, are mandatory, it is probable that the treatment of the option contracts adopted by both Bell Group and Adsteam would continue to be the practice. Phillip Hancock, of Murdoch University, is the author of AARF Discussion Paper No.14, “Financial Reporting for Financial Institutions and Accounting for Financial Instruments “. References Accounting standards Board and Public Sector Accounting Standards Board, Proposed Statement of Accounting Concept ED 42C “Definition and Recognition of Assets”, December 1987. Accounting Standards Board and Public Sector Accounting Standards Board, Proposed Statement of 352

Accounting Concept ED 42D “Definition Recognition of Assets”, December 1987.

and

Accounting Standards Board and Public Sector Accounting Standards Board, Statement of Accounting Concept 2 “Objective of General Purpose Financial Reporting”, August 1990. Accounting Standards Board and Public Sector Accounting Standards Board, Statement of Accounting Concept 3 “Qualitative Characteristics of Financial Information”, August 1990. Hancock, P.J., “Financial Reporting for Financial Institutions and Accounting for Financial Instruments”, Discussion Paper No. 14, Australian Accounting Research Foundation, December 1990.

© 1991 by CPA Australia

353

Philip Brown, H.Y. Izan and Alfred L. Loh Fixed Asset Revaluations and Managerial Incentives The study attempts to explain why Australian companies revalue their fixed assets, when a revaluation, by itself, has no discernible direct effect on cash flows and is costly to carry out. A revaluation is hypothesised to affect contracting and political costs. It may also help resolve problems associated with information asymmetries, or be used to signal information to investors. The results support the proposition that economic forces help explain the decision to revalue assets. Key words: Fixed assets; Incentives; Management; Revaluation; Valuation. An asset revaluation refers to the restatement of the asset’s book value (carrying amount) to approximate some current value. Revaluation involves an accounting entry which, by itself, has no discernible direct effect on the firm’s cash flows, except for the cost of undertaking and recording the revaluation. Why, then, have asset revaluations been commonplace in Australia and, for that matter, elsewhere? Earlier studies have looked at asset revaluations from an information content perspective.1 They have attempted to identify abnormal returns on the firms’ securities around the time the market became informed that a revaluation had taken place, with mixed results. Brown and Finn (1980) suggested that the question of whether asset revaluations per se affect share prices could better be 354

answered if it were understood why revaluations were made. That is, the incentives that motivate managers voluntarily to revalue their firms’ assets should be explicated, so that the wealth effects of their decisions can be understood. We argue that a revaluation affects contracting and political costs. Hence managers are not indifferent to how and when they revalue their firms’ assets. We also argue that an asset revaluation may help resolve problems caused by information asymmetries, in that insider-managers can signal important information which they hold, by revaluing the assets they manage. Background and Hypotheses The Nature of an Asset Revaluation and its Accounting Effects The Accounting Standards Review Board, in ASRB 1010,2 defines a revaluation as establishing a revised carrying amount for a class of non-current assets, other than by way of depreciation. When a revaluation is made, either a revaluation increment or a decrement can occur. This study examines why firms revalue their fixed assets upwards. It does not address the issue of downward revaluations. Our analysis is confined to fixed asset revaluations because the economic incentives are more clearly specifiable.3 Focusing on fixed asset revaluations yields a more homogeneous sample and so allows meaningful relationships to be more easily specified and tested.

355

The costs of a revaluation include: fees charged by the valuer4; the opportunity and out-of-pocket costs of the time directors spend in reviewing the figures to be reported and discussing them with the auditors; record-keeping costs; and the costs charged by the auditors for the additional review. These costs are presumed to be sufficiently large to influence management’s revaluation decision. A similar position was taken by Watts (1977, p. 67): ‘Market values of assets for use in monitoring covenants are costly to obtain… Revaluations would occur when the effect on agency costs of the disparity between market and book value exceeded the cost of revaluation’ (p. 67). A revaluation of fixed assets increases the book value of total tangible assets and, therefore, of total assets. Shareholders’ equity increases by the same amount, because the bookkeeping ‘credit’ is made to asset revaluation reserve. Before July 1987, the balance in the asset revaluation reserve could be used to capitalize an issue of tax-free bonus shares. As depreciation reflects the most recent carrying amount, a revaluation of depreciable assets affects future profit calculations. Net profit can again be affected when the depreciable asset is sold. Accounting standards require: (a) that a gain or loss on the sale of a previously revalued asset be determined as the difference between the sale proceeds and the carrying amount of the asset at the time of sale; and (b) that the gain or loss be brought to account in determining the profit

356

or loss for the financial period in which the disposal occurs. Revaluation of a non-depreciable asset can, similarly, affect the net profit when the asset is sold. In the short run, and except for the revaluation costs, if firms revalue their nondepreciable assets there need only be a ‘mainly balance sheet effect’; that is, an increase in the book value of tangible assets with net profit unaffected. If firms revalue depreciable assets, there is a net increase in tangible assets, which is accompanied by lower periodic profit figures. Asset Revaluation in the Contracting and Political Process Contracting Costs. Since the revaluation of an asset affects accounting numbers, the decision by management to revalue assets changes the firm’s financial profile. In particular, it changes the debt to total tangible assets ratio, commonly used in debt contractual arrangements. Australian debt covenants typically limit total liabilities and secured liabilities to fractions (depending on the type of debt issued) of total tangible assets. Violation of a limit imposes costs on the firm, because its investment and financing choices become sub-optimally 5 constrained. Furthermore, a technical default can trigger costly debt repayment or renegotiation costs.6 Since a revaluation increases the book value of total tangible assets, it loosens debt covenant restrictions; a formal limit on leverage will therefore increase the probability of a revaluation, to avoid these costs. This reasoning leads to the following hypotheses:

357

H1: The higher the ratio of debt to total tangible assets, the more likely a firm will revalue its assets. H2a: A firm with a debt covenant in place is more likely to revalue than a firm without a debt covenant. Most previous studies of accounting choice have not determined the financial constraints of each covenant. Rather, the debt/equity ratio was assumed to be an adequate proxy for them all. Ball (1985) claimed that the use of actual leverage to proxy for closeness to debt covenant constraints has led to a bias towards finding significant results. For this reason, we test directly the significance of the firm’s proximity to violating its specific constraint.7 The closer the firm is to a violation, the higher the ‘debt-proximity ratio’ (defined as the ratio of debt to total tangible assets, divided by the debt constraint of the type of debt in place) and the greater the probability that the firm will revalue.8 H2b: The higher the debt-proximity ratio, the greater the probability a firm will revalue. Political Costs. Accounting numbers may be used to justify corporate actions in a politically regulated environment. The Prices Justification Tribunal was established in 1973 to force firms, especially large ones, to pay closer attention to pricing policies, and to promote awareness of public responsibilities. Nieuwenhuysen and Daly (1977, p. 32) documented that the Tribunal evaluated applications to increase prices by making ‘a comparison of a company’s profitability with the average for the industry… Companies considered to be

358

highly profitable were required to absorb a larger proportion of cost increases’. Underlying the political process theory is the crucial assumption that regulators and other interested parties do not incorporate into their decisions the substantive effect of an accounting change. In relation to the Prices Justification Tribunal, Leech (1975, p. 157) suggested that, despite the different ways in which profitability ratios were calculated, ‘the Tribunal either considers the difference unimportant, that there is no theoretical justification for preferring one to the other, or that, in the name of expediency, it is best to accept whatever figures the company can supply!’ When larger firms report ‘high’ profits, their profit reports are more likely to be noticed by regulators and others who may have incentives and the capacity to reallocate resources away from them. Under such circumstances, larger firms have greater incentives to adopt income-reducing procedures and to cut the expected loss from regulation. If it is assumed that managers’ incentives are aligned more closely with the interests of shareholders,9 a manager might undertake a revaluation to lower the firm’s return on equity because, other things being equal, a lower return can reduce political costs and increase the value of the firm. For instance, Watts (1977, p. 69) suggested that the adoption of replacement cost depreciation by BHP in the 1970s was ‘most likely due to that corporation’s size and sensitivity to the political process’. In BHP’s case, the change to replacement cost depreciation resulted in lower profit figures

359

(due to higher depreciation charges) and lower rates of return (through lower profits and a higher asset base).10 It is possible that a revaluation would so increase the size of the asset base that it would draw political attention to the firm. Such an effect would mitigate the benefits from a reduced accounting rate of return. Over time, a firm’s accounting return on equity (ROE) will increase due to inflation, other things being equal. This increase in ROE may result, for example, from the realization of inventory holding gains or from the asset base being recorded at historical cost. Larger firms, being more under public scrutiny, will be more likely to revalue their assets in order to report lower rates of return. H3: Larger firms are more likely to revalue than smaller firms. Size, however, is a noisy proxy for political cost (Ball and Foster, 1982). As it can proxy for other effects (such as industry membership and leverage), the results for this variable must be interpreted with care. Another proxy for political costs is the strike-proneness of the industry in which the firm operates. The high incidence of strikes, particularly in the 1970s (a sub-period covered by this study), was partly due to pressure being applied by unions in the belief that employers were unduly reluctant to accede to wage increases, despite the historically higher accounting returns on equity. Chambers (1973) suggested that a revaluation may relieve companies of misinformed pressure for higher wages.11 Management would find it 360

easier to explain and justify the firm’s financial position, even if the union could figure the effect of the revaluation. H4: Firms operating in strike-prone industries are more likely to revalue than firms operating in other industries. Perry (1979) identified certain industries which have been prone to strikes: coal mining, waterfront, metal trades and building and construction. Firms in these industries may be more strongly motivated to revalue their assets, to avoid ‘excessive’ wage demands. Asset Revaluation and Information Asymmetry Information asymmetry arises when one party to a potential transaction has information not held by another. Myers and Majluf (1984) argued that, in the presence of information asymmetry, a firm with sufficient ‘financial slack’ can avoid higher external financing costs and can take on profitable new projects which otherwise might be passed over. Financial slack can take the form of liquid assets such as cash and marketable securities, and reserve borrowing capacity.12 Reserve borrowing capacity is enhanced by increasing the book value of total tangible assets. The increase can result from retained profits (net assets added as a result of the firm’s income-producing activities and dividend policy), or from increases in specific asset prices. The latter source of reserve borrowing capacity can be accessed, at a cost, through an asset revaluation. With information asymmetry, it is generally assumed that outsiders cannot observe corporate characteristics in

361

sufficient detail to calculate the values of the securities (e.g., Ross 1977). Managers of firms with undervalued shares have an incentive to expend additional resources, such as by the payment of a higher dividend or meeting the costs of a revaluation (as suggested, e.g., by Standish and Ung, 1982), to signal the undervaluation. Financial Slack. As we discussed above, reserve borrowing capacity can be accessed by a revaluation of assets. The ‘worth’ of such a revaluation is related to the size of the slack created, the size and composition of existing slack, the opportunity cost of maintaining slack in the form of liquid assets such as cash and marketable securities, the growth opportunities of the firm and the method by which the revaluation is done. The size of the slack that can be created depends on the disparity between the book and market values of the assets. The greater this disparity, the greater the probability of a revaluation. In the case of property, for example, this disparity depends on the firm’s current investment in property, the length of time since the property was last revalued, the overall inflation rate since the last revaluation, the specific price increases for property relative to inflation generally, and the location and nature of the property (which determine specific property values). The disparity between book and market values increases with the length of time and the annual increase in specific prices since the last revaluation. Because it is difficult to construct a price index that appropriately captures increases in property values, we do not introduce the inflation rate directly into our model. But,

362

other things being equal, we expect a greater frequency of revaluations during a higher inflationary period than at other times (as stated in H10 below). H5: The longer the time since the last asset revaluation, the greater the probability that the firm will revalue this year. We conjecture that property values, relative to other fixed assets in place, are likely to be more strongly correlated with inflationary change. If our conjecture is correct, a significant disparity between balance sheet and current values is more likely for property than for other fixed assets. In addition, since the land component of property is non-depreciable and the depreciation rate for buildings is lower than, say, for plant and equipment, the disparity between the book and market values of property is likely to become relatively larger. Hence, a higher property investment implies more opportunity for revaluation. A corollary is that a revaluation is more likely to be of property than of plant and equipment. H6: Firms are more likely to revalue property than plant and equipment. Revaluations can be carried out with economies of scale. For instance, the rate recommended by the Australian Institute of Valuers for ‘bulk’ valuations is lower than that for a single valuation. Other things being equal, the higher the property investment, the greater the revaluation frequency (i.e., the number of times a firm revalues within a given time frame).

363

H7: The higher a firm’s relative dollar investment in property, the greater the probability the firm revalues this year. The location and nature of the firm’s property should provide a more accurate determination of the value of the particular property and hence the size of the slack. Unfortunately, the requisite information is not published and the issue is not considered further. It can be expected that larger firms have a greater ability to revalue. They have larger pools of assets with greater dollar values, as in the case of a firm with a relatively large investment in property. This means that larger firms can more frequently access slack sufficient to offset revaluation costs. Firm size thus jointly proxies for (a) potential political costs, and (b) the availability of slack that can be accessed. To have more ‘value’ for restoring debt capacity, the revaluation should be done by an independent valuer rather than the directors, for the same reason as that suggested by Healy and Lys (1986) regarding choice of auditor. They argued that Big Eight auditors are likely to be particularly valuable to clients who anticipate raising capital in national or international financial markets, since their reputations lower the information costs of investors. Except for companies with debt covenants, where the valuations in most cases must be carried out by an independent valuer approved by the trustee for the debtholders, management can decide between having a directors’ or an independent valuation, in which event the incremental benefits and costs of an independent

364

valuation over a directors’ valuation would be important considerations. The acceptancce of a directors’ valuation would likely depend on, among other factors, the credibility of management. Because it is difficult to observe the circumstances in which a directors’ valuation is optimal for the intended purpose, we have no hypothesis on the method by which a revaluation is carried out. We do attempt to gain some insight into whether the method of revaluation is relevant to the revaluation decision, by re-estimating the model with the revaluation method taken into account. Financial slack, as mentioned before, consists of liquid assets and reserve borrowing capacity. Other things being equal, the smaller the existing financial slack, the greater the potential gain to the firm from revaluing its assets to restore borrowing capacity. However, in times of higher inflation, if the slack consists largely of cash and marketable securities, the firm could have an incentive to revalue in order to hold slack in the form of reserve borrowing capacity, because the opportunity cost of holding slack in the form of cash and marketable securities increases with inflation. H8: Firms are more likely to revalue when they have lower holdings of cash and marketable securities relative to total tangible assets, than when they have higher holdings relative to total tangible assets. H9: Given high inflation, a firm with higher holdings of cash and marketable securities relative to total

365

tangible assets is more likely to revalue than a firm with lower holdings relative to total tangible assets. This cost effect is examined by choosing two periods, one (1974–7) of relatively high and the other (1984–6) of relatively low inflation. The annual inflation rate (measured by changes in the Consumer Price Index) was about 14 per cent in the 1974–7 period, and 6.6 per cent in the 1984–6 period. Controlling for inflation by having two shorter sub-periods also helps to control for inter-temporal instability of financial statement numbers over time. H10: Due to the effect of inflation on the movement of property values (and hence the amount of slack that could be created), firms revalue more frequently during periods of higher inflation. The size of existing slack varies inversely with the ratio of total debt to total tangible assets. The higher the ratio, the greater the probability that the firm will revalue assets to restore borrowing capacity. However, different firms may have different ‘desired’ levels of slack, even if they have the same ratios. That is, the debt/total tangible assets ratio could be an inadequate proxy. But, since we do not know each firm’s desired level of slack, no adjustment can be made. The failure to make an adjustment will reduce the explanatory power of the model. To discriminate between ‘financial slack’ and ‘default avoidance’ explanations of an asset revaluation (because a revaluation to avoid or minimize the probability of default must increase borrowing capacity), we adopt the

366

following procedure: (a) we examine how revaluers with debt covenants differ from those without covenants, and (b) we test whether revaluations are related to changes in debt subsequent to the revaluation. Revaluers with no debt covenants might have revalued to increase borrowing capacity rather than to lower the probability of default. Further, if default avoidance was the dominant reason for the revaluation, debt would not be expected to increase significantly after a revaluation. Firms with more growth options will value financial slack more highly because of the greater probability of viable projects being passed up, due to higher external financing costs. Consequently, they would be more likely to revalue in order to have sufficient slack, to avoid potential underinvestment. H11: Firms with more growth options are more likely to revalue than firms with fewer growth options. Creating Tax-Free Reserves for a Bonus Issue. Standish and Ung (1982) suggested a revaluation might signal an increase in management’s forecast of future profits, which would make increased dividends possible. It might also signal that the firm has the potential for raising further debt. They proposed that a revaluation plus a bonus issue enabled a company to increase its dividend payout without necessarily increasing the dividend rate, thereby avoiding adverse political criticism for reporting what may be taken as an excessively high rate of return on assets. Whatever the reason for a bonus issue, Ball, Brown and Finn (1977) showed that bonus issue announcements convey ‘good’ news to the share market, probably via their connection

367

with the information content of subsequent dividend changes. Based on the preceding observation, it is predicted that, before July 1987, firms with low tax-free reserves are more likely to have revalued their assets in order to create sufficient tax-free reserves, thus giving them the flexibility to declare future bonus issues if and when they were desired for signalling purposes. H12: Firms are more likely to have revalued when they had lower tax-free reserves than when they had higher tax-free reserves. A related view can be taken, ex post. When a firm declares a bonus issue, one can argue ex post that the firm would have created, earlier, a sufficient level of tax-free reserves. H13: Firms that declare bonus issues are more likely to have revalued, to create sufficient tax-free reserves. While each revaluation would have increased the level of tax-free reserves, the firm’s desired level cannot be specified because we lack a theory which explains each firm’s policy on bonus issues. Thus the relationship between the revaluation decision and whether a firm made a bonus issue (H13) is likely to be stronger than that between revaluations and the level of reserves (H12). Defending Against a Takeover Bid. When a firm receives a takeover bid, its management might face a wealth loss because of the threat to job tenure if the takeover is successful. Management of the target firm might have incentives to fend off the bid. A common defence (Casey and Eddey, 1986) is to signal the informed value of the 368

firm’s assets to its shareholders, through an asset revaluation. Firms which had received takeover bids during the review year are, therefore, more likely than other firms to have undertaken a revaluation (to fend off the threat). Firms which had successfully defended themselves against a takeover in the previous year are also more likely to revalue their assets because, as documented in Walter (1984), a further bid is now more likely. H14: Firms under threat of a takeover are more likely to revalue than firms not under threat. Sample Properties Our sample is confined to firms listed on the Industrial Board of the Australian Associated Stock Exchanges (but excluding utilities, banks, finance and investment companies), and for which annual reports are available in the Australian Graduate School of Management Annual Report File. Two random samples, of 204 and 206 firms, were selected for the 1974–7 (higher inflation) and 1984–6 (lower inflation) periods respectively. The firms selected in the second period are independent of those in the first, to avoid a survival bias. The firms were then placed into the revaluer/ non-revaluer categories depending on whether a revaluation took place in each review year. Random sampling, which avoids the estimation problems associated with choice-based sampling typical of many studies on accounting policy choice,13 is possible here because there is a ‘reasonable’ representation for the two groups, revaluer and non-revaluer.

369

The definitions of the explanatory variables and their predicted relationships with the revaluation decision are given in Table 1. Except for the level of tax-free reserves (we have two variables RES/CAP1 and RES/CAP2) and the level of cash and marketable securities to total tangible assets (CMS/TTA, our proxy for liquid assets), the relationship is predicted to be positive. Table 1 Variable Definitions

370

371

Table 2 shows the frequency of revaluations, the size of the revaluation increments relative to total tangible assets, the types of assets revalued and the methods by which they were revalued. In sum, in the first period there were 816 cases spread over the four years. In the second period, due to the non-availability of 1986 annual reports for 11 firms (most were taken over), there were 607 cases spread over the three years. In the shorter 1984–6 period, about two-thirds of the firms in the sample revalued at least once; and in the longer 1974–7 period, it was about two-thirds again. About 83 per cent of revaluations in 1974–7, and 90 per cent in 1984–6, had a less than 10 per cent effect on total tangible assets. In both periods, about 90 per cent of revaluations were of real property. About 70 per cent of the property revaluations were carried out by independent valuers, compared with 20 and 53 per cent for plant and equipment in 1974–7 and 1984–6 respectively. Table 2 Frequency, Method and Size of Revaluation

372

a

Out of 204 firms in 1974–7 and 206 firms in 1984–6.

b

As some firms revalued both types of assets simultaneously, the total number will not equal the number of revaluers (209 in 1974–7 and 190 in 1984–6). Results Univariate Results Within each sub-period, statistical analysis was performed on both the individual year and the pooled sample of all years. Estimation precision is enhanced by the increase in the pooled sample size.14 For brevity, only pooled results are discussed here. As the distributions of some variables are markedly

373

non-Normal, the Mann-Whitney U-test was chosen to test for differences in the continuous explanatory variables. Revaluers were more levered than non-revaluers in both periods, and were significantly closer to violating their debt constraints (see Table 3). Table 4 shows that, in both periods, higher proportions of revaluers had debt covenants, compared with non-revaluers. Revaluers were significantly larger in size, as expected. The results for labour-related political pressures are mixed. In 1974–7 revaluers had a relatively higher frequency of strikes, but in the 1984–6 period the difference was reversed in sign (though it was not significant). In both periods revaluers had lower time lost due to strikes, but the difference was not significant. Revaluers had higher investments in property; and they revalued more frequently, though again the difference was not significant. They also had lower (albeit statistically insignificant) levels of cash and marketable securities, relative to total tangible assets. In 1974–7 both groups had approximately the same rates of asset growth, while in 1984–6 revaluers had higher growth, whether measured by ASSETS or by P/E multiple. In 1974–7 revaluers had (insignificantly) lower levels of tax-free reserves available and in 1984–6 the means were approximately the same. About 31 per cent in 1974–7 and 28 per cent in 1984–6 of revaluers declared a bonus issue in the revaluation year (BONUS 1), compared with 12 per cent in 1974–7 and 19 per cent in 1984–6 of non-revaluers.15 Revaluers received takeover bids twice as frequently, in both periods. Thus all statistically significant differences in the means of the variables are in the hypothesized directions.

374

Table 3 Univariate Test of The Relationship Between Variables for Revaluer and Non-Revaluer

a

Based on Mann-Whitney U-test.

b

Only firms with debt covenants are included in the tests. c

Negative relation in 1984–6; but can be positive or negative in 1974–7 Table 4

375

Univariate Test of The Relationship Between The Binary Explanatory Variable for Revaluer and Non-Revaluer

a

One degree of freedom.

On the whole, the univariate results generally support our arguments that firms revalue to increase financial slack (or to avoid violating a debt covenant), for signalling purposes and to avoid political costs. Multivariate Probit Results Estimates of the coefficients of a multivariate Probit model are reported in Table 5.16 Except for the ratio of

376

cash and marketable securities to total assets (CMS/ TTA), and the time between successive revaluations (NYEARS) in 1984–6, all coefficients have their expected signs. Table 5 Probit Analysis

377

Significant coefficients were obtained for the leverage variable (DEBT/TTA), debt-proximity ratio (DEBT/ COV), size, relative investment in property (PPTY), ratio of reserves to paid-up capital (RES/CAP1), the payment of a bonus issue (BONUS1), and the receipt of a takeover bid (TAKEOVER). The growth variable, ASSETS, is significant in the 1984–6 period, while NYEARS is significant in the 1974–7 period. The remaining variables, CMS/TTA and the frequency of strikes (FREQUENCY), are not significant in either period. It is difficult to interpret the results for CMS/ TTA for the first period, because we predict a positive or a negative effect could have occurred within that period only. The multivariate and univariate results are essentially consistent, except for RES/CAPI and FREQUENCY. In the univariate tests, RES/CAP1 is not significant in either period (but it is significant when other variables are controlled for), and FREQUENCY is significant in the first period. The Probit model in both periods is significant at less than the 0.001 level. However, its explanatory power is modest, in that the adjusted /?-square is about 12 and 10 per cent for 1974–7 and 1984–6 respectively. Classification accuracy, assessed bytheLachenbruch test, was 75 per cent for 1974–7 and 70 per cent for 1984–6, results which are significant at less than the 0.001 level when compared to the chance classification model.17 Taken together with the univariate results, it is reasonable to conclude that financial slack, signalling

378

reasons, contracting and political costs are statistically significant explanations for the revaluation decision.18 Other Results Revaluation and Debt Covenants. As argued above, if the reason for a revaluation is opportunism to avoid the potential violation of a debt covenant restriction because of the operating and financial risks of existing investments and debt contracts, it is plausible to expect little increase in debt after a revaluation. Alternatively, a revaluation could plausibly be equilibrium restoration of the kind rationally expected by experienced borrowers and lenders. To investigate the issue, the percentage increase in total liabilities (INCR) in the year following the review year was calculated for each observation and the sample was then partitioned into firms with and without debt covenants in place. For firms with a decrease in total liabilities, the variable INCR is truncated at zero. The results (for 1974–7 only) are reported in Table 6. Panel A shows the means of the DEBT/TTA ratio and of the percentage increase in total liabilities for revaluers and non-revaluers. The mean percentage increase in total liabilities in the year following the review year for non-revaluers is only one per cent higher than for revaluers; the difference is significant at the 0.10 level. Although revaluers were more highly levered than non-revaluers (the average DEBT/TTA for revaluers was 5 per cent higher than for non-revaluers) they could still assume substantial amounts of debt, possibly with the help of the revaluation.

379

Table 6 Leverage and Percentage Increase in Total Liabilities 1974–7a

a

The results for 1984–6 are very similar and for brevity are not reported here. They are available from the authors. b

Percentage increase in total liabilities.

Panel B shows that revaluers with no debt covenants (REVALUER/NO COV) had lower leverage but higher increases in leverage than revaluers with debt covenants (REVALUER/COV). That the REVALUER/COV group had lower INCR could be because of their higher leverage and the presence of formal constraints which may have been more binding than informal ones.

380

However, this conclusion is conjectural in view of the statistically insignificant difference. Panel C shows non-revaluers by whether or not they had debt covenants. Non-revaluers without a debt covenant (NON-REVALUER/NO COV) had lower leverage, but higher increases in leverage, than non-revaluers with debt covenants (NON-REVALUER/COV). (The difference in the respective rates of increase is not significant in 1984–6.) Again, that firms with covenants had a lower mean percentage increase in liabilities might be attributable to their higher initial leverage and to the more binding nature of their formal borrowing constraints. The above observations suggest that a firm with a debt covenant typically does not revalue to avoid technical default on past borrowings. Rather, any revaluation is to restore the firm’s borrowing capacity so that it can assume additional liabilities in the following year. Inflation and the Frequency of Revaluation. H10, which predicts that firms revalue more frequently during periods of higher inflation, is not supported. About two-thirds of all firms revalued once in the four years 1974–7 (the higher inflation period) and once in the shorter period 1984–6. Furthermore, the average time elapsed between successive revaluations was 5.4 years in the first period and 3.9 years in the second period, results that again do not support the hypothesis. With hindsight, it could be argued that since a revaluation exercise is costly, a firm would not revalue (even if it could have a larger revaluation increment)

381

unless it would benefit from doing so; for example, when there is a need for reserve borrowing capacity, or when there is strong economic growth (as in the 1984–6 period). A possible counter-argument is that the CPI does not adequately capture movements in property values, such that a comparison between these two periods (high and low inflation, based on changes in the CPI) is irrelevant. Multinomial Logit Results. McMillan (1988) observed that directors’ valuations (in contrast to independent valuations) were associated with negative cumulative average sharemarket residuals for the thirteen months preceding and including the announcement month of the revaluation. In view of her findings that directors’ and independent valuations might be intrinsically different, we fitted a Multinomial Logit model, which allows the dependent variable to take multiple discrete values, compared to the Probit model’s zero-one dichotomy. The dependent variable was assigned a value of unity if the revaluation was carried out by an independent valuer (N = 146 in 1974–7), a value of two if it was carried out by directors (N = 63), and zero if the firm did not revalue at all (N – 607). The logit model was significant at < 0.001 (x2 = 160.3), with a prediction accuracy of 76.1 per cent, compared with 59.1 per cent for the chance model. The results (see Table 7) show that the revaluation method is possibly related to the reason for the revaluation. Firms that are more highly levered and closer to violating debt covenant restrictions tend to have revaluations done by independent valuers; the results are intuitively reasonable, as debt covenants normally

382

require that revaluations be independent. Firms that were larger and those that were more strike-prone tended to have directors’ valuations. Firms facing takeover threats were more likely to have revaluations done by independent valuers; this observation is consistent with the former National Companies and Securities Commission’s requirement that a target company’s disclosure of the market value of its assets be supported by an expert’s report. Table 7 Coefficients of Multinomial Logit Models: 1974–7a

383

a

The results for 1984–6 are very similar and for brevity are not reported here. They are available from the authors. b

Includes cases where a revaluation results from valuations of some assets by the directors and others by an independent valuer. As for the other reasons for a revaluation, there appears to be no significant difference in the choice between the two valuation methods. Conclusions This study investigated whether variables proxying for contracting/political costs and information asymmetry/ signalling explanations could explain why firms voluntarily revalue their fixed assets. Revaluers were more highly levered, closer to violating their debt covenant constraints, larger, had relatively higher property holdings and lower tax-free reserves. Firms which declared bonus issues were also more likely to revalue, as were those facing a takeover, presumably to frustrate the bidder. In the first period, revaluers operated in industries more prone to strikes while in the second period they enjoyed stronger growth prospects. Thus, the results are consistent with the proposition that an asset revaluation lowers the probability of a wealth transfer arising from contracting and political costs and conveys a signal to the users of financial statements. Despite the statistical significance of the model, the level of the R-squared suggests it is far short of providing a complete explanation of the revaluation decision.

384

Misspecification of the theoretical constructs is probably one reason for the modest explanatory power: for example, we used imperfect proxies for the strike-proneness of the industry, the threat of a takeover, the level of tax-free reserves relative to paid-up capital, and financial slack. Other relevant variables, such as the influence of the general state of the economy on the revaluation decision, have not been explicitly modelled. For example, it is likely that the level of economic activity affects the firm’s incentives to revalue, through its effect on the demand for slack, the profitability of the business, the firm’s industrial climate and so on. We leave these issues for future research. References Allen, D., Debt Capacity: Australian Company Practices, Mimeo, University of Western Australia, Department of Accounting and Finance, 1990. Ball, R., Commentary On: Corporate Pension Funding: A Test of Accounting Choice and Finance Theories, Conference on Accounting for Post-Employment Obligations, University of Illinois, 1985. Ball, R., P. Brown and F. Finn, ‘Share Capitalization Changes, Information and the Australian Equity Market’, Australian Journal of Management, October 1977. Ball, R., and G. Foster, ‘Corporate Financial Reporting: A Methodological Review of Empirical Research’, Supplement to Journal of Accounting Research, 1982. Brown, P., and F. Finn, ‘Asset Revaluations and Stock Prices: Alternative Interpretations of a Study by Sharpe

385

& Walker’, in R. Ball, P. Brown, F. Finn and R. Officer (eds), Share Markets and Portfolio Theory, University of Queensland Press, 1980. Casey, R., and P. Eddey, ‘Defence Strategies of Listed Companies Under the Takeover Code’, Australian Journal of Management, December 1986. Chambers, R., Securities and Obscurities: A Case for Reform of the Law of Company Accounts, Gower Press, 1973. Craswell, A., ‘An Examination of Alternative Hypotheses Concerning Depreciation of Buildings’, Abacus, March 1986. Emanuel, D., ‘Asset Revaluations and Share Price Revisions’, Journal of Business Finance and Accounting, Spring 1989. Ford, G., and J. Hearn, ‘Industrial Conflict: An Overview’, in G. Ford, J. Hearn and R. Lansbury (eds), Australian Labour Relations: Readings, Macmillan, 1980. Healy, P. and T. Lys, ‘Auditor Changes Following Big Eight Mergers with Non-Big Eight Audit Firms’, Journal of Accounting and Public Policy, Winter 1986, Kelly, L., ‘The Development of a Positive Theory of Corporate Management’s Role in External Financial Reporting’, Journal of Accounting Literature, Spring 1983.

386

Leech, S., ‘The Prices Justification Tribunal and Profitability Assessment’, The Chartered Accountant in Australia, October 1975. McMillan, M., ‘Asset revaluations in Australia’, draft PhD dissertation, September, University of Queensland, 1988. Myers, S., and N. Majluf, ‘Corporate Financing and Investment Decisions when Firms Have Information that Investors Do not Have’, Journal of Financial Economics, June 1984. Nieuwenhuysen, J., and A. Daly, The Australian Prices Justification Tribunal, Melbourne University Press, 1977. Palepu, K., ‘Predicting Takeover Targets: A Methodological and Empirical Analysis’, Journal of Accounting and Economics, March 1986. Perry, G„ ‘Trends in Australia Strike Activity: 1913–1978’, Australian Bulletin of Labour, 1979. Ross, S., ‘The Determination of Financial Structure: The Incentive-Signalling Approach’, Bell Journal of Economics, Spring 1977. Sharpe, I., and R. Walker, ‘Asset Revaluations and Stock Market Prices’, Journal of Accounting Research, Autumn 1975. Standish, P., and S. Ung, ‘Corporate Signalling, Asset Revaluations and the Stock Prices of British Companies’, The Accounting Review, October 1982.

387

Walter, T., ‘Australian Takeovers: Capital Market Efficiency and Shareholder Risk and Return’, Australian Journal of Management, June 1984. Watts, R., ‘Corporate Financial Statement, A Product of the Market and Political Processes’, Australian Journal of Management, April 1977. Whittred, G., and K. Chan, ‘Asset Revaluations and the Management of Financial Slack’, Working Paper, University of New South Wales, July 1987. Whittred, G., and I. Zimmer, ‘Accounting Information in the Market for Debt’, Accounting and Finance, November 1986.

1

Sharpe and Walker (1975), Standish and Ung (1982) and Emanuel (1986) studied revaluation announcements by firms in Australia, New Zealand and the United Kingdom respectively. PHILIP BROWN is KPMG Peat Marwick Professor of Accounting and H. Y. IZAN is Professor of Accounting and Finance, Department of Accounting and Finance, University of Western Australia; AFRED L. LOH is a Lecturer in Accounting, Nanyang Technological University, Singapore. This paper has benefited from the valuable comments of David Emanuel, Graham Peirson and Ian Zimmer and workshop participants at the University of Western Australia, Australian Graduate School of Management and the University of Queensland. Financial support has

388

been provided by Assistance Bureau.

the

Australian

Development

2

ASRB 1010, Accounting for the Revaluation of Non-Current Assets, which supersedes the Statement of Accounting Standards, AAS 10, of the same title, applies to the financial years of a company that ends on or after 30 September 1987. Before this date but after 1 July 1981, AAS 10 was operative. Prior to AAS 10, the relevant standard was AAS 1, Profit and Loss Statements. The basic reporting requirements for upward asset revaluations, the focus of this paper, remained the same over the period we studied (1974–86). 3

The economic incentives for revaluation of other assets are less clear. For example, debt covenants and industry norms exclude intangible assets in determining the borrowing power of a firm; and investments are said to be less capable of supporting debt as they are easily converted from ‘in-place’ to ‘out-of-place’ assets. 4

The fee recommended by the Australian Institute of Valuers for a property with an appraised value of $30,000 was about 0.5 per cent of the appraised value, and for a property with an appraised value of $2 million, about 0.2 per cent. 5

Whittred and Chan (1987) found that potential violation is more likely to be associated with the total liabilities constraint than with secured liabilities. The potential violation of the latter constraint is not examined in this study. 6

It has been argued that the higher renegotiation costs of public debt might provide an incentive for the firm to 389

revalue rather than renegotiate. However, the question of public versus private debt is not examined further as there are major difficulties in delineating public and private debt (see Kelly, 1983). 7

See Table 1. We assume the constraint to be whichever is the most binding: 60, 75 or 80 per cent of the total tangible assets for debentures, convertible notes and unsecured notes respectively (refer Whittred and Zimmer, 1986). 8

Since revaluation of a depreciable asset lowers reported profits, it might be argued that management could be concerned with how profit reductions would affect debt contracts. However, few Australian debt contracts contain an interest coverage (see Whittred and Zimmer, 1986) and when they do, continuing compliance is not usually required. Further, some trust deeds permit the writing back to profit of depreciation charged in excess of the amount allowed for tax purposes, as long as the write-back represents an increase in the net book value of tangible assets (see Whittred and Zimmer, 1986). This adjustment effectively negates the profit effect of increased depreciation, for debt contracting purposes. 9

It is plausible that management might be discouraged from undertaking a revaluation because it would result in a lower accounting rate of return and thereby adversely affect management compensation. Because it is difficult to observe the nature of management compensation plans in Australia and, in particular, how an adjustment for a revaluation is incorporated into the plans, we are unable to analyse their effects.

390

10

An interesting question that arises is why many firms have not provided for depreciation on buildings as required by accounting standards (see Craswell, 1986). Some large firms possibly choose not to depreciate buildings (although depreciation lowers accounting rate of return and might affect political costs) for contracting reasons, especially if the firms are highly geared. That is, unless their trust deeds permit the write-back of the ‘excess’ depreciation, firms might be unwilling to provide for depreciation on buildings, and prefer to settle for a qualified audit report. By extending Watts’ argument one could predict that larger firms are more likely to revalue depreciable assets (in particular, plant and equipment) than smaller firms, because of the more pronounced (negative) effect on return. As only a small proportion of fixed asset revaluations were of plant and equipment, the prediction was not tested. 11

Ford and Hearn (1980) alleged that the Australia-wide strike against General-Motors Hoiden in 1964 resulted from the inadequate understanding by the union of the firm’s financial position, in particular of its profit levels. 12

Based on 132 responses to a mail questionnaire conducted towards the end of 1989, Allen (1990) found that the average short-term debt capacity, in the form of lines of credit and bill facilities, is 24 per cent of total debt levels. Anecdotal evidence that reserve borrowing capacity is useful is also found in Sabco’s 1988 Annual Report. The management (having carried out a major revaluation) reported that ‘gearing has been reduced from 47% to 39%. This has enhanced the ability of the

391

Company to take advantage of future expansion opportunities as they arise.’ 13

In the absence of appropriate modification to the estimation procedure, Palepu (1986) showed that choice-based sampling yields biased coefficient estimates, overstates the model’s predictive ability and impairs the validity of hypothesis tests. 14

There are gains from pooling the observations into one sample for estimation only if the relationship among the variables is stationary within that period. As observations are pooled across a relatively short period, the non-stationarity problem (other than perhaps for the CMS/TTA variable in the 1974–7 period) should not be serious. 15

The second variable to denote a bonus issue, BONUS2, differs from BONUS 1 in that a second fiscal year is scanned to determine if, ex post, a bonus issue was made. The results for BONUS2 are likewise significant at < .01. 16

The estimated coefficients from the Probit model are not estimates of the marginal probabilities that a firm will revalue. To obtain the probability of a revaluation conditional on certain characteristics of the firm, as described by the explanatory variables, we need to compute Pj = Prob. (I* ≤ ), where I* is a Normal variable, with a mean of 0 and a variance of 1; Xj is the vector of characteristics for firm j; and β is the vector of estimated coefficients. P„ the probability of a revaluation for firm j, has a monotonic relationship with , such that we can describe the direction of the change in Pj, as

392

each independent variable is varied, by the sign of the variable’s coefficient. In the multivariate model, we have chosen to omit the COVENANT variable; the debt proximity ratio (DEBT/ COV) is used instead as it provides us with more information than the dichotomous COVENANT variable. 17

The expected accuracy rate from the proportional chance model is 61.9 per cent for 1974–7 and 57 per cent in 1984–6. This is calculated as (n1/n)2 + (n2/n)2, where n1 and n2 are the numbers of revaluers and non-revaluers respectively and n = n1 + n2. 18

Tests on each year of the sub-period samples provided some additional confirmation of these associations. Results are available from the authors. © 1992 by Accounting Foundation, The University of Sydney

393

Part III Standard-Setting and Regulation

394

Philip Brown, Stephen L. Taylor and Terry S. Walter The Impact of Statutory Sanctions on the Level and Information Content of Voluntary Corporate Disclosure This article examines the effect of statutory civil and criminal sanctions on voluntary corporate disclosures by firms listed on the Australian Stock Exchange (ASX). Apart from direct investigation of the quantity of voluntary disclosure, we also investigate several possible consequences of altered corporate disclosure policies, namely properties of analysts’ forecasts, the degree to which share prices anticipate the information content of periodic earnings reports, and the relationship between volatility and corporate disclosures. Results suggest that, post-sanctions, any increase in voluntary disclosure is confined to smaller firms and those which performed relatively poorly. Moreover, analysts’ earnings forecasts did not become more accurate or less diverse following the introduction of statutory sanctions, and there was no statistically significant increase in the weight placed on each disclosure’s ability to explain return volatility. There is some evidence that share prices have anticipated earlier the value relevant components of annual periodic accounting data, although this result is again confined to smaller firms, Although the tests used are not independent and have a limited time period post-sanctions, the results cast doubt on the extent to which the imposition of substantive civil or criminal sanctions affects corporate disclosure policy. Key words: Disclosure; Information; Sanctions.

395

This paper investigates the impact of significant statutory civil and criminal sanctions on both the quantity and timeliness of voluntary disclosures made by firms listed on the Australian Stock Exchange (ASX).1 Following a number of high profile corporate failures during the latter half of the 1980s, and prosecutions which followed, considerable criticism was levelled at the quality and frequency of Australian corporate reporting,2 especially the way in which the ASX enforced listing rules relating to the maintenance of an informed market.3 Reviewing events leading up to the legislative action, Brown et al. (1998) argue that the process which followed these criticisms was, in many respects, consistent with Peltzman’s (1976) description of the political process, whereby legislative action is provoked by an alleged ‘crisis’. In this case, the alleged crisis was a lack of confidence in the timeliness of corporate reporting, leading to demands for legislative intervention. In its final form, the legislation focused on the extent to which statutory sanctions could provide support for existing disclosure requirements (i.e., the introduction of substantial civil and criminal penalties). This received bipartisan political support, as well as the endorsement of the Australian Securities Commission (ASC, now Australian Securities and Investment Commission) and the ASX.4 However, given relatively constant requirements for timely disclosure by ASX-listed firms, the assumption accepted by all these groups was that the statutory imposition of substantial civil and criminal penalties would result in significant changes in corporate disclosure policies, especially in respect of voluntary disclosures.5 Put simply, the 396

legislation which took effect from 5 September 1994 was intended to enhance the ‘enforceability’ of existing ASX disclosure rules.6 The extent to which corporate disclosures can be influenced by the threat of civil and criminal penalties has been debated extensively (Friend, 1976). For example, Stigler (1964) compares the post-listing perfonnance of United States equity issues before and after the Securities Exchange Commission (SEC) was given control over the registration of new issues (via the Securities and Exchange Act, 1934). Although Stigler concludes the legislation was ineffective, his empirical method is relatively naive, requiring an assumption that market-adjusted post-listing stock returns have no ‘time-specific’ elements, and can be compared purely on the basis of a legislative intervention date (Friend, 1976). Benston’s (1973) investigation of ‘disclosure improvements’ following the 1934 Securities and Exchange Act also relies on the critical assumption that a proxy for the extent of corporate disclosure (i.e., price volatility) is otherwise temporally constant. Investigation of the effect of statutory civil and criminal sanctions on ASX-listed firms’ disclosures also needs to make temporal comparisons across a common intervention date. Methods employed in this research typically involve the use of several ‘controls’ absent in early United States-based studies such as Stigler (1964) and Benston (1973), as well as utilizing a variety of (non-independent) proxies for the effect of any change in disclosure timeliness. Apart from measures of the quantity of voluntary disclosure made by ASX-listed

397

firms, we also investigate post-sanctions changes in each of the following indicators; attributes of analysts’ earnings forecasts, share price anticipation of periodic earnings reports, and the relationship between corporate disclosures and price volatility. The initial focus on voluntary disclosures differs from much of the extant literature. Rather than modelling the cross-sectional determinants of a single disclosure type (e.g., earnings forecasts), we try to capture the total extent of voluntary disclosures, using one of two proxy measures. These are the number of disclosures flagged as price sensitive by the ASX, and the number of disclosures assigned an ASX code consistent with them being ‘non-routine’. Irrespective of the proxy used, the results suggest that any increase in voluntary disclosure is confined to relatively small firms, and those that performed relatively poorly. Likewise, tests which focus on the consequences of an increase in timely voluntary disclosure also fail to yield results consistently supportive of statutory sanctions having significantly affected corporate disclosure policies. After controlling for firm and time specific influences, analysts’ forecasts of net profit after tax are more (rather than less) biased, less (rather than more) accurate, and no less dispersed following the introduction of statutory sanctions. These results do no support the claim that the imposition of substantial statutory civil and criminal penalties affected the timeliness of voluntary disclosures by ASX-listed firms. In contrast, there is some evidence that post-sanctions

398

stock prices show relatively earlier anticipation of earnings results. Using an approach similar to Ball and Brown (1968), it appears that the value-relevant component of the preliminary final statement is anticipated earlier, although this effect is confined to smaller firms, which are less likely to have significant institutional shareholders or analyst following. However, when the relation between monthly stock price volatility and disclosures designated by the ASX as ‘price sensitive’ is examined, there is no evidence of a statistically significant change in the strength of this relationship. Overall, our results are best described as mixed, although the short time period since the introduction of statutory sanctions may be insufficient to observe any significant impact. It is also important to recognize that our four basic types of test involve overlapping time periods and sample fìrms, and therefore cannot be viewed as independent tests of any relationship between the imposition of substantial statutory civil and criminal sanctions and the extent of timely, voluntary disclosure. Subject to these limitations, the results at least cast doubt on the extent to which variation in the quantity and timeliness of corporate disclosure is significantly related to enforcement mechanisms directed at the flow of irregular information. Variations in economic incentives (including ‘cultural’ factors) may be substantially more important determinants of voluntary corporate disclosure than variations in regulatory or enforcement procedures and penalties. Hypotheses

399

Voluntary Disclosures The primary hypothesis addresses whether substantive statutory civil and criminal sanctions have a significant impact on the frequency of voluntary disclosure. However, tests of this hypothesis require adequate controls for ‘other’ influences on disclosure policy, which may not be constant across the period we examine. Models of voluntary disclosure policy typically assume that disclosure of information is costly, recognizing that possible ‘frictions’ result in conflicting incentives to disclose. For example, Verrecchia (1983) shows that managers will exercise discretion in disclosing or withholding information about firm value, reflecting ‘proprietary’ disclosure costs.7 Empirical evidence also supports the proposition that firms face potentially conflicting incentives in deciding whether voluntarily to disclose some information. On the one hand, firms can reduce the cost of capital by increasing their level of disclosure, especially forward-looking information (Healy et al, 1995; Botosan, 1997). However, product market considerations may discourage disclosure, and support for this view is provided by Lee et al. (1995) and Clarkson et al. (1994). Recognition of these potentially conflicting incentives yields two important lessons. First, one would not expect to observe all firms following the same disclosure policies. Second, tests of temporal differences in disclosure need to control for both firm specific and broader economic influences such as variations in the level of capital market activity.

400

Our interest in the distinguishing characteristics of firms which voluntarily disclose information is analogous to Ruland et al. (1990), who examine characteristics of listed U.S. firms making earnings forecasts. They argue that management possession of good news, variations between analyst and management expectations (i.e., correction or confirmation), new capital offerings and the level of ‘inside’ ownership will differ systematically between forecasters and other firms. With the exception of the good news hypothesis, Ruland et al. find results consistent with their hypotheses. However, unlike Ruland et al., the focus here is on a measure of total voluntary disclosure, rather than a single type of voluntary disclosure such as an earnings forecast. While Lang and Lundholm (1993) examine a number of characteristics associated with differential disclosure, their measure of ‘comprehensive disclosure’ is as assessed by financial analysts. Hence they do not model disclosure per se, but rather analysts’ perceptions of disclosure. They find that disclosure is an increasing function of past, current and future performance, consistent with the good news hypothesis. They also find that disclosure ratings increase with firm size and the issuance (or pending issuance) of securities, and decrease with the volatility of prior periods’ performance. Consistent with Ruland et al. (1990) and Lang and Lundholm (1993), other studies also find evidence of planned capital raisings influencing disclosure policy (Clarkson et al., 1994; Frankel et al., 1995; Healy et al., 1995) In summary, the theory and evidence on voluntary disclosure suggest that disclosure is associated with firm

401

size (large firms disclose more information), whether the firm has good news or bad news to report (disclosure is positively related to the presence of good news), whether the firm is engaged in ai equity issue (more information is disclosed around the time of an equity issue) and whether the firm has analysts’ earnings forecasts (firms with an established analyst following are expected to disclose more information). Also, controls are used for the firm’s industry because ASX listing requirements result in more frequent reports for some industries. For example, mining and exploration companies must file a quarterly activity report. Banking and finance companies have different regulatory regimes and very different capital structures, and investment companies are presumably driven more by market-wide events than are other companies. Another reason for controlling for industry is that many ‘other factors’ cluster by industry, and a simple industry dummy variable can serve as a ‘fixed effects’ control for these factors. In seeking to control for the known determinants of voluntary disclosure, the following multiple regression model is employed:

where:NUMDOCS

= a measure of the extent of voluntary disclosure by firm i in period k, where к is either the pre-sanctions period or the post-sanctions period;

402

SANDUM

= a dummy variable that takes the value 0 if the observation is drawn from the pre-sanctions period, and 1 for the post-sanctions period;

SIZE

= the natural logarithm of the firm’s total market capitalization, at August 1992 if the observation is drawn from the pre-sanctions period, and at August 1994 for the post-sanctions period;

LNDDUM1–5 = an industry dummy variable that takes the value 1 if the firm is in industry group 1 to 5, and 0 otherwise; INDDUM16–19= an industry dummy variable that takes the value 1 if the firm is in industry group 16 to 19, and 0 otherwise; LNDDUM20

= an industry dummy variable that takes the value 1 if the firm is in industry 20, and 0 otherwise;

ISSUEDUM

= a new issue dummy variable which takes the value 1 if the firm made a rights issue in the period, and 0 otherwise;

NEWSDUM

= a dummy variable which takes the value 0 if the firm had ‘bad’ news in the period, and 1 403

otherwise. Good news is measured as the market-adjusted cumulative return for the firm pre- or post-statutory sanctions. If this market adjusted return (CAR) is positive, the firm is classified as having generally revealed good news; and BARDUM

= a dummy variable which takes the value 1 if the firm is included in the BAR CEP survey of analysts’ earnings forecasts, and 0 otherwise

Properties of Analysts’ Forecasts Although tests of the primary hypothesis control for several documented determinants of corporate disclosure policy, such models remain relatively ad hoc. Moreover, ‘direct’ tests of voluntary disclosure decisions do not capture the effect of such decisions. Hence, consideration is given to indicators of the extent to which statutory sanctions may have resulted in more timely voluntary disclosures. Analysts’ forecasts are one such indicator, and have been widely used as a proxy for investor beliefs (Abarbanell et al, 1995). Our tests are based on the idea that more timely disclosures will improve forecast accuracy and reduce analyst disagreement. However, while there is a very large literature on the accuracy of analysts’ forecasts and the determinants of analysts’ disagreement, there is less evidence on the relationship between

404

corporate disclosure policy and the properties of analysts’ forecasts.8 Detailed evidence on the relationship between corporate disclosure policy and analysts’ forecast properties is provided by Lang and Lundholm (1996).9 They argue that the effect of additional or more timely disclosure on the dispersion of analysts’ forecasts depends on whether they have a common forecasting model and the extent to which their forecast information differs. If analysts have common firm-provided and private information, but place differing weights on information from each source, then additional disclosure will increase the dispersion in analysts’ forecasts. However, Lang and Lundholm show that analysts’ forecasts become less diverse as firms increase their disclosure (as measured by analyst rankings), consistent with the hypothesis that increased disclosure should result in less (rather than more) dispersion in analysts’ forecasts. Provided also that information distributed by sell-side security analysts is not a substitute for enhanced corporate disclosure, it is expected that, post-sanctions, analysts’ forecasts should become less biased, more accurate and less dispersed. Once again, hypothesis testing based on temporal intervention requires the use of sufficient control variables. Consistent with the existing literature (Brown, 1994), the accuracy and consensus of forecasts are expected to increase with the extent of analyst following. Accordingly a control for analyst following is included. There is also a control for firm size, because larger firms should have forecasts that are more accurate and have greater consensus. Shares with higher volatility indicate

405

greater uncertainty in the market about that stock, which implies less accurate forecasts and greater analyst disagreement. This suggests a need to control for volatility. The accuracy of last year’s forecast is also included, because we expect this year’s forecast to be more accurate and have greater consensus among analysts if last year’s was more accurate. Also, the opportunity to short a stock when adverse information is uncovered increases analysts’ incentives to collect information and can result in more accurate forecasts and greater agreement among them. Controls for whether the stock has options traded and whether it can be short sold are also included. Industry effects can play a part because the earnings of companies in some industries are easier to forecast than others, for reasons including differences in product maturity, and operating and financial leverage. Finally, in tests of forecast accuracy or bias, a control for analyst disagreement is incorporated, because analysts disagree more when there is greater uncertainty, which implies that their consensus forecasts are then likely to be less accurate.10 Hence, the following models are tested:

406

where:ERROR

= a measure of the error in the consensus analysts’ forecast for firm i in period k, as recorded by BAR CEP;

PERIOD

= the number of months from the forecast date until the month in which the company files its preliminary final statement with the ASX;

NUMAN

= the number of separate analyst forecasts contributed to BAR CEP;

DISAN

= the standard deviation across analysts’ forecasts of net profit after tax for firm i in a given month;

SIZE

= the natural logarithm of the firm’s market capitalization (in $ millions);

PREVERROR = last year’s forecast error, measured at the corresponding month in the previous year; PREVABSERR = the absolute value of last year’s forecast error, measured at the corresponding month in the previous year; VOL

= the variance of the monthly rate of return on the firm’s shares.

407

INDDUM1–5

= an industry dummy variable that takes the value 1 if the fimi is in ASX industry group 1–5, and 0 otherwise;

INDDUM16–19= an industry dummy variable that takes the value 1 if the film is in ASX industry group 16–19, and 0 otherwise; INDDUM20

= an industry dummy variable that takes the value 1 if the firm is in ASX industry group 20, and 0 otherwise,

OPTIONS

= a dummy variable that takes the value 1 if the firm has ASX traded options in the same period;

SHORT

= a dummy variable that takes the value 1 if the firm is approved by the ASX for short-selling;

SANDUM1

= a dummy variable that takes the value of 1 if the fiscal year straddles the introduction of statutory sanctions (i.e., years which end between July 1994 and May 1995), and 0 otherwise; and

SANDUM2

= a dummy variable that takes the value of 1 if the fiscal year ends June 1995 or later

Stock Prices and Disclosure

408

The relationship between stock price movements and periodic earnings releases has been extensively documented (Brown, 1994) and it is well accepted that these statutory reports are anticipated (i.e., much of the related stock price movement occurs prior to the report’s release). Because variations in corporate disclosure may be evident in the behaviour of stock prices, attention is directed to the relationship between share prices and periodic accounting reports as well as the relationship between volatility and information releases. If firms voluntarily disclose information in a more timely fashion, then stock prices will show relatively earlier anticipation of the information contained in periodic accounting reports. In addition, postsanctions, corporate disclosures designated as ‘price-sensitive’ by the ASX should be more closely associated with variations in stock price volatility. It is difficult to measure the nature of the news in a financial statement as complex as a half yearly report (HYR) or a preliminary final statement (PFS). The view taken here is that, embedded somewhere in the financial statements, is all the information released to the market since the date of the previous report.11 That is, we take the value-relevance of the firm-specific information contained in a financial statement to be accurately measured by the market-adjusted price movement in the period leading up to the report’s announcement.12 For HYRs, price movements are analysed over the six months to the report’s announcement, and for PFSs, price movements over twelve months. The conjecture is that if statutory civil and legal sanctions lead to earlier disclosures of price-sensitive information, then that

409

information would have been reflected in share prices earlier. For tests of the relationship between stock price volatility and corporate disclosure, there is once again difficulty in controlling for ‘other effects’. The approach taken is to control for variations in market volatility, as well as the number of price-sensitive disclosures, with the expectation that an interaction effect (i.e., the product of a post-sanctions dummy and the quantity of disclosure) should be significantly positive. Hence, the following model is tested:

where:VOL

= a measure of monthly price volatility,

NUMDOCS

= the number of disclosures in a given month flagged as price sensitive by the ASX,

ASX

= the absolute value of the monthly rate of return on the ASX All Ordinaries Accumulation Index,

MSCI

= the absolute value of the monthly rate of return on the Morgan Stanley Capital International World Index,

410

SP

= the absolute value of the monthly rate of return on the S and Ρ 500 Index, and

SANDUM

= a fixed-effects dummy variable equal to 1 if the measure of volatility (VOL) is for a post-sanctions month, otherwise zero.

Data Our initial data set comprises the complete set of announcements made by ASX-listed films through Signal G electronic records, provided by Securities Industry Research Centre of Asia-Pacific (SIRCA). SIRCA has been receiving Signal G data, which contain the full electronic text of announcements made by Stock Exchange Automated Trading System (SEATS) listed firms, since 29 August 1992 through to our cut-off date of 31 March 1996.13 While there were 1,474 firms listed on the ASX at some time during this period, our tests require the availability of monthly trading data. This requirement reduces our sample size for disclosure tests to 463 firms which survived for the entire period and which traded at least once in each month. The extent of voluntary disclosure is measured in two ways, both of which are simple document counts. First, we count the number of disclosures labelled by ASX staff as price sensitive.14 While this includes regular reports, based on discussions with ASX staff, this is presumed to include any material, voluntary disclosures which would normally be flagged as price sensitive. 411

Second, we count disclosures within the two-digit category most likely to capture irregular disclosure. While the ASX attaches two-digit event codes to all Signal G transmissions, our focus is on item 14 (‘Other’) announcements.15 Analysts’ consensus forecasts, the standard deviation across analysts’ forecasts, the number of analysts who contributed net profit after tax (NPAT) forecasts, actual NPAT and the announcement month were all extracted from the monthly BARCEP publication. NPAT, as defined by BARCEP, is before extraordinary and abnormal items and after tax and preference dividends. The pre-sanctions sample is drawn from financial years ending between August 1992 and June 1994 and the post-sanctions sample from financial years ending between June 1995 and December 1995. For consistency reasons BARCEP-reported NPAT actuals are also used (Philbrick and Ricks, 1991). We are able to identify ninety-two firms which are within our initial sample described above and which also are covered by BARCEP. Market capitalization data were supplied by the ASX for all index stocks beginning December 1989 and for all listed stocks for the period July 1995 to March 1996. Gaps in the market capitalization data sourced from the ASX were filled using data extracted from the ASX’s Statex database on-line at the University of Western Australia. Return volatility was measured by the monthly discrete return variance over the period January 1992 to March 1996, estimated using last-trade-for-the-month prices and dilution factors supplied by SIRCA. Industry numbers were supplied by the ASX. Stocks with put or call options traded on the

412

ASX were identified from newspapers, and stocks eligible for short selling were identified from the daily share quotation sheets published by the ASX. The release date for the HYR and PFS are identified from Signal G. Because the minimum monthly price and trading period requirement is shorter than for our tests of disclosure quantity, our sample size is accordingly larger, with a maximum sample size (postsanctions) of 737 firms. For tests which use price volatility as the dependent variable (i.e., tests of equation (3)), firms were required to have at least twelve monthly returns between January 1992 and August 1994 (pre-sanctions) and another twelve between September 1994 and March 1996 (post-sanctions). There are 727 ASX-listed firms which satisfied this requirement. Evidence of Corporate Disclosures Total Disclosures Table 1 contains preliminary descriptive results for the frequency counts of documents disclosed to the ASX. It contains two sets of results. Panel A summarizes disclosures made by the 1,474 firms, split by ASX two-digit reporting code, for the pre-sanctions and post-sanctions periods. It is a simple count of all documents released by all firms for which we have electronic records.16 The pre-sanctions period in Panel A is defined as 29 August 1992 to 4 September 1994 (approximately twenty-four months), while the post-sanctions period is from 5 September 1994 to 31 March 1996 (approximately nineteen months). Despite

413

having a shorter post-sanctions observation period, there is a substantial increase in total disclosures. The pre-sanctions total document count is 52,741, while post-sanctions it is 77,723. However, there is a substantial decrease in the number and frequency of code 14 Other’ reports (i.e., one of our proxies for the quantity of voluntary disclosure). Table 1 Pre-sanctions period and post-sanctions period document count split by ASX report code for all 1,474 ASX-listed-firms (Panel A), and for the standardized pre-sanctions period (5 September 1992 to 31 March 1993 and 5 September 1993 to 4 September 1994) and post-sanctions period (5 September 1994 to 31 March 1996) split by price-sensitive and non price-sensitive documents (Panel B)

414

Panel B of Table 1 controls for seasonal patterns in reporting by ASX firms by standardizing the period for our document count. The pre-sanctions period is defined as 5 September 1992 to 31 March 1993 and 5 September 1993 to 4 September 1994 (a total of approximately nineteen months) and the post-sanctions period as 5 September 1994 to 31 March 1996 (again, approximately nineteen months). In defining periods this way maximum use is made of available post-sanctions observations, subject to the seasonality constraint. Panel B results

415

show that the introduction of statutory sanctions was associated with a substantial increase in total documents, rising from 48,091 to 63,235. However, the post-sanctions increase in price-sensitive documents (our second proxy for total voluntary disclosures) is relatively low compared to non price-sensitive documents. Determinants of Disclosure Table 2 contains the results of our tests of equation (1). Results are reported using two definitions of NUMDOCS. First, NUMDOCS is defined as all disclosures flagged as potentially ‘price sensitive’ by the ASX. Alternatively, exclusive focus is on item 14 (‘Other’) disclosures, which are most likely to be disclosures made voluntarily. We also report two modifications of equation (1), first by excluding firms with an analyst following (as indicated by their inclusion in BAR CEP reports), and second, by excluding firms classified as having released predominantly good news. Because firms included in BARCEP must have an analyst following, these firms are likely to voluntarily provide more financial as well as non-financial information. Accordingly, the introduction of statutory sanctions may have less impact on these firms’ voluntary disclosures. For United States firms, Lang and Lundholm (1996) show that ‘more informative’ disclosure policies result in a larger analyst following, consistent with the view that firms use voluntary disclosure as a means of generating analyst following, and ultimately reducing their cost of capital. Table 2 416

Multiple regression disclosure model where the dependent variable is a measure of voluntary disclosure by the 463 ASX listed firms which survived and traded in each month in both the pre-sanctions period (5 September 1992 to 31 March 1993 and 5 September 1993 to 4 September 1994) and the post-sanctions period (5 September 1994 to 31 March 1996). Each firm has two observations in the regression, one being the pre-sanctions document count and one being the post-sanctions document count. Two measures of voluntary disclosure are used, namely the number of price-sensitive documents, and the number of ASX category 14 documents

Sanctions dummy (SANDUM) takes the value 0 for the period prior to 4 September 1994, and 1 thereafter. Firm size (SIZE) is the natural logarithm of the firm’s market 417

capitalization measured at 31 August 1992 for the pre-sanctions period and 31 August 1994 for the post-sanctions period. Industry dummy variables take the value 1 if the firm is a member of the ASX two-digit industries in the above table (i.e., INDDUM 1–5, INDDUM 16–19, and INDDUM 20), and 0 otherwise. The equity issue dummy (ISSUEDUM) takes the value 1 if the firm made a rights issue in the pre-sanctions or postsanctions period respectively, and 0 otherwise, The good news dummy (NEWSDUM) is 1 if the abnormal return (from a ‘zero-one’ market model) in the pre-sanctions period and post-sanctions period is positive, otherwise it is 0. The BARCEP dummy (BARDUM) is assigned 1 if the firm is included in the BARCEP consensus analyst forecast database, and 0 otherwise. Statutory sanctions may also differentially impact on firms depending on whether they possess good or bad news. Although there is some theoretical and empirical evidence suggesting that good news is more likely to be voluntarily disclosed, Skinner (1994, 1997) presents evidence that, among United States firms, bad earnings news is more likely to be pre-empted than good earnings news, and that this pre-emption reduces the expected costs of shareholder litigation. Although shareholder litigation costs are probably lower in Australia than in the United States, the introduction of statutory civil and criminal sanctions may have a greater impact on irregular disclosure of bad news vís-à-vis good news. Brown et al. (1998) argue that the introduction of statutory sanctions was largely motivated by political

418

concern that listed firms forthcoming with bad news.

were

not

sufficiently

Table 2 shows that, as expected, the number of documents disclosed is positively and significantly related to the size of the firm, the mining industry dummy (Industry 1 to 5) and whether the firm is included in the BARCEP survey.17 When NUMDOCS is defined as the number of ASX-designated price-sensitive documents, there is a significant, positive equity issue coefficient. Of central interest to our test is the statutory sanctions dummy variable. The estimated coefficient in the regression using the price-sensitive disclosures is insignificant. It is, however, positive, suggesting an insignificant increase of 0.78 documents per firm in the period following the introduction of statutory sanctions relative to the earlier period. It is significantly negative for the ‘Other’ disclosures, with an estimated 7.6 fewer disclosures per company. However, when estimates of equation (1) are confined to firms without a BARCEP-repoited analyst following, there is some evidence of a post-sanction increase in disclosures flagged as price sensitive, although not for the regression using item 14 disclosures as the dependent variable. Likewise, when equation (1) is estimated solely for finns classified as ‘bad’ news firms, the results provide some support for the view that the statutory sanctions had some effect on the disclosure of bad, vis-à-vis good news.18 There is a statistically significant positive coefficient for the sanctions dummy when NUMDOCS is measured as price sensitive disclosures, but this reverses when item 14 (‘Other’)

419

documents are used as the measure of voluntary disclosure. Discussion Several aspects of the results reported in Table 2 are worthy of further comment. First, it is clear that results reported in earlier studies of voluntary disclosure (e.g., size, equity issuance and industry effects) are also evident in our results. The relatively consistent findings of the control variables having the predicted sign adds to our confidence that equation (1) is well specified, relative to its limited theoretical underpinnings. Overall, Table 2 provides only limited evidence at best of any impact from civil and criminal sanctions. There is some evidence of increased voluntary disclosure among smaller (i.e., non-BARCEP) and for ‘bad news’ firms, which is consistent with the argument, favoured by supporters of statutory sanctions, that these firms are typically less forthcoming with timely voluntary disclosures. As noted in our earlier discussion, although much of the extant analytical and empirical disclosure literature predicts the revelation of good news, other research (e.g., Skinner 1994, 1997) finds some support for the early release of bad news in an attempt to minimize expected litigation costs. The introduction of statutory civil and criminal sanctions may therefore affect the timely revelation of bad news more significantly than good news. However, we caution against concluding that sanctions have had the ‘desired’ effect. First, it is noted that the results are not consistent for our two measures of

420

voluntary disclosure (i.e., item 14 and price sensitive). This may reflect an increasing degree of vigilance by the ASX in identifying price sensitive documents and/or a greater effort at classifying releases differently from ‘Other’. Both possibilities are consistent with the decline in item 14 (‘Other’) disclosures reported in Table 1. Second, it must be remembered that, particularly by international standards, the market capitalization of non-BARCEP firms is very small. Hence, the economic significance of any modification to voluntary disclosure policy among smaller ASX listed firms is doubtful. Analysts’ Forecasts Descriptive Statistics Using data for ninety-two BARCEP firms, Table 3 describes several properties of analysts’ forecasts, for both the pre- and post-sanctions periods. Three measures of forecast error (mean error, mean absolute error, and mean square error) are reported, as well as a measures of analyst disagreement, which is the standard deviation of analysts’ forecasts. All measures use the BARCEP forecast of net profit after tax (NPAT), as well as actuals reported by BARCEP (for consistency purposes). All forecasts and actuals are deflated by market capitalization.19 Table 3 Mean forecast error, mean absolute forecast error, mean square forecast error and analyst disagreement for Australia-domiciled BARCEP companies, pre-sanctions and post-sanctions

421

* The difference between the metric pre-sanctions and post-sanctions is statistically significant at 0.1, based on a Mann-Whitney two-tailed U-test. The numbers in the column headed ‘Month’ indicate forecasts made eleven, ten, etc. months before the company announced, via its preliminary final statement filed with the ASX, its actual net profit after tax (NPAT) for the financial year. A consensus forecast error for a given company in a given month is defined as the company’s actual NPAT less the mean of the analysts’ NPAT forecasts for that company that month, divided by the company’s equity capitalization thirteen months

422

before the report was announced, The disagreement measure is the standard deviation across analysts’ forecasts of NPAT for that company that month, again divided by the company’s equity capitalization. The pre-sanctions period includes reports for fiscal years ending between August 1992 and June 1994; the post-sanctions period includes fiscal years ending between June 1995 and December 1995. Table 3 supports several conclusions. First, the classical, optimistic bias of sell-side analysts is manifest in the first seven months of the pre-sanctions period and throughout the post-sanctions period. Second, as the announcement date approaches, the bias towards optimism declines, forecasts become more accurate and analysts tend to agree more on their forecasts. Third, compared to the pre-sanctions period, post-sanctions analysts’ forecasts are more optimistic, more accurate and more in agreement.20 Increased accuracy and greater agreement are predicted outcomes if the statutory sanctions significantly affected corporate disclosure policies in the desired manner. It must be noted, however, that these results do not control for known determinants of analyst forecast accuracy. Regression Tests Results from fitting OLS regressions for equations (2a) and (2b) are provided in Table 4. The first column reports evidence on the determinants of bias in analysts’ forecasts (i.e., the mean error). Most importantly, it is evident that, post-sanctions, analysts’ forecasts are significantly more optimistic (the coefficient for the postsanctions dummy has a t-statistic of –5.66). 423

Table 4 Regression estimates of the determinants of the forecast error, absolute forecast error, square forecast error and analyst disagreement for Australia-domiciled BARCEP companies, pre-sanctions and post-sanctions

See Table 3 for definitions of forecast error and analyst disagreement. The regression is fitted by ordinary least squares (OLS) to pooled cross-section and time-series data for each company, The forecast period (PERIOD) is the number of months from the forecast date until the month in which the company files its PFS with the ASX; the number of analysts (NUMAN) is the number of separate forecasts contributed to BARCEP; the disagreement 424

among analysts (DISAN) is the standard deviation across analysts’ forecasts; a company’s size (SIZE) is measured as the material logarithm of its equity market capitalization in A$ million; the error (PREVERR) and absolute error (PREVABSERR) of last year’s forecast are measured at the corresponding month in the year· prior to the measurement of the dependent variable; industry dummy variables (INDDUM 1–5, INDDUM 16–19, INDDUM 20) take the value 1 if the firm is a member of the relevant ASX two-digit industry group and 0 otherwise; the existence of ASX options trading (OPTIONS) and approved ÀSX short-selling (SHORT) are dummies which take the value 1 when these are available, and 0 otherwise; return volatility (VOL) is the sample variance of the share market monthly rate of return on the company’s ordinary shares; the existence of statutory sanctions is denoted by the dummy variables SANDDUM 1 and SANDDUM 2, which take the value 1 if the financial year straddles the introduction of statutory sanctions (SANDUM1) or if the financial year ends June 1995 or later (SANDUM2). Confidence in the result is enhanced by the intuitively sensible coefficients on many of the control variables. Overall, analysts were not overly optimistic, as the constant term is not significantly different from zero. There is only weak evidence that analyst optimism increased with the length of the forecast period, while any bias towards optimism appears to decline with the extent of analyst following. The results also suggest that optimism is positively related to analyst disagreement and the size of last year’s forecast error, and negatively related to firm size. Analysts’ forecasts are more 425

optimistic for stocks that were eligible for short selling or which had options traded on them, and less optimistic for resource stocks and industries 16–20 compared with other stocks. Columns 2 and 3 of Table 4 report tests examining the size of the forecast error (absolute error and squared error respectively). Both measures provide results consistent with the size of analysts’ forecast errors having increased significantly post-sanctions. The results also suggest that, as expected, forecast accuracy is inversely related to the length of the forecast period, and positively related to the extent of analyst disagreement. Also, as expected, accuracy improves with firm size, but declines with the size of the previous year’s forecast error. Finally (and somewhat surprisingly) we find that the accuracy of analysts’ forecasts increased with return volatility but was lower for stocks with options traded on them. The fourth column of Table 4 provides results relating to the extent of analyst disagreement. There is no evidence of any significant change in analyst disagreement post-sanctions (the coefficient is negative but not significantly different from zero). As expected, analyst disagreement increases with both the forecast period length and the prior year forecast error, and declines with increases in analyst following and decreased optimism in prior year forecasts. There is also evidence that analyst disagreement was higher for the resources and financial sectors (industries 1–5 and 16–19) than for other industries. Discussion

426

In summary, despite the initial favourable evidence in Table 3, when other factors known to influence analysts’ forecasts are controlled for it seems that statutory sanctions did not have the expected effect if they had significantly improved the quality of the information available to security analysts.21 From Table 4, it is apparent that post-sanctions, analysts’ profit forecasts have become more, not less, optimistic; less, not more, accurate; and there has been no detectable convergence in analysts’ beliefs. Obviously, the strength of these conclusions depends on our ability to successfully control for ‘other’ determinants of the forecast error, its size and the extent of analyst consensus. The results in Table 4 suggest that this has been achieved, as almost all of the control variables have coefficient signs in the expected direction, and many are statistically significant. We therefore conclude that analysts’ forecasts have not changed post-sanctions in a manner consistent with statutory sanctions increasing the flow of timely, voluntary disclosures. Share Price Evidence Anticipation of Profit Reports Tables 5 and 6 report tests of the effect of statutory sanctions on the extent to which periodic profit reports are anticipated. The ASX’s All Ordinaries Accumulation Index was applied to end-of-month share prices to calculate market-adjusted returns. Long positions were taken on ‘good news’ stocks (i.e., those that gained relative to the market over the twelve months leading up to their PFS or the six months leading up to their HYR) and short positions in the ‘bad news’ stocks, each

427

position being opened at the start of the holding period (i.e., at the end of month −12 in the case of a PFS, where −12 denotes 12 months prior to month 0, which is the announcement month itself). Cumulative market-adjusted returns were calculated from the end of month −6 (−12) to the end of month t, t = −5,…, 0 (month t, t = −11,…, 0) for each stock with a complete set of monthly share prices over the six (twelve) months leading up to an HYR (PFS) announcement month. Portfolios were then formed; for instance, companies that reported a PFS between 1 January 1993 and 4 September 1994 formed a ‘pre-sanctions’ portfolio and those that reported between 1 July 1995 and 29 March 1996 formed a ‘post-sanctions’ portfolio,22 Portfolio members’ cumulative market-adjusted returns were averaged, and for PFSs the ratio of the average return from month –12 to month t to the average return from month −12 to month 0 was then calculated (the starting point was month −6 for HYRs). This ratio is reported in Tables 5 (HYRs) and 6 (PFSs). The statistical significance of the difference between the average degree of anticipation reflected in market prices was assessed using standard resampling methods. Hence, Tables 5 and 6 report respectively the average proportion of the full period’s return that was reflected in prices by the end of the indicated month. Pre-sanctions and post-sanctions reports are compared for all firms, and then for BARCEP firms only. Table 5

428

Evidence on differences in the rate at which the market anticipates the information content of half-yearly reports filed by companies listed on the ASX

The metric which is the basis of Table 5 is the proportion of the full period’s value-relevant information that is reflected in share prices by the end of the indicated month. For half-yearly reports, the full period is assumed to be from the end of month –6 (i.e., six months before the announcement month) to the end of month 0 (the announcement month itself). The columns headed N1 and N2 indicate the number of cases in each set (e.g., there were 1,059 cases in Set 1, Pre-sanctions, All Firms; and 1,186 cases in Set 2, Post-sanctions, All Firms). The columns headed A V1 and A V2 contain the mean metrics for Set 1 and Set 2 respectively; X = AV1 – AV2 denotes the difference between the means for Sets 1 and 2; and p(X > C) is the relative frequency (resampling probability) with which X exceeded the similarly calculated difference between the mean metrics for two sets, of size N1 and N2 respectively, formed by successive random drawings sampling with replacement

429

from the combination of Sets 1 and 2. The resampling probability is the result of 10,000 trials. Table 6 Evidence on differences in the rate at which the market anticipates the information content of Preliminary Final Statements filed by companies listed on the ASX

See Table 5 for an explanation of how this table was constructed. The results for HYRs (Table 5) are as follows. First, averaged over all firms, the mean level of anticipation was not significantly different pre- and post-sanctions; 430

for example, by the end of month –3,51.8 per cent of the information was reflected in pre-sanctions prices while 48.6 per cent was reflected in post-sanctions prices, but this difference of 3.2 per cent could easily have been a chance result (probability = 0.165). Second, for the BARCEP firms, the degree of anticipation post-sanctions has been significantly less except in the last month before the earnings disclosure, when although anticipation was on average less post-sanctions, the difference between the pre- and post-sanctions levels was not statistically significant. A clearer picture emerges when returns are calculated relative to PFS announcements (Table 6). The main results are as follows. First, averaged over all firms, the mean level of anticipation was significantly higher post-sanctions from month –9 onwards; for example, by the end of month –3,75.3 per cent of the information was reflected in pre-sanctions prices while 94.5 per cent was reflected in post-sanctions prices, and this difference of –19.2 per cent was most unlikely to have been a chance result (probability < 0.001). Second, for the BARCEP firms, the degree of anticipation post-sanctions was not significantly greater, except possibly in the last two months before the disclosure (probabilities < 0.117 and 0.102 respectively). In summary, there is some evidence that the share market has derived a benefit from sanctions in the sense that share prices on the whole have anticipated to a greater extent the information content of PFSs. The validity of this conclusion is

431

tempered by our assumption that the information content of PFSs and market-adjusted share returns are isomorphrc, as well as evidence that the effect of sanctions appears to have been concentrated among firms of less interest to institutional investors. Volatility and Disclosure Tests of equation (3) are reported in Table 7. Our measure of monthly price volatility is the absolute value of the monthly discrete rate of return on the stock.23 The first seven numeric columns contain selected percentile values, across the 727 companies, of the probability of observing the estimated regression coefficient if its true value were zero. For instance, in half the regressions fitted there was at least a 0.986 probability that the true value of the regression constant term was not zero (first row of numbers, fiftieth percentile). The last three numeric columns give an alternative way of looking at the same findings. They contain the percentile of the distribution when the t-probability reached the indicated value. For instance, for only 19 per cent of companies was there a less than 0.90 probability that the true value of the regression constant term was not zero (first row, eighth column of numbers). Table 7 Distributional statistics from the regression of a company’s monthly share volatility on the number of price-sensitive documents filed with the ASX that month and three indexes of market-wide volatility

432

A company’s monthly share volatility is measured by the absolute value of its discrete share market rate of return that month. ASX refers to the ASX’s All Ordinaries Accumulation Index, MSCI refers to Morgan Stanley Capital International’s World Index, and SP refers to the Standard and Poor’s 500 Index. NUMDOCS and SANDUM are as described in Tables 2 and 4. Overall, the regression model performs poorly. For instance, the median adjusted R2 is 0.027. Table 7 shows that the relationship between individual stock volatility and index volatility was negligible and could easily have been a chance result (see especially the percentile figures in the fifth, sixth and seventh rows and the last three columns). There is some evidence that monthly share price volatility was related to the number of price-sensitive documents filed with the ASX; the relationship was significant at the 10 per cent level or better for 26 per cent of the companies, at the 5 per cent

433

level or better for 19 per cent of the companies, and at the 1 per cent level or better for 11 per cent of the companies. But the evidence is weak. Moreover, there is also weak evidence in Table 7 that, holding the other variables constant (in particular, the number of price-sensitive documents filed each month with the ASX), the post-sanctions period may have been marked by higher volatility at the company level, which is in contrast to (unreported) simple tests of market volatility.24 One interesting aspect of the results in Table 7 is that, despite the greater number of price-sensitive documents filed post-sanctions, their weighting in explaining share return volatility has not increased. The coefficient of the variable that reflects the interaction between the post-sanctions period and the number of price-sensitive documents is not statistically significant. Conclusions Differences in corporate disclosure policies, and especially the extent of voluntary forward-looking disclosure, are often attributed (at least partially) to differences in regulation and enforcement procedures. Politicians and corporate regulators frequently express the view that substantial civil or criminal penalties will significantly affect corporate behaviour. The introduction of such sanctions in Australia with respect to existing ASX disclosure rules provides an excellent opportunity to test this proposition, as the key requirement of the relevant ASX listing rule (i.e., the maintenance of an informed market) was not changed. Rather, legislative action was in the form of providing

434

statutory civil and criminal sanctions to ‘enforce’ ASX listing mies. However, identifying the effect of such legislative action on corporate disclosure policies is a difficult task. Essentially, tests involve the assumption that sufficient controls can be put in place to be confident that a temporal intervention (i.e., the date at which statutory sanctions become operative) is sufficient to distinguish the effect of these sanctions. Given the relatively ad hoc nature of empirical models of voluntary disclosure, we also turn to several indirect indicators (i.e., possible consequences) of increased timely, voluntary disclosure. These are the extent of disagreement among, and accuracy of, analysts’ earnings forecasts, the extent to which the value-relevant components of periodic accounting reports are anticipated, and the linkage between corporate disclosures and share price volatility. The results are best described as mixed. First, although total disclosures increased post-sanctions, disclosures classified as ‘price-sensitive’ by the ASX only became more frequent for firms without a large analyst following and for firms which are more likely to have revealed relatively bad news. Second, after controlling for other intervening variables such as firm size, industry group and trading-based incentives to collect information, the estimates suggest that forecasts became less accurate and that analyst disagreement remained at about the pre-sanctions level. Third, post-sanctions, share prices on the whole have anticipated earlier the value-relevant components of a company’s preliminary final statement; however, the information advantage is concentrated

435

among companies of less interest to institutional investors. Fourth, despite the increase in the frequency with which companies have filed price-sensitive documents post-sanctions, the weight placed on each document’s ability to explain return volatility is not statistically different from the pre-sanctions period. Caution is needed in interpreting these results. The four principal tests are clearly not independent. Moreover, the time span since the introduction of sanctions until the end of our data effectively includes only one reporting year for companies with 30 June fiscal year ends (i.e., about 75 per cent of Australian firms), while commonalities across firms in the samples further reduce the statistical reliability of the effects observed. Nonetheless, it seems reasonably evident that there is not, at this point, strong evidence of statutory civil and criminal sanctions having a marked effect on corporate disclosure policies. References Abarbanell, J. S., and B. J. Bushee, ‘Fundamental Analysis, Future Earnings and Stock Prices’, Journal of Accounting Research, Spring 1997. Abarbanell, J. S., W. N. Lanen and R. E. Verrecchia, Analysts’ Forecasts as Proxies for Investor Beliefs in Empirical Research’, Journal of Accounting and Economics, Vol. 20, No. 1,1995. Alford, Α., J. Jones, R. Leftwich and M. Zmijewski, ‘The Relative Informativeness of Accounting Disclosures in Different Countries’, Journal of Accounting Research, 31 (Supplement), 1993. 436

Bagįnski, S., and J. Hassell, ‘The Market Interpretation of Management Earnings Forecasts as a Predictor of Subsequent Financial Analyst Forecast Revision’, The Accounting Review, January 1990. Ball, R., and P. Brown, ‘An Empirical Evaluation of Accounting Income Numbers’, Journal of Accounting Research, Autumn 1968. Benston, G. J., ‘Required Disclosure and the Stock Market: An Evaluation of the Securities Exchange Act of 1934’, American Economic Review, Vol. 63, No. 1,1973. Botosan, С., ‘Disclosure Level and the Cost of Equity Capital’, The Accounting Review, July 1997. Brown, P., Capital Markets-Based Research in Accounting: An Introduction, Coopers and Lybrand Accounting Research Methodology Monograph No. 1, Coopers and Lybrand, 1994. Brown, P., S. L. Taylor and T. S. Walter, Enhanced Statutory Disclosure: Much Ado About Nothing?, working paper, University of Sydney, 1998. Clarkson, P., J. Kao and G. Richardson, ‘The Inclusion of Forecasts in the MDA Section of Annual Reports: A Voluntary Disclosure Perspective’, Contemporary Accounting Research, Fall 1994. Darrough, M. N., and N. M. Stoughton, ‘Financial Disclosure Policy in an Entry Game’, Journal of Accounting, and Economics, Vol. 12, No. 1, 1990.

437

Frankel, R., M. McNichols and G. P. Wilson, ‘Discretionary Disclosure and External Financing’, The Accounting Review, January 1995. Friend, I., ‘Economic Foundations of Stock Market Regulation’, Journal of Contemporary Business, Summer 1976. Frost, C. A., ‘Disclosure Policy Choices of U.K. Firms Receiving Modified Audit Reports’, Journal of Accounting and Economics, Vol. 23, No. 2, 1997. Healy, P., K. Palepu and A. Sweeney, Do Firms Benefit From Expanded Voluntary Disclosure?, working paper, Harvard Business School, Harvard University, 1995. Jennings, R., ‘Unsystematic Security Price Movements, Managerial Earnings Forecasts, and Revisions in Consensus Analyst Earnings Forecasts’, Journal of Accounting Research, Spring 1987. Lang, M., and R. Lundholm, ‘Cross-Sectional Determinants of Analyst Ratings of Corporate Disclosures’, Journal of Accounting Research, Autumn 1993. Lang, M., and R. Lundholm, ‘Corporate Disclosure Policy and Analyst Behaviour’, The Accounting Review, October 1996. Lee, P., S. L. Taylor and T. S. Walter, The Voluntary Disclosure of Forecast Data by Australian IPOs, working paper, University of Sydney, 1995.

438

Peltzman, S., ‘Toward a More General Theory of Regulation’, Journal of Law and Economics, February 1976. Philbrick, D. R., and W. E. Ricks, ‘Using Value Line and IBES Analyst Forecasts in Accounting Research’, Journal of Accounting Research, Autumn 1991. Ruland, W., S. Tung and N. George, ‘Factors Associated with the Disclosure of Managers’ Forecasts’, The Accounting Review, July 1990. Skinner, D., ‘Why Firms Voluntarily Disclose Bad News’, Journal of Accounting Research, Spring 1994. Skinner, D., ‘Earnings Disclosure and Stockholder Lawsuits’, Journal of Accounting and Economics, Vol. 23, No. 3, 1997. Stigler, G. J., ‘Public Regulation of Securities Markets’, Journal of Business, January 1964. Verrecchia, R., ‘Discretionary Disclosure’, Journal of Accounting and Economics, Vol. 5, No. 1, 1983. Verrecchia, R., ‘Endogenous Proprietary Costs Through Firm Interdependence’, Journal of Accounting and Economics, Vol. 12, No. 1, 1990. Waymire, G., ‘Additional Evidence on the Accuracy of Analysts’ Forecasts Before and After Voluntary Management Earnings Forecasts’, The Accounting Review, January 1986 PHILIP BROUN is Professor of Accounting in the Department of Accounting and Finance at the University of Western Australia. STEPHEN TAYLOR and TERRY 439

WALTER are each Professor of Accounting in the Department of Accounting at the University of Sydney. Financial support for this research was provided by the Companies and Securities Advisory Committee, the Australian Securities Commission and the Australian Stock Exchange. We acknowledge the programming assistance of Jennifer Cross and Franc Carter, as well as the research assistance of Mark O’Donnell and Peter Lambousis. We are grateful for comments made by workshop participants at the Australian Graduate School of Management, and the following universities: Melbourne, New South Wales and Sydney. We have also benefited from suggestions made by Greg Clinch, Graeme Dean and an anonymous referee. 1

These changes were enacted via the Corporations Law Reform Act 1994 (Cth), and became effective on 5 September 1994. 2

Typical of these criticisms is the testimony before the House of Representatives Standing Committee on Legal and Constitutional Affairs (the Lavarch Committee) offered by a senior investment banker who stated: The lack of meaningful disclosure has enabled the management of certain companies to undertake transactions that have caused a significant transfer of wealth from shareholders to management and selected large shareholders associated with management… During this process shareholders have had limited opportunity to intervene and question … the motives of the directors of companies in which they own shares. 440

The investment banker argued that disclosure by listed firms was lacking in both detail and timeliness, a view supported by several other witnesses appearing before this enquiry. (Evidence presented by Mr M. McComas, Director, County NatWest Australian Corporate Services Ltd, 3 September 1990.) 3

Prior to 1992, ASX Listing Rule 3A(1) required companies to notify the exchange ‘immediately’ of any information about their activities likely to have a material effect on share price, or to avoid the creation of a false market in their shares. This requirement was amended to coincide with the requirements of s. 1022 of the Corporations Law dealing with information required in prospectuses, so that under 3A(1) firms were required to reveal information which: investors and their professional advisers would reasonably require, and reasonably expect to be disclosed to the market, for the purpose of making an informed assessment of: a. the assets and liabilities, financial position, profits and losses, and prospects of the listed company; and b. the rights attaching to securities of the listed company. 4

In conjunction with the legislation, Listing Rule 3A (1) was again amended, presumably to clarify listed entities’ obligations in light of the introduction of statutory civil and criminal penalties. The revised version of 3A (1) operative from September 5,1994 states:

441

A listed company shall immediately notify the Exchange of any information concerning the company of which it is or becomes aware and which a reasonable person would expect to have a material effect on the price or value of securities of the company. This requirement does not apply if each of the following conditions is and remains satisfied: i. a reasonable person would not expect the information to be disclosed; and ii. the information is confidential; and iii. one or more of the following conditions apply: a. it would be a breach of a law to disclose the information; b. the information is, or is part of, an incomplete proposal or negotiation; c. the information comprises matters of supposition or is insufficiently definite to warrant disclosure; d. the information is generated for the internal management purpose of the company; or e. the information is a trade secret.

For the purpose of this listing rule, the company becomes aware of information where a director or executive officer has, or ought reasonably to have, come into possession of the information in the course of the performance of duties as a director or executive officer. 5

The term ‘voluntary disclosure’ is used to describe those disclosures made to comply with ASX Listing Rule 3A(1). In the same manner as Frost (1997), we argue that disclosures made to comply with 3A(1) are also often described as voluntary, while determining whether information is price sensitive (and hence subject

442

to 3A(1)) is judgmental, resulting in such disclosures being effectively discretionary. 6

Specifically, these changes are contained in s. 1001A of the Corporations Law. 7

Extensions of this approach include Darrough and Stoughton (1990) and Verrecchia (1990). 8

A small number of studies examine the relationship between properties of analysts’ forecasts and management’s earnings forecasts (e.g., Waymire, 1986; Jennings, 1987; Bagįnski and Hassell, 1990). 9

Lang and Lundholm (1996) also investigate the relationship between corporate disclosure policy and the extent of analyst following. 10

That is, analyst disagreement is used as an explanatory variable when we are predicting forecast accuracy, but not vice versa as forecast accuracy is unknown when the forecast is made. 11

This view, commonly used in capital markets research in accounting, dates from the earliest studies (e.g., Ball and. Brown, 1968). They showed that the earnings figure alone accounts for half the value of all information that comes to the market annually. 12

A similar approach is used by Alford et al. (1993) to compare the information content of periodic accounting reports across several countries. 13

There are some instances where Signal G records are missing from SIRCA ‘s database. Data for 30 November 1993, 20 December 1993 and the period from 13 January

443

1994 to 20 January 1994 are missing. It is unlikely that these missing observations would substantially alter our results and conclusions. 14

A summary of the ASX’s procedures in transmitting announcements is available from the authors. 15

These two-digit codes are a standardized classification system for all announcements. Some announcements have more than one code, as releases often contain information on several different topics. For example, the PFS is normally assigned a 03 code (Periodic Reports), a 08 code (Notice of Meeting) and a 10 code (Dividend Announcement). The reliability of these electronic records was tested for a stratified random selection of small, medium and large firms (thirty in total) by comparing the electronic records with the hard copy documents held in the ASX Collection at the University of Sydney. In general, we find the electronic records to be both accurate and complete. Full details of these tests are available from the authors. 16

Because a single document can contain several different ASX announcement codes, Table 1 reports a count of total ASX codes, rather than documents. 17

Our approach assumes that the coefficients γ2 through are stable over time, A substantial drift in some or all of these coefficients may affect the significance of the SANDUM coefficient. However, the relatively high explanatory power of the model reduces the likelihood that this can occur 18

Thirty-eight firms are classified as good news in both periods, while 323 firms have good news in only one 444

period. There are 51 firms classified as bad news in both periods, while 425 firms have bad news in only one period. 19

For example, the absolute forecast error at month –11 averaged 0.019, which means that analysts’ average error was about 2 per cent of market capitalization. 20

However, as Table 3 indicates, the pre- and post-sanctions accuracy rates were not statistically different until the last month before the company released its annual results. 21

Of course, our tests assume that analysts will take full advantage of more timely voluntary disclosures. However, Abarbanell and Bushee (1997) provide evidence consistent with United States security analysts failing to efficiently impound ‘fundamental’ signals into their earnings forecasts. To the extent this is also true of Australian analysts, then our tests are biased against finding improvements post-sanctions. 22

The period from 5 September 1994 to 31 December 1994 was excluded for HYRs because reports covering this period overlapped the introduction of statutory sanctions. 23

Extreme returns, defined to be those less than –80 per cent and greater than 400 per cent, were winsorized to those values. 24

These results are available on request from the authors. © 1999 by Accounting Foundation, The University of Sydney 445

Do Better-Governed Australian Firms Make More Informative Disclosures? WENDY BEEKES AND PHILIP BROWN*

Abstract: We investigate whether and if so, how, corporate governance ‘quality’1 is related to the information flows from a company and how the share market and its agents respond. Specifically, we study links between the ‘quality’ of a firm’s corporate governance (CGQ) and the informativeness of its disclosures. We employ six indicators of informativeness. They include document counts, properties of analysts’ forecasts and a ‘timeliness’ metric, in the spirit of Ball and Brown (1968), that reflects the average speed of price discovery throughout the year. Our results suggest the answer to our question is Ύes’: better-governed firms do make more informative disclosures. Keywords: corporate governance quality, disclosure frequency, analysts’ forecasts, price discovery, timeliness 1. Introduction Widespread community concerns expressed since the Enron debacle about inadequate standards of corporate governance have not been confined to North America. Governance codes have been introduced throughout Europe; and in March 2003 the Australian Stock Exchange (ASX) also produced a set of governance

446

principles and guidelines.2 The introduction of such codes of practice and companies’ response to them has created a rich area for accounting research. While there are many facets of corporate governance that can be and indeed have been investigated, we focus on an area that has not been fully researched so far. We investigate whether a listed company’s corporate governance ‘quality’ is related to the information it reports to the share market and how the market and its agents respond. Specifically, we study links between a firm’s corporate governance ‘quality’ rating (CGQ) and the informativeness of its disclosures. Our study is motivated by two propositions embedded in the ASX statement of principles: better-governed firms make more timely and more ‘balanced’ disclosures of both good and bad news. We treat these propositions as maintained assumptions which, when coupled with findings in the research literature, lead us to predict better-governed firms make more informative disclosures, with implications for analysts’ earnings forecasts and ultimately, market efficiency. We investigate six indicators of informativeness: (i) the frequency with which the firm makes price-sensitive announcements to the share market; (ii) the extent of its security analyst following; (iii) the accuracy, (iv) bias and (v) disagreement in the analysts’ earnings forecasts for that firm; and (vi) the speed with which its share price reflects value-relevant information. Our measure of CGQ is sourced from the Horwath-University of Newcastle Corporate Governance Report (Horwath Report; Horwath, 2002). The measure in the Horwath

447

Report is primarily structural in nature, reflecting among other things the independence of the board, its Chair and its principal committees. We employ a novel, intra-year ‘timeliness’ metric, in the spirit of Ball and Brown (1968) Alford et al. (1993) and Brown et al. (1999), to capture the average speed of price discovery throughout the year. Controls are added to our models for other factors (such as the firm size and the volatility of share returns) that could explain at least part of the cross-sectional variance in the dependent variables. In brief, we find strong support for the proposition that better-governed Australian firms do make more informative disclosures. This suggests corporate governance is influential and contrasts some of the prior literature, where findings have been mixed (Larcker et al., 2005). Clearly our results could have wide ranging policy implications. Moreover, if the same proposition is true in other countries, potentially the ‘quality’ of a firm’s corporate governance will be valued more highly by share market participants worldwide. We contribute to the literature on corporate governance and timeliness of information in several ways. First, we consider whether corporate governance has any impact on the informativeness of disclosures to the market. Second, we use a novel approach to measure the timeliness of this disclosure. Third, we provide some insight into whether better corporate governance was in fact associated with more timely and more balanced disclosures prior to the introduction of the ASX principles of good corporate governance.

448

The remainder of our paper is organised as follows. Section 2 reviews related literature and discusses evidence on whether corporate governance has any impact on firms’ activities. Section 3 outlines the ASX principles of good corporate governance. Section 4 describes our data and research method. Section 5 presents the main results while Section 6 summarises additional tests that demonstrate the robustness of our results. Section 7 concludes the paper. 2. Some Related Studies A key research question is: does corporate governance really make any difference to a firm’s performance and actions? In a recent, comprehensive US study, Larcker et al. (2005) ask, Ήow important is corporate governance?’ To answer this question, they first needed to define and measure corporate governance. They do this by ‘going back to the drawing board’, collecting information on 39 measures reflecting various aspects of corporate governance for 2,106 firms. Principal components analysis is used to reduce the 39 measures to 14 governance factors, which are then correlated with indicators in six areas where the quality of corporate governance might matter: accounting accruals, accounting restatements, class action lawsuits, Tobin’s Q, and future operating and abnormal share market performance. Larcker et al. conclude: the typical structural indicators of corporate governance used in academic research and institutional rating services have some ability to explain managerial decisions and firm performance and valuation (p. 4, emphasis in original). 449

There have been a number of earlier studies of possible linkages between corporate governance and future investment and operating performance, with inconsistent results (Larcker et al. 2005, p. 1). For instance, Gompers et al. (2003) suggest that the corporate governance of a firm, more specifically the level of shareholder versus managerial rights, influences stock returns. They create a governance index, the G-index, for US companies from Investor Responsibility Research Centre data, awarding one point for each provision that increases managerial rights relative to shareholder rights. A portfolio that was long in firms with strong shareholder rights and short in firms with strong managerial rights would have generated returns averaging 8.5% per year during the 1990s. Gompers et al. (2003) provide several explanations for this result, concluding weak corporate governance was a contributor. However Core et al. (2006) find the Gompers et al. result is probably period specific. Indeed, when the study period is extended for four more years (until December 2003), Core et al. (2006, Table VIII) find the cumulative difference in the return on firms with strong versus weak corporate governance actually declines. Interestingly, as part of their study, Core et al. regress analysts’ earnings forecast errors on the firm’s G-index, (the log of) its market capitalisation and (the log of) its book-to-market ratio. They conclude any bias in their sample of analysts’ forecasts is unrelated to the G–index. They interpret this result as indicating sell-side analysts understand the implications of weak governance for operating performance and adjust their forecasts accordingly.

450

Other evidence from Brown and Caylor (2004) provides additional insight into the association between the G-index and firm performance. They compare the G-index with Gov-score, an index they constructed from Institutional Shareholder Services data based upon firms meeting minimum standards of corporate governance. Their findings suggest a positive association between better governance (as measured by Gov-score) when firm performance is measured by return on equity, net profit margin, Tobin’s Q, dividend yield and share repurchases but not when it is measured by growth in sales. This is in contrast to results for the same sample of firms using the G-index, which suggests better governance is positively associated with firm performance for sales growth only. They attribute this difference to the narrow focus of the G-index, which primarily reflects anti-takeover measures. Other studies using the G-index to measure corporate governance include Defend et al. (2005), Cremers and Nair (2005) and Klock et al. (2004). Defend et al, (2005) report the appointment of a financial expert to the audit committee is associated with abnormal returns when the appointment is announced and where there are strong shareholder rights. Cremers and Nair (2005) investigate the complementarity of internal and external governance mechanisms and find greater shareholder rights (proxied by greater vulnerability to take-over) are associated with greater profitability and positive long-term abnormal returns when there is strong internal corporate governance. Klock et al. (2004) conclude corporate governance influences the cost of debt, firms with greater takeover defences (i.e., greater

451

managerial rights) paying less when they borrow. These studies suggest there may be complementarity or substitutability among governance mechanisms used by companies. Much of the research in this area is US based, but two Australian studies are also worth noting. The first is noteworthy because it uses the same CGQ ratings as we do, namely those published annually in the Horwath Reports. In that paper, Linden and Matolcsy (2004) find no evidence that a firm’s CGQ is related either to its investment performance (as measured by return on the share market) or operating performance (as measured by an accounting rate of return). A second paper, by Matolcsy et al. (2004), uses a self-constructed corporate governance database and asks whether additional external directors are associated with a greater market value of equity, other things equal. Generally the answer is they are not, an exception being where growth options are a relatively high component of the firm’s value. Theory and prior evidence indicate a firm’s disclosure policy is related to a number of factors, including its size and growth opportunities, the incentives of its managers (Noe?1999), proprietary costs (Verrecchia, 1983) and litigation costs (Skinner, 1994; and Skinner, 1997). A firm’s governance structure also can influence the nature of its disclosures: on one view, a well-designed governance structure can help ensure the firm’s disclosure policy is ‘optimal’ (Shleifer and Vishny, 1997; and Core, 2001), However, the evidence on the link between corporate governance and disclosure is mixed. Coulton et al.

452

(2001) examine the influence of corporate governance mechanisms on CEO compensation disclosures following a change in reporting provisions in Australia. They find no evidence that corporate governance characteristics of the firm influence the level of disclosure and conclude corporate governance is either ineffective, or does not influence compensation disclosure. Leung and Horwitz (2004) examine voluntary segment data disclosures in Hong Kong and conclude non-executive directors have a positive impact on the level of voluntary disclosure, but only when there is low executive stock ownership. Eng and Mak (2003) find the number of outside directors is inversely related to disclosure levels in Singapore. Wright (1996) investigates disclosure quality in the US by examining reports of the Association for Investment Management and Research Corporate Information Committee (AIMR) and Securities and Exchange Commission Accounting and Auditing Enforcement Releases (AAER). Firms with more ‘grey’ directors tend to be associated with lower analyst evaluations and in addition, ‘grey’ directors are more heavily represented on audit committees of firms subject to SEC enforcement actions. Substantial shareholders can strengthen corporate governance by actively monitoring the firm and may influence disclosure levels (Healy et al.?1999); but their presence can reduce the informativeness of disclosures where disclosure policies and substantial shareholders are substitute mechanisms. Ajinkya et al. (2005) examine the link between outside directors and institutional ownership, and the

453

propensity to make earnings forecasts. They find these governance factors increase the likelihood, frequency and accuracy of managers’ earnings forecasts. The ASX principles are based on the premise that corporate governance quality influences the level and timeliness of company disclosure in particular ways: higher quality corporate governance leads to more timely and more ‘balanced’ disclosures of both good and bad news, which if true would have consequences for the frequency of disclosures and for properties of financial analysts’ forecasts. For example, Lang and Lundholm (1996, p. 467) find US ‘firms with more informative disclosure policies have a larger analyst following, more accurate analysts’ earnings forecasts (and) less dispersion among individual analysts’ forecasts’.3 Further, if the ASX’s Corporate Governance Council is correct in its view that higher CGQ leads to more ‘balanced’ disclosures and if, contrary to Core et al. (2006), Australian analysts do not fully understand the implications of poorer governance for biased disclosures, then their forecasts are likely to be more optimistically biased for poorer-governed firms. Another implication we test is that higher CGQ leads to faster price discovery and to a more efficient market in the company’s shares. We refer to this notion as the ‘timeliness’ of the information that becomes embedded in price, and predict higher CGQ is associated with more timely price discovery. Much of the research on the timeliness of information has to date focused on how quickly value-relevant information embedded in price is recognised in reported earnings per share, as inferred

454

from a reverse regression of earnings on returns (Basu, 1997). In investigating the link between corporate governance and timeliness as defined by Basu (1997), the literature has focused upon the board of directors (a primary internal governance mechanism; Fama and Jensen, 1983) and especially upon the board’s composition and the role of non-executive directors. The empirical evidence on the link between non-executive board membership and accounting timeliness in the Basu sense is mixed. Beekes et al. (2004) examine the link between earnings timeliness and conservatism, and board composition, for UK companies between 1993 and 1995. They find firms with greater non-executive director representation adopt a more conservative approach with respect to the timeliness of bad news recognition in earnings. In contrast, Bushman et al. (2004) only find evidence of a greater proportion of experienced outside directors in firms with low earnings timeliness, for a sample of Fortune 1,000 firms in the US. Using a similar research method, Basu et al. (2001a and 2001b) and Chung et al. (2003) find big-five auditees are more conservative.4 There has been some debate in the literature about the appropriateness of the Basu model and issues of reverse causality appear problematic, which raises a question as to the model’s reliability (Bushman et al., 2004; Dietrich et al., 2004; and Givoly et al., 2004). We use a different method to assess timeliness, one based upon ideas set out in Ball and Brown (1968). As explained more fully in Section 4,

455

we measure timeliness by the accuracy with which the firm’s share price reflects the net pricing effect of all value-relevant information made available over the year. 3. The ASX Principles of Good Corporate Governance Following the enactment of the Sarbanes-Oxley Act in the US in 2002 and the adoption of various governance initiatives in the UK, in March 2003 the ASX’s Corporate Governance Council issued a report, titled ‘Principles of Good Corporate Governance and Best Practice Recommendations’. The recommendations are ‘based on the objectives of restoring investor confidence and promoting transparency to enable shareholders to make comparative investment decisions.’5 The report espouses 10 principles, including timely and balanced disclosure of all material matters concerning the company (Principle 5) and a respect for shareholders’ rights, including the right to be properly informed (Principle 6). Principle 5 envisages a system whereby ‘all investors have equal and timely access to material information concerning the company – including its financial situation, performance, ownership and governance’ and ‘company announcements are factual and presented in a clear and balanced way. ‘Balance’ requires disclosure of both positive and negative information’ (p. 35). Principle 6, which is related to Principle 5, ‘means that a company should empower its shareholders by communicating effectively with them (and) giving them ready access to balanced and understandable information about the company and corporate proposals’ (p. 39).

456

The Corporate Governance Council comprises representatives of 21 of the most influential professional and commercial organisations in Australia. They represent the interests of accountants, actuaries, major corporates, bankers, security analysts, portfolio managers, lawyers, property managers, retail investors, share and derivatives traders and the ASX itself. The Council’s report acknowledges such a diverse range of organisations can hold different views, but they did at least agree on the need to issue a set of defensible corporate governance guidelines: Despite their differing perspectives, the Council’s overriding mission was commonly held: to develop and deliver an industry-wide, supportable and supported framework for corporate governance which could provide a practical guide for listed companies, their investors, the wider market and the Australian community (Foreword). The guidelines, which are summarised in Appendix A, are applied in a similar manner to the Cadbury Code in the UK: on a ‘comply or explain’ basis.”6 4. Data and Research Method We take the position that, despite the negative findings of many past studies, there might still be some truth to the view that CGQ matters to financial markets, given the resources devoted to promoting better corporate governance and the ubiquitous nature of the investment community’s concern. Thus we ask the following questions. Is it in fact the case that ‘better-governed’ firms – meaning those whose governance structures 457

more closely conform to the ASX guidelines – are ‘more transparent’ and make ‘more timely’ disclosures that are ‘better balanced’ in terms of the release of good and bad news? If so, is the variation in CGQ, which reflects the extent of conformity with those guidelines, apparent in the outcomes of security analysts’ decisions and in the speed of share price discovery? To answer these questions, we investigate whether higher CGQ firms make more frequent disclosures of price-sensitive material, whether more analysts follow these firms, whether the analysts’ earnings forecasts are on average less biased, more accurate and less disperse, and whether, presumably as a consequence, share price discovery is faster. Our sample of companies comprises the 250 Australian firms rated in the 2002 Horwath Corporate Governance Report, which pre-dates the introduction of the ASX Principles. This sample period has two key advantages: first, widespread concerns about deficiencies in corporate governance suggest a corporate governance equilibrium of the kind envisaged by Bushman et al. (2004, p. 169) did not exist in Australia in 2002; and second, it was prior to any necessary disclosures relating to the ASX Guidelines.7 In the remainder of this section we explain how we measure CGQ and other key variables. (i) Measuring CGQ We use the 2002 (first) edition of the Horwath Report as our source of a measure of CGQ.8 It contains corporate governance rankings for Australia’s top 250 companies 458

by market capitalisation as at 30 June, 2001. The rankings are based on information about the Board and its principal committees that is contained in the companies’ 2001 Annual Reports and related party disclosures. The index is calculated in the same manner for all companies irrespective of their size. The Horwath Report contains each company’s ranking and the rankings are also summarised according to a five-star system. The award of a five star rating indicates that the firm’s ‘corporate governance structures were outstanding. The structures met all best practice standards and could not be faulted’. At the other extreme, a one-star rating indicates the company’s ‘corporate governance structures were lacking in several areas’ (2002 Horwath Report, p. 23). Just nine of the 250 companies surveyed were awarded a five-star rating while 13 companies, 5.2% of the sample, were rated one-star.9 Some, but not all, of the companies with an excellent corporate governance rating under the Horwath Report were large (Horwath, 2002). Companies with a five-star rating were governed by an independent Board of Directors and associated Board committees, which met on a regular basis. Also transactions with related parties were disclosed fully and in a transparent manner. An example of a company with a five-star corporate governance rating is David Jones, a department store company. ‘Its Board had a clear emphasis on independence, with seven independent directors (including the Chairman) and only one non-independenť (Horwath, 2002, p. 16). Companies with a one-star rating had poor governance structures,

459

which typically included no independent director and few, if any, board committees, as was the case for Cape Range Wireless and Ecorp. A large company with a one-star corporate governance ranking is Harvey Norman, which had no independent board representation (Horwath, 2002). Thus despite the prior existence of other governance codes, which tend to induce uniformity, firms during this period had very different governance structures and substantial variation exists within our sample. The detail of the model used to derive the ratings is proprietary information but the 2002 Horwath Report provides the following broad description: The corporate governance assessment model developed in the research is based upon factors identified in national and international best practice guidelines and research studies … Central to the model is the need for companies to have appropriate levels of independence on the Board of Directors and their associated committees. The model considers objective factors based on publicly disclosed information pertaining to the existence and structure of a company’s Board of Directors, audit committee, remuneration committee, and nomination committee (p. 9). Further Information about the Horwath model was provided in the 2004 Horwath Report, extracts of which are set out in Appendix B. The Horwath measure of CGQ is not all-encompassing. It focuses primarily on board and committee 460

independence and meeting frequency. Nevertheless, the independence of the board of directors and associated committees has been highlighted as very important for good corporate governance (Cadbury Report, 1992).10 For convenience, we transform the raw rating score to range from 1 to 100, where better corporate governance is reflected in a higher score. (ii) Frequency of Disclosure A history of company announcements is sourced from ASX via the Securities Industry Research Centre of Asia-Pacific (SIRCA). The ASX listing rules include a requirement that price-sensitive information be released via the ASX. In brief, when a company forwards a document for release to the market, an ASX executive vets the document. If it is suitable for release, the document is classified into one or more report types (to capture the nature of the message) and it is also classified according to whether it is price-sensitive or not. Trading is halted typically for about 10 minutes when a price-sensitive document is released. Although there are grounds for believing the ASX is conservative in the sense that it is more likely to classify a non-price-sensitive document as price-sensitive than vice versa, two recent studies of the ASX’s history have concluded the ASX ‘typically gets it right’: order flow and trades are significantly different when the ASX releases to the market a document it classifies as price-sensitive, but they are not significantly different when the document is classified as not price-sensitive (Brown et al., 2004; and Brown et al, 2005).

461

We measure the frequency of disclosures by the number of price-sensitive documents released by the company over the 250 trading days ended two weeks (10 trading days) after the company’s Preliminary Final Statement (PFS)’ which is regarded as the primary announcement of the year’s results.11 A total of 1,226 PFSs were released by the 250 sample firms for which we had sufficient disclosure information and share return data, of which 601 releases were made between January 2000 and June 2003.12 We control for statutory reports, which all firms must file, by including a constant term in the regression model. We also include an industry dummy variable for firms operating in the natural resource sector because mining and exploration companies are required to report quarterly whereas other companies must report at least semi-annually. Since large firms are subject to greater public scrutiny and are therefore likely to disclose more frequently, we control for firm size. Our other control variable is a dummy variable to denote good news, which is defined to occur when the company outperforms the market over the same 250-day period.13 (iii) Timeliness (and the Speed of Price Discovery) We build on Ball and Brown (1968), Alford et al. (1993) and Brown et al. (1999) to measure the timeliness with which a firm’s share price reflects the net effect of all value-relevant information impounded in share price over the year. The year is defined to be the period of 250 trading days ended two weeks after the release of the firm’s PFS, which encapsulates the announcing firm’s operating, investment and financial performance in the financial year just ended.

462

Ball and Brown (1968) referred to the timeliness of the annual income number in terms of the extent to which share price fails to reflect information in the income number prior to its release, concluding that: the annual income report does not rate highly as a timely medium, since most of its content (about 85 to 90 per cent) is captured by more prompt media (p. 176). Alford et al. (1993) adapted the Ball and Brown approach to compare the timeliness of accounting information across countries. Brown et al. (1999) applied a similar method, estimating at a portfolio level the extent to which Australian share prices have anticipated earlier the value relevant components of periodic accounting information following the introduction in 1994 of statutory sanctions. Those sanctions reinforced pre-existing stock exchange regulations requiring prompt disclosure of price-sensitive information. We further develop the Ball and Brown concept, to examine timeliness at the individual firm level. Our measure addresses the question: How accurately does the share price observed throughout the year approximate the market’s valuation of the share two weeks after the Preliminary Final Statement has been released? As in Ball and Brown (1968) and Brown et al. (1999), prices are explicitly market-adjusted. Continuously compounded returns are used, although our metric (Mc)

463

is not especially sensitive to either of these two matters. Specifically, Mc is defined in this paper as:

where Pt is the market-adjusted share price, which is observed at daily intervals from day −249 until day 0.14 The intuition behind our metric is simple. Suppose price tracked the market index from day −249 until day 0, when it fell from P−249 to P0. In this case, the speed of adjustment is slow and Mc is ‘close to’ the absolute value of the market-adjusted continuously compounded return over the 250 trading days. If, however, price changed to P0 on the first day of trading (day −249) and then simply tracked the market index for the remaining 249 days, the speed of adjustment is at a maximum and Mc is identically zero. While the metric is simple in concept, it is not without its complications. For example, idiosyncratic share price volatility will tend to inflate the metric when calculated at the individual firm level15 but its effect will be dampened when the metric is calculated for portfolios, as in previous studies. Nonetheless, if firms with higher CGQ do release price-sensitive information in a more transparent and timely fashion, then our metric should capture, albeit with substantial error, this feature. Daily share prices are sourced from SIRCA’s Core Research Database, while the market index we use is the Australian All-Ordinaries Accumulation Index. The principal explanatory variable is the Horwath CGQ

464

rating, and the control variables are the firm’s size,16 a good news dummy variable that indicates whether the share’s price rose (=1) or fell (=0) relative to the market and a dummy variable to denote firms operating in the natural resource sector. (iv) Analysts’ Earnings Forecasts We use the I/B/E/S international summary database as our source of analysts’ annual EPS forecast data, for the 2002 financial year, extracting monthly information on firm size, the consensus earnings forecast, analyst disagreement (measured by the standard deviation across analysts’ forecasts for the same forecast horizon and fiscal year), analyst following and actual EPS for cases where at least four analysts contributed their forecasts and where the forecast horizon was at most 11 months. The forecast error is measured by the mean forecast17 less the actual I/B/E/S figure, the difference being deflated by the share price on the I/B/E/S cut-off date 12 months before the announcement. The standard deviation across analysts is likewise deflated by share price. A total of 1,571 analysts’ forecasts of annual EPS for financial year 2002 met our sample selection criteria for inclusion in the models. Four regression models are estimated (shown below),18 the dependent variables being the price-deflated EPS forecast error (bias), the absolute value of the forecast error (accuracy), the analyst disagreement variable, which is also price-deflated, and the number of analysts contributing a forecast (a measure of analyst following). Each model is fitted with a constant term and, following

465

Brown et al. (2001), the control variables include the firm’s size (measured by the natural log of the market value of the firm’s equity on the I/B/E/S forecast date), the previous year’s forecast error over the same forecast horizon and its absolute value, the share’s volatility,19 three industry dummy variables, whether the share is traded on the options market, whether it is eligible for short selling and the forecast horizon. The three industry dummy variables (Resource, Finance, Property) are included to control for industry specific factors. For example, the ASX requires more frequent reporting by mining and exploration companies. The finance sector is subject to a different regulatory regime and we allow for the possibility that property investment companies are affected by economic circumstances differently from other companies. A dummy variable is included to denote the opportunity to short sell a stock because, as Brown et al. note: (t)he opportunity to short a stock when adverse information is uncovered increases analysts’ incentives to collect information and can result in more accurate forecasts and greater agreement among them (1999, p. 144). The same argument applies to stocks traded on the options market. Horizon is included to acknowledge the length of time until the firm files its PFS with the ASX. Naturally, nearer to filing, we expect more accurate forecasts, as more information will be in the public domain about the company’s prospects and performance. The exact specifications are set out below:

466

where: Bias:

Accuracy: Disan:

Analyst: Size:

signed Forecast Error (FE). Forecast error (FE) is defined as the mean forecast EPS less EPS as reported by I/B/E/S, and is deflated by share price a year before the announcement month; absolute value of the FE, deflated by share price a year before the announcement month; level of disagreement, measured by the standard deviation across analysts’ forecasts for that firm-month as reported by I/B/E/S, deflated by price; number of analysts contributing to the forecast as reported by I/B/E/S; firm size, proxied by the natural log of the firm’s market capitalisation (in AUD

467

million) as reported by I/B/E/S on the forecast cut-off date; PrevFE: last year’s FE, for the same firm and same forecast horizon; ABSPrevFE:absolute value of PrevFE; Vol: volatility, calculated from daily returns in the 60 trading days ended the day before the I/B/E/S forecast date; Resource: Dummy variable coded 1 for natural resources industry, 0 otherwise; Finance: Dummy variable coded 1 for finance industry, 0 otherwise; Property: Dummy variable coded 1 for property industry, 0 otherwise; Options: Dummy variable coded 1 for firms traded on the options market, 0 otherwise; Short: Dummy variable coded 1 for firms that the ASX permits to be sold short, 0 otherwise; Horizon: Forecast horizon, measured by the number of months from the forecast date until the company releases its Preliminary Final Statement to the ASX; CGQ: company’s ranking in the Horwath 2002 report re-scaled to range from 0 to 100, where companies with higher ranks are reported to have higher quality corporate governance structures. In our initial investigations we have opted for a set of relatively simple, parsimonious, single-equation models. Nonetheless we include analyst disagreement and analyst following as explanatory variables in the bias and 468

accuracy models, and analyst following in the analyst disagreement model.20 Each model includes the CGQ variable, the specific predictions being that higher CGQ is associated with a higher analyst following, with less biased and more accurate forecasts, and with less disagreement among the analysts. 5. Results (i) Properties of the Sample Descriptive statistics for the main variables are reported in Table 1 (document count and timeliness models) and Table 2 (analysts’ forecasts); bivariate relationships are reported in Table 3 (document count and timeliness) and Table 4 (analysts’ forecasts). The sample of analysts’ forecasts is restricted to 1,571 cases where at least four analysts’ forecasts were embodied in the I/B/E/S summary measures for 1 to 11 months before the month in which the firm released its 2002 PFS to the ASX, and where similar I/B/E/S summary data were available for the same forecast horizon and the previous financial year. The sample of timeliness metrics and document frequency counts is restricted to 1,226 cases where we could calculate both measures over the 12 months leading up to a PFS release date and where other required data were available. Table 1 Descriptive Statistics for the Variables in the Document Count and Timeliness Models

469

Notes: The sample (N = 1,226) is constructed from the set of 250 Australian firms rated in the 2002 Horwath Corporate Governance Report. All documents is the annual total number of documents filed with the ASX for release to the market over the year, PS documents is the number that were judged by the ASX to be price-sensitive, and log PS docs denotes the natural logarithm (log) of the number of PS documents. Size is proxied by the log of the firm’s market capitalisation (in AUD million) as at June 2002. The good news dummy has a value of one if the market-adjusted return over the 250 trading days ended 10 days after the release of the firm’s preliminary final statement (PFS, which is when the firm first announces its financial results for the year) is positive and is otherwise zero. The timeliness metric is the average daily absolute difference between the log of share price that day and the log of share price 10 days after the release of the firm’s PFS; the deflated timeliness metric is the timeliness metric divided by one plus the absolute rate of return on the share over the 250-day period used to calculate the share’s timeliness metric. Resource dummy has the value of one if the firm is classified as operating in the natural resource sector and is otherwise zero. CGQ is the company’s ranking in the Horwath 2002 report re-scaled to range between 0 and 100, where companies with higher ranks are

470

reported to have higher quality corporate governance structures. Table 2 Descriptive Statistics for the Variables in the Models Used to Explain Properties of Analysts’ Forecasts of Annual EPS

Notes: The variables are based on models specified by Brown, Lee, Taylor and Walter (2001), with the addition of a CGQ measure. The sample (N = 1,571) is constructed from the set of Australian firms with I/B/E/S annual EPS forecasts, for the 2002 financial year, for a forecast horizon (i.e., the lag in months from the forecast date to the EPS announcement date as reported by I/B/E/S) of 1 to 11 months, and where at least 4 analysts contributed their forecasts to the summary file, The sample firms were also rated in the 2002 Horwath Corporate Governance Report. Forecast error (FE) is defined as the mean forecast EPS less EPS as reported by I/B/E/S, and

471

is deflated by share price a year before the announcement month. Bias is the signed FE, accuracy is its absolute value and disagreement is the standard deviation across analysts’ forecasts for that firm-month, also deflated by price. Size is proxied by the natural log of the firm’s market capitalisation (in AUD million) as reported by I/B/E/S on the forecast cut-off date. Last year’s FE is for the same firm and same forecast horizon, volatility is calculated from daily returns in the 60 trading days ended the day before the I/B/E/S forecast date and dummy variables are used to designate firms in certain industries (natural resources, finance and property stocks) and firms traded on the options market or that the ASX permits to be sold short. CGQ is the company’s ranking in the Horwath 2002 report re-scaled to range from 0 to 100, where companies with higher ranks are reported to have higher quality corporate governance structures. Table 3 Bivariate Relationships Among the Variables in the Document Count and Timeliness Models (N = 1,222; see for the definitions of the variables)

472

Table 4 Bivariate Relationships Among the Variables in the Analyst Forecast Models (N = 1,571; see for the definitions of the variables)

473

Companies in the sample released a minimum of 1 and a maximum of 788 documents over the 250 days, the median being a rate of 1 document a week (Table 1), About 1 in 5 documents was judged by the ASX to be price-sensitive, the maximum number in any one year being 203 (BHP Billiton),21 The raw timeliness metric ranged from 0.017 to 1.789 and from 0.017 to 0.961 when scaled by one plus the absolute value of the share’s relative price movement. Market capitalisation as of mid-2002 ranged from AUD 22.5 million (Aurora Gold) to AUD 53.8 billion (National Australia Bank). About half the companies lost ground relative to the market in the sample period (mean good news dummy is 0.50, median is 0). The CGQ measure, which ranges between 0 and 100, is slightly asymmetric as both the mean and median are about 55. Analysts’ forecasts were biased upwards relative to actual EPS, with the average forecast error being 5.66% of the base price, which is the share price 12 months before the actual EPS was announced (Table 2). The median forecast error was 0.28% of the base price. Daily share price volatility ranged from less than 1% to 7.5%,

474

the average being 2%. Property shares account for 30% of the sample, finance companies 16% and natural resource firms 13%. Almost 40% of the sample relates to shares traded on the options market and 75% to shares eligible for short selling. Table 3 explores bivariate relationships among variables in the document count and timeliness models. Panel A contains product-moment correlations between the continuous variables, Panel B contains medians and Wilcoxon test statistics when the continuous variables are compared for the two sub-sets of each categorical variable, and Panel C contains the cross-tabulations for the two categorical variables. In brief, Panel A shows the two timeliness metrics are strongly correlated (r = 0.93) with each other and negatively correlated, as expected, with firm size and CGQ. The frequency of price-sensitive announcements is positively correlated with firm size and CGQ and weakly correlated with timeliness. Panel B reveals disclosure frequency is unrelated to whether shareholders received good or bad news over the year. Natural resource firms make more frequent disclosures, as is expected under the ASX’s listing rules. Table 4 explores in a similar fashion bivariate relationships among the variables in the four analysts’ forecast models. There are too many pairs to comment in detail so we refer only to the more salient results. Accuracy and bias are correlated 0.96, suggesting there was a remarkably consistent, strong upward bias in EPS forecasts for fiscal 2002. CGQ is negatively correlated

475

with forecast bias and the size of the forecast error (accuracy), weakly negatively correlated with analysts’ disagreement and positively correlated with analyst following; all correlations have the expected sign. The forecast horizon has little bearing on the properties of analysts’ forecasts of EPS in fiscal 2002. We take up this issue again, after discussing the multivariate results, to which we now turn. (ii) Frequency of Disclosure and Timeliness The single-equation models used to capture the relationship between CGQ and (1) disclosure frequency and (2) timeliness were estimated by OLS. The natural log of the number of price-sensitive disclosures was used as the dependent variable to mitigate the severe skewness in the raw document count. Note that the regressors were transformed to have mean zero and standard deviation one, so that the constant term in the regression is identically the mean of the dependent variable. The regression estimates are reported in Table 5. Table 5 OLS Regression Estimates for the Document Count and Timeliness Models

Notes: 476

See Table 1 for the definitions of the variables. All regressors are normalised, so that the intercept is the mean of the dependent variable and each coefficient indicates the change in the dependent variable predicted for a one-standard deviation change in the regressor, other things equal. White’s heteroscedasticity adjustment is applied when calculating the t-statistics and the t-probability is for a two-tailed test. As expected, better governed firms report more frequently to the market: a one standard deviation increase in CGQ leads to an estimated increase of 0.068 in the log of the number of price-sensitive documents released to the ASX, which is equivalent to about one more document in raw terms. Natural resource firms release on average 2.2 more documents annually than firms in other industries, once we control for firm size, good/bad news and CGQ.22 A one standard deviation increase in (the log of) firm size translates into an average of one-and-a-third more documents released annually. How well the share does relative to the market appears to have little bearing on the frequency of price-sensitive disclosures by the top 250 Australian companies. That is not to deny investment performance might have a lot to do with the frequency of disclosures by say small, start-ups, which are not in our sample, or by firms doing either extremely well or extremely poorly. Timeliness is also related to CGQ. Recall that the smaller the value of the timeliness metric, the faster value-relevant news is incorporated into share price. Table 5 shows value-relevant news is priced more

477

rapidly when the firm has a better corporate governance structure, when it is larger, and when the company outperforms the market. All three effects are of the same order of magnitude. It does not seem to matter whether the firm operates in the natural resource sector, even though these firms are required to report more frequently. The deflated timeliness results, where the raw timeliness metric is scaled by a measure related to volatility, are stronger than the undeflated results but the statistical inferences are much the same. The ASX Corporate Governance Council’s Principle 5 emphasises ‘balance’ in the disclosure of good and bad news. Table 5 shows good news is reflected in share prices sooner. When the product of the good news dummy and CGQ is added to the regressors in Table 5 to capture the interaction effect, its coefficient is insignificant in the document count regression (t-statistic = 0.964) but it is positive and significant in the timeliness regression. The coefficients in the augmented regression are −0.073 (t-statistic = 3.369) for the good news dummy, −0.058 (t-statistic = 4.799) for CGQ and 0.055 (t-statistic = 3.369) for the interaction term. These estimates are consistent with the Council’s thesis that better-governed firms are more balanced in the extent to which good and bad news are reflected in share prices on a timely basis. (iii) Analysts’ Earnings Forecasts: Bias, Accuracy, Disagreement and Analyst Following Four separate, single equation models were estimated by OLS. The results are given in Table 6. The models explain between 40% and 70% of the within-sample

478

variance in the dependent variables and the F-statistics show that the models are, overall, strongly significant. Table 6 OLS Regression Estimates: Models of Bias, Accuracy and Disagreement in Analysts’ Forecasts, and the Number of Analysts Following the Share

Notes: The regression models are based on models specified by Brown, Lee, Taylor and Walter (2001), with the addition of a CGQ measure to the regressors. The sample is constructed from the set of Australian firms with I/B/E/S annual EPS forecasts, for the 2002 financial year, for a forecast horizon (i.e., the lag in months from the forecast date to the EPS announcement date as reported by I/B/E/ S) of 1 to 11 months, and where at least 4 analysts contributed their forecasts to the summary file. The sample firms also were rated in the 2002 Horwath Corporate Governance Report. See Table 2 for the definitions of the variables. All regressors are normalised, so that the intercept is the mean of the dependent variable and each coefficient indicates the change in the dependent variable predicted for a 479

one-standard deviation change in the regressor, other things equal. White’s heteroscedasticity adjustment is applied when calculating the t-statistics and the t-probability is for a two-tailed test. Our predictions with respect to CGQ are confirmed for the bias, accuracy and analyst following regressions in that the coefficients of CGQ are all in the expected direction and statistically significant. Bias diminishes with higher CGQ (coefficient = −0.021, t-statistic = −4.937); accuracy improves with higher CGQ (a one standard deviation increase in the CGQ measure reduces the absolute value of the forecast error by 1.7% of the base price; t-statistic = −3.995); and analyst following increases, albeit somewhat marginally, with CGQ (coefficient = 0.239, t-statistic = 4.34). The bivariate correlations in Table 4 indicate higher CGQ is associated with less analyst disagreement, as predicted. However, the multivariate results for analyst disagreement are inconsistent with our prediction and, assuming they are correct, we are left with the unanswered question, why should analysts disagree more on their forecasts for better-governed firms?23 As already noted, analysts’ forecasts in 2002 were strongly biased upwards, on average, as reflected in the constant term (0.057, t-statistic = 14.398). Bias increased with analyst disagreement, the size of last year’s corresponding forecast error and share price volatility; and it was higher for resource shares and shares traded on the options market. It was lower for shares that could be sold short, and it decreased with firm size, the absolute value of last year’s corresponding forecast error

480

and for better-governed firms. Analyst following, the length of the forecast horizon and whether the firm was in the finance or property industry had no significant effect on forecast bias. Forecast error averaged 6.9% of the base price. It increased in absolute value with analyst disagreement, share price volatility and for shares traded on the options market. The size of the error decreased with firm size, and it was lower for shares that could be sold short and for better-governed firms. The size of the forecast error is only weakly related to last year’s signed forecast error (coefficient = −0.01, t-statistic = −1.824) and it is unrelated to last year’s unsigned forecast error, analyst following, the firm’s industry and the forecast horizon. Analysts disagreed more when last year’s signed forecast error was greater, when the share was more volatile, when it was a natural resource or finance firm, when it could be sold short, when the forecast horizon was longer and when it was better-governed. They disagreed less when the firm had a larger analyst following and when it was larger or a property stock. The absolute value of last year’s forecast error and whether the share was traded on the options market appear not to have been significant factors. Finally, analyst following averaged 9.6 analysts per company. The number increased with firm size, last year’s forecast error, volatility, for firms in the natural resource sector, for shares that were traded on the options market, for shares eligible for short selling, for longer forecast horizons and for better-governed firms.

481

Analyst following was lower where last year’s forecast error was greater in absolute value. 6. Robustness Tests One set of robustness tests involved verifying the statistical inferences by resampling and bootstrapping the data underlying the regressions in Tables 5 and 6. All inferences were robust to these procedures. A second set of tests relates to the analyst forecast models. There are 17 cases in the sample where the absolute value of the EPS forecast error exceeds the share price on the base date. If we truncate (winsorise) these forecast errors so their absolute value equals the base price, no inference with respect to the CGQ variable in any of the four analyst forecast models changes. If instead we censor (delete from the sample) the 17 outliers, the influence and statistical significance of the CGQ variable in the bias, disagreement and analyst following regressions are essentially the same; in the accuracy regression, the coefficient of CGQ retains its negative sign but its significance is weakened (one-tailed p = 0.04). If we extend the time period sampled from 2002 to also include fiscal 2000 and 2001, on the assumption that the CGQ variable is sufficiently intertemporally stable and the models’ parameters are stationary, the censored sample size increases to 4,062 cases and the CGQ variable retains its sign and significance in each of the four regression models.24 Importantly, the coefficient of the forecast horizon is positive (i.e., the forecast error increases in absolute value with the length of the forecast horizon) and highly significant (p < 0.001).

482

The undeflated and deflated timeliness models were further investigated by substituting raw returns for market-adjusted returns, confining releases to the 601 cases relating to the period January 2000 to June 2003 inclusive, adding volatility to the set of regressors,25 and winsorising/censoring the tails of the distributions (top and bottom 2.5% of cases) of the timeliness metric, the number of price-sensitive documents and the market-adjusted return over the 250-day period. Substituting raw returns for market-adjusted returns lowers the undeflated timeliness model’s explanatory power (adjusted R2 = 0.069, down from 0.092) but leaves the CGQ inferences unchanged. Confining the releases to 2000–2003 halves the sample (N = 601), increases explanatory power (adjusted R2 = 0.121) and no inference regarding the regressors is changed. Adding volatility to the regressors increases explanatory power (adjusted R2 =0.170) but again no inference regarding the regressors is changed. Winsorising/censoring the tails does not change the inferences in any substantive way for the overall sample but the combination of censoring outliers, restricting the sample period and adding volatility leaves CGQ with a negative sign but no longer statistically significant (one-tailed p = 0.135). The document count model was similarly investigated for robustness to using the raw document count rather than its natural logarithm, to confining releases to the period January 2000 to December 2003 and to winsorising/censoring the tails of the distributions of the timeliness metric, the number of price-sensitive documents and the market-adjus ted return over the 250-day period. Using the raw document count

483

(estimated for both Poisson and Negative Binomial models) does not change any of the inferences; for example, the t-statistic for CGQ is 4.201 for the Poisson model and 3.125 for the Negative Binomial, compared with 3.455 in Table 5. Winsorising/censoring the tails does not change the inferences in any substantive way for the overall sample but the combination of restricting the sample period and censoring renders the CGQ variable insignificant (one-tailed p = 0.27). The significance of the other explanatory variables (size, good news dummy and natural resource sector dummy) is unaffected by any of these changes. Our final set of robustness tests explores alternative proxies for firm size and the effect of adding financial leverage as a control variable.26 The variables investigated were: for size, log total assets, log total revenue (both sourced from Aspect Financial) and the firm’s size rank according to the Horwath 2002 report; and leverage based on book values of debt and equity (also sourced from Aspect Financial). Our findings can be summarised as follows. For the document count regression, the significance of the CGQ variable is robust to substituting log assets (CGQ t = 3.828) or log total revenue (CGQ t = 5.422) for log MVE, and when leverage is added as a control variable (CGQ t = 3.792). Similarly, for the timeliness regression, the significance of CGQ is robust to substituting log total assets (CGQ t = −3.476) or log total revenue (CGQ t = −3.419), and when leverage is added as a control (CGQ t = −4.32). For the analyst forecast bias regression, the significance of CGQ is robust to the size variable being log total assets (CGQ t = −2.188), log total revenue (CGQ t =

484

−2.355) or the Horwath size rank variable (CGQ t = −3.523), and to adding leverage as a control (CGQ t = −1.972). For the analyst accuracy regression, the significance of CGQ is robust to log total revenue (CGQ t = −1.679) and the Horwath size rank (CGQ t = −3.903) but it is not robust to the size variable being log total assets (CGQ t = −0.685) or to adding leverage as a control (CGQ t = −0.043). For the analyst disagreement regression, the significance of CGQ is robust to the size variable being log total assets (CGQ t = 4.564), log total revenue (CGQ t = 3.872) or the Horwath size rank variable (CGQ t = 2.994), and to adding leverage as a control (CGQ t = 4.986). Lastly, for the analyst following regression, CGQ remains significant when the size variable is log total assets (CGQ t = 4.147), log total revenue (CGQ t = 3.895) or the Horwath size rank variable (CGQ t = 6.687), and when leverage is added as a control variable (CGQ t = 4.363). In summary, there is some evidence that the smaller sample sizes observed in shorter time periods can affect the statistical reliability of the disclosure frequency and timeliness models. There is also evidence that different proxies for firm size, and the addition of a control variable for leverage, likewise can affect the results for one of the analyst forecast models. However, our findings are, on the whole, robust to the matters investigated. 7. Conclusions Following the recent interest in corporate governance post-Enron and the introduction of corporate governance guidelines in Australia, we test the thesis that

485

better-governed firms make more informative disclosures. To do this, we consider the number of price-sensitive documents released to the share market, four properties of analysts’ EPS forecasts, and the speed with which share price reflects the net effect of value-relevant information impounded in share price over the year. We dub this last measure ‘timeliness’, after Ball and Brown (1968). In brief, we find better-governed Australian firms do make more informative disclosures. Five, specific predictions were supported: better-governed firms make more price-sensitive disclosures, they have a larger analyst following, analysts’ consensus forecasts for better-governed firms are less biased and more accurate and, presumably as a consequence, value-relevant information about better-governed firms is more timely in the sense that price discovery is faster.27 The one case where the results were contrary to expectation is where analysts appear to disagree more about the future earnings of better-governed firms. Our timeliness metric requires further investigation, which we are doing in ongoing work. Endogeneity issues are also being investigated. In addition, our study is being replicated for other countries, including the UK, Canada and the US.28

Appendix A The ASX’s ‘Best Practice Recommendations’

486

1.1 Formalise and disclose the functions reserved to the board and those delegated to management. 2.1 A majority of the board should be independent directors. 2.2 The chairperson should be an independent director. 2.3 The roles of chairperson and chief executive officer should not be exercised by the same individual. 2.4 The board should establish a nomination committee. 2.5 Provide the information indicated in Guide to reporting on Principle 2. 3.1 Establish a code of conduct to guide the directors, the chief executive officer (or equivalent), the chief financial officer (or equivalent) and any other key executives as to: 3.1.1 the practices necessary to maintain confidence in the company’s integrity 3.1.2 the responsibility and accountability of individuals for reporting and investigating reports of unethical practices. 3.2 Disclose the policy concerning trading in company securities by directors, officers and employees. 3.3 Provide the information indicated in Guide to reporting on Principle 3. 4.1 Require the chief executive officer (or equivalent) and the chief financial officer (or equivalent) to state in writing to the board that the company’s financial reports present a true and fair view, in all material respects, of 487

the company’s financial condition and operational results and are in accordance with relevant accounting standards. 4.2 The board should establish an audit committee. 4.3 Structure the audit committee so that it consists of: • only non-executive directors • a majority of independent directors • an independent chairperson, who is not chairperson of the board • at least three members. 4.4 The audit committee should have a formal charter. 4.5 Provide the information indicated in Guide to reporting on Principle 4. 5.1 Establish written policies and procedures designed to ensure compliance with ASX Listing Rule disclosure requirements and to ensure accountability at a senior management level for that compliance. 5.2 Provide the information indicated in Guide to reporting on Principle 5. 6.1 Design and disclose a communications strategy to promote effective communication with shareholders and encourage effective participation at general meetings. 6.2 Request the external auditor to attend the annual general meeting and be available to answer shareholder questions about the conduct of the audit and the preparation and content of the auditor’s report.

488

7.1 The board or appropriate board committee should establish policies on risk oversight and management. 7.2 The chief executive officer (or equivalent) and the chief financial officer (or equivalent) should state to the board in writing that: 7.2.1 the statement given in accordance with best practice recommendation 4.1 (the integrity of financial statements) is founded on a sound system of risk management and internal compliance and control which implements the policies adopted by the board 7.2.2 the company’s risk management and internal compliance and control system is operating efficiently and effectively in all material respects. 7.3 Provide the information indicated in Guide to reporting on Principle 7. 8.1 Disclose the process for performance evaluation of the board, its committees and individual directors, and key executives. 9.1 Provide disclosure in relation to the company’s remuneration policies to enable investors to understand (i) the costs and benefits of those policies and (ii) the link between remuneration paid to directors and key executives and corporate performance. 9.2 The board committee.

should

establish

a

remuneration

9.3 Clearly distinguish the structure of non-executive directors’ remuneration from that of executives.

489

9.4 Ensure that payment of equity-based executive remuneration is made in accordance with thresholds set in plans approved by shareholders. 9.5 Provide the information indicated in Guide to reporting on Principle 9. 10.1 Establish and disclose a code of conduct to guide compliance with legal and other obligations to legitimate stakeholders. Appendix B Extracts from the 2004 Horwath Report The following extracts explain the basis of the Horwath corporate governance ratings. (M)any of the factors considered in the Horwath Report are similar to the ASX Guidelines. (2004 Horwath Report, p. 9). The model considers objective factors based on publicly disclosed information pertaining to the existence and structure of a company’s Board of Directors and associated committees, the level of perceived independence of the company from the external auditors, and disclosures relating to the existence of a code of conduct, risk management and share trading policy (2004 Horwath Report, p. 25). Consistent with the requirements of the ASX Corporate Governance Council it was considered worthwhile to expand the analysis (in 2004) to consider whether companies had a formal code of

490

conduct (recommendation 3.1) and whether they had rigorous and documented policies on risk management (recommendation 7.1) and share trading (recommendation 3.2). These factors were not considered in the 2003 or 2002 Horwath Reports (2004 Horwath Report, p. 53). (A)n independent director is defined as someone who is not a member of management and who: • Is not a substantial shareholder of the company, or otherwise associated directly or indirectly with a substantial shareholder of the company • Has not been employed in an executive capacity by the company in the last decade • Is not an original founder of the company • Is not a principal of a professional adviser to the company • Is not a significant supplier or customer of the company, or otherwise associated directly or indirectly with a significant supplier or customer of the company • Has no significant contractual relationship with the company, outside of their directorship • Is free from any interest or relationship, which could, or could reasonably be perceived to, materially interfere with the director’s ability to act in the best interests of the company • Has been a director for ten years or less (2004 Horwath Report, p. 24). References

491

Ajinkya, B., S. Bhoraj and P. Sengupta (2005), ‘The Association between Outside Directors, Institutional Investors and the Properties of Management Earnings Forecasts’, Journal of Accounting Research, Vol. 43, No. 3 (June), pp. 343–376. Alford, A., J. Jones, R. Leftwich and M. Zmijewski (1993), ‘The Relative Informativeness of Accounting Disclosures in Different Countries’, Journal of Accounting Research, Vol. 31, No. 3 (Supplement), pp. 183–223. ASX Corporate Governance Council (2003), Principles of Good Corporate Governance and Best Practice Recommendations (Sydney: Australian Stock Exchange). Ball, R. and P. Brown (1968), ‘An Empirical Evaluation of Accounting Income Numbers’, Journal of Accounting Research, Vol. 6, No. 2 (Autumn), pp. 159–78. Basu, S. (1997), ‘The Conservatism Principle and the Asymmetric Timeliness of Earnings’, Journal of Accounting and Economics, Vol. 24, No. 1 (December), pp. 3–37. Basu, S., L. Hwang and C.L. Jan (2001a), ‘Auditor Conservatism and Quarterly Earnings’, Working Paper (Baruch College, City University of New York). Basu, S., L. Hwang and C.L. Jan (2001b), ‘Differences in Conservatism Between Big Eight and Non-Big Eight Auditors’, Working Paper (Baruch College, City University of New York). Becker, C.L. M.L. Defond, J. Jiambalvo and K. Subramanyam (1998), ‘The Effect of Audit Quality on 492

Earnings Management’, Contemporary Accounting Research, Vol. 15, No. 1 (Spring), pp. 1–24. Beekes, W. P. Pope and S. Young (2004), ‘The Link Between Earnings Timeliness, Earnings Conservatism and Board Composition: Evidence From the UK’, Corporate Governance: An International Review, Vol. 12, No. 1 (January), pp. 47–59. Brown, L.D. and M.L. Caylor (2004), ‘Corporate Governance and Firm Performance’, Working Paper (Georgia State University). Brown, P., C. Forbes and M. Wee (2004), ‘Liquidity Differences Surrounding Announcements Classified by ASX According to their Market Sensitivity’, 2004 AFAANZ Conference Proceedings (Accounting and Finance Association of Australia and New Zealand, CD-ROM). Brown, P., B. Howitt and M. Wee (2005), ‘Order Flow and Price Effects Surrounding an ASX Announcement’, 2005 AFAANZ Conference Proceedings (Accounting and Finance Association of Australia and New Zealand, CD-ROM). Brown, P., S. Taylor and T. Walter (1999), ‘The Impact of Statutory Sanctions on the Level and Information Content of Voluntary Corporate Disclosure’, Abacus, Vol. 35, No. 2 (June), pp. 138–62. Brown, P., P. Lee, S. Taylor and T. Walter (2001), ‘Are Security Analysts “Walked Down” to Beatable Forecasts?’ Working Paper (University of Western Australia).

493

Bushee, B. and C. Noe (2000), ‘Corporate Disclosure Practices, Institutional Investors, and Stock Return Volatility’, Journal of Accounting Research, Vol. 38, No. 3 (Supplement), pp. 171–202. Bushman, R., Q. Chen, E. Engel and A. Smith (2004), ‘Financial Accounting Information, Organizational Complexity and Corporate Governance Systems’, Journal of Accounting and Economics, Vol. 37, No. 2 (June), pp. 167–201. Butler, M. A. Kraft and I. Weiss (2003), ‘The Effect of Reporting Frequency on the Timeliness of Earnings: The Cases of Voluntary and Mandatory Interim Reports’, Working Paper (University of Rochester). Cadbury Report (1992), Report of the Committee on the Financial Aspects of Corporate Governance (London: Gee). Chung, R., M. Firth and J-B. Kim (2003), ‘Auditor Conservatism and Reported Earnings’, Accounting and Business Research, Vol. 33, No. 1, pp. 19–32. Collett, P. and S. Hrasky (2005), ‘Voluntary Disclosure of Corporate Governance Practices by Listed Australian Companies’, Corporate Governance: An International Review, Vol. 13, No. 2 (March), pp. 188–96. Core, J. E. (2001), ‘A Review of the Empirical Disclosure Literature: Discussion’, Journal of Accounting and Economics, Vol. 31, Nos. 1–3 (September), pp. 441–56. Core, J. E., W.R. Guay and T.O. Rusticus (2006), ‘Does Weak Governance Cause Weak Stock Returns? An 494

Examination of Firm Operating Performance and Investors’ Expectations’, Journal of Finance, Vol. LX1, No. 2 (April), pp. 655–87. Coulton, J. C. James and S. Taylor (2001), ‘The Effect of Compensation Design and Corporate Governance on the Transparency of CEO Compensation Disclosures’, Working Paper (University of Technology, Sydney). Cremers, K.J. and V.B. Nair (2005), ‘Governance Mechanisms and Equity Prices’, Journal of Finance, Vol. LX, No. 6 (December), pp. 2859–94. Daley, L.A., D.W. Senkow and R.L. Vigeland (1988), ‘Analysts’ Forecasts, Earnings Variability, and Option Pricing: Empirical Evidence’, The Accounting Review, Vol. 63, No. 4 (October), pp. 563–85. Defond, M. and K. Subramanyam (1998), ‘Auditor Changes and Discretionary Accruals’, Journal of Accounting and Economics, Vol. 25, No. 1 (February), pp. 35–67. Defond, M., R.H. Hann and X. Hu (2005), ‘Does the Market Value Financial Expertise on Audit Committees of Boards of Directors’, Journal of Accounting Research, Vol. 43, No. 2 (May), pp. 153–93. Dtrich, J.R., K.A. Muller III and E.J. Riedl (2004), ‘Asymmetric Timeliness Tests of Accounting Conservatism’, Working Paper (Ohio State University). Eng, L.L. and Y.T. Mak (2003), ‘Corporate Governance and Voluntary Disclosure’, Journal of Accounting and Public Policy, Vol. 22, No. 4 (July/August), pp. 325–45.

495

Fama, E. and M. Jensen (1983), ‘Separation of Ownership and Control’, Journal of Law and Economics, Vol. 26, No. 2 (June), pp. 301–25. Givoly, D., C. Hayn and A. Natarajan (2004), ‘Measuring Reporting Conservatism’, Working Paper (http://www.papers.ssrn.com/paper. taf?abstract_id=606841). Gompers, P., J. Ishii and A. Metrick (2003), ‘Corporate Governance and Equity Prices’, The Quarterly Journal of Economics, Vol. 118, No. 1 (February), pp. 107–55. Healy, P., A. Hutton and K. Palepu (1999), ‘Stock Performance and Intermediation Changes Surrounding Sustained Increases in Disclosure’, Contemporary Accounting Research, Vol. 16, No. 3 (Fall), pp. 485–520. Horwath (2002) Corporate Governance Report (Sydney, Australia: Horwath (NSW) Pty Limited). Horwath (2003) Corporate Governance Report (Sydney, Australia: Horwath (NSW) Pty Limited). Horwath (2004) Corporate Governance Report (Sydney, Australia: Horwath (NSW) Pty Limited). Kim, J-B., R. Chung and M. Firth (2003), ‘Auditor Conservatism, Asymmetric Monitoring, and Earnings Management’, Contemporary Accounting Research, Vol. 20, No. 2 (Summer), pp. 323–59. Klock, M.S., S.A. Mansi and W.F. Maxwell (2004), ‘Does Corporate Governance Matter to Bondholders?’

496

Journal of Financial (forthcoming).

and

Quantitative

Analysis

Lang, M. and R. Lundholm (1993), ‘Cross-sectional Determinants of Analyst Ratings of Corporate Disclosures’, Journal of Accounting Research, Vol. 31, No. 2 (Autumn), pp. 246–71. Lang, M. and R. Lundholm (1996), ‘Corporate Disclosure Policy and Analyst Behavior’, The Accounting Review, Vol. 71, No. 4 (October), pp. 467–92. Larcker, D., S. Richardson and I. Tuna (2005), ‘How Important is Corporate Governance?’ Working Paper (http://ssrn.com/abstract=595821). Leung, S. and B. Horwitz (2004), ‘Director Ownership and Voluntary Segment Disclosure: Hong Kong Evidence’, Journal of International Financial Management and Accounting, Vol. 15, No. 3 (Autumn), pp. 235–60. Linden, P. and Z. Matolcsy (2004), ‘Corporate Governance Scoring Systems: What Do They Tell Us?’, Australian Accounting Review, Vol. 14, No. 1 (March), pp. 9–16. Matolcsy, Z., D. Stokes and A. Wright (2004), ‘Do Independent Directors Add Value?’, Australian Accounting Review, Vol. 14, No. 1 (March), pp. 33–40. Noe, C. (1999), ‘Voluntary Disclosures and Insider Transactions’, Journal of Accounting and Economics, Vol. 27, No. 3 (July), pp. 305–26.

497

Shleifer, A. and R. Vishny (1997), ‘A Survey of Corporate Governance’, Journal of Finance, Vol. 52, No. 2 (June), pp. 737–83. Skinner, D. (1994), ‘Why Firms Voluntarily Disclose Bad News’, Journal of Accounting Research, Vol. 32, No. 1 (Spring), pp. 38–60. Skinner, D. (1997), ‘Earnings Disclosures and Stockholder Lawsuits’, Journal of Accounting and Economics, Vol. 23, No. 3 (November), pp. 249–82. Verrecchia, R. (1983), ‘Discretionary Disclosure’, Journal of Accounting and Economics, Vol. 5, pp. 179–94. Wright, D. (1996), ‘Evidence on the Relation Between Corporate Governance Characteristics and Quality of Financial Reporting’, Working Paper (University of Michigan). * The authors are respectively from Lancaster University; and the Universities of New South Wales and Western Australia. They gratefully acknowledge the comments of Jeff Coulton, Tom Smith, Stephen Taylor, Martin Walker, Steve Young and an anonymous referee, as well as participants in research workshops at ANU, Lancaster, Macquarie, Monash and UNSW. The paper has also benefited from comments received at the JBFA Conference in May 2005, and at the 2005 annual meetings of the American Accounting Association, the Accounting and Finance Association of Australia and New Zealand, and the European Accounting

498

Association. The authors are especially indebted to Jim Psaros for providing corporate governance ratings data. Address for correspondence: Wendy Beekes, Department of Accounting and Finance, The Management School, Lancaster University, Lancaster LAI 4YX, UK. e-mail: [email protected] 1 Our measure of corporate governance ‘quality’ increases as additional, common corporate governance standards are met. For convenience, we describe firms with higher ratings as firms with higher corporate governance ‘quality’. 2 ASX Corporate Governance Council (2003). 3 The case is strengthened by noting first that security analysts are represented on and support the activities of the ASX Corporate Governance Council and second that portfolio managers, who buy analysts’ services, are not disinterested in corporate governance matters. 4 Big five auditees are also found to engage in less earnings management and are generally considered to report ‘higher quality’ accounting numbers (Becker et al, 1998; and Defond and Subramanyam, 1998). However, the big-five auditors’ attention has been found to centre on constraining income-increasing earnings management activity (Kim et al., 2003), possibly because of asymmetric litigation risk. 5 ASX Corporate Governance Council, March 2003, Principles of Good Corporate Governance and Best Practice Recommendations, Frequently Asked Questions, p. 7.

499

6 For further discussion of the development of corporate governance guidance and disclosure requirements in Australia, see Collett and Hrasky (2005). 7 Specifically, Bushman et al. (2004) adopt the view that where accounting does not reflect ‘news’ on a timely basis, such as in intangible rich firms, higher CGQ would be employed by a firm to mitigate agency costs. This would suggest that there is an endogenously determined trade off between the use of governance mechanisms and the timeliness of information flows. 8 The report is copyright by the University of Newcastle and is based on research by Jim Psaros and Michael Seamer at the University of Newcastle Business School. 9 Linden and Matolcsy (2004) report that 73 companies in the 2002 Horwath Report had a rating of 1–2 stars, 92 companies were rated 2.5–3.5 stars; and 85 companies 4–5 stars. 10 To the extent that the Horwath measure excludes important indicators of CGQ, our tests are biased against rejecting the null hypothesis. We leave the development of a more comprehensive measure of CGQ to future research. 11 Relative to other countries, Australian companies make extensive disclosures in their PFSs. For example, they include standard form Income Statements, Balance Sheets, Statements of Cash Flows, dividend announcements and details of any completed or planned capital raisings.

500

12 See Section 6 for further discussion on the sample of PFSs. 13 In other words, we make no allowance for differences in firms’ systematic risk. 14 The expression on the R.H.S. of equation (2) can be simplified but others have found the measure easier to understand in this form. The end points of the summation are somewhat arbitrary; for example, we could end the summation at the close of trading on day t = −1 (recall that day t = 0 is 10 trading days after the announcement). 15 To cater for this effect we investigate a model where the timeliness metric is deflated by one plus the absolute rate of return on the share over the 250-day period used to calculate the timeliness metric. We also control for the firm’s volatility in the timeliness regression. See Section 6. 16 For the document count and timeliness regressions, size is measured by log MVE, where MVE is the firm’s market capitalisation, which was sourced from the Australian Graduate School of Management’s share price-price relative file. For the analyst forecast regressions, size is MVE as reported by I/B/E/S – see below. 17 We reach the same conclusions if we use the median rather than the mean forecast. 18 The regression models are based on those in Brown et al. (1999) and Brown et al. (2001).

501

19 Deciding whether to include volatility as a control variable is somewhat problematic, because volatility is an indicator of disclosure informativeness (Lang and Lundholm, 1993; and Brown et al., 1999); which, in the context of our study, means it potentially proxies for CGQ. We find CGQ is a statistically significant determinant of bias, accuracy, disagreement and analyst following when volatility is included among the regressors. Volatility can also reflect aggressive trading by institutions attracted to firms that have improved their disclosure rankings (Bushee and Noe, 2000), which we see as an outcome of higher CGQ. 20 The significance of CGQ does not turn on the inclusion of these variables. 21 The minimum number of price-sensitive disclosures was one (two cases), which indicates a deficiency in our data source. All listed companies must make at least two price-sensitive disclosures each financial year, namely their half-yearly reports and PFSs. 22 Calculated as e(0.303/0.375), where the within-sample standard deviation of the dummy variable is 0.З75 and its regression coefficient 0.303. This estimate is in line with the ASX listing requirement of quarterly reporting for companies in this sector and half-yearly reporting for others. 23 While this result is contrary to our expectation, prior evidence (Daley et al., 1988) suggests more disclosure can result in greater analyst disagreement. 24 The coefficients of CGQ and their corresponding two-tailed p-values are −0.555E-02 (p 1; after that point, operating cashflow decreases at an increasing rate. For instance, if the discounted rate of return on goodwill is 10% per annum (ie, p = 0.10), then operating cashflow attributable to goodwill increases over the first 10 years, it is the same in years 10 and 11, and after year 11 it declines at a rate increasing in t. The higher the rate of return, the earlier will be the point where cashflow starts declining. We might also note that the annual amortisation charge under the ISOYD method is less than under the straight line method until after the mid-point of the amortisation period. Of course, the carrying amount of goodwill is always greater under the ISOYD method once amortisation begins. F.H. Faulding & Company’s Method F.H. Faulding & Company’s notes to its 1994 annual accounts state that the amount of goodwill amortisation increases by 8% each year. This method, which is 761

unusual, is said to reflect “increasing profitability over the longer term”. I interpret increasing profitability to mean increasing operating cashflows from the investment. Let’s assume a new business is acquired and goodwill is recognised at the beginning of the first financial year. Then a more general statement of Faulding’s accounting rule is the simple relation At=(1 + r)At−1, 1< t≤N, where r is the rate of increase in the amortisation charge. Specifically, r = 8%, according to Faulding’s 1994 annual accounts. It then follows that the RHS of equation (4) is just (r − ρ) A 1(1 + r)t−2 /G0. This expression looks complex but it is not, really. Two things stand out. First, whether Faulding’s rule is consistent with growing, flat or declining expected future benefits depends on whether the annual rate of increase in amortisation r exceeds, equals or is less than the yield p to Faulding on its investment in goodwill. One can hardly imagine that Faulding often contemplates a major acquisition where the discounted rate of return on goodwill is expected to be less than 8% pa. Yet that’s what our model implies would have to be the case if Faulding’s method were to be consistent with increasing operating cashflows attributable to goodwill. Second, the absolute difference in operating cashflows from year to year increases over time under Faulding’s method, except for the pathological case where r = p. Figure 1 summarises the pattern of expected future benefits over time for each of the three amortisation rules considered. To generate Figure 1, acquired goodwill is assumed to cost $1 million and yield 10%

762

over 20 years. For Faulding’s method, amortisation is assumed to grow at 8% annually. Figure 1 captures most of the intertemporal cashflow relationships considered so far, including the downwards concavity of cashflows under the ISOYD method and the monotonically declining operating cashflows under the straight line method and under Faulding’s method. FIGURE 1: COMPARISON OF ANNUAL OPERATING CASH FLOWS FOR THREE AMORITISATION METHODS; AMORTISATION OF $1 MILLION GOODWILL YIELDING 10% PA OVER 20 YEARS.

Figure 2 summarises the three methods’ respective goodwill carrying amounts over time, under the same assumptions of a $1 million investment that yields 10% pa and is amortised over a 20-year life. Clearly, the straight line method is the most conservative of the three, the ISOYD method is the least conservative, and Faulding’s method lies somewhere in between. FIGURE 2: COMPARISON OF CARRYING AMOUNTS FOR THREE AMORITISATION METHODS; AMORTISATION OF $1 MILLION GOODWILL YIELDING 10% PA OVER 20 YEARS.

763

To gain some additional perspective on the carrying amounts, we can calculate goodwill’s “half-life” under each method; ie, how long it takes to amortise half of the goodwill’s original cost. Obviously it takes 10 years under the straight line method; it takes 14 years under the ISOYD method and it also takes 14 years under Faulding’s method. Under US accounting principles (APB17, paragraph 29) goodwill may be amortised over at most 40 years, which for the straight line method gives it a maximum half-life of 20 years in the US. Goodwill in Practice Most Australian companies amortise goodwill using the straight line method. A few have used the ISOYD method at one time or another to amortise at least part of their goodwill. For instance, 1994 users of the ISOYD method include Amcor, Brambles Industries, Email, Mayne Nickless, National Foods, Pacific Dunlop and Southcorp Holdings. The 1994 users list excludes Foodland Associated and Gibson Chemical Industries, both of whom discontinued using the ISOYD in 1994 and switched to straight line for all of their goodwill amortisation. Foodland Associated reported that the switch to straight line reduced its 1994 profit by $829,000 while Gibson Chemical described the effect on

764

its bottom line as “not material”. In contrast, Email first adopted the ISOYD method in its 1993 accounts. Previously (“in poor years”, according to Email’s 1993 accounts) it had used straight line. We can use data from the profit and loss account, balance sheet and cashflow statement to approximate the effect of a change in a company’s goodwill amortisation method on its financial profile, but it is difficult to know precisely what the effect would be unless the company discloses enough information directly. Both Southcorp Holdings and Pacific Dunlop provided information in their 1994 annual accounts that allows us to compare the impact on profit if they, too, were to abandon ISOYD and switch to straight line. Southcorp Holdings simply reported that the switch would increase the amortisation charge and take $3.277 million off its $49.808 million bottom line. Pacific Dunlop’s disclosure was more detailed. Pacific Dunlop’s quarrel with the goodwill accounting standard dates back to before ASRB 1013 was introduced. ASRB 1013 first applied to Pacific Dunlop’s 1988 accounts. However, Pacific Dunlop wrote off goodwill in 1988 after the National Companies and Securities Commission (forerunner to the Australian Securities Commission) issued an order granting Pacific Dunlop’s application to depart from the standard that year. Its 1989 accounts disclosed a change in goodwill accounting policy, from direct write-off as an extraordinary item to capitalisation and amortisation using the “Inverted Sum of the Years Digits Method” (that is, the ISOYD). The accounting change reduced

765

1989’s consolidated operating profit by $748,000 but cut the bottom-line loss by $384 million. The flow-on effect was to boost shareholders’ funds by the same $384 million. Consequently, Pacific Dunlop took a “double hit” on a key performance ratio, operating profit to shareholders’ funds. That ratio dropped by about a third, from 37% before the policy change to 24% after it. In 1992, goodwill was written down by $96 million at the same time as brandnames worth $136 million were reinstated on the balance sheet after previously having been written off. The net figure, $40 million, was treated as an abnormal gain in Pacific Dunlop’s profit-and-loos account. Following the ASC’s initial stoush with Pacific Dunlop over its use of the ISOYD method, the 1994 accounts explored the bottom-line effect of goodwill accounting in more detail than any similar disclosure by an Australian company of which I am aware. Footnote 14 disclosed total capitalised goodwill of $706 million. The amortisation charge reported under the ISOYD method was $24 million; it would have been $38 million had Pacific Dunlop adopted straight line over 20 years ($19 million over 40 years, the maximum permitted under US and Canadian standards), and nil if there was a direct write-off as allowed under UK and German standards. As a benchmark, Pacific Dunlop’s 1994 bottom line was $303 million. Conclusion The model we have set out is quite general. It can be used to identify the operating cashflow pattern implicit

766

in any arbitrary cost allocation scheme. The inverse sum-of-the-years’-digits method, which has attracted much public attention and is used by few, is consistent with an acquisition’s expected future benefits growing at a decreasing rate for the first few years, levelling off, then declining at an increasing rate as the end of the amortisation period approaches. The straight line method implicitly assumes that expected future benefits decline throughout the amortisation period by a constant absolute amount each year, while “Faulding’s method” typically assumes expected future benefits either increase or decrease at an accelerating rate. Constant expected future benefits are a special case of Faulding’s method, which is then identical to the annuity method. Our model assumes that goodwill is the firm’s sole asset. In essence, we assume that the tangible and identifiable intangible assets that are recognised on the balance sheet along with goodwill are also planned to be depreciated, amortised, or even revalued upwards each year in such a way that their accounting rate of return is p. Also, our analysis has been strictly in terms of a problem in initial cost allocation. There is always the requirement that the carrying amount of goodwill be reviewed annually and written down “to the extent that future benefits are no longer probable” (AASB 1013, clause .36). If goodwill is amortised by the ISOYD rather than the straight line method, then, other things being equal, annual profit is greater until after the mid-point of the amortisation period. Strictly speaking “other things being equal” is not in the spirit of our theoretical model,

767

which presupposes that the choice of cost allocation method is determined by the pattern of expected future benefits coupled with the requirement that the accounting rate of return be constant over the asset’s life. In other words, the underlying pattern of operating cashflows is different for each method. In practice, a few companies have reported the difference a switch in method either did or would make to the bottom line. For those that have, the difference ranges from “immaterial” to a reduction of 4.6% if Pacific Dunlop were to switch from ISOYD to straight line in 1994 and 6.6% for the same switch by Southcorp Holdings. While goodwill remains on the balance sheet, its carrying amount is more conservative under the straight line method than under either of the other two methods we analysed. Indeed, the difference in goodwill’s carrying amount under the ISOYD and straight line methods reaches its maximum at the mid-point of the amortisation period. Direct write-off would be the most conservative method of all, but it is not allowed under current Australian standards. Philip Brown is KPMG Professor of Accounting at The University of Western Australia. The views expressed in this article do not necessarily represent the views of KPMG. The author is indebted to Greg Clinch and Malcolm Miller for their helpful comments. References Accounting Principles Board, 1970, “APB Opinion No. 17—Intangible Assets”, American Institute of Certified Public Accountants, New York.

768

Australian Accounting Standards Board, 1988, AASB 1013 Accounting for Goodwill. Anton, H., 1956, “Depreciation, cost allocation and investment decisions”, Accounting Research, 7, pp. 117–134. Bierman, H., 1958, “A theory of depreciation consistent with decision-making”, The Controller, 36, pp. 370, 371, 393. Bierman, H., 1961, “Depreciable assets—timing of expense recognition”, Accounting Review, 36, pp. 613–618. Financial Accounting Standards Board, 1990, “Present value-based measurements in accounting”, Discussion Memorandum, 7 December. Lorig, A.N., 1962, “On the logic of decreasing charge depreciation”, Accounting Review, 37, pp. 56–58.

© 1995 by CPA Australia

769

Valuing Executive Stock Options: Performance Hurdles, Early Exercise and Stochastic Volatility Philip Browna, b, Alex Szimayerc a

Australian School of Business, University of New South Wales, Sydney, 2052, Australia b

UWA Business School, University of Western Australia, Crawley, 6009, Australia c

Department of Financial Mathematics, Fraunhofer ITWM, Fraunhofer-Platz 1, 67663 Kaiserslautern, Germany Abstract Accounting standards require companies to assess the fair value of any stock options granted to executives and employees. We develop a model for accurately valuing executive and employee stock options, focusing on performance hurdles, early exercise and uncertain volatility. We apply the model in two case studies and show that properly computed fair values can be significantly lower than traditional Black–Scholes values. We then explore the implications for pay-for-performance sensitivity and the design of effective share-based incentive schemes. We find that performance hurdles can require a much greater fraction of total compensation to be a fixed salary, if pre-existing incentive levels are to be maintained.

770

Key words: Executive stock options; Performance hurdles; Early exercise; Stochastic volatility JEL classification: G13 doi: 10.1111/j.1467-629x.2008.00259.x

1. Introduction Executive and employee stock options (ESOs) are popular and constitute a substantial component of executive pay (see Hall and Murphy, 2002; Frederic W. Cook & Co., 2006). ESOs are more complex than exchange-traded options and raise difficult valuation issues. These issues cannot be brushed aside because under the international accounting standard, Share-Based Payment (IFRS 2; International Accounting Standards Board, 2004), introduced on 1 January 2005 and the basis of Australian accounting standard AASB 2, the fair value of ESOs, measured at their grant date, must be recognized in the employer’s financial statements. Similar requirements have applied in the USA under US Financial Accounting Standards Board’s FAS 123R (Financial Accounting Standards Board, 2004), since 15 June 2005. We provide a framework for accurately valuing ESOs from the employer’s viewpoint, which is required by both IFRS 2 and FAS 123R. Traditional valuation of ESOs relies on Black and Scholes (1973) and mild extensions of it. This framework has inherent shortcomings when valuing ESOs. More accurate valuations will benefit firms and their remuneration advisers when determining

771

appropriate compensation arrangements, and accountants and auditors when assessing the options’ fair values and measuring the amount of compensation expense to be recognized in the financial statements. In particular, we address the effects of performance hurdles, early exercise and stochastic volatility on the ‘fair value’ of an ESO. Performance hurdles can enhance the incentive capacity1 of ESOs but detract from their value (see Johnson and Tian, 2000; Calvet and Rahman, 2006). Performance hurdles increase their transparency from the shareholder’s perspective, because vesting is directly linked to achieving predetermined targets such as growth in earnings-per-share (EPS) or outperformance in terms of total shareholder return (TSR). Performance hurdles are common in Australia and in European countries but much less common in the USA, apparently for an accounting reason.2 We cater for different types of performance hurdle by including either other asset price processes or non-market variables such as accounting measures. ESOs are unlisted securities and typically cannot be sold by executives, but they can be exercised early (i.e. before they mature). The ability to exercise early can have a significant effect on ESO valuation and a large proportion of ESOs are exercised early (see Huddart and Lang, 1996; Bettis et al., 2005; Armstrong et al., 2006; Boyd et al., 2007). Accounting standards require the effect of expected early exercise to be taken into account when applying a pricing model.3 Models

772

allowing for early exercise typically find ESO value is substantially lower (see Cuny and Jorion, 1995; Carr and Linetsky, 2000; Armstrong et al., 2006). We model the effects of early exercise and employee departure by random times that can be driven by market conditions (see Lando, 1998). The future volatility of the underlying stock is an essential input for ESO valuation models. ESOs are typically long-lived, and often with a time to maturity of up to 10 years. Assuming a constant volatility as in Black and Scholes (1973) and most ESO valuation models is potentially a source of valuation error. Stochastic volatility models capture the uncertainty of the future volatility (see Hull and White, 1987; Heston, 1993). Moreover, these models can produce the skewness observed in stock returns and can explain the volatility smile (or smirk) of exchange traded options. This is relevant for ESOs, which are typically at-the-money call options, and is likely to result in a lower value compared to constant volatility models (see Duan and Wei, 2005). Our model incorporates stochastic volatility of asset prices, following Heston (1993). Heston’s approach is appealing because of its simplicity and richness. Our model can be implemented using tree-based methods, as outlined by Nelson and Ramaswamy (1990) and Leisen (2000). Two Australian case studies are used to illustrate the method and elicit key observations on the contracts and their fair values, in comparison with more ‘conventional’ valuations. We then investigate how these

773

factors affect the pay-for-performance relationship and their implications for risk-based incentive contracts. We also explore the extent of bias in Black–Scholes valuations relative to the valuations derived from what we argue is a more appropriate model. This issue is important to accounting researchers investigating the value-relevance of accounting numbers, because the extent of the ESO valuation bias depends on the nature of the contract and varies across firms. For the two case studies we find value reductions of 26 and 38 per cent, respectively, compared to the reported Black–Scholes values. The value reduction caused by the actual performance hurdles is marginal, with values of 2 and 6 per cent, respectively. In response to this observation, that some performance hurdles have been set ‘almost at ground level’, we then specify more effective performance hurdles that reduce ESO value from 14 to 60 per cent. The inclusion of stochastic volatility for a share with a highly variable volatility decreases the ESO value by up to 20 per cent for an out-of-the-money ESO, by 13 per cent if it is at-the-money and by 7 per cent if it is in-the-money. Allowing for employee departure reduces the value by around 26 per cent. Our model can deal with performance hurdles based on the stock’s TSR performance measured against a largish group of companies. In general, this might pose problems of high dimensionality, where traditional valuation approaches are prone to break down. We base our approach on Merton’s intertemporal capital asset pricing model (1990), which relies on classical portfolio selection theory (see Markowitz, 1952; Sharpe, 1964).

774

Furthermore, we assume that the idiosyncratic risk components of the stocks are pairwise independent. This assumption provides a meaningful and parsimonious ‘single index model’ setup that reduces the complexity of the model so that it is numerically tractable. Our results contribute to ESO valuation in practice, especially for small-cap stocks. Typically, these stocks have highly fluctuating volatility; and stochastic volatility models are more descriptively accurate than the Black–Scholes constant volatility assumption. The systematic upward bias of constant volatility models ranges from 5 to 15 per cent for most ESOs issued by small-cap stocks. The sensitivity of the ESO with respect to the inclusion of performance hurdles can be significant (with value reductions of up to 60 per cent), but it is typically less than one would naively expect. ESOs of small-cap stocks frequently include performance hurdles based on the stock’s TSR performance measured against a largish group of companies (e.g. stocks in the small-cap stock index of the relevant country). The proposed ‘single index model’ setup should appeal to practitioners because the required parameters (index β’s and idiosyncratic risk components) are easily estimated, transparent and familiar, and the valuation procedure itself is straightforward. Pay-for-performance sensitivity of ESOs is an important issue both for remuneration consultants and for research into the link between pay and performance. In general, we find that performance hurdles increase the incentive capacity of ESOs and can mitigate their dilution effect.

775

In a case study we find a more effective performance hurdle would allow the company to increase the cash (fixed salary) component of remuneration from 70 to 76 per cent while retaining an identical incentive level. At the same time the potential dilution effect of the ESO decreases by 6 per cent.4 Because our primary focus is on fair value for accounting purposes, we adopt the employer’s perspective.5 Other ESO characteristics, such as resetting of the strike price or reload features (see e.g. Brenner et al., 2000), are not included in our framework, in line with accounting standards.6 The rest of the present paper is organized as follows. The ESO valuation model is outlined in Section 2. The ESO valuation formulas are derived in Section 3. The two case studies are presented in Section 4, which also investigates pay-for-performance sensitivity. Section 5 concludes the paper. 2. Valuation Model Our model setup contains a dividend yield, which is the main driver for a well-diversified holder of an American option to exercise the option early. The stock price S(t) is the underlying of the ESO. The dynamics of S(t) are in the broad framework of Black and Scholes (1973), but are extended for stochastic volatility σ(t), following Heston (1993) and Fouque et al. (2000). The risk-free asset is introduced via a money market account that grows at a non-stochastic but potentially time-dependent rate r(t) > 0.7 2.1. Stock Price Dynamics 776

The risk-neutral dynamics of the stock price by assumption satisfy

where r(t) is the non-stochastic risk-free rate at time t, q ≥ 0 is the dividend rate of the stock, and Ws(t) is a Wiener process. The volatility σ(t) is a mean-reverting process, as proposed by Heston (1993),8 with dynamics

Therefore, σ2(t) describes a mean-reverting process with reversion rate κ > 0, mean level θ, ‘volatility’ η > 0 and initial value σ2(0). The process driving σ2(t) is a Wiener process Wσ(t). To allow for possible leverage effects, Ws(t) and Wσ(t) can be correlated with the coefficient ρsσ, with | ρsσ | 0, when the performance hurdle is tested and, depending on the outcome, a specific number of options vest. The second stage spans the time from vesting date T* to maturity date T > T*.10 Vesting hurdles typically fall into two categories. First, performance can be measured by accounting variables such as growth in EPS, net income or revenue over a specific time period. This type of performance hurdle is said to be tested against non-market conditions. Second, 777

the vesting hurdle can be specified in terms of absolute or relative share price performance. If the TSR over the vesting period is sufficiently high or outperforms the TSRs of companies of a nominated peer group, then the ESO vests. This type of performance hurdle is said to be tested against market conditions.11 The two performance hurdle categories have different characteristics when applying option pricing theory for valuation. Accounting variables are not traded assets so that ‘vesting risk’ cannot be hedged. One way to deal with this setting is to introduce risk premium parameters linking the real-world dynamics of the accounting variables to their risk-neutral ones.12 But paragraph 20 of IFRS 2 requires the use of ‘the best available estimate of the number of equity instruments (ESOs) expected to vest’, indicating that probabilities used for evaluating the performance hurdle (risk-neutral probabilities) ought to coincide with the real-world probabilities. In other words, the additional risk premium parameters for ‘vesting risk’ are redundant. Hurdles based on share price performance are specified using variables that arise from assets traded in financial markets. Risk-neutral valuation then can be applied although the market price model, as specified in equation (1) and (2), should be augmented for the dynamics of the assets against which the company’s share price performance is assessed. Again, the presence of stochastic volatility might result in an incomplete market model. We deal with this matter as outlined in the previous section. 2.2.1 Non-Market Conditions

778

According to accounting standards, the value of an ESO for accounting purposes should ignore any non-market conditions. Instead, the value in the absence of the performance conditions is to be multiplied by the number of ESOs expected to vest rather than the actual number granted. The expected number is calculated by multiplying the actual number initially granted by the probability that the options will vest. Further adjustments should be made as that probability is revised with experience, but the initial valuation of each ESO is retained (see e.g. IFRS 2, para. 19). Nonetheless, our method can accommodate options subject to a performance condition based on an accounting variable.13 Accounting variables are reported discretely in time and usually applied on an annual basis. Depending on disclosure requirements, some are also reported more frequently (e.g. quarterly or half-yearly). Assume the accounting variable A(t) is embedded in continuous time and that the dynamics of A(t) are described by a potentially mean-reverting process X(t) that is transformed by g to obtain

and

where Wx(t) is a Wiener process that is potentially correlated with the driver of the stock price process Ws(t) (i.e., psX dWx(t) = dWx(t) dWs(t)), but is independent of the driver of the stock price volatility Wσ(t), and a, b and c are real constants.

779

Paragraph 19 of IFRS 2 implies that performance hurdles based on non-market conditions are to some extent unrelated to market conditions affecting the ESO value.14 Our proposed setup includes the independent case (by specifying PSX = 0) but also allows for a more general framework, which would be a more realistic description. The proportion p of ESOs that vest at T* is determined when testing the performance hurdle:

where p is a functional C(0, T*) → [0, 1] and C(0, T*) is the space of continuous functions (i.e. the sample space where A(t) lives). Consider the following example. Example 1 (average EPS growth). The accounting variable describing the performance A(t) is EPS. It is well known that EPS is a ‘sticky’ variable exhibiting high autocorrelation. Therefore, it is reasonable to model A(t) by a geometric Brownian motion. Set g(x) = ex and assume that the dynamics of X(t) satisfy

resulting in

The EPS growth (on log-scale) in year n is G(n) = log(A(n)/A(n−1)). The average EPS growth from year T0 to year T* is then

780

The proportion p of the ESOs that vest at T* depending on A(t) can be written as

where G* is the hurdle for the average EPS growth. 2.2.2. Market Conditions The performance hurdle that is tested against market conditions is usually based on outperforming a relevant market index, or a peer group of companies (usually quantified by TSR). In the present paper, we focus on performance measured against an index where stochastic volatility is included.15 Outperformance of an Index. The stock price process is extended to dimension 2, (S(t), I(t)) where S(t) represents the stock price of the company granting the ESO and I(t) represents the index required for the performance hurdle. Equation (1) is extended to

where q1 is the dividend rate of the index, rI(t) the risk-free rate that applies to the index (we assume r1 = r), and (Ws(t)WI(t)) is a two-dimensional Wiener process with independent components Ws(t) (representing idiosyncratic risk) and WI(t) (index, or market-related risk). Furthermore, WI(t) is independent of Wσ(t)16

781

We follow Johnson and Tian (2000), where the stock price risk can be decomposed into two independent components: risk spanned by the market index, quantified by ß, and idiosyncratic risk. Furthermore, Johnson and Tian (2000) is extended by allowing for stochastic volatility of the idiosyncratic risk component. Outperformance of a peer group of stocks. Frequently, the peer group consists of companies listed in broad equity indices (e.g. Australian Stock Exchange (ASX) 50, Standard & Poor’s (S&P) 500 (USA) or Financial Times Stock Exchange (FTSE) 100 (UK)). When the peer group comprises more than a few stocks, it can be extremely difficult to estimate or calibrate the model because of the abundance of parameters.17 Our specification reduces the complexity to an analytically tractable extent. The input parameters are just the companies’ betas and volatilities. Assume the dynamics for each stock price can be written as a risk-neutral version of the intertemporal capital asset pricing model (Merton, 1990, p. 373) with a further requirement that the companies’ idiosyncratic risks be mutually independent:

where N is the number of stocks in the group, βn is each stock’s index beta, qn its dividend yield, sn its idiosyncratic risk, and Z(t) = (Z1(t), …, ZN(t)) is an N-dίmensional Wiener process. Assume also that the index is the only joint risk factor in stock prices so that

782

the idiosyncratic risk factors Z1(t), …, ZN(t) are pairwise independent and also independent of WI(t). The performance over the timespan [0, t] is typically defined by the TSR (i.e. for the stock price process Sn(t)):

where the stock price Sn(t) is adjusted for the (continuous) dividend rate qn earned over this period to . It is clear from the specific definition of the TSRs that the dividend yields qn can be dropped from the formulation when the underlying data is given as the TSR series of each stock. The same argument applies for the index I(t), and when dealing with a total return index the parameter qI (dividend yield of index) is redundant. The proportion of TSRs in the comparator group that are below the threshold x is

and the expected proportion of TSRs that are below the threshold x given I(t) is

The following proposition states that in expectation the quantiles (or percentiles) of the TSR of the peer group can be expressed in term of the TSR of the index I(t). Proposition 1 The expected proportion of TSRs below the threshold x given I(t), EP(x; I(t)), is a distribution function in x, for fixed I(t), and 783

where Φ is the standard normal cumulative distribution function and

The proof is available from the authors on request. Proposition 1 states the expected proportion EP(x; I(t)) of stocks in the peer group that are TSR-outperformed by the company granting the ESO, given the information S(t) and I(t), where

, is the TSR

corresponding to S(t). Vesting conditions are typically not linearly related to P(x) (and, hence, P(x) cannot be replaced by EP(x; I(t)) when calculating the ESO value using conditional expectations). For testing performance hurdles, more information on the conditional distribution of P(x) is needed than the conditional expected value EP(x, I(t)). Proposition 2 The proportion of TSRs below the threshold x, P(x), conditioned on I(t) has mean µP(x; N) = EP(x, I(t)) and variance and

784

where EP and dn are given in Proposition 1. Provided that following limiting behaviour applies

, the

where Z is a standard normal random variable. The proof is available from the authors on request. Proposition 2 conveys two statements. First, testing the performance hurdle is feasible using univariate integration of the standard normal pdf: condition on I(t) and S(t), compute µP(x) and (x), and then perform a numerical integration using the standard normal pdf to obtain (p(P(x)) | S(t), I(t)), where p is the proportion of ESOs that vest as a function of the performance measure P(x). The approximation should work well for small numbers in the peer group, for example, N ≈ 10. Second, if the number of stocks is large, for example, N = 50, then the upper bound for the standard deviation is useful: . Taking the normality into account the expected quantity EP(x; I(t)) is an accurate proxy for P(x), with a 95 per cent confidence interval:

This result indicates that for large N, P(x) can be replaced by its conditional expected value EP(x; I(t)), leading to a welcome simplification.

785

The proposed framework provides a tractable valuation model. The parameters specifying the model are meaningful and manageable. The benefit of an index framework is that the number of parameters does not increase dramatically when the number of stocks in the peer group increases. The formulas required for this setup might appear to be complex but they are analytically tractable as a closed-form expression is provided. 2.3. Departure and Early Exercise Although the lifespan of the ESO can be separated into two periods (the vesting period and the period when the ESO has potentially vested and can be exercised), we specify a single random time τ. During the vesting period [0, T*], τ models the departure, and during the ‘active life’ of the option, [T*, T], τ describes early exercise. A vested ESO can be exercised early and, hence, is an American option. In a no-arbitrage pricing framework, the exercise time of an American option is solely determined by the market conditions, exercise occurring when the share price crosses a specific boundary (the critical share price S*(t), which is a deterministic function of time for the Black–Scholes model). However, ESO holders are in general undiversified and their exercise behaviour is largely driven by risk-aversion and liquidity needs (see e.g. Huddart, 1994). In the present paper, we adopt a reduced-form approach for modelling departure and early exercise following Cuny and Jorion (1995) and Carr and Linetsky (2000). The time of departure or early exercise τ is

786

determined by the actual market conditions but retains a degree of unpredictability. This is achieved by specifying τ as the first jump of a Cox process. The time of the first jump of a Cox process τ can be described as a generalized exponential stopping time with intensity λ(t), where λ(t) is a stochastic process (see e.g. Lando, 1998). In our setting, λ(t) is a function of the actual market conditions where the market conditions at time t (complete history of factors driving the market over the time period [0, t]) are formalized by the information sets FT, 0 ≤ t ≤ T. Given the entire market information FT, the distribution of the stopping time satisfies

Boyd et al. (2007) provide guidance on the factors driving early exercise behaviour in Australia, where the dividend yield is the most significant factor, followed by the volatility of the underlying and the employment level of the person to whom the ESO is granted. Their findings are consistent with US studies. Armstrong et al. (2006) propose a proportional hazard model of the early exercise intensity, which is explained by factors including stock price history, vesting, dividends, volatility and other behavioural effects. Their approach fits well in our valuation model and their empirical findings could be used to value US ESOs. 3. ESO Valuation

787

Recall that we have separated the lifespan of the ESO into two periods: the vesting period [0, T*] and the period [T*, T] when the ESO potentially has vested and can be exercised. The ESO is valued recursively. First, we derive the value of the vested ESO and then performance hurdles and departure are included to derive the ESO value at grant date. 3.1. Valuation of a Vested ESO Recall that λ is a function of actual market conditions. At time t, the value of a vested but not yet exercised ESO can be written as a function of the current values of the state variables Y(t) = (S(t), σ(t)):

where is the conditional expectation given S(t) = x, σ(t) = s on the set {τ > t}, and τ ∧ T is the minimum of the random time of early exercise and the maturity date. The price c(t, x, s) can be simplified using the Cox property of the early exercise time τ by conditioning on the entire information of the state process Y(t) over the period [0, T], This approach ‘averages out’ the uncertainty of τ resulting in

788

The above representation has two components. The first component gives the value of the vested option at time t, conditional on being exercised at maturity date (when in the money). Therefore, the expression is weighted by the probability that the option is not exercised early in [t, T], .

The

second

component

represents the contribution to the option value for the early exercise case: for every time t* in [t, T], the discounted early exercise value exp max (S(t*)–K, 0) is weighted by the probability of early exercise at the specific time point λ(t*) dt* conditional on having no early exercise prior to t* (i.e. by multiplying on

.

3.2. Incorporating Performance Hurdles

789

Two types of vesting are common. ‘Cliff vesting’ describes a performance hurdle tested once, at time T*; ‘graded vesting’ is when it is tested more than once. 3.2.1. Cliff Vesting The performance hurdle is tested at the potential vesting date T*. The proportion of ESOs that vest is the stochastic quantity p depending on the specification of the performance hurdle, p can be determined at vesting date T*, and can be written as

for performance hurdles written on non-market conditions and market conditions, respectively (see Section 2.2). The following representation of the ESO value can be established (from Section 3.1):

Applying iterated expectation conditioning on the information generated by the processes S(t), σ(t), and I(t) up time T*, FT*, yields

where TSR is the total shareholder return of the company granting the ESO over the period [0, T*].

790

Following Proposition 2, P(x) is asymptotically normally distributed given I(t), so that for each fixed set of values for S(T*) and I(T*), p(P(TSR)) can be evaluated by a univariate integration using the normal pdf. If the number of stocks in the peer group is rather large, for example, N ≥ 50, the normally distributed performance measure P(x) (conditional on S(T*) and I(T*)) can be approximated by its expected value EP(x; I(T*)), and

3.2.2. Graded Vesting The performance hurdle can be tested at different points in time, . The random variable pk. specifies the proportion of ESOs that vest at time

.

The quantity is a functional of either the non-market conditions specified by A(t) or of market conditions given by Pκ(x). Testing the performance hurdle at ,1 < l ≤ k, may depend on the current state of A(t) or Pk(x) and on the outcome of past testing. Most specifications of ‘graded vesting’ can be written as a portfolio of k ESOs with vesting date and performance hurdle pk. Therefore, for an ESO

791

with ‘graded vesting’ we can rely on the analysis in the previous section. 4. Implementing the Approach: Two Case Studies 4.1. Preliminaries To demonstrate how the approach can be implemented, we conduct two Australian case studies: Metcash Limited and Amcor Limited. In the first case, Metcash granted an ESO package totalling 510 000 options to an executive director, Bernard J. Hale, on 2 September 2004. The ESOs include performance hurdles specified by a target average compound growth of EPS. Metcash disclosed, in its 2005 annual report, an ESO value of $A0.97, which was reported to have been calculated using the Black–Scholes model. In the second case, Amcor granted 5 588 200 ESOs to employees on 2 August 2004. The performance hurdle applicable to the Amcor ESOs is based on the company’s TSR measured relative to a peer group of 46 stocks selected from the ASX 100. The option value was disclosed in its 2005 annual report as $A1.03 and was said to have been calculated using Black–Scholes, with an adjustment for expected life. Black–Scholes, on which both valuations were based, does not take into account directly the three matters we are investigating; namely, performance hurdles (market and non-market), non-zero probability of early exercise and stochastic volatility. A priori it is clear each component lowers the ESO value. The non-zero probability of not meeting performance hurdles potentially reduces the option’s payoff and, therefore,

792

the value of the ESO. Likewise, early exercise typically leads to a lower valuation. Stochastic volatility as modelled by Heston (1993) caters for the leverage effect and, hence, an increasing stock price is associated with decreasing volatility. Once the stock price moves the call option into a profitable payoff region, volatility tends to drop and, hence, so does the time value of the call. Therefore, Heston’s stochastic volatility model, which we adopt, produces lower call option values than Black–Scholes for at-the-money calls and more pronounced lower values for out-of-the-money calls. 4.2. Implementation The stochastic volatility model is implemented by a tree-based method that discretizes the state space of stock price and the volatility process and then calculates the transition probabilities between the states (see Nelson and Ramaswamy, 1990; Leisen, 2000). The key idea is to describe the stochastic system as specified in equations (1) and (2), or alternatively equations (6) and (7), as a homoscedastic system following a suitable transformation. The state space of the transformed system is then discretized to form a multivariate version of the binomial tree; the distance of the tree branches is designed to match the variance of the driving process (which is a constant depending neither on time nor space), and the transition probabilities are determined to match the local drift (and, hence, can be state dependent). The intensity of early exercise and departure λ(t) is a function of the state variable Y(t). For Metcash, the EPS A(t) is included following Example 1. For Amcor, the value of the index

793

I(t) formed from the 46 comparator stocks is included and the approximation in Proposition 2 is applied. Backward induction is used when valuing the option. First, the tree is initialized at maturity date by writing the option payoff at the maturity nodes. Next, when rolling backward through the tree, the discounted average option value of the successor nodes is calculated at each node (using risk-neutral probabilities and risk-free discounting), and this quantity is then weighted by the probability of no early exercise (i.e. multiplied by e−λ(t)∆t). Furthermore, the value in the case of early exercise at the current node is added, weighted by the respective probability λ(t) ∆t. This procedure will reproduce the representation of c(t, x, s) as above when passing through the limit Δt → 0. The pricing algorithm was implemented using Matlab 7. For Metcash, the time horizon of 6 years was discretized into 1500 steps, corresponding to 250 share trading days per annum. For Amcor, the performance hurdle was path dependent (due to a retesting feature of the performance hurdle); here, the time horizon of 6 years was discretized into 312 steps, corresponding to 52 weeks per annum. 4.3. Performance Hurdles Based on EPS: Metcash Metcash is a marketing and distribution company operating in the food and consumer goods categories. Metcash had a market capitalization of $Al.4bn at the end of its 2005 financial year and it is included in the ASX 100 Index. Metcash’s history can be described as a history of change.18 A company with a high degree of activity in a competitive segment of the market is at odds

794

with the Black–Scholes model, which assumes a constant volatility. Rather, a stochastic volatility model as outlined in Section 2.1 would seem to be a more suitable framework. The 510 000 options awarded to Metcash executive director Bernard J. Hale on 2 September 2004 mature 6 years later. Their exercise price was set to $A2.43 and the spot price of the share at issue date amounted to $A2.75. The vesting condition (apart from tenure requirements) was specified by the company’s EPS growth. The ESOs vest on 2 September 2007 if EPS grows on average by 20 per cent per annum measured relative to the 2001 EPS of 3.93¢.19 Although Metcash disclosed an ESO value of $AO.97 in its 2005 annual report, it did not disclose the assumed parameters of Black–Scholes (i.e. the interest rate, dividend yield and volatility). Examining Metcash’s stock price and dividend history, we assumed the values r = 5 per cent, q = 3 per cent and σ = 36 per cent, which together reproduce the disclosed option value. We adopt these values for the interest rate and dividend yield. The stochastic volatility model is specified such that the expected long-run volatility coincides with the Black–Scholes volatility of 36 per cent. In particular, the parameter choice was: mean reversion rate ĸ = 1, mean level of the variance rate θ = 0.1296 equals volatility assumed for Black–Scholes), ‘volatility of the volatility’ η = 0.5, σ(0) = 20 per cent (equals the 1 year historical volatility preceding the option grant), and correlation pSσ = −0.75.20 795

Metcash’s performance hurdle is based on EPS growth. At grant date EPS was 16.10¢; the EPS hurdle at vesting date was determined as 11.73¢, which is derived from the required annual average growth of 20 per cent and EPS at the reference date in 2001 of 3.93¢. EPS is assumed to follow geometric Brownian motion (see Example 1), with parameters determined from historical data and analyst forecasts: growth rate 10 per cent, volatility 30 per cent and correlation with the driver of the stock price, Ws(t)), of 40 per cent. Departure and early exercise are modelled by the constants λDeparture = 0.10 and λEarly Exercise = 0.33. That is, λ(t) = λDepaıtuıe prior to the vesting date (t ≤ T*) and λ(t) = 1S(t) ≥ x λEarly Exercise + 1s(t) < xλDeparture between vesting and maturity date (T* < t ≤ T). The intensity is constant before vesting date and path dependent during the exercisable life of the option. If the option is out-of-the-money (S(t) < X), the departure intensity applies, and if the option is in-the-money (S(t) ≥ X), the early exercise intensity applies. The parameter choice for the intensities assumes values consistent with Boyd et al. (2007) for the Australian market.21 The intensity process λ(t) is path dependent because it is driven by the stock price S(t), and this setting could be generalized to incorporate other factors and more complicated functional forms than the one assumed here. Developing the model along these lines would require further research on the estimation of the relevant parameters for the Australian market, possibly as in Armstrong et al. (2006) for the USA. Table 1 reports the values of the Metcash ESOs. Stochastic volatility reduces the ESO value by 7 per

796

cent, from the Black–Scholes value of $A0.97 to $A0.90. The reduction is in line with expectations since the option grant was roughly at-the-money and stochastic volatility accounts for the smile effect (i.e. the implied Black–Scholes volatility reduces with increasing strike price), and the same applies to the option price when compared to constant volatility Black–Scholes.22 The American option value using Heston’s stochastic volatility framework is $AO.92 and increases the European option by approximately 2 per cent. Table 1 Executive and employee stock options valuation: Metcash

This table displays the prices of the executive and employee stock options (ESOs) on Metcash with time to maturity of 6 years, vesting period of 3 years (granted 2 September 2004). Metcash spot was $A2.75, exercise price X = $A2.43, assumed risk-free rate r = 5 per cent, dividend yield q = 3 per cent. Volatility σBS = 36 per cent assumed constant for Black–Scholes or alternatively Heston’s specification assumed with θ = 0.1296, ĸ=l, η = 0.5, ρsσ = −0.75, and σ(0) = 20 per cent. Performance

797

hurdle based on earnings-per-share (EPS) growth; EPS at grant date 16.10¢, hurdle at vesting date 11.73¢. EPS assumed to follow geometric Brownian motion with growth rate 10 per cent, volatility 30 per cent and correlation with stock return of 40 per cent. Departure and early exercise are modelled by constants λDeparture = 0.10 = λEarly Exercise and λEarly Exercise = 0.33, where λ(t) = λDeparture for 0 ≤ t ≤ T*, and λ(t) = λ·Eaily Exercise + 1S(t)≥X λDeparture, for T* < t ≤ T. The performance hurdle reduces the Heston value by an additional 6 per cent. The further reduction is expected although, at first sight, the percentage is not in line with the probability of the ESO not vesting, which is 12 per cent. This can be explained by the 40 per cent positive correlation of EPS and the driver of the stock price Ws(t). Therefore, the performance hurdle tends to be met if the option component is in the money at the vesting date. Most of the time, the option is of relatively little value if the performance hurdle is not met. The discrepancy between value reduction due to the presence of a performance hurdle and the vesting probability sheds some light on paragraph 19 of IFRS 2. It requires reporting the effect of a non-market-based performance hurdle, such as a preset minimum rate of growth in EPS, on ESO value. Vesting conditions are taken into account by ‘adjusting the number of equity instruments’ and this value is initially based on ‘the best available estimate of the number of equity instruments expected to vest’ (here: 88 per cent). However, the performance hurdle only reduces ESO value by 6 per cent. For performance hurdles based on non-market conditions, the IFRS 2 approach might produce ESO 798

values that are too low. A ‘value-weighted’ expected proportion of ESOs that will vest (here: 94 per cent) produces a more appropriate valuation for accounting purposes. Departure is the main driver for the difference between the European call option price and the ESO value in the chosen setup. The ESO holder, B. J. Hale, is assumed to leave Metcash with an annual rate of λDeparture = 0.10 corresponding to an expected time of 10 years with Metcash following the ESO grant date. Therefore, the probability of departure before vesting is 1 − e−0.1×3 = 26 per cent. It is worthwhile noting we assume departure is introduced into the option pricing framework by working with historical probabilities that are not adjusted for a potential market price of risk. This approach is reasonable for companies granting ESOs to a wide range of employees. However, in the case of a company granting ESOs to a small number of employees, and mostly CEOs or directors, the diversification concept might not apply, and the introduction of a market price of risk component seems appropriate. In our setting, this would correspond to reducing the departure rate, which would in turn produce a higher ESO value and, hence, a more conservative profit figure. Early exercise accounts for a reduction of the ESO value of 5 per cent, which is the smallest value reduction in our setting. In general, early exercise sacrifices time value of the option. However, for a company with a relatively high dividend yield, the impact of time value sacrifice is mitigated.

799

The ESO value based on our model is $A0.60 compared to the reported value of $A0.97 (Table 1). We value the 510 000 ESOs at $A305 100 compared to $A494 700 disclosed in the annual report. The overall reduction, which is due to our allowance for the performance hurdle, possible departure and early exercise, and stochastic volatility, totals 38 per cent and indicates substantial overvaluation in the annual report. The overvaluation is still substantial if departure, being the factor carrying most weight, is specified more conservatively. When setting λDeparture = 0.05, the ESO package value totals $A354 500 or 28 per cent below the reported value. When departure is eliminated, λDepaıture = 0, we obtain $A411 800, which still constitutes a substantial reduction (17 per cent). Metcash’s ESOs were granted well in the money. The strike price was set to $A2.43, whereas the spot at grant date was $A2,75, implying a risk-neutral forward at maturity of $A3.10. The impact of stochastic volatility on the valuation of call options depends crucially on the option’s moneyness. In additional analysis (available from the authors on request), we have varied the strike price to explore the effect of stochastic volatility for an ‘at-the-money’ and ‘out-of-the-money’ specification; setting the strike price to $A3.00 and $A3.50, respectively. The value reduction of Heston’s call option value compared to Black–Scholes increases with the strike: a reduction of 13 per cent for the ‘at-the-money’ specification and 18 per cent for ‘out-of‫־‬the-money’ respectively. The observed effect transmits to the ESO value; for a strike price of $A2.43 we observed a value reduction compared to Black–Scholes of 38 per cent, for

800

a strike price of $A3 the reduction is 44 per cent, and for a strike price of $A3.50 it is 49 per cent. 4.4. Performance Hurdles Based on Market Conditions: Amcor Amcor is a global packaging company operating in five geographical regions: Australasia, North America, Latin America, Europe and Asia. Its headquarters are in Melbourne, Australia. It is one of the top global packaging companies by sales, and its market capitalization was $A5.9bn at the end of the 2005 financial year. Amcor is included in the ASX 50 Index. On 2 August 2004, Amcor granted 5 588 200 ESOs to employees with time to maturity of 6 years. The ESOs were granted exactly at the money when the spot price was $A6.84. The vesting condition (apart from tenure requirements) was formulated in terms of TSR outperformance of a specified group of 46 ASX 100 stocks. At vesting date (2 August 2007), the TSR of Amcor and of the stocks in the comparator group are to be measured with reference to their prices on 1 July 2004. If Amcor’s TSR outperforms at least half of the stocks in the comparator group, then the ESOs vest. If not, then the hurdle can be retested each month for 2 years following the vesting date (i.e. until 2 August 2009). In 2005, Amcor disclosed a value of $A1.03 per ESO, which was adjusted for the expected lifetime of each option to incorporate the effect of early exercise. The parameters used for the valuation were volatility σвs = 20 per cent, dividend yield q = 4.70 per cent, risk-free rate r = 5.60 per cent and an expected life of 5 years. The valuation took no account of the performance hurdle.

801

The performance hurdle requires the prices of 47 stocks (Amcor and the 46 stocks in the comparator group) to be modelled simultaneously. The index we use is formed as an equally weighted total return index of the 46 comparator stocks. The index volatility is then estimated along with the volatility components of the idiosyncratic risks for the stocks forming the index, and their index betas as well. The ESO was valued using the model setup outlined in Section 2.2.2 (Outper-formance of a peer group of stocks). The dividend yield and risk-free rate were as stated in Amcor’s 2005 annual report (i.e. q = 4.70 per cent and r = 5.60 per cent). Amcor’s overall volatility is decomposed into an idiosyncratic component (which is stochastic) and a systematic component introduced by the index ß. The parameter choice was K = 1, θ = 0.0361, η = 0.25 and pSσ = −0.75, and the initial value of the idiosyncratic volatility component was set at 19 per cent. Based on the 5 year history preceding the ESO grant, the remaining parameter estimates were index volatility σ1 = 10 per cent, and Amcor’s index ß = 0.65. The volatilities and betas of the stocks in the comparator group were estimated from historical data over 5 years to 2 August 2004 sourced from Datastream. The TSR is measured with respect to the end of the financial year and, hence, the reference value is $A7.10 (Amcor’s share price on 1 July 2004). Departure and early exercise were modelled in a similar fashion to Metcash (i.e. by constants λDeparture = 0.10 and λEarly Exercise = 0.33; see Section 4.2 for details).

802

Table 2 reports the valuation results for the Amcor ESO. Amcor disclosed an ESO value of $A1.03, which was said to be based on Black–Scholes adjusted for the option’s expected lifetime. Alternatively, we can calculate the ESO value using Black–Scholes with no adjustment for expected life, resulting in $A1.12. This figure appears to be a more appropriate benchmark when investigating the effects on ESO valuation of including stochastic volatility. The value based on Heston’s model is $A1.07, which is 5 per cent below the Black–Scholes value (but 4 per cent above the Black–Scholes value when adjusting for expected life of the ESO). The observed value reduction is in line with expectation and of similar magnitude to that in the previous case study of Metcash ESOs. However, the American option value is about 5 per cent higher than the European value within the stochastic volatility framework, which is more pronounced than for Metcash. A likely reason for this finding is that the dividend yield for Amcor is higher (i.e. q = 4.7 per cent) compared to Metcash’s dividend yield of 3 per cent. Table 2 Executive and employee stock options valuation: Amcor

803

a

Adjusted for expected lifetime, as in Annual Report. This table displays the prices of the executive and employee stock options (ESOs) on Amcor with time to maturity of 6 years, vesting period of 3 years (granted 2 August 2004). Amcor’s spot was $A6.84, exercise price X = $A6.84, assumed risk-free rate r = 5.60 per cent, and dividend yield q = 4.70 per cent. Volatility σвs = 20 per cent assumed constant for Black–Scholes or alternatively Heston’s specification assumed with θ = 0,0361, k = 1, η = 0.25 and psσ = –0.75. Departure and early exercise are modelled by constants λDeparture = 0.10 and λEsrly Exercise 0.33 =, where λ(í) = λDsparture for 0 ≤ t ≤ T*, and λ(t) = 1s(t)≥x λEarly Exercise +1s(t) < X, λDeparturefor T*< t ≤ T. Performance hurdle based on total shareholder return outperformance of 50 per cent of the 46 stocks in the comparator group. Index approximation was applied with σI = 10 per cent and Amcor’s index ß = 0.65. The effects of employee departure and early exercise are of similar magnitude to those for Metcash. The approximation of replacing the performance measure P by its expected value EP was analysed separately. The outperformance hurdle of 50 per cent was varied within 1 standard deviation (2 per cent); hence, it was set to 48 804

and 52 per cent. The effect on the ESO value is below 0.50 per cent and we may conclude that the approximation is robust and acceptable. It is worth noting that the value adjusted for early exercise, $A1.04, is within 1 per cent of the figure disclosed in Amcor’s annual report, $A1.03. From this observation one might argue that the lifetime adjustment can work reasonably well; however, it captures only one aspect important for ESO valuation (namely, early exercise). It does not adjust for stochastic volatility and performance hurdles. Including these factors results in an ESO value of $A0.76, which is a reduction of 26 per cent compared to the valuation disclosed by Amcor. Most of the value reduction is caused by including employee departure in the setup. The somewhat surprising result is that Amcor’s performance hurdle had little effect (only 2 per cent) on the ESO value. But the probability of not vesting is rather high, at 27 per cent. The seemingly different quantities can be explained by the positive correlation of the vesting event and the option value at vesting date (see also Cuny and Jorion, 1995). If the option is valuable then the performance hurdle is more likely to be met and conversely, most non-vesting situations correspond to relatively small values of the option component. In additional analysis (available from the authors on request), we have modified Amcor’s performance hurdle. First, the retesting feature is left out. The loss of the retesting facility produces a non-vesting probability of 45 per cent, which is, as expected, substantially higher

805

than the 27 per cent in the original setting (retesting). The value reduction due to the inclusion of the performance hurdle increases from 2 to 14 per cent. The second alternative performance hurdle investigated is to require Amcor to TSR-outperform 75 per cent of the stocks in the comparator group (originally 50 per cent) and the retesting feature is again left out. The probability of not vesting increases from 27 to 83 per cent. The value reduction due to the performance hurdle increases sharply, from 2 to 60 per cent. Compared to the original specification of the performance hurdle, the two investigated alternatives provide a more effective performance hurdle. In the original specification, the 50 per cent hurdle provided an at most mediocre challenge, and this was weakened further by the retesting feature. Once more effective performance hurdles are introduced, a direct impact on the ESO value can be observed. The performance hurdle, together with the probability of departure, are the main factors driving the ESO value. The ESO value without retesting is $A0.67, which is 35 per cent below the disclosed Black–Scholes value. A 75 per cent TSR-outperformance hurdle produces an ESO value of $A0.32 (69 per cent below the disclosed Black–Scholes value). 4.5. Implications for Pay-for –Performance Our ESO valuation model provides key insights useful for answering questions such as, ‘How can we design remuneration packages that will achieve a specific pay-for-performance target?’ To keep the analysis simple we assume the ESO’s worth to the employee

806

equals its cost (‘fair value’) to the employer. We consider an annual remuneration package with total present value W comprising a share-based payment with value v and a cash (or bond) component with value b. Therefore,

The share-based payment v can be written as N × c, where N is the number of ESOs and c denotes the value of a single option. The sensitivity of the remuneration W with respect to a change in the underlying stock price is

where the cash component b is insensitive to a change in S (i.e. ∂b/∂S = 0). The target incentive can be expressed as the elasticity of the remuneration package with respect to the company’s stock price

, calculated at the grant date. In

the following discussion we set this ratio to the value 1. From this and the ESO delta ∂c/∂S, the number of ESOs N that will achieve the target incentive level can be written as

The number of ESOs increases proportionally with the total value of the remuneration package W and reciprocally with the ESO’s delta, ∂c/∂S. If the company selects its stock as the share-based component, then c = 807

S and ∂c/∂S = 1, Hence, N = W/S, and the entire remuneration package consists of shares only (as a result of setting the target incentive level to 1). We refer to the latter as the base case. In Table 3 a range of possible remuneration packages for Amcor are considered. The table gives the weights of the components, which can be either cash, shares or ESOs. In addition to the base case and the ESO actually granted by Amcor, several combinations of cash and ESOs are explored. We investigate ESOs with no TSR-performance hurdle and vary the moneyness: in-the-money with a strike price of $A5, at-the-money with a strike price of $A6.84, and out-of-the-money with a strike price of $A9. The higher the strike, the more leveraged the option position. Therefore, the employee can receive a substantial cash component within the salary package; 59, 70 and 78 per cent, respectively. The weight of the cash component increases, compared to the 70 per cent of the at-the-money specification, once performance hurdles are introduced. For the 50 per cent TSR hurdle, the figure increases to 76 per cent, and to 86 per cent for the 75 per cent TSR hurdle, respectively. For a risk-averse employee, a large cash component appears to be more favourable. The inclusion of performance hurdles achieves this outcome while maintaining the target incentive level. Table 3 Pay-for-performance sensitivities, Amcor, in stochastic volatility setting

808

This table displays a range of possible Amcor remuneration packages when setting the target incentive level to 1. The remuneration package can include cash, stock and executive and employee stock option (ESO) components. For each package the components are given in percentage terms. All possible packages are written on the Amcor stock with grant date of 2 August 2004, time to maturity of 6 years, and vesting period of 3 years (if applicable). The hurdle was retested on a monthly basis over a 2 year period following the vesting date. Amcor’s spot was $A6.84, assumed risk-free rate r = 5.60 per cent, dividend yield q = 4.70 per cent. The volatility is assumed to follow Heston’s specification with θ = 0.0361, κ= 1, η = 0.25, and ρsσ = –0.75. Departure and early exercise are modelled by constants λDeparture = 0. 10 and λEarly Exercise = 0.33, where λ(t) = λDeparture, for 0 ≤ t ≤ T*, and λ(t) = 1s(t)≥XλEarly Exercise + λDeparture for T* < t ≤ T. Possible performance hurdles are based on total shareholder return (TSR) outperformance of 46 stocks in the comparator group; index approximation was applied with σ1= 10 per cent and Amcor’s index ß = 0.65. The last column quantity in Table 3, the maximal potential dilution, refers to the number of instruments that need to be granted to the employee measured 809

relative to 1 unit of stock, as in the base case.23 Dilution will only take effect where the company granting the ESOs does not engage in any form of hedging. The company can hedge the exposure itself by buying back stock on the market, or if the company effectively buys the ESOs from a third party such as a financial institution. Therefore, the following discussion applies when the company decides not to pursue any such hedging activity. From the shareholder’s perspective, granting ESOs with a higher strike price to compensate for an increased cash component sought by a risk-averse employee will come at the expense of more options being granted, which later might lead to dilution of the shares. The higher the strike price (and, therefore, the higher the cash component), the more pronounced is the potential dilution effect. However, the inclusion of performance hurdles reduces dilution by maintaining a higher proportion of the remuneration package in the form of cash. A possible reason for this result is that the ESOs might have effectively a twofold strike price; first, the traditional strike price, and, second, the performance hurdle. If the stock price increases, the call option becomes more valuable as the underlying stock moves more into the money, and the same applies to the probability of meeting the performance hurdle. The two effects combine, thereby forming a higher sensitivity of the ESO value to changes in the stock price, compared to the plain call option sensitivity. For example, the highest cash component, 86 per cent, can be obtained by granting ESOs with strike price $A6.84 and a 75 per cent TSR performance hurdle. This still exceeds the cash

810

component of the out-of-the-money call (strike price $A9.00). However, the dilution measure is 3.01, whereas it is 4.34 for the out-of-the-money call without a performance hurdle. 5. Summary and Conclusion We have proposed a valuation model for ESOs that addresses three potentially major determinants of ESO value: performance hurdles, the probability of departure and early exercise, and stochastic volatility. The model was implemented for two Australian case studies, enabling us to explore how these three factors affect ESO value, the pay-for-performance relationship and risk-based incentive contracts. The first case study, Metcash, focused on stochastic volatility and performance hurdles based on non-market conditions. We found that stochastic volatility can reduce the ESO value by 10 per cent when the option is granted at the money; for out-of-the-money options, the value reduction was close to 20 per cent. The non-market condition type of performance hurdle was modelled allowing for correlation with the stock price and the outcome was somewhat at odds with accounting standards. For non-market conditions, accounting standards require vesting conditions and the valuation of the ESO to be treated separately: the ESO value is to be calculated without taking any non-market performance hurdles into account and the result multiplied by the number of options expected to vest. Accounting standards are biased in that the expected proportion of ESOs that will not vest is typically higher than the effective value reduction caused by the performance

811

hurdle. Therefore, accounting standards can result in ESO values that are too low. We suggest valuing the ESO for both situations, with and without performance hurdles. The ratio of these numbers can then be interpreted as the ‘value-weighted’ expected proportion of ESOs that will vest, and these numbers will in turn produce a more accurate ESO value. The second case study, Amcor, emphasized performance hurdles based on market conditions. In particular, we proposed a model for dealing with performance hurdles that are based on the stock’s TSR-performance measured against a largish group of companies. The high dimensionality can cause traditional valuation approaches to break down. We based our model on Merton’s intertemporal capital asset pricing model and assumed that the idiosyncratic risk components of the stocks were pairwise independent. The parameters of the models are index ß’s and volatilities, quantities that are transparent and well understood. Furthermore, we proposed an approximation procedure that massively reduced the model complexity to at most dimension three (stock, stochastic volatility, and market/index), an approximation that was found to work well. Applying our model to the Amcor ESOs, we observed that performance hurdles can be structured, as Amcor did, so that they have relatively little impact on ESO value. The performance hurdle of the Amcor ESO was not achieved with a probability of 27 per cent, but the value reduction caused by the performance hurdle was as little as 2 per cent. We investigated this effect further and found it most likely was caused by the strong

812

positive correlation of achieving the performance hurdle and the option payoff. Therefore, we were led to the conclusion that performance hurdles need not decrease the ESO value as much as one would naively expect. However, their presence has a substantial effect on pay-for-performance sensitivity. We then compared remuneration packages consisting of various ESO specifications. Here we assumed that the employee receives both cash (i.e. fixed salary) and a share based payment, keeping the total value of the remuneration and the pay-for-performance sensitivity (incentive level) identical for all specifications. Because of their risk-aversion, employees typically prefer a higher cash proportion in their remuneration package. We observed that increasing the cash component was closely related to increasing the ESO strike price, which resulted in a higher potential dilution effect as more instruments would have to be granted. We found that the inclusion of performance hurdles in general increases the incentive capacity of ESOs and, assuming the company granting the ESOs does not hedge its exposure, performance hurdles can substantially mitigate the potential dilution effect. References Accounting Principles Board, 1972, APB Opinion No. 25: Accounting for Stock Issued to Employees (APB, New York). Armstrong, C. S., A. D. Jagolinzer, and D. F. Larcker, 2006, Timing of employee stock option exercises and the valuation of stock option expense, working paper

813

(Graduate School of Business, Stanford University, Stanford, CA). Bates, D. S., 2006, Maximum likelihood estimation of latent affine processes, Review of Financial Studies 19, 909–965. Bettis, J. C., J. M. Bizjak, and M. L. Lemmon, 2005, Exercise behavior, valuation, and the incentive effects of employee stock options, Journal of Financial Economics 76, 445–470. Black, F., and M. Scholes, 1973, The pricing of options and corporate liabilities, Journal of Political Economy 81, 637–654. Boyd, T., P. Brown, and A. Szimayer, 2007, What determines early exercise of employee stock options in Australia? Accounting and Finance 47, 165–186. Brenner, M., R. K. Sundaram, and D. Yermack, 2000, Altering the terms of executive stock options, Journal of Financial Economics 57, 103–128. Calvet, A. L., and A. H. Rahman, 2006, The subjective valuation of indexed stock options and their incentive effects, Financial Review 41, 205–227. Carpenter, J. N., 1998, The exercise and valuation of executive stock options, Journal of Financial Economics 48, 127–158. Carr, P., and V. Linetsky, 2000, The valuation of executive stock options in an intensity-based framework, European Finance Review 4, 211–230.

814

Cuny, C. J., and P Jorion, 1995, Valuing executive stock options with endogenous departure, Journal of Accounting and Economics 20, 193–205. Duan, J.-C., and J. Wei, 2005, Executive stock options and incentive effects due to systematic risk, Journal of Banking and Finance 29, 1185–1211. Financial Accounting Standards Board, 2004, Statement of Financial Accounting Standards No. 123R ‘Share-based Payment’ (FASB, Norwalk, CT). Fouque, J. P., G. Papanicolaou, and K. R. Sircar, 2000, Derivatives in Financial Markets with Stochastic Volatility (Cambridge University Press, Cambridge, UK). Frederic, W. Cook & Co. 2006, The 2006 Top 250: Long-term Incentive Grant Practices for Executives (Frederik W. Cook & Co. New York). Hall, B. J., and K. J. Murphy, 2002, Stock options for undiversified executives, Journal of Accounting and Economics 33, 3–42. Heston, S. L., 1993, A closed form solution for options with stochastic volatility with applications to bond and currency options, Review of Financial Studies 6, 327–343. Huddart, S., 1994, Employee stock options, Journal of Accounting and Economics 18, 207–231. Huddart, S., and M. Lang, 1996, Employee stock option exercises: an empirical analysis, Journal of Accounting and Economics 21, 5–43.

815

Hull, J., and A. White, 1987, The pricing of options on assets with stochastic volatility, Journal of Finance 42, 281–300. International Accounting Standards Board, 2004, International Financial Reporting Standard 2 ‘Share-based Paymenť (IASB, London, UK). Johnson, S. A., and Y. S. Tian, 2000, Indexed executive stock options, Journal of Financial Economics 57, 35–64. Kulatilaka, N., and A. J. Marcus, 1994, Valuing employee stock options, Financial Analysts Journal 50, 46–56. Lando, D., 1998, On Cox processes and credit risky securities, Review of Derivatives Research 2, 99–120. Leisen, D. P. J., 2000, Stock evolution under stochastic volatility: a discrete approach, Journal of Derivatives 8, 9–27. Markowitz, H., 1952, Portfolio selection, Journal of Finance 7, 77–91. Merton, R. C., 1990, Continuous Time Finance (Blackwell, Oxford, UK). Nandi, S., 2000, How important is the correlation between returns and volatility in a stochastic volatility model? Empirical evidence from pricing and hedging in the S&P 500 index options market, Journal of Banking and Finance 22, 589–610. Nelson, D. B., and K. Ramaswamy, 1990, Simple binomial processes as diffusion approximations in 816

financial models, Review of Financial Studies 3, 393–430. Sharpe, W. F., 1964, Capital asset prices: A theory of market equilibrium under conditions of risk, Journal of Finance 19, 425–442. Stein, E., and I. Stein, 1991, Stock price distributions with stochastic volatility: An analytic approach, Review of Financial Studies 4, 727–752

We acknowledge the helpful comments of Tristan Boyd, John Gould. Steven Huddart. Ross Mailer, Tom Smith, Stephen Taylor and participants in the FIRN Symposium on Valuing Executive Share Options held in Sydney, Australia, on 13 February 2006. We also thank the UWA Business School for financial support. Received 16 May 2007; accepted 15 October 2007 by Robert Faff (Editor). 1

The incentive capacity of ESOs is the elasticity of the value of the total remuneration package including ESOs with respect to the company's stock price, measured at grant date. 2

Under previous US GAAP (APB Opinion No. 25, issued October 1972), options that did not vest until they satisfied performance conditions would have to be recognized as an expense, when they vested, equal to their intrinsic value. It is perhaps unsurprising, then, mat performance vesting conditions have been unpopular among US companies, with their history of strident

817

opposition to treating ESOs as an expense. According to Frederic W. Cook & Co. (2006), 88 per cent of the top 250 US firms had ESO plans (down from 95 per cent in 2004) but only 3 per cent had performance-related vesting conditions (down from 6 per cent in 2004). Despite the trend, Frederic W. Cook & Co. (2006) forecasts vesting conditions will become increasingly common in the USA as well. 3

IFRS 2, Appendix B (para. B9).

4

The analysis here is based on the simplifying assumption that the ESO values from the company's and employee's perspective coincide. Our approach can be in principle extended to include a valuation method for the ESO holder that takes account of risk-aversion. Although this might be an important topic, it is beyond the scope of the present paper. 5

We acknowledge that, when studying the incentive effects of options (including any ability employees might have to affect the evolution of the stock price or to profit from their private information) and early exercise behaviour, the employee's perspective is more relevant. This is often addressed in a utility-maximization framework (see Huddart, 1994; Kulatilaka and Marcus, 1994; Carpenter, 1998; Hall and Murphy, 2002). However, valuing ESOs from the employee's perspective are beyond the scope of our paper. 6

IFRS 2 (para. 27) and Appendix B (para. B42) require the ESO value to be adjusted when any option terms are altered in a manner that benefits the employee. Any reload feature is not taken into account when valuing

818

ESOs for accounting purposes. 'Instead, a reload option shall be accounted for as a new option grant, if and when a reload option is subsequently granted' (IFRS 2, para. 22). 7

The term structure of interest rates is not deterministic but stochastic. However, by assuming independence of bond and stock markets, the stochastic term-sUucture collapses eventually to the non-stochastic forward rate curve, which can be used to price options. 8

The setup can be modified to accommodate other stochastic volatility models (e.g. those of Hull and White (1987) and Stein and Stein (1991)). For more details see Fouque et al. (2000, Ch. 2.3, p. 42). 9

The specification of a stochastic volatility introduces a risk in the market model that is not present in Uaded assets in the first instance. The non-traded risk is represented by Wa (r) and the financial market is incomplete. In general, options cannot be priced uniquely and risk premium parameters need to be introduced relating the real-world market dynamics and the risk-neutral dynamics. Alternatively, auxiliary assets can be introduced into the market. 10

This description applies to performance hurdles with so-called 'cliff vesting', where the ESOs can vest only at one single point in time. In practice, performance hurdles are often tested multiple times, which is refened to as 'graded vesting'. In essence, an ESO with 'graded vesting' can be stripped to a portfolio of ESOs with 'cliff vesting'. 11

See IFRS 2 (para. 19) and Appendix A.

819

12

Arguably, accounting variables are correlated with the prices of assets traded in financial markets. The problem of market incompleteness persists as long as the variation of the accounting variables cannot be explained fully by a set of financial market variables, which is undoubtedly the case. 13

We ignore any managerial incentives to 'manage' the accounting numbers. 14

The assumption of independence of market conditions and non-market conditions is suggested by IFRS 2, Para. 19: '(v)esting conditions, other than market conditions, shall not be taken into account when estimating the fair value of the shares or share options at the measurement date'. The treatment required by IFRS 2 implies that the ESO value can be written as the product of the expected number to vest and the ESO value of the vested instruments. In an option pricing framework, such a decomposition arises in the case where performance hurdles and market conditions determining the ESO value are independent. In a case study (Section 4.1) we scrutinize this assumption by allowing non-market and market conditions to be dependent. 15

It is also possible to specify an indexed ESO (see Johnson and Tian, 2000). Then, the exercise price is not fixed at grant date but is dynamically expressed in terms of the relevant index. 16

A two-dimensional stochastic volatility model could be developed, following Fouque etal. (2000). 17

A multidimensional Black–Scholes model incorporating the ASX 50 would have 1325 parameters: 820

50 dividend yields, 50 volatilities and 1225 for the correlation structure. 18

To provide more background, Metcash experienced severe financial difficulties in 1997 and was acquired by Metro Cash & Carry South Africa in 1998. Up until May 2005, the South African company retained control but Metcash is now Australian-owned. Metcash later launched a successful takeover bid in conjunction with Woolworths for rival Foodland Associated. Foodland Associated controlled a small share of the Australian supennarket industry, which was in contrast to its market share of approximately 75 per cent in New Zealand. These factors have had an impact on the dividend policy of Metcash and have also had a significant impact on trading in Metcash securities. 19

EPS is adjusted for dilution with respect to the 2001 reference date. 20

The issue of estimation is beyond the scope of this paper and, hence, we have used parameters that are close to others reported in the literature, such as Nandi (2000) and Bates (2006), who investigated the stochastic volatility model of Heston (1993) for the S&P 500 index. Bates (2006) gives in Table 7: K = 5.94, Ve = 12.5%, r) = 0.315, pSa = −0.597. We have adjusted those estimates to reflect the high fluctuations in Metcash's past volatility. 21

Boyd et al. (2007) found that the typical Australian ESO has a vesting period of 2 years, maturity of 5 years and an average lifetime of 40 months. The assumed

821

parameter choice produces an expected lifetime of the typical Australian ESO of 40 months. 22

The value reduction can be decomposed into two components. The first component is attributed to the leverage effect due the negative correlation of psσ = −0.75. The second component captures the time dependence of the expected volatility (or volatility term structure) inherent in Heston's stochastic volatility model. For Metcash an initial spot volatility of σ0 = 20 per cent is assumed, and a long-run mean of . The average volatility over the life time of the ESO is then and results in a Black–Scholes price of $AO.934. The two components contribute equally to the value reduction. 23

Quantity equals N when setting the total remuneration to the value of the stock (i.e. W = S).

© 2008 The Authors; Journal compilation © 2008 AFAANZ

822

Index A AAA (American Accounting Association) 11, 22–23, 51, 264, 281–282 AAER (Accounting Releases) 236

and

Auditing

Enforcement

AARF (Australian Accounting Research Foundation) 175, 179, 181, 271, 280–282, 319, 321, 326, 330, 344 AAS (Australian Accounting Standard) 184 AASB (Australian Accounting Standards Board) 16, 269, 271, 275–282, 317–339, 365 AASC (Australian Accounting Standards Committee) 270–272, 328, 335 ABA (Australian Bankers’ Association) 329 abandonment option 155, 164 abnormal performance index see API Accounting and (AAER) 236

Auditing

Enforcement

Releases

Accounting and Finance Association of Australia and New Zealand see AFAANZ accounting: courses 2, 16, 48; information 241, 264–265, 274, 287–289, 292–293; numbers 8, 51, 53, 60, 149, 185–186, 237, 264, 293, 303–304, 309, 391, 395; policies 51, 303; quality 302, 303, 309

823

accounting standards: Clayton’s 279; costs and benefits 297–313, 338; domestic 17, 20, 175–182, 184, 187, 257, 264, 269–276, 297–313, 315–344, 345–380; economic consequences 269–270, 297; harmonization 16, 263, 270, 272, 276, 282, 287–295, 326–327; impossibility of Pareto optimality 264, 333; international 17, 271–276, 297–313, 315–316, see also AAS; IFRS; politics and process 315–344; principles-based 365–366; research opportunities 267, 308–309; responsiveness 336–337; role 22, 298–299, 315–319 Accounting Standards Board see ASB Accounting Standards Review Board see ASRB accuracy of analysts’ forecasts see forecasts, analysts acquisitions 277–278, 302, 361, 368, 375–384 actuaries 132, 238 Adelaide Steamship (Australian company) 21, 175–182; adopters; voluntary 300, 303, 306–307; mandatory 300–302, 306, 329 adoption of: LIFO 52; U.K. government initiatives 238; Uniform Companies Acts (Australia) 69; replacement cost depreciation 186; equity accounting 279; IASC standards 272, 299, 316, 323–324, 326, 329–339; IFRS 297–309, 345–348, 356, 362–363, 367–370 AFAANZ (Accounting and Finance Association of Australia and New Zealand) 18 agency costs see cost, agency

824

AGSM (Australian Graduate School of Management) 6–7, 9, 11, 151, 192, 242, 268 AICD (Australian Institute of Company Directors) 329, 334 Amcor (Australian company, ESOs) 403–414 American Accounting Association see AAA American option 393, 400, 406, 410 analysts see security analysts analyst forecasts see forecasts announcements: price-sensitive 13, 18, 234, 247, 270; types 13, 85, 96–97, 98, 104, 105, 109–110, 112–113, 119 annual report 32–44, 52–54, 63–69, 73–80, 115, 179–180, 188, 192, 239, 267, 276, 289–290, 304, 350–357, 403–410 anomalies 9, 19, 50–51, 55–56, 60, 145–151, 170 API (Abnormal Performance Index) 36–42, 66–69, 77–82 arbitrage 151, 347, 400 ASA (Australian Shareholders Association) 279, 336 ASB (Accounting Standards Board, U.K.) 320, 331–334 ASC (Australian Securities Commission; predecessor to ASIC) 208, 276, 278–280, 327–328, 332–333, 349–350, 386–387

825

ASIC (Australian Securities and Investments Commission; formerly ASC) 16, 317–319, 321–322, 327, 332, 334, 345–380 ASRB (Accounting Standards Review Board) 184, 317–321 assets: acquisition 96, 104, 361, 368, 375, 378–380, 383–386; disposal 96, 104; revaluation of 11, 183–192, 195, 198–203, 274–275, 289–291, 339; traded 393–394 ASX (Australian Securities/Stock Exchange): announcements process 13, 18, 85–120, 209–221, 240–245, 292, 367; announcement types 85, 87, 96–98, 114–120; Corporate Governance Council 233, 237–238, 252, 259; corporate governance principles 233–239, 252–258; intraday stock and flow trading metrics 95–96; listing rules 89, 114–118, 146, 153, 160, 166, 208–211, 229, 240, 243, 247, 252, 258, 270; query (of a listed company) 96, 104–106, 112–113, 118, 218; regulatory environment 13, 208, 318–319, 327–334, 3149, 361; support for international standards 17, 271–273, 325–334; surveillance (see ASX, regulatory environment and surveillance, ASX); trading platform (SEATS) 85–120, 216 audit committee 235–236, 240, 257, 354, 356 auditor: audit costs 128, 184; Big N auditor 148, 189, 237, 302, 352, 356; choice of 189, 309; independence 259, 298, 307–308, 354, 366; other issues 258, 301, 307–310, 347, 349, 351, 354–368, 390 Australian Accounting Research Foundation see AARF Australian Accounting Standards Board see AASB 826

Australian Accounting Standards Committee see AASC Australian Associated Stock Exchanges (merged to form ASX) 132, 178, 192 Australian Bankers’ Association (ABA) 329–330 Australian GAAP 273, 289, 304 Australian Graduate School of Management see AGSM Australian Institute of Company Directors (AICD) 329, 334 Australian Institute of Valuers 184, 189 Australian Research Council 18–19 Australian Securities and Investments Commission see ASIC Australian Securities Commission see ASC Australian Securities Exchange see ASX Australian Shareholders Association see ASA Australian Stock Exchange see ASX B B/P (book-to-price) ratio 154, 159–169 banks 128, 192, 266, 268, 299–300, 305–306, 373 BARCEP Australian database 212–227 BCA (Business Council of Australia) 326, 334 Bell Group/Bell Resources (Australian companies) 21, 175–182

827

better-governed firms see corporate governance BHAR (buy-and-hold abnormal return) 155–169 BHP (Australian company) 20, 65, 103, 175–182, 186, 245, 330 bid and ask prices 58, 88, 290–291, 306–309 Black-Scholes (option valuation) 21–22, 176, 389–411 bonus share issue 65, 133, 176, 180, 184, 191–195, 198, 203 book-to-price ratio see B/P book value 8, 126, 152–164, 173, 183–188, 193, 256, 268, 280, 289, 304–305, 309, 375 borrowing 52, 128, 159, 166, 184, 188–191, 201, 300, 373 Business Council of Australia (BCA) 326, 334 C calendar time abnormal return see CTAR call option 20, 176–180, 217, 391, 403, 407–408, 413 CAMAC (Corporations and Markets Committee; formerly CASAC) 16, 270, 333

Advisory

CAO (Company Announcements Office of ASX) 89 capital gains tax 10, 14 CAPM see pricing model, capital asset CASAC (Companies and Committee) see CAMAC 828

Securities

Advisory

C.C.A. (Current Cost Accounting) 8, 125–129, 186 Center for Research in Security Prices see CRSP Centre for Research in Finance see CRIF CESR (Committee of European Securities Regulators) 346–347, 363–369 CLERP (Corporate Law Economic Reform Program, Australia) 269–272, 315–316, 325–339, 350, 362, 367 coefficients, earnings response 51, 56, 265, 289, 310 Committee of European Securities Regulators see CESR Companies and Securities Advisory Committee see CAMAC compound interest method 8, 383–384 confirmation effects 29, 63 continuous disclosure 150, 264, 270, 367 convertible securities 58, 65 corporate governance 351–352, 356, 362–363

18–19,

233–261,

307–309,

Corporate Governance Council see ASX Corporate Governance Council Corporate Law Economic Reform Program see CLERP Corporations and Markets Advisory Committee see CAMAC Corporations Law (Australia) 146, 150, 153, 208–209, 270, 278–279, 328, 333

829

cost of capital 17, 56, 126, 211, 219, 271, 293, 297–310, 323, 337 costs: adverse selection 306, 309; agency 22, 184, 239, 298, 316–319, 339; compliance 17, 186, 258, 288, 291–292, 298–308, 322–344, 345–367; contracting 11, 183–186, 198, 203, 278–279; executive remuneration 21, 54, 58, 128, 258, 282, 389–415; litigation 219–221, 236, 358; multiple standards 271–273; information/ preparation 189, 287, 327, 337; political 11, 52, 183–203, 294; proprietary 150, 160, 170, 210, 236, 256; regulation 300, 338, 350; replacement 52, 126–128, 186 (see also C.C.A.); standard setting 300, 319, 326; transaction 21, 41–42, 73, 91, 135–136; see also costs and benefits costs and benefits: accounting regulation 269; asset revaluations 183–204; disclosure of executive remuneration 258; enforcement 283, 300–302, 305–308; foreign listing 291; harmonization of standards 272–273, 288, 292–293; IAS/IFRS adoption 297–313, 326; international diversification 288; new standards 125–128, 278, 339; stock exchange trading halt 89 CRIF (Centre for Research in Finance, AGSM) 151–159 CRSP (Center for Research in Security Prices, Univ of Chicago) 5, 34–36, 48–49, 76, 257 CTAR (calendar time abnormal return) 148 Current Cost Accounting see C.C.A. D

830

debt 12, 52, 58, 147–203, 236, 256, 308–310, 316, 383 depreciation 8, 127, 184, 186–189, 378–380, 384, 387 disagreement: among academics 10, 60; among analysts see forecasts, analysts disclosure: document 96, 104–106, 116; frequency 18, 233–256; policies 18, 236, 309; public 87, 93–94, 150, 293; voluntary 11; other 52, 128, 177–182, 202, 266, 271; see also continuous disclosure discretionary accruals 148, 302 discriminant analysis 69 dispersion of disagreement

forecasts

see

forecasts,

analysts,

dissemination server (of ASX) 90–91, 104, 119 dividends: announcement 42–45, 54, 65, 98, 112–113, 117, 216–218, 241; corporate policy 10, 188, 191; drop-off ratio 14; forecast 14–15, 191; imputation tax credit 14; irrelevance 278; reinvestment plan 151; yield 55, 235, 393–412 documents: announcement document 13, 85–120, 393–412; count of number 216–230, 233–256, 281 drift: API/stock price 38–41, 51, 54–55, 64, 68; earnings 40, 76, 81; post-earnings announcement (PEAD) 5, 55–56, 68 E early exercise, of options see ESOs earnings management 54, 147–148, 237, 302–303, 307 831

EMH (Efficient Market Hypothesis) 28, 31, 50, 55, 59–61, 72, 94, 131–132, 148, 151, 237, 288 employee stock options see ESOs enforcement: ASIC and FRRP compared 345–380; ASX see ASX regulatory involvement; benefits 210, 229, 298, 300–302, 305–308; difficulties 150; mechanisms 17, 210, 229, 291, 320, 325, 345–380 EPS forecasts see forecasts equity: accounting 263–264, 278–279; instruments 394, 407; SEO underperformance see SEOs ESOs (executive/employee stock options): case studies Amcor and Metcash 403–414; hedging risk 394, 412–414; valuation 21–22, 266, 389–415 exchange-traded options 52, 176, 389 executive share options see ESOs F fair value 21–22, 266, 308, 389–415 FASB (Financial Accounting Standards Board) 263–266, 274–275, 280–282, 317–327, 334, 383, 387, 390 FEE (Fédération des Experts-Comptable Européens) 324, 347, 356, 365, 367 Financial Accounting Standards Board see FASB financial instruments 175–182, 266, 281, 317, 360, 367–368 Financial Integrity Research Network see FIRN

832

Financial Reporting Council see FRC Financial Reporting Panel see FRP Financial Reporting Review Panel see FRRP Financial Services Authority see FSA financial slack 188–203 firms, loss-making 145–173, 208, 380, 386; see also timeliness, accounting FIRN (Financial Integrity Research Network) 21 focal points 15 forecast error see forecasts, accuracy forecast horizon 12, 154, 243–248, 254–255 forecasts, analysts: accuracy 207–231, 233–261, 270, 301–302; bias 233–261, 301; disagreement 201, 207–231, 233–261, 270, 301, 303, 307; following 233–261, 301; managers 15, 53, 213, 374; naïve models 31–54, 64–76, 81; time-series 53–54, 63 FRC (Financial Reporting Council) 270–272, 320–322, 328–339, 350–366 FRP (Financial Reporting Panel, Australia) 367 FRRP (Financial Reporting Review Panel, U.K.) 322, 345–380 FSA (Financial Services Authority) 350–366 fundamental value 101, 149, 151, 159, 171

833

G globalization of capital markets 60, 297 good and bad news 5, 18, 32, 42, 52–53, 64–67, 72–78, 100, 191, 211–212, 219–230, 234–255, 277, 304 goodwill 8, 263–264, 274–278, 290, 301–304, 368–369, 374–379, 383–387 groups: comparator (ESO valuation) 22, 398, 404, 408–412; liaison 319, 321 growth options 191, 236 H half-yearly reports see HYRs harmonization of accounting standards 16–17, 263–264, 270–282, 287–288, 291–293, 323–339 haziness (of value) 15 hedge: currency 58; risk with ESOs see ESOs Horwath corporate governance reports 18, 234–247, 258–259 HYRs (half-yearly reports) 65–69, 77, 119, 215–217, 225–226 I IAS (International Accounting Standards) 270–282, 288, 299–310, 315–332, 390

16–17,

IASB (International Accounting Standards Board) 21, 299–310, 316, 348, 356, 365–366

834

IASC (International Accounting Standards Committee) 270–274, 284, 300, 315–344 ICAA (Institute of Chartered Accountants in Australia) 318–319, 326, 330–337 ICAEW (Institute of Chartered Accountants in England and Wales) 297 IFAC (International Federation of Accountants) 324, 331 IFRIC (IFRS Interpretations Committee) 365–368 IFRS (International Financial Reporting Standards) 16, 23, 297–313, 316, 345–370, 390–407 IFRS Interpretations Committee see IFRIC immediacy in trading 15, 106 imputation tax credit see dividends incentives: accounting quality 303, 310; agents 316; policy choice 11, 51, 183–204, 303, 306–307; disclosure 210–211, 213, 230, 236; earnings management 307, 395; information acquisition 87, 213, 243, 271, 293; regulators, researchers and standard setters 266 Income Information (II: Ball and Brown definition) 42–44 income numbers 3, 27–46, 49–51, 63, 68, 241 independence: auditor 298, 307–308, 316, 338, 366; board/director 234, 239–240, 257–258, 319, 326; judicial 307; regulator 346–349, 364–369; standard setter 328, 331–337, 355; valuer 189, 195, 201–202, 374 inflation 8, 60, 125–128, 187–192, 201, 287–288, 299 835

information: advantage 87, 148, 150, 230, 292–293; asymmetry 87–111, 148–150, 188–203, 288, 291–292, 301; inside 13, 150; private 5, 87, 91, 93–94, 106, 146, 150, 213, 292, 392; transfers 51–59, 265, 290, 337 informed traders see investors, informed/uninformed initial public offering (IPO) 15, 154 Institute of Chartered Accountants in Australia see ICAA Institute of Chartered Accountants in England and Wales see ICAEW Institute of Public Accountants see IPA interest: groups 315, 317, 319, 325–339; public 315, 317, 331, 335, 360–362 interest-rate swaps 58 International Accounting Standards see IAS International Accounting Standards Board see IASB International Accounting Standards Committee see IASC international enforcement models compared 345–380 International Federation of Accountants see IFAC International Financial Reporting Standards see IFRS International Organization of Securities Commissions see IOSCO

836

interpretations: alternative 349, 365–366; conflicting 365; cross-country differences 347, 370; demand for 366; of future 29; see also IFRIC intrinsic value 15, 20, 154, 390 inverse sum-of-the-years’-digits method see ISOYD Investor Responsibility Research Centre (IRRC) 235 investors: institutional 136, 210, 228, 230, 302, 305; informed/uninformed 13, 42, 86–109, 150, 175, 292 IOSCO (International Organization Commissions) 274, 323, 347, 366, 369

of

Securities

IPA (Institute of Public Accountants; formerly NIA, National Institute of Accountants) 319, 333, 337 ISOYD (inverse sum-of-the-years’-digits method) 8, 277–278, 383–387 L leases 11, 28, 52, 374, 380 Leverhulme Trust 18 limit order 13, 87–99, 106–108 liquidity 133, 272, 288–294, 297, 305–307, 323, 400; traders 91, 292 litigation 54, 237, 374; see also costs, litigation logit (logistic regression) 53, 201 loss-making SEO firms 145–170 M

837

management compensation 52, 54, 58, 128, 186, 236; see also ESOs management of earnings see earnings management managers dividend forecasts see forecasts, managers mandatory: adoption IAS/IFRS 298, 300–310, 329, 340; disclosures 59, 182 mark-to-market accounting 275, 317, 328 market: competitive 54, 59, 61, 288, 323, 404; conditions for vesting see ESOs valuation; corporate control 5, 10; efficient see EMH; experts 5, 131–144; informed 179, 208, 229; liquidity 294, 305–307; misvaluation 145–173; operators 299, 306–307, 351; options see options, exchange-traded; secondary 271, 288, 298, 316 Mathews Committee 15, 22, 267 Metcash (Australian company, ESOs) 403–414 Miller-Modigliani (corporate finance) 9, 48, 56, 63, 155 minority shareholders 299, 304–307 multi-period earnings forecasts 51, 57–58 N national enforcement bodies, comparison of 345–380 National Institute of Accountants (NIA) see IPA Net Information (NI: Ball and Brown definition) 42–44 New York Stock Exchange see NYSE

838

New Zealand Accounting Standards Board 282 non-market conditions for vesting see ESOs valuation non-renounceable rights see rights issues NYSE (New York Stock Exchange) 29, 31, 35, 48, 55, 59, 63–64, 68, 76, 88, 93, 108, 273, 291–294 O on-market trades see trades, on-market operating cash flow 383–387 options: abandonment 155, 164; agreements 20, 175–179; exchange traded 52, 58, 176, 213–224, 243–254, 389; executive/employee see ESOs; BHP shares 20–21, 175–182 outsiders 126, 188, 298, 308, 316 overvaluation of SEO firms see SEOs P parsimonious 53, 245, 392 payment, share-based 21, 390, 411, 414 performance, stock TSR see ESOs performance hurdle see ESOs PFS (Preliminary Final Statement) 87, 97, 103–108, 210–230, 241–246 Pierpont (author’s pen-name) 20 placement of shares see SEOs

839

political: costs 11, 52, 183–204, 294; forces 125, 219, 287–288, 299, 308, 310, 315, 338–339, 345–347, 365; process 185–186, 208, 315–344, 348–349, 362–363, 369 portfolio managers 57, 131–132, 237–238, 288 predictive ability 28, 72, 83, 149, 155, 161, 192 Preliminary Final Statement see PFS preliminary report (of EPS) 34, 41, 77 price discovery 5, 18, 89, 108, 233–241, 257, 288 Prices Justification Tribunal (Australia) 186 pricing model, capital asset (CAPM) 32, 50, 154, 391, 397, 414 private information see information, private product life cycle 59 prospectus 15, 116, 150, 160, 166, 208, 267, 351 PSASB (Public Sector Accounting Standards Board) 281, 318, 321, 327–328, 331, 333 public sector 321, 328, 331, 333 Q Quality: accounting numbers/reports 59–60, 208, 225, 236–237, 265, 272–279, 281, 287–294, 301–309, 317, 331, 345–380; accounting standards 293, 299, 305, 323, 329, 331–332, 347; see also corporate governance quarterly earnings 4, 54, 63, 71–83 R

840

random walk 4, 23, 53, 64 rate, risk-free 32, 133, 154, 181, 393–412 recognition (accounting term) 179–182, 224, 237, 279, 302–303, 345–380, 387 recommendations: ASX Corporate Governance Council 238, 257–259; Australian Securities Commission 276; Dearing Committee (U.K.) 354; FRC 329; investment experts 131–144; Mathews Committee 15–16; reductio ad absurdum argument 12 regulation/regulators: accounting standards 17, 21, 128, 175, 181, 263–270, 277, 280–283, 298–300, 308, 315–344, 345–380; companies 186, 211, 229, 243, 263–264, 269–270, 279, 315; incentives 266, 310; markets 8, 16–17, 58, 150, 210, 270, 292, 304, 307, 309, 347–351; stock exchange 114, 145, 177, 241, 291; theories 315 reload option see ESOs 392 resampling 11, 13, 102, 226, 254 research and development (R&D) costs 28, 256, 263–264, 275–276, 289, 308, 375 residual income model see RIM/RIV resource firms 152, 156, 159, 164–165, 169, 223, 225, 241–255 return, calendar-time abnormal see CTAR returns-earnings relation 265, 277, 291, 304–305 revaluation see assets, revaluation of

841

rights issues (SEOs) 109, 145–170, 212, 220 rights of minority shareholders 299, 304–307 RIM/RIV (residual income model/valuation) 145–173 S sanctions, statutory 207–231, 241, 270, 322, 352 seasoned equity offering see SEOs SEATS (Stock Exchange Automated Trading System) see ASX, trading platform Securities Industry Research Centre of Asia-Pacific see SIRCA security analysts: expertise 51, 63, 225, 291–294; recommendations 136, 239; research interests 76, 126; other 132, 225, 237–238 SEOs (seasoned equity offerings) 19–20, 145–173 share-based compensation see SEOs shareholder rights shareholders

235–236;

see

also

minority

share purchase plans (SPPs) see SEOs Signal G (ASX) 88, 91, 116, 216–217 SIRCA (Securities Industry Research Centre of Asia-Pacific) 14–15, 18, 95, 151–153, 216–217, 240–242, 267 slack, financial 188–191, 195, 198, 203 standards see accounting standards standard

842

setters/setting standards

see

regulation/regulators,

accounting

statutory sanctions see sanctions, statutory stochastic volatility see ESOs, valuation Stock Exchange Automated Trading System (SEATS) see ASX, trading platform stock split 33, 50, 76 straight-line method 277, 384–387 surveillance: ASX 13; ASIC 348–349, 352, 357, 360–362, 369; FRRP 348, 355–357, 362–363, 369 T takeovers 13, 86, 88, 115, 119, 149, 192–203, 236, 277–278, 350, 404 timeliness: accounting 42–44, 289–290, 302–309; Ball and Brown measure 8–9, 18, 42–44, 210, 233, 241; corporate governance 233–261; disclosures 207–210; price discovery 18, 233–261 timing of equity issues 145 Total Information (TI: Ball and Brown definition) 42–44 total shareholder return see ESOs trades, on-market 100–103, 108 TSR (total shareholder return) see ESOs U UIG (Urgent Issues Group) 282, 321

843

underperformance of new issues see SEOs undervaluation 150, 188 uninformed traders see investors, informed/uninformed usefulness 27–28, 44, 49, 60, 71, 83, 277, 291, 304 US GAAP 289–294, 304–308 V value reduction see ESOs value-relevant 155, 210, 226–230, 234–257, 266–277, 288–291, 304–305 vesting see ESOs voluntary: adoption IAS/IFRS 300–307; compliance with regulator 358, 361; disclosures 11, 207–231, 236 W windows of opportunity 146–148, 156, 170

844