Facts of Life: The Social Construction of Vital Statistics, Ontario 1869-1952 9780773564244

Table of contents :
Contents
List of Tables
List of Figures
Acknowledgments
Prologue
1 Statistics as Culture: The Origins of Ontario Civil Registration
2 The Evolution of Ontario Civil Registration, 1869–1950
3 Death in Ingersoll, 1880–1972: A Case-Study Approach to the Revision of Defective Mortality Statistics WITH KEVIN McQUILLAN
4 The Completeness of Birth Registrations, 1900–60
5 Getting Born: How Definition Influenced Statistics for Infant Deaths and Stillbirths, 1926–51
6 Fatal Pregnancies in Ontario, 1920–35: A Study of the Nature of Statistics for Deaths by Cause
7 Whose "Facts of Life"? The Problem of Residence in Vital Statistics, 1920–43
Conclusion
APPENDICES
A: Ontario-Quebec Differences in the Application of Canada's 1932 Statistical Definitions for Live Birth and Stillbirth: Discussion of the Evidence
B: Ontario Death-Certificate Forms
C: Definitions of Selected Medical Terms
D: Rules for Choice of Causes of Death in the Dominion Bureau of Statistics, 1933
E: Problems Encountered in Linking Death Registrations to Published Statistics
F: The Evidence of Residence Classification Practices in Vital Statistics, 1930–36
G: Hospital Beds and Reallocations for Residence: Ontario in the 1930s
Notes
References
Index
A
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P
Q
R
S
T
U
V
W
Y

Citation preview

Facts of Life

This page intentionally left blank

Facts of Life The Social Construction of Vital Statistics, Ontario 1869-1952 GEORGE EMERY

McGill-Queen's University Press Montreal & Kingston • London • Buffalo

McGill-Queen's University Press 1993 ISBN 0-7735-1111-3

Legal deposit third quarter 1993 Bibliotheque nationale du Quebec

Printed in Canada on acid-free paper This book has been published with the help of grants from the Social Science Federation of Canada, using funds provided by the Social Sciences and Humanities Research Council of Canada, and from the J.B. Smallman Publication Fund, Faculty of Social Science, University of Western Ontario.

Canadian Cataloguing in Publication Data Emery, George Neil, 1941Facts of life: the social construction of vital statistics, Ontario, 1869-1952 Includes bibliographical references and index. ISBN 0-7735-1111-3

1. Ontario - Statistics, Vital. 2. Ontario - Population Statistics, I. Title. HA39-C3306 1993 304.6'09713'021 C93-090237-8 Typeset in Palatino 10/12 by Caractera production graphique inc., Quebec City.

Contents

List of Tables vii List of Figures xi Acknowledgments xiii Prologue 3 1 Statistics as Culture: The Origins of Ontario Civil Registration 17 2 The Evolution of Ontario Civil Registration, 1869–1950 31 3 Death in Ingersoll, 1880–1972: A Case-Study Approach to the Revision of Defective Mortality Statistics WITH KEVIN MCQUILLAN 50 4 The Completeness of Birth Registrations, 1900–60 72 5 Getting Born: How Definition Influenced Statistics for Infant Deaths and Stillbirths, 1926-51 98 6 Fatal Pregnancies in Ontario, 1920–35: A Study of the Nature of Statistics for Deaths by Cause 116 7 Whose "Facts of Life"? The Problem of Residence in Vital Statistics, 1920–43 137 Conclusion

155

vi Contents APPENDICES

A Ontario-Quebec Differences in the Application of Canada's 1932 Statistical Definitions for Live Birth and Stillbirth: Discussion of the Evidence 165 B Ontario Death-Certificate Forms

167

c Definitions of Selected Medical Terms 169 D Rules for Choice of Causes of Death in the Dominion Bureau of Statistics, 1933 170 E Problems Encountered in Linking Death Registrations to Published Statistics 180 F The Evidence of Residence Classification Practices in Vital Statistics, 1930-36 184 G Hospital Beds and Reallocations for Residence: Ontario in the 1930s 186 Notes

193

References Index 239

223

Tables

2.1 A Comparison of Age-Specific Death Totals, 1891 and 1901 40 2.2 Number of Enumerated Deaths per 100 Registered Deaths in Ontario, 1880-81, 1890–91, 1900–01 42 2.3 Census Deaths, Registered Deaths, and Registered Deaths as a Percentage of Census Deaths 42 3.1 Published Statistics for Ingersoll Deaths by Place of Occurrence and Place of Residence 57 3.2 Ingersoll Area, 1930–32, 1940–42: Deaths by Place of Residence and Place of Occurrence 59 3.3 Sekar-Deming Estimates of Completeness, Ingersoll-Area Mortality Data, 1871–75 to 1891–95 61 3.4 Sekar-Deming Estimates of Completeness, Mortality Data for Selected Population Groups in the Ingersoll Area, 1871–90 62 3.5 Ingersoll-Area Persons in the Manual-Simple Occupation Category as a Proportion of Classifiable Persons: 1871 and 1881 Enumerated Census Populations and Death Registrations for 1871-74 and 1881-84 63 3.6 1871 and 1881 Enumerated Census Populations for Ingersoll Area: Age Composition by Occupational Categories 64

viii Tables

3.7 Age-Specific Death Rates for the Ingersoll Area, 1871–90 to 1950–72 67 3.8 Expectation of Life at Specified Ages, Ingersoll Area and Ontario, 1881-1900 to 1950–72 68 4.1 Searches, Certificates Issued, and Birth Certificates Issued 77 4.2 Stated Purposes for Searches of Vital Records, 1926 77 4.3 Delayed Registrations for Selected Cohorts by Period When Registration Was Granted 80 4.4 Delayed Registrations for Selected Cohorts by Age Category 81 4.5 Proportions by Cohort of Delayed Registrations Granted by 1986 That Were Issued before 1945 84 4.6 Estimates of Incompleteness in Birth Registrations for the 1900 Cohort, Assuming Different Survival Rates 86 4.7 Estimates of Unregistered Births for the 1900, 1905, and 1930 Cohorts as of 1986 86 4.8 Accuracy of Step 3 Estimates for Predicting Known Values for Previous Age Categories 88 4.9 Cases Added in Step 3 as Percentages of Birth Registrations and Estimated Birth Totals 88 4.10 Ontario Live Births: Published Totals and Step 4 Estimates 90 4.11 Predicted and Actual Numbers of Delayed Registrations of Births in the Sample 92 4.12 Proportions of Fathers' Occupations in Different Occupational Categories, for 1920: Birth Registrations and Delayed Registrations 96 5.1 Ontario and Quebec Reported Rates for Stillbirths, Deaths in the First Day of Life, Neo-natal Deaths, and All Infant Deaths, 1926–50 101 5.2 Ontario Stillbirths Registered as Births and as Deaths 103

ix Tables

5.3 Canada, 1930: Provincial Registrars' Answers to Questions Concerning Gestational Age and Criteria for Life 105 6.1 Deaths Reported for the Puerperal-State Class for 1920 and 1930 122 6.2 Percentage of International Assignments to the Puerperal-State Class in 1931 for 1,073 United States Deaths Associated with Pregnancy 124 6.3 Deaths of Women Not Classified to Pregnancy or ChildBearing but Returned as Associated Therewith, Ontario 1930 126 6.4 Documentary Basis for the Identification of Deaths as Maternal 128 6.5 Four Registered Deaths for 1920 and Matching Descriptions in Published Statistics for the PuerperalState Class 130 6.6 Maternal Deaths by Link-Group Status, 1920 and 1930 131 6.7 Ontario Maternal Deaths: Published Totals and Totals Found in Contemporary Investigations 133 7.1 Non-resident Births and Deaths Reported for Ontario in Vital Statistics and the DBS Special Reports, 1930–36 143 7.2 Non-resident Vital Events per 1,000 Registered Vital Events 144 7.3 Sanatoria, Asylums, and Hospitals for Incurables in Ontario, 1930–36: All Deaths and Non-resident Deaths per 1,000 Population in Municipality of Occurrence 145 7.4 Ontario: Four Greater Cities and Their Satellite Communities, 1931 147 7.5 Surplus Non-resident Events in Vital Statistics Relative to the DBS Special Report Totals 148 7.6 Effect of Registration Data Errors on the Birth and Death Rates Calculated for Ontario's Five Largest Cities, 1930-32 149

x Tables

7.7 Births and Deaths by Occurrence and Residence: Mean Values Calculated from Data in DBS Special Reports for the Period 1930-36 150 7.8 1933 Ontario Hospital Beds per 1,000 Population

151

7.9 Origins of Canadians Who Had Non-resident Vital Events in Ottawa, and Number of Residents of Hull and Pontiac Counties, Quebec, Who Had Vital Events outside Their Home Counties 152 7.10 Non-resident Births and Deaths Involving Residents of Other Canadian Provinces 152 7.11 Estimated Numbers of Ontario Residents Who Gave Birth or Died in Quebec and Quebec Residents Who Gave Birth or Died in Ontario; Percentages of Ontario Totals for Quebec Resident Events in Ottawa and the Four Towns of Cornwall, Pembroke, Renfrew, and Hawkesbury 153 E.1 Puerperal Categories and Subcategories for Which the Registrar-General's Annual Report for 1930 Gives Conflicting Information 181 F.1 Vital Statistics Data for Deaths and Non-resident Deaths in Public Institutions, 1931-36; and Non-resident Deaths in Vital Statistics and the DBS Special Reports, 1930-32 and 1935-36 185

Figures

3.1 The Study Area: Ingersoll, North Oxford Township, and West Oxford Township 52 3.2 Ingersoll-Area Mortality File: Composition by Documentary Source 56 3.3 Crude and Adjusted Crude Death Rates for the Ingersoll Area 65 3.4 Female Advantage in Life Expectancy at Birth, Ingersoll and Ontario 70 4.1 Registered Live Births Occurring in Hospitals

74

4.2 Licensed Motor Vehicles per 1,000 Ontario Population 76 4.3 Registered Live Births per 1,000 Ontario Population 79 4.4 Estimated Incompleteness in Live Birth Registrations 89 4.5 1920 Cohort: Proportions of Delayed Registrations by Age Categories 93 4.6 Ratio of Male to Female Registered Births, Ontario and England, 1881-85 to 1956–60 95 7.1 Registered Live Births and Deaths in Public Institutions 140

This page intentionally left blank

Acknowledgments

I thank the Aid to Scholarly Publications program and McGillQueen's University Press for arranging two appraisals and four reader evaluations of my work. I am indebted to each of the four readers for committing the time to do the work and completing it within a reasonable time. Due to their criticism, the book has an additional chapter and is more sharply focused, integrated, and clearly written than the version first appraised. I also thank Susan Kent Davidson for her meticulous copy-editing, and Philip Cercone, executive director and editor of McGill-Queen's University Press, for noting merit and potential in the manuscript as first appraised and guiding the writer through the shoals of the appraisal process. Portions of the work have appeared in scholarly journals. Chapters i and 2 draw from "Ontario's Civil Registration of Vital Statistics, 1869-1926: The Evolution of an Administrative System," Canadian Historical Review 64, no. 4 (Dec. 1983): 468-93. The basis for chapter 3, written with Kevin McQuillan, is "A Case Study Approach to Ontario Mortality History: The Example of Ingersoll, 1881-1972," Canadian Studies in Population 15, no. 2 (1988): 135-58. Chapter 4 revises "Incomplete Registration of Births in Civil Systems: The Example of Ontario, Canada, 1900-1960," Historical Methods 23, no. i (Winter 1990): 5-21. Through their criticisms of the article manuscripts, the journal editors and anonymous reviewers contributed substantially to the development of this book. Like other authors, I have accumulated personal debts along the way. My major obligation is to Kevin McQuillan, my collaborator for

xiv Acknowledgments the Ingersoll mortality case study. Kevin introduced me to historical demography and demographers' techniques for dealing with difficult empirical problems. He read most of my work in raw form. More than he knows, his offhand comments gave me the confidence to tackle difficult problems and helped me to place my work in perspective. At the eleventh hour, after the appraisal process was complete, Bruce Curtis, of the Department of Sociology, Wilfrid Laurier University, introduced me to a burgeoning literature on capitalist state formation. His own work especially helped me to sharpen the argument in chapter i and portions of later chapters. Suzanne Shiel and the Population Studies Centre at the University of Western Ontario provided helpful guidance on bibliographic and archival resources pertaining to my interests. The centre also has given the writer and other researchers at Western a supportive, interdisciplinary environment for research in historical demography. I have long-term debts to Dick Alcorn and colleagues on the Landon Project at Western during 1977 and 1978. Dick introduced me to social-science research methods, mainframe computing skills, and the immense pleasure of interacting with kindred spirits on interdisciplinary problems. I have a special debt to Jose, Maria-Therese, Karine, and Josee-Nadine Igartua for their friendship and hospitality. Had they not prodded me into a series of mid-life Frenchlanguage immersions, I would never have discovered the extraordinary richness of la documentation frangaise for historical demography. I thank three deputy registrars-general - E.W. Pike, Rosemarie E. Gage, and Despina H. Georgas - for granting me access to the Records Room of the Ontario Registrar-General's office, where the registrations then were stored; and Dale Demarche and the Records Room staff for their assistance and toleration of whatever disruption my presence created for them. Ed Phelps, the Regional Collection librarian at Western and a motherlode of information on archival sources, made it possible to assemble materials for the Ingersoll case study. Many other library staff guided me through government publications or found hard-to-locate publications through interlibrary loan. Helen and Wilfred Ziegler made me welcome in their home for the research in Toronto. Both gave me their interest and encouragement. Wilf, a medical doctor, also shared his expertise on pregnancyassociated causes of death. Several colleagues criticized portions of the manuscript; I particularly thank Jack Blocker, Jr., J.J. Talman, Jose Igartua, Peter Neary, Ben Forster, Ian Steele, Roger Hall, and Jim

xv Acknowledgments Woycke. My sister, Elaine Balpataky, read the second appraised version for faults of style. Finally I thank my family for their support through the years: my late father, Dr George H. Emery, and mother, Jean (O'Neil) Emery; my children Herb, Neil, Alexandra, and Nicholas; and my wife Anne Sawarna.

This page intentionally left blank

Facts of Life

This page intentionally left blank

Prologue

Vital statistics - statistics of birth, marriage, and death - are a documentary staple for demography and important areas of social history. Yet the data collected for historical and late-twentieth-century third-world populations are suspect for statistical purposes, and scholars have differing views about their research value. Some take the statistics at face value, on the assumption that they are reasonably good and better than nothing at all. Others reject them on the grounds that they mislead and cannot be corrected. A middle-ground approach, the best of the three, is to use the data, but with revision for errors and lacunae. If limited to technical adjustment, even the third approach underestimates the problems with historical vital statistics. All statistics are socially constructed. What they contain and omit reflects the concerns of their collectors and the particular arrangements for collection. Their meaning also varies according to the social conditions in which they are collected and the purposes for which scholars use them. Because statistics are cultural phenomena, they require social interpretation, not mere technical correction. Yet if the problems of vital statistics are greater than believed, so too is their research potential, provided that scholars build on their social properties. This book explores the social construction of Ontario vital statistics. To this end, the term social construction is broadly defined to include an external component - the ideologies, concerns, and processes "out there" in society and acting on civil registration officials; and an internal component - the way officials went about organizing

4 Facts of Life

civil registration, which was far from straightforward and powerfully influenced the statistics. In other words, the book treats statutes, regulations, the content of registration forms, definitions of critical terms, and the like as an integral part of the social construction, not as technical background material. From its emphasis on the internal component, the book derives much of its originality. By showing the pertinence of details such as the content of the registration forms or contemporary definitions for life at birth, it is unique in the literature on civil registration. As a way of developing the theme of social construction, the book provides empirical illustrations of how incomplete registration and aspects of definition influence the data. It is not a manual for historical demographers. It is, however, a critical appraisal that attempts to sensitize analysts to the problems they will encounter, and to suggest strategies for responding. As discussion of Ontario-Quebec differences in infant death statistics shows, the form of the problems encountered varies from place to place, but the problems are constant. To elaborate the direction and contribution of the book, the balance of the prologue introduces the pertinent definitions and documentary sources for vital statistics; discusses the surprisingly limited use made of vital statistics in current historical research; summarizes the demographic literature on data problems in vital statistics; discusses scholarly research agendas as a dimension of the data problems; and outlines the scope of the book and the organization of the material. DEFINITIONS AND

DOCUMENTARY

SOURCES

Vital events are "facts of life." They include live birth, death, and foetal death (stillbirth or abortion). Because societies traditionally have produced offspring through family formation, marriage, divorce, adoption, legitimation, recognition, annulment, and legal separation are also considered vital events.1 By contrast, migration is not a vital event, although it is an important demographic fact. Vital statistics is the English equivalent to the French mouvement de la population, where mouvement refers to change, not migration. The interplay of vital events may, for example, change the age-sex composition of population, or its composition by marital status. Ratios of vital events to population, known as vital rates, describe the interplay. The crude mortality rate, for example, is the annual number of deaths per 1,000 population. Because the age-sex composition of population influences the crude rate, a more refined measure is the

5 Prologue

age-sex-specific mortality rate - the annual number of deaths in a given age-sex group per 1,000 population in that group. Historical sources for vital events have included church parish registers of baptisms, marriages, and burials; bills of mortality; mortality schedules in manuscript censuses; cemetery registers; newspaper death notices; and civil registration. Increasingly after the midnineteenth century, civil registration became the principal documentary source in most countries. In Quebec government simply acquired copies of church records for its vital records. Other provinces developed secular systems for collection because comprehensive church records were lacking. Beginning in 1869, for example, Ontario registered vital events through its existing network of municipal organization. USES OF VITAL STATISTICS

IN

SCHOLARLY RESEARCH

Vital and census statistics are documentary staples for demography, the science of population. In general, the census reports a population's stock or structure at points in time. Vital statistics report its flow of vital events over time. By relating the flow to the stock, one obtains the vital rates. A period (cross-sectional) analysis compares vital rates for an entire population at different points in time. By contrast, a cohort (longitudinal, generation) analysis examines vital rates for one birth group (e.g., persons born in 1905) over time. Similarly, a period life table summarizes mortality experience for an entire population at a given moment; specifically, it reports expectation of life at different ages, were each age group to experience the survival rates of the older groups at the same point in time. By contrast, a generation life table reports the expectation of life at different ages for one birth group, based on its actual survival rates at different ages. Through demography, vital statistics is intimately linked to many aspects of social experience. Thus government in late twentiethcentury societies relies heavily on vital statistics for policy planning in areas such as public health, public schooling, management of the labour force, and budgeting for pension benefits. Similarly, historians use them in research areas such as family history and the history of public health. More generally, an understanding of demography affects all knowledge in history. French Annaliste scholars in particular view historical populations as totalities in which the resource base, demography, economy, society, politics, and belief systems interconnect.2

6

Facts of Life

Except for Quebec Roman Catholic populations, Canadian scholars make remarkably little use of historical vital statistics.3 "Blind spots" in their mainstream research traditions doubtless contribute to this neglect. Few historians know the quantitative techniques for demography, and few demographers show interest in historical periods. Further, limitations of the documentary sources discourage both demographers and historians from venturing into historical demography for English Canadian populations. Simply put, it is difficult to assemble records of vital events that are adequate for statistical purposes. Church records, for example, yield poor coverage of Ontario vital events. Unlike Quebec, which had a dominant church with uniform standards of record-keeping, Ontario has had a plethora of denominations and record types, not to mention numbers of unchurched. Extant church records for the town of Ingersoll illustrate the problem. Of the five denominations or denominational groups (e.g., Presbyterian denominations, Methodist denominations) represented, the Roman Catholic and Anglican churches alone recorded church ceremonies (infant baptism, marriage, and burial) that can be linked to vital events (birth, marriage, and death). Even here, the persons involved in the ceremonies are tricky to place in the town because the parish boundaries for each church changed over time and usually differed markedly from municipal boundaries. Population censuses are a more promising source. Although these hold no direct evidence for birth or marriage events, Jacques Henripin, for example, estimates births from census data for children in the 0-4 and 5-9 age groups. Similarly, one can estimate nuptiality from "the proportions ever-married in critical age-groups (15-19, 20-24, etc.)."4 Nevertheless, census data (population stock) have limitations for estimating population trends (flow). For example, the effects of migration, between-censuses mortality, and the underenumeration of children in censuses make Henripin's census-based estimates problematic. Furthermore, manuscript censuses are needed for recordlinkage of individuals in successive censuses, and these are unavailable until 1852 and subject to restricted access after 1901. Finally, mortality statistics in the 1911 and earlier censuses are unreliable, having been obtained through the flawed method of a retrospective question.5 Given the deficiencies and limitations of church and census records, one might expect that Canadian scholars would turn with alacrity to vital statistics collected through civil registration. Yet civil registration does not rate a mention on Chad Gaffield's 1982 list of

7 Prologue routinely generated sources that inform the literature.6 Various factors explain its omission. First, continuous civil registration is of recent origin; it dates only from 1869 in Ontario, and it awaited the twentieth century in several provinces. Second, the coverage and reliability of the data are suspect for the early years of civil registration, and it is tricky to determine when the data become reliable. Published nineteenth-century death rates for Ontario, for example, appear to show rising mortality levels, whereas they actually show improvement in the registration of deaths. Third, research access to registrations is restricted. Fourth, published materials give little guidance on how to interpret summary statistics that officials calculated from the registrations. Fifth, the literature on historical vital statistics is sparse. A newly abundant literature on the nineteenth-century statistical movement, census enumerations, and church parish records provides a startingpoint for a critical assessment of vital statistics.7 Few historical studies treat civil registration, however, and none of them for twentiethcentury periods.8 As discussed below, some demography texts discuss twentieth-century statistics, but their treatment is general. Finally, other than the writer's published articles, little on Canadian civil registration has appeared since the 19305.9 DEMOGRAPHERS AND VITAL

HISTORICAL

STATISTICS

Widely used textbooks on demography summarize the demographic literature on historical vital statistics. Some texts ignore data problems, perhaps because the authors' research concentrates on developed Western populations in recent time periods, for which the data are believed to be relatively good.10 Other texts do discuss data problems, and some of their material deals with, or is pertinent to, historical contexts. Shryock, for example, notes that the civil registration of vital events in the United States only gradually became complete and that the coverage for marriage and divorce lagged behind that for deaths and births.11 He discusses five aspects of data quality: i) the collector's definition of the vital event (e.g., live birth or stillbirth); ii) the completeness of the registration coverage; iii) the accuracy of allocation by place (where the vital event occurred); iv) the accuracy of allocation by time (when it occurred); and v) the accuracy of classification of vital events by demographic and socio-economic characteristics (e.g., birth by the mother's age). Although all five problems are important in "statistically underdeveloped areas," the problem of

8

Facts of Life

registration coverage "looms largest." Then, as the first, second, and fourth problems are overcome, the third and fifth become the major ones. Allocation of vital events by place of occurrence, for example, becomes problematic as urban hospitals emerge as central sites for birth and death and rural dwellers acquire easy transportation to them.12 Some text discussion of systematic data errors is pertinent to historical records, although not referring specifically to them. Barclay, for example, notes the low reliability of ages declared for the very old and the very young (0-4); and "age-heaping" - the tendency of people to report ages ending in zero and five, which are right on the boundaries between standard age-group categories.13 Barclay, Shryock, and the United Nations publication Manual X note social influences on demographic data.14 Their examples include the false declaration of ages that have legal or ceremonial significance (e.g., voting, pension, or legal working age); the failure of parents to register a birth until a birth certificate is required for the child's school registration; and the low coverage of civil registration for births and infant deaths in societies with high infant mortality (parents of a dead infant may regard registration as futile). Similarly, Shryock urges consideration of the legal and administrative frameworks: is registration compulsory? are the administrative arrangements sufficient for enforcement throughout the jurisdiction? The literature offers good general advice on the revision of data errors. Barclay, for example, urges caution because in a given context a defective statistic ma}' be preferable. With a registration of births that is 99 per cent complete and a census enumeration only 90 per cent complete, the birth rate would be 1.10 times its "correct" level, or ten per cent too high. The crude birth rate would be more accurate if birth registration were more defective (that is, if it were more comparable to the census data). In this case, a revision of the figure for registered births, making it 100 per cent complete, would only increase the error of the birth rate. When there is doubt about the meaning of a contemplated revision, it is better not to adopt it. One solution is to make alternative calculations, showing the effect of adopting or rejecting the revision. When the amount of the adjustment rests on a subjective judgement, it is often helpful to calculate separate results based on extreme high and low assumptions about the defect. For example, the most extreme assumption about the actual sex ratio of yearly births may have so little effect that the original figure remains as satisfactory as the revision.15 Because the demographic literature on defective data is directed primarily at twentieth-century third-world populations, its value for

9

Prologue

historical studies is general and restricted in scope. Here standard techniques for detecting and revising defective data divide into two general groups.16 In the first group are sample surveys and sample registrations, which generate new data through interviews with living persons. Through comparison with the sample data, the researcher estimates the coverage and accuracy of the data collected in a traditional system (census, vital statistics). Sampling can also obtain data for questions that the traditional system omits, or modify the wording of a question in the traditional system so as to obtain more satisfactory responses. Unfortunately, sample-interview techniques are inapplicable to most historical populations, whose individuals are deceased. Techniques in the second group make use of existing data. As noted above, for example, Henripin estimated births from census data. Some techniques involve the use of models, or sets of models that have been calculated from (supposedly) reliable data for other populations. Regional model life tables, for example, report survival rates by age group and sex for different world regions and mortality levels. Users of these tables assume that a structured interrelationship exists among survival rates for the different age-sex groups. Thus, if the rates are "known" for some groups (i.e., calculated from mortality data that are thought to be reliable), the model table can be used to estimate the survival rates for groups whose data are suspect (e.g., infants, old people). First, one selects the model life table(s) that best fit(s) the groups whose survival rates are "known." Second, one adopts the model-table rates for the other groups as the estimated rates for corresponding groups in the study population. In the process one obtains an estimate of biased incompleteness in the study data, and also an estimate of the crude mortality rate. Although clearly useful in historical demography, techniques in the second group have limited range. As Shryock notes, they were designed primarily for census data, which tend to be more reliable than vital statistics in third-world populations and hence easier to work with.17 Consequently, techniques designed for vital statistics have developed slowly. In 1958 Barclay lamented that "it is difficult to devise tests of vital registration. [Compared to censuses] There are fewer internal comparisons, for vital statistics are not collected at one time and do not refer to the same universe. An independent standard is necessary, and this is usually a census (which has its own defects). The two sets of data do not correspond closely enough to permit many definite tests. And so in practice most items of vital statistics are used with very little scrutiny."18 Eighteen years later Shryock observed that the "completeness of [historical] death registration in the United States is not known and can only be surmised;

10 Facts of Life

no systematic national test has ever been conducted."19 In fact those gloomy assessments reflect a neglect of historical demography more than thb properties of historical vital statistics. Had more scholars tackled historical demographic problems, techniques for handling defective vital statistics would have followed.20 Finally while the demographic literature acknowledges social influences on civil registration and other data, it stops short of viewing them as social constructions with both the external and internal dimensions.21 As Kevin McQuillan, my collaborator for chapter 3 and a demographer, notes, "Demography has long prided itself on being the scientific side of sociology [on the assumption that] we deal with hard, easily observable, easily quantifiable facts." SCHOLARLY RESEARCH AGENDAS AS A D I M E N S I O N OF D A T A P R O B L E M S

Like the demographic literature, "the early work of quantitative historians tended to 'reify' quantitative data and to treat empirical data as effectively synonymous with underlying concepts ... The early forays into quantification, in short, did not free historians from the assumption of nineteenth century 'scientific' historiography that data are somehow self-interpreting."22 Thus, by combining quantitative data with explicit models and rigorous hypothesis testing, quantitative historians hoped to move history towards knowledge in the natural sciences, which they perceived as objective and advancing cumulatively towards truth.23 In the 19905 the objectivist understanding of scientific knowledge has few supporters. In current wisdom, "facts" are not "hard facts." Rather, the facts used to test theory are themselves products of theory, which, in turn, is imbibed from culture.24 In recent years, for example, cold-war antipathy to "communism" attracted North American scholars to non-Marxist alternatives such as "modernization" theory and French Annaliste historiography.25 Similarly, vigorous empiricism in English, and American historiography functioned as a prophylactic against Marxism, much of which works from "nonsensible realities" (e.g., relations of production) not readily open to empirical verification.26 From this perspective, Ian Tyrrell suggests, quantitative methods "reinvigorated" the mainstream liberalempiricist research tradition just when it was becoming moribund. As expected, the discipline of history has moved closer to the natural sciences, but not for the expected reason. The prime influence was not change in historical studies but rather the triumph of cognitive relativism in the philosophy of science. Like history and the social sciences, the natural sciences were increasingly perceived as

ii Prologue culture-driven, sometimes irrational activities rather than objective enterprises standing independently of culture. Even objectivity, it turns out, was a historical social construct, not a timeless verity. Civil registration data for deaths by cause illustrate the problematic status of truth and progress in historical knowledge. Here one must consider how physician, popular, and scholarly understandings of disease have changed over time and varied at particular points in time.27 When Ontario civil registration was introduced in 1869, physicians held a systemic humoral understanding of disease, according to which local illness (i.e., specific to some part of the body) was merely a local manifestation of a general disequilibrium in the body, provoked by environmental influence (e.g., seasonal change in the climate), developmental crisis (e.g., teething, menstruation), or mental distraction. Following the discovery of pathogenic bacteria in the i88os, a microbiological or specific-disease-entity understanding gradually took hold. Physicians increasingly differentiated infectious diseases, viewing each as the result of infection from a particular pathogenic bacterium or virus rather than a general body state. In the meantime, developments in surgery (the introduction of antiseptic and aseptic procedures and anaesthetics) gave physicians access to the inside of the body, thereby adding to the information for diagnosis. These developments were diffused unevenly among provincial physicians and the public, and their implications sometimes poorly understood. Only in the 19205, for example, did physicians fully appreciate that water and milk supplies could transport pathogenic micro-organisms over long distances, and that a municipal-level approach to public health was therefore inadequate.28 In light of these changes, what did a particular physician mean by the causes he reported? Dropsy (fluid retention), for example, was a cause from a humoral standpoint, but only a symptom in the specificdisease-entity understanding. Did a rise in the reported incidence of a disease (e.g., cancer, tuberculosis) show an actual increase in the incidence, or merely enhancements in the diagnostic capacities of physicians? How reliably could one convert reported causes for one era into the terminology in vogue for another?29 Based on circumstantial evidence, for example, undulant fever equated to the microbiological term brucellosis, a generalized infection resulting from contact with livestock. Yet other reported causes, such as convulsions, dropsy, and inflammation, were ambiguous in specific-disease-entity terms. An important cognitive issue is whether historical change in medical understanding represents progress towards truth. In his 1972 study, for example, William Rothstein assumed that his own

12 Facts of Life

microbiological understanding of disease was more truthful than the knowledge available to mid-nineteenth-century American physicians. In his interpretation, the physicians did not know causes, so their therapies aimed at alteration of the patient's symptomatic display. For example, they used bloodletting to lower a fever (a symptom) without affecting its cause (a particular micro-organism).30 As Charles Rosenberg showed in 1979, however, the physicians knew causes, but their understanding of causes differed from that of Rothstein's time. From the perspective of social medicine, the earlier humoral view also possessed an asset - a systemic view of disease - that the supporters of the microbiological view often lacked. In a series of publications dating from 1955 to 1988, for example, Thomas McKeown emphasized improving nutrition as the principal cause of Britain's mortality decline, not interventions from medical science. Here McKeown argued from the systemic premise that host resistance to infectious disease determined morbidity and mortality levels.31 Recent work rejects McKeown's "invisible hand of rising living conditions" explanation for mortality decline and insists upon the historical importance of interventionist public-health measures centred on hygiene and the isolation of infected persons.32 It also detects vested interest in McKeown's interpretation, which was calculated to direct state health-service resources into preventive humanist rather than curative technical medicine. For our purposes, three generalizations are important. First, understandings of disease causation change over time and vary at points in time, both in historical populations and among scholars who study them. Second, a historical change regarded as progressive by its supporters (e.g., the shift from the humoral to a microbiological understanding) eventually may be perceived as partly wrong-headed (e.g., the loss of the systemic component). Thus historians should resist the temptation to reify current wisdom. As Larry Laudan argues, we cannot know whether knowledge in science moves closer to truth.33 Instead, we should judge progress in terms of capacity to solve problems, while recognizing that culture continually changes what people recognize as a problem and what constitutes a solution. Effectively, knowledge progresses from a previous state but not towards anything, as in a truth-based epistemology. Third, because scholarly research agendas arise from culture, they are inseparable from the social construction of historical "facts." In the examples discussed above, the scholar's culture and the historical culture interact to determine what gets recognition as causes of death and the salience of particular causes.

13 Prologue SCOPE OF THE I N V E S T I G A T I O N

To summarize, the book explores the social construction of Ontario vital statistics, where social construction is given a broad definition (with external and internal components) and multiple-level expression (including goals, definitions, and organizational arrangements of the collectors, change in the conditions for the data collection, and interpretation through the filter of scholarly research agendas). Following an overview of the book's theme in chapters \ and 2, chapters 3 to 7 provide empirical illustrations of data problems and suggest strategies for dealing with them. The study begins in 1869, when Ontario introduced civil registration, and ends about 1952, the approximate endpoint for the empirical illustrations. The book treats many aspects of the civil registration data, but not all. First, the ground covered reflects the nature of the evidence and the writer's research interests. Thus marriage statistics receive less attention than statistics for births and deaths. Similarly, the book treats deaths-by-cause statistics in detail for pregrancy-related causes but not for other causes that held greater historical importance (e.g., tuberculosis, epidemics of acute infectious disease). Second, as noted above, the book focuses on the internal components of social construction while giving a general treatment to external social influences on the data. Third, the book is primarily about vital statistics, not the administrative procedures used to obtain them. Thus, while the work makes extensive use of civil registration history for context and evidence, its selection and organization of material differs from what a study of administrative history would require. In response to a reader's criticism, for example, I completed an indepth social history of the central offices connected with Ontario vital statistics (the Registrar General's Branch and the Dominion Bureau of Statistics). The research findings showed how the office work was transformed by office machines, the feminization of the staff and increase in staff numbers, and the introduction of scientific management programs. In each office a handful of males filled the professional and supervisory posts, while the routine clerical tasks and counting were left to machines and a legion of single women. In the Dominion Bureau of Statistics, francophones were poorly represented until the 19408, and even then continued to be underrepresented in the professional and supervisory positions. I found no evidence, however, that the gender and ethnic arrangements made a difference to the issues the book is treating. While in principle more women or francophones in leadership posts might have led to different definitions or reporting procedures, I was unable

14

Facts of Life

to demonstrate this. The chapter 5-6 findings for infant and maternal death statistics, moreover, emphasize the role of non-Canadian developments and public-health groups. Here again it is doubtful that a change in central-office leadership would have mattered much. Thus, after a brief existence as a draft chapter 3, the office-history materials were either incorporated into other sections of the manuscript or scrapped. Their retention would not have deepened the analysis yet would have blurred the focus of the book by introducing new issues. ORGANIZATION OF THE MATERIAL

The organizing question concerns what one must learn about Ontario vital statistics before using them in historical research. To this end, chapters i and 2 provide an overview of processes by which the data were socially constructed and of the data problems arising. Chapters 3 to 7 provide empirical illustrations of how incomplete registration and aspects of definition influence the data, and suggest strategies for dealing with such problems. In the process the later chapters extend the analysis of the social construction of data to particular statistical areas. Chapter i introduces civil registration as cultural expression. First, it puts the origins of Ontario civil registration in the context of the nineteenth-century statistical movement, which included the emergence of civil registration in other countries. Second, it analyses the social background to Ontario's introduction of civil registration in 1869. Chapter 2 describes the evolution of Ontario civil registration, first from its inception in 1869 to 1920, and then during the period 192150, when Ontario was part of a Canadian national registration area. Especially for the first of the two periods, the chapter explores features of the statistics collected, such as temporal change in the completeness and quality of provincial registrations; why the coverage of registration became complete for deaths more quickly than for births; and how marriage registrations presented an early version of the residence problem in vital statistics. Chapter 3, written with Kevin McQuillan, gives a case study of the completeness of mortality registrations for the period 1869-1972. First, it shows how to revise the low death totals by combining death registrations with reports of local deaths in other sources (a cemetery register and newspaper death notices). Second, it documents how deaths occurring in the study area (the basis for reportage in civil registration) became somewhat different from deaths of study-area residents

15 Prologue (the information required for calculation of death rates). Third, it discusses the strengths and limitations of the case-study approach in relation to what standard estimation techniques have shown, the data problems encountered in the study, and intrinsic properties of the case study. Chapter 4 examines provincial birth registration in the period 190060. First, it shows how government's enactment of legislation with age and naturalization requirements caused an increase of applications for birth certificates and late registration of birth. Second, it uses evidence from late registrations and generation life tables to estimate the completeness of birth registration for the period. Third, it analyses how the completeness of registrations for the 1920 cohort is biased by geographical region, mother's ethnicity, father's social class, and infant's sex. Chapter 5 examines the influence of definition on published statistics for infant and foetal death. First, it describes contemporary interest in these statistics and the arrangements for their collection (statutory requirements, content of the registration forms). Second, it discusses issues of definition for live birth, stillbirth, and infant death and how Canadian officials responded to them. Third, it shows how Ontario-Quebec differences in reported infant death and stillbirth rates partly reflect differences of definition and different applications of national standard definitions. More generally, the chapter emphasizes the social nature of the statistics. Whether Ontario mortality statistics are complete for infant deaths depends partly on how the persons reporting the deaths defined infant life and partly on the analyst's evaluation of the contemporary definition(s). Chapter 6 explores the nature of published statistics for deaths by cause, using Ontario statistics for pregnancy-related deaths for the period 1920-35 as an example. To place the statistics in historical context, it describes maternal mortality as a public-health priority and the different classification systems used for medical statistics. Second, it shows how published statistics based on classification by underlying cause underreport maternal deaths relative to a totalmentions count. Third, it identifies the factors responsible: a physician's misreporting of the underlying cause on the death certificate, the design of the death-certificate form, and the nature of Canada's underlying-cause classification. Fourth, it considers whether the published statistics for both underlying-cause and total-mentions tabulations became more reliable after 1935. Chapter 7 examines allocation by place as a growing problem in vital statistics because of changes in social conditions during the 19205 and 19303. Civil registration recorded a vital event by its

16 Facts of Life

municipality of occurrence. With the spread of motor-vehicle transportation and the emergence of hospitals as central sites for birth and death, however, the events occurring in a municipality increasingly differed from the events of its residents. The eventual solution, adopted in provincial statistics for 1944, was to "reallocate" the registrations by municipality of residence. This chapter examines the residence problem for the previous quarter-century. First, it discusses the social forces that created the problem. Second, it examines how officials dealt with the problem, through the collection and compilation of information about residence; by devising standard definitions of residence for inmates of public institutions and residents of foreign jurisdictions; and through efforts to improve the reliability of the information collected about residence. Third, it uses the information that officials collected to generalize about the extent of the residence problem, or whose "facts of life" the published statistics were reporting. For some eighty years after 1869, Ontario's civil registration system faced classic problems: an incomplete coverage of the registrations for the province's vital events; data errors and omissions in the returns; the growing unreliability of the returns for populations defined by residence; and the slow development of standard international definitions, registration forms, and statistical practices. By 1950, as the empirical illustrations in chapters 3 to 7 show, officials had largely tamed those problems. Their solutions, like the vital statistics, were expressions of culture, to be understood and interpreted according to a context.

i Statistics as Culture: The Origins of Ontario Civil Registration

During the 19603 and 19705 quantitative social-science historians throughout Britain and North America broke with the traditional focus on "bishops, generals, and kings" to rewrite history "from the bottom up." Using manuscript censuses and other routinely generated quantitative data, they laid bare, as never before, the lives of ordinary people in past times. Various influences fuelled the new interest the expansion of university enrolments and a consequent broadening of the social constituency for professional history; the inability of elitist theory to explain contemporary social unrest (e.g., the civil rights movement in the United States); and access to mainframe computers and packaged statistics programs, which enabled historians to analyse vastly greater quantities of data than manual calculation permitted. There was much in the new research direction to master, including mainframe computer skills, social-science theory, a vast broadening of the historian's subject-matter, and new source materials. Perhaps because so much was on their plate, and so much of novelty, quantitative historians tended to treat their data as a value-free contrast to the "subjective" literary sources on which "traditional" historians relied. Although recognizing that quantitative data held random and systematic errors, the "new" historians believed that the flaws were detectable, measurable, and correctable. Thus, by combining corrected data with explicit models and rigorous testing of hypotheses, they could ground history in objectivity, thereby purging it of moralism and opinion. In the process, history would become a social

i8 Facts of Life

science - that is, it would move closer to knowledge in the natural sciences, which reputedly was "objective" and advanced towards "truth" in rational, cumulative fashion, independently of culture. "Traditional" historians predictably dismissed the claims made for quantitative social-science history, but their innumerate and atheoretical research tradition muted their criticism. The dominance of the objectivist understanding of history and quantitative data was short-lived. Philosophers questioned the rationality and objectivity of natural science and its putative independence of culture. The thrust: of social history, moreover, was to give everything a social interpretation. Thus quantitative historians increasingly viewed their data as historical social constructs that required interpretation, not just technical adjustment. Before using the data, one had to learn the context for their collection, the social background and motives of the collectors, the cultural history of the classification used for published summary statistics, and more generally, how the social properties of the data connected to the historian's research agenda - which also was culture-driven. In light of the changes outlined above, Ontario vital statistics require a social interpretation before one can use them in research. As a first step towards this end, this chapter describes the wider context for Ontario's introduction of civil registration in 1869: the nineteenth-century statistical movement and the introduction of civil registration in other countries. It then examines local conditions that led government in Ontario to enact the compulsory civil registration of births, marriages, and deaths. THE NINETEENTH-CENTURY STATISTICAL

MOVEMENT

The introduction of civil registration was a general phenomenon in Western societies during the nineteenth century.1 It first occurred on the European continent, where civil authorities secured access to comprehensive parish records of dominant state churches. Subsequently, secular systems of civil registration were started, in England and Wales in 1837, Massachusetts in 1842, Scotland in 1855, and Ireland in 1864. In the fashion of continental Europe, Canada East registered vital events through an ecclesiastical system that dated from 1621. By contrast, Nova Scotia took England's secular system as its model when introducing civil registration in 1864. Although only vital statistics, not all government statistics, are at issue here, secular civil registration systems developed within the framework of a larger statistical movement. English interest in

19 Statistics as Culture

statistics greatly increased during the 18305 and 18405 with the founding of statistical societies and government statistical departments, and the beginning of "modern" census enumerations.2 Participants in the statistics movement believed that statistics revealed patterns, or natural laws, in otherwise complex events and that knowledge of these laws could be used to rationalize affairs of government. In England these notions were fostered greatly by the example of the insurance industry, which grew enormously during the nineteenth century and became based upon the use of life tables and actuarial principles.3 In the meantime, pressure to rationalize government affairs had developed from unprecedented fiscal and material demands on government during the Napoleonic wars, and subsequently from an increase of government functions in response to problems associated with urban and industrial growth. Supporters of the statistical movement ostensibly sought value-free information for the identification of natural laws; in practice they used statistics to "prove" culture-bound preconceptions.4 During the 18305 and 18405 England's middle class used the movement as a vehicle for social legitimacy (based on a claim of "expertise") against the forces of aristocratic exclusion (based on ascribed status).5 Their leadership in the movement also expressed anxiety about workingclass violence and unrest. Manufacturers, for example, expected statistics to vindicate the factory system of production. In their view lower-class poverty and unrest were rooted in environmental conditions caused by urbanization, not industrialization, and these conditions were identifiable through statistics and reversible through sanitary reform, moral reform, and education. More generally, English middle-class persons expected statistics to preserve social harmony by providing a basis for ameliorative policies and by disproving misinformation, which, they believed, was the principal source of class tension. Supporters of an emergent statistical movement in the United States also believed that "certain knowledge," derived from "objective, authentic facts," was indispensable for wise government. Yet American empiricism too was "freighted with unacknowledged values" that informed what was counted and how.6 For example, the selection of sex, race, and nativity categories for organizing statistics reflected heightened gender, race, and ethnic distinctions. Thus, hitherto acceptable levels of liquor consumption were redefined as excessive when native-born Americans linked them to English and Irish immigrants and threats to industrial work discipline. Similarly, a consistent theme in the collection of public-health statistics for Massachusetts was a belief that public-health problems originated in moral

2O Facts of Life

weaknesses and unhygienic practices of immigrants.7 Socially determined systematic data errors also compromised the objectivity of descriptive facts. The coverage of urban slum dwellers in census enumerations, for example, reflected how the enumerators were paid and whether they were students and retired persons or experienced social workers. During the late nineteenth century a professional middle class reinforced and reshaped popular interest in statistics.8 Professionals derived their livelihoods from mental labour rather than manual labour or control of capital, and they sought to raise their occupational status by-basing it upon accredited systematic bodies of knowledge. From this perspective professionals criticized laissez-faire features of capitalist industrialism and called for government regulation, guided by "experts" and informed by statistics. Thus, while government interest in statistics was recommended for the public good, it also promised jobs and prestige for middle-class professionals. Statistics as Coercion. The nineteenth-century state, Philip Abrams argues, was an idea or myth - a "distinctive collective misrepresentation of capitalist society" that acted to legitimate politically organized subjection by concealing it.9 The state in myth was the unifying symbol of government agencies that had no unity; in this role it masked the coercion of those agencies by "overcrediting" them as the "integrated expression of the common good." The success of the "state-formation project" amounted to a cultural revolution.10 The general populace earlier had perceived government interventions, accurately, as coercion by dominant social groups. By contrast, they now accepted the interventions as natural, "common-sense" responses to demonstrated public need. Along the way, democracy tended to co-opt dominated groups, allowing them to participate in their own subordination. Central government used statistics to legitimize its coercive interventions in local affairs. As its officials claimed, government alone, through its systematic accumulation of "value-free" statistics, possessed the "scientific" knowledge required for policy development. By contrast, alternative information, such as the idiosyncratic, personal knowledge of local notables, was inadequate and unreliable.11 Vital statistics contributed notably to state formation in Ontario. As elaborated in chapter 2, for example, Ontario government used mortality statistics for deaths by cause to justify its imposition of public-health requirements on municipal populations. To give a second example, Ontario's various Marriage Acts defined what a

21 Statistics as Culture

marriage was, and the civil registration of marriages established whether a particular marriage had occurred. In official rhetoric both measures served "the public good" by buttressing the rights of weaker parties (women and children) in institutions that were natural and desirable (marriage and family).12 Yet they worked coercively against unions that were outside the legal definition of marriage and did not fit the statistical categories based upon it (marriages of minors, closely related persons, persons already married, persons of the same sex, common-law marriages). Even legally sanctioned marriage, moreover, was a historical, not a natural institution. As such, a feminist scholar recently has judged, it was "a particular, exploitative relationship between men and women in which the wife provide[d] unpaid domestic and sexual services, childbearing and rearing, and wage-earning and contribution to the household income when convenient ... supposedly in exchange for protection, assured upkeep and some rights to children."13 We need not subscribe to conspiracy theories to appreciate the coercive aspects of state formation and statistics. Recognition of a common human tendency to generalize a "public interest" from our own group's interests suffices. We should also treat dominance as complex, fluid, and historical. Groups compete for power and face resistance from within as well as from other groups. The outcome of their struggles varies by time and place, and the nature of the dominance (e.g., by class, sex, age, ethnicity) varies by issue. THE ORIGIN OF CIVIL REGISTRATION IN ENGLAND AND MASSACHUSETTS

The immediate cause of civil registration legislation was dissatisfaction with traditional sources of vital statistics - bills of mortality and church registers of baptisms, marriages, and burials - for legal purposes (e.g., proof of residence, ancestry, personal relationships).14 In England church records of dissenters lacked legal status, and Church of England parish registers were incomplete for Anglicans and without central supervision or collection. Religious dissenters were the chief political force behind passage of a registration act. Support also came from various medical men, who sought reliable death records to study laws of morbidity and mortality, and actuaries and statisticians, who wanted a sound actuarial basis for government annuities, life insurance, and friendly-society sick benefits.15 An earlier development - the English Poor Law amendment of 1834 - made a secular collection of vital statistics possible without exorbitant expense. For the purpose of eliminating relief outside the

22

Facts of Life

workhouse, the amendment established the first English example of a strong central authority that held supervisory powers over local authorities. Consequently, centrally administered local officials - Poor Law commissioners - were available for appointment as local registrars.16 While initiated primarily for legal reasons, civil registration developed through alliance with the public-health movement. A desire to reduce costs caused Poor Law officials to interest themselves in public health.17 Especially in urban populations, they supposed, sanitary reforms could reduce morbidity and mortality among heads of families, and this reduction would lessen the number of dependents and destitute on relief. With such reasoning, the Poor Law Commission formally investigated English sanitary conditions, and its reports laid the groundwork for England's Public Health Act of 1848. Throughout the nineteenth century sanitary reform remained the focus of the public-health movement, and mortality statistics were part of its empirical justification. Thus registration improved more quickly for deaths than for births.18 Just as civil registration was strengthened by the public-health movement, so that movement was a by-product of urban and industrial growth.19 Although poor sanitation generally characterized nineteenth-century societies, it was especially evident in large urban places, where problems of sewage disposal and water supply became acute. Furthermore, while the introduction of water closets eased domestic sewage problems, it magnified the public problem by vastly increasing liquid sewage and overwhelming traditional sewagedisposal methods. THE ORIGINS

OF ONTARIO

REGISTRATION,

CIVIL

18699

The interrelationships among civil registration, public health, and urban and industrial growth largely explain why Ontario derived its statistical practices from England and Massachusetts, whose economies were more developed. Ontario's Vital Statistics Act of 1869 was modelled on England's Vital Statistics Act of 1836, and its publichealth legislation during the i88os was informed by English and Massachusetts experience.20 The inchoate state of its economy also explains why Ontario began with a low-cost, loosely enforced registration system and developed it gradually. As Thomas S. Shenston, clerk for Brant County, explained in 1869, "it is simply impossible in a young country like this" to obtain records accurate enough for

23 Statistics as Culture statistical purposes "without incurring more expense than the ratepayers would submit to."21 In Upper Canada as in England, government tried first to obtain vital statistics from church records. The Upper Canada Marriage Act of 1831 required clergy of specified denominations to keep a marriage register and report marriages to a clerk of the peace. In 1847 a Census Act for the Province of Canada established a Board of Registration and Statistics, required clergy to report baptisms, marriages, and burials quarterly (amended to annually in 1849) to clerks of the peace, and directed the clerks to forward their returns to the provincial board. The act also required a household head to report a vital event directly to the nearest township clerk where a clergyman of his denomination was not resident locally. The 1848 Upper Canada census provided the province's first enumeration of births, marriages, and deaths. In the meantime, marriage legislation in 1847 and 1857 extended to ordained clergy of all Christian denominations the right to perform and the obligation to report marriages.22 As a select committee of the Legislative Assembly on births, marriages, and deaths in Upper Canada learned in 1865, however, this legislation was ineffective.23 Officials of the provincial statistics board testified that no registration system existed for births and deaths, and that the law was inadequate even if enforced. Providing for a denominational rather than a provincial registration system, the law "missed vast numbers who are practically unconnected with any church ... numerous deaths of infants who die at an early age, and are buried without register or record of any kind; the many persons who die in outlying parts of the country and are buried on the farm which they occupied ... [and] the children ... of Baptists." In Canada East civil registration was relatively successful.24 There, in a system dating from the seventeenth century, priests kept duplicate registers of births, marriages, and burials, and once a year they presented the registers to civil authorities, who authenticated them for legal purposes and retained one of the duplicates. The system worked because good Roman Catholic parish records existed in a population that was largely Roman Catholic. In Canada West, by contrast, denominational pluralism precluded comprehensive parish records. Thus the 1865 select committee recommended a new system "based on the English Acts, and adapted to the Municipal system of Upper Canada." Municipal acts passed during the 18405 made the proposed system feasible.25 In the fashion of the 1834 English Poor Law reform, this legislation established centrally administered local bureaucracies that could do the work.

24

Facts of Life

The 1869 Ontario Vital Statistics Act originated as a private member's bill from Timothy Blair Pardee, soon to become provincial secretary but then a back-bencher. Pardee's bill passed unanimously and without debate. Judging by later remarks from two provincial secretaries, Peter Gow in 1871 and Pardee in 1872, provincial politicians regarded the importance of vital statistics as self-evident, based on the experience and practice of other countries, especially England.26 The officials who administered civil registration saw their work as a disinterested response to public need. As Hugh S. Crewe, chief clerk of the registrar-general's central office in Toronto and the provincial inspector of vital statistics, observed in 1881: Vital statistics furnish the most unerring lessons as to the health, prosperity, and morals of the people; they teach the influence of marriage on illegitimacy and morality, the vital force of children, the duration of life with its expectation and value for all ages and races, the influence of meteorology, occupation, and locality in generating disease and improving health, and thereby the removal of unfavourable conditions, often where least expected, and the approach of morbific storms, by ignorance of which cities and even nations have been destroyed. They serve alike to guide the resident and the immigrant, the capitalist and the labourer, the politician and the statesman, the moralist and the scientist... their establishment has become a test of the degree of civilization reached by a people and their rulers.27 In 1891 Dr Peter Bryce, then deputy registrar-general and Secretary of the provincial Health Board, recalled that Ontario had established civil registration primarily for legal purposes, "in order that questions of inheritance arising in the matter of wills, estates, titles, etc., may be speedily and properly settled."28 Even in 1869, however, provincial officials looked forward to the day when the civil registration coverage of Ontario's vital events would suffice for statistical purposes.29 By studying population in relation to natural increase (excess of births over deaths), for example, they hoped to estimate population for intercensal years. They also viewed the birth rate as an indicator of national vitality and the marriage rate as a measure of prosperity, a factor in the birth rate, and an indicator of the influence of marriage upon illegitimacy and morals. Through accurate mortality returns officials hoped as well to ascertain "exact rates applicable to insurance of life," government annuities, and "annuities granted by public companies." As the registrar-general noted in 1870, Canadian actuaries then were dependent upon statistics derived from the life experience of populations of other countries.

25 Statistics as Culture

Above all, vital statistics had application for public-health purposes - therein lay their most important and far-reaching "beneficent effects." In Ontario as elsewhere, government officials sought to discover natural laws of mortality through compilation of mortality statistics. More specifically they expected analysis of mortality by occupation, locale, and meteorological and other factors to show the "productive loss to the community" of persons dying from diseases that were preventable through sanitary reform. In this regard, nineteenth-century public-health movements were informed primarily by the miasmatic theory of disease, according to which diseases were transmitted by noxious, invisible atmospheric gases whose generation was increased by filth. Thus public-health advocates regarded death rates by age, cause of death, and other factors as indicators of environmental conditions and the "quality"of the people. Through analysis of mortality and morbidity statistics, they intended to trace the path of epidemic disease and identify areas where sanitary reform was needed. Sanitary precautions in turn were expected to lower morbidity rates for large cities and "to prevent or arrest the progress of the most dangerous epidemics throughout the breadth of our land."30 The emphasis on sanitary reform lasted until the early twentieth century, when the spread of the germ theory of disease and the development of antitoxins shifted the focus of public health to immunology.31 Special-interest groups had little to do with Ontario's enactment of civil registration. Ontario jurists, for example, limited comment on vital statistics to an editorial in the March 1870 issue of the Upper Canada Law Journal, which was written at the registrar-general's request. The editor acknowledged the importance of authentic records of vital events for legal purposes and the difficulty and expense of obtaining such records in Ontario. Yet although the act declared that certificates issued by the registrar-general were prima facie evidence in provincial courts, the editor doubted that this was so, given the absence of adequate safeguards against mistakes and fraud in the returns.32 The Ontario insurance industry also was indifferent to vital statistics, perhaps because insurance premiums were calculated from life tables based on insured lives, which had lower mortality than the general population.33 During the i88os certain Ontario medical men showed considerable interest in the registration of deaths by cause. In 1881 a Canadian Medical Association committee sought signatures in Ontario's five largest cities for a "memorial of the general practitioners of the Province of Ontario," addressed to Sir John A. Macdonald and calling for

26

Facts of Life

the establishment of a federal bureau of statistics.34 The following year the president of the Ontario Medical Association urged doctors to co-operate with provincial efforts to obtain vital and morbidity statistics.35 These medical spokesmen urged the importance of mortality statistics for research and also for preventative medicine based on sanitary reform and the inspection of immigrants. In the absence of a microbiological understanding of disease, they inferred causes from correlations between mortality rates and observed characteristics of the environment.36 Of course, statistics and appeals to science also enhanced the influence and prestige of the medical profession. Like their counterparts in Quebec, these Ontario physicians were in quest of a state-run public-health system in which medical public-health officers were the administrators and sole advisers.37 However, medical lobbying was not in evidence in 1869 when Ontario passed its Registration Act. In his report for 1870 the registrar-general noted the "unreasonable opposition" of medical men to the act; this he was at a loss to explain, though he noted one physician's objection to filling out registration forms without payment. In 1873, after a Wentworth County doctor was fined $54.75 for failing to report a cause of death, the editor of the Canada Lancet argued that the law was "very unjust towards medical practitioners" and that friends of the; deceased should bear the entire responsibility for registering a death and obtaining the medical certification of death from a doctor.38 In 1877 the division registrar for Ingersoll initiated eleven court actions, involving all but one of the town doctors, to secure medical reports on cause of death.39 In 1881 one Dr Sullivan was unable to obtain any signatures from Kingston, other than his own, for the 1881 memorial of Ontario medical practitioners to the Dominion government.40 This resistance to registration law was not surprising. First, the public-health movement was less developed in the i86os and 18705 than in the i88os, and physicians were less important in it. Second, Ontario physicians were a disparate group, comprising eclectic and homeopathic as well as regular licensed physicians, and many of them had been trained primarily through apprenticeship, with little formal training.41 Perhaps also the resisting physicians perceived the public-health movement as a harbinger of state medicine and a threat to their professional autonomy.42 In any case, the Ontario government at Confederation seemed unlikely to enact recommendations of the 1865 Province of Canada select committee on births, marriages, and deaths. Little public pressure existed for such action, and under section 6 of the British North

27 Statistics as Culture America Act, the census and vital statistics were a Dominion government responsibility. For that matter, the provincial government preferred national civil registration. As the registrar-general's annual reports for 1869 and 1870 later elaborated, foreign inquiries about vital statistics were normally addressed to the Dominion government, and only a national system could ensure uniform methods of collecting and reporting them. For population analysis, vital statistics belonged with federal immigration and census statistics, and only at the national level were statistics responsibilities large enough to warrant a separate department with permanent, specialized officials. Appropriately, the Dominion government gave indications of action on civil registration. In 1867 it placed functions of the old provincial Board of Registration and Statistics under the Ministry of Agriculture. Subsequently, the ministry assumed all costs of Nova Scotia's registration system and part of the costs of the Quebec system, and it compiled vital statistics for certain Quebec counties.43 In the 1871 census the ministry also enumerated mortality in unprecedented detail on a separate mortality schedule. Yet no national system was attempted after 1867. Dominion officials feared the expense involved in a country of Canada's size and with its population dispersal.44 Until 1905 the Ministry of Agriculture had no permanent statistics office, and its resources for statistics were exhausted by the demands of census enumerations. The 1871 census, for example, was a large undertaking; the enumeration and compilation of the data lasted seven years and cost $511,330 - an amount equal to 2 per cent of the 1871 budget and supplementary estimates.45 In the circumstances, the department created a temporary statistics office for each enumeration, later phasing it out as compilation of the data was completed. Little changed thereafter. "As administered," the Dominion statistician later recalled, the 1879 Census and Statistics Act was "mainly a census Act, and took little cognizance of general statistics."46 Such attention as was given to general statistics went to immigration, natural resources, and production, which were major Dominion government responsibilities. In the absence of national registration, the Ontario government's 1869 Registration Act was logical. Though constitutionally a federal government responsibility, vital statistics were linked closely to provincial government concerns such as property rights and public health. Even the administrative apparatus recommended by the 1865 select committee - Ontario's network of municipal organization - was a provincial responsibility. Moreover, population dispersal was less serious a physical obstacle to civil registration in Ontario than in the nation at large.

28

Facts of Life

Adopting a model suggested by the select committee four years earlier, Ontario based its system on the province's existing network of municipal organization.47 Under the act as passed and amended in 1869, the provincial secretary was the registrar-general and the central office for civil registration was the Registrar-General's Branch of the Provincial Secretary's Department. A single clerk, Hugh S. Crewe, initially manned the central office, with the assistance of temporary extra writers. Within six years Crewe was to become chief clerk, presiding over a permanent staff of seven. County and municipal clerks served respectively as district and division registrars. Municipal registrars recorded vital events on forms supplied by the central office, and in January of each year forwarded the completed returns to the county registrars, who checked them for legibility, error, and missing information before forwarding them to the central office. Initially the act entitled the local registrar to a ten-cent fee from the person registering the vital event, but the amendment abolished the fee and directed the municipalities to pay their clerks "just" sums for registration work. The Branch paid the salaries of the county registrars; these amounted to $3,7OO-$3,75O annually during the 1870-75 period, an average of $93 for each of the forty registrars who forwarded returns.48 As amended in 1869, the act required the father, mother, or person representing the parents to register a birth within thirty days, although registration of a birth was allowed within a period of one year. A clergyman was to register a marriage he had solemnized within ninety days. An occupier of the house in which a death occurred (including keepers of gaols, refuges, and other public institutions) was obliged to register the death within ten days of interment. Additionally, an attending physician was to report the death and also the cause within ten days (changed to thirty by the amendment) of the death. The amendment relieved physicians of the responsibility to report births they had attended. It also required them to report a death they had attended to the division registrar in their own municipality of residence, not the municipality in which the death had occurred. Where the municipalities differed, the county registrar was to link the information in the occupier's and physician's report. Finally, the act provided a schedule of fines for persons who neglected or refused to perform their statutory obligations, and made the municipal registrar responsible for initiating prosecution. The birth registration reported the infant's name, sex, and date of birth, the parents' names, the father's "rank or profession," and the name of the accoucheur (attendant). The marriage registration reported name, age, place of residence, place of birth, religious

29 Statistics as Culture

denomination, marital status, and names of parents for both bride and groom, along with the clergyman's name and the groom's "rank or profession." The death registration reported the decedent's name, date of death, sex, age, religion, "rank or profession," place of birth, name of attending physician (if any), certified cause of death, and duration of illness. Economy was an outstanding feature of the system. Through to 1900 its cost to the provincial treasury rarely exceeded $10,000, and its central office staff numbered no more than ten.49 Yet its legislative provisions were the most advanced in Canada. Whereas Ontario law provided for the centralized collection and analysis of individuallevel data, the Quebec government collected only abstracts, showing for each locality the number of marriages and the number of births and deaths for each sex.50 Not until 1893 did Quebec introduce medical certification for cause of death. Only in 1924 did Quebec extend its registration requirements to cover persons missed by parish registers and provide its central authority, the provincial Health Board, with copies of the individual-level records. Elsewhere, Nova Scotia discontinued the registration of births and deaths in 1876, and permanent civil registration in Maritime Canada awaited the twentieth century.51 The advanced state of Ontario's civil registration legislation relative to Quebec's possibly reflected differences of economic development in the two provinces.52 More urbanized and integrated than Quebec, Ontario was quicker to experience public-health problems on a province-wide basis and hence to perceive a need for mortality statistics. Moreover, anti-statism was less of an obstacle to legislative innovation in Ontario than in Quebec, where the social-service functions of the Roman Catholic Church were becoming entrenched.53 Even so, in 1886 the Quebec provincial government enacted a state system of public health, shortly after comparable Ontario legislation in 1882 and 1884.^ Thus a possibly decisive difference between the two provinces was that Ontario, lacking vital records for any purpose, was free to legislate civil registration for all purposes; by contrast, Quebec's ecclesiastical system produced good vital records for legal purposes, and the provincial government may have been reluctant to tamper with its system to accommodate the goals of the public-health movement. CONCLUSION

Like the larger international statistical movement, Ontario's introduction of civil registration was a by-product of nineteenth-century capitalist industrial economic development. On a general level it

30 Facts of Life

evidenced government's growing thirst for information as regions and individuals in society became more interconnected and interdependent. On an immediate level it responded to anticipated changes in social and environmental conditions, such as the aggravation of public-health problems in cities. To its partisans civil registration was a disinterested government response to a self-evident public need. Yet it also served persons from dominant social groups (the middle class, males, professionals) who sought to normalize certain of their values and advance their perceived self-interests. To these ends supporters of civil registration coerced, educated, and concealed through the rhetoric of "public interest," "natural laws," and "value-free" statistics. They also met resistance, including some from the same groups that championed civil registration. Thus, while some physicians promoted compulsory death registration for public-health purposes, other physicians ended up in court for refusal to fill out the death certificates.

2 The Evolution of Ontario Civil Registration, 1869-1950

Between 1869 and 1920 the Ontario government developed a centralized system for registering births, marriages, and deaths. Although the registration of a vital event was legally compulsory under the 1869 Registration Act, the law was widely ignored during the nineteenth century. As officials learned, the coverage and quality of registrations depended on the adequacy of the supporting legal and administrative arrangements. Even more critical was whether the person legally responsible for registering a vital event had practical incentive to do so. In the circumstances, officials believed, the returns for deaths and marriages were incomplete for cities until the i88os and the entire province until 1900. Worse still, the birth returns remained incomplete until the 19205. This chapter describes the efforts of provincial officials to improve the coverage and quality of civil registration through five Registration or Vital Statistics acts (1869, 1875, 1896, 1908, and 1919). Second, it summarizes problems with the province's birth, marriage, and death statistics between 1896, the year of the last major changes to the province's Vital Statistics Act, and 1920, the year Ontario joined a Canadian national registration area. Third, it discusses how provincial civil registration changed during the period 1920-50, the era of national civil registration. C I V I L R E G I S T R A T I O N U N D E R T H E 1869 A C T

For years after 1869 Ontario's civil registration of vital events was unsatisfactory. Although its legislative provisions surpassed

32 Facts of Life

Quebec's, Ontario was less successful in obtaining registrations. Working from English experience (e.g., a crude death rate of 21 per 1,000), the registrar-general estimated that registrations for 1870 captured only a fifth of the province's deaths, a third of its births, and two-thirds of its marriages. He also found that county registrars submitted returns from the municipal registrars without correction, even though the information in them was frequently incomplete or incorrect.1 The data had little value, he concluded, other than to show the relative efficiency of the various municipal clerks. In explanation he cited public ignorance of or indifference to the law, a lack of support from the medical profession and clergy, and the failure of municipal registrars to enforce the law beyond registering vital events brought directly to their attention. In fairness, municipal registrars had little incentive to go after defaulters. In 1871 the registrar-general acknowledged that "in a small community, the Division Registrar, who is also the Clerk of the Municipality, would naturally shrink from instituting legal proceedings against his neighbours, and particularly in cases where the offending party happens to be a prominent medical man."2 In addition, township clerks were poorly paid and part time (usually farmers), and clerks of urban municipalities, though more apt to be full time, were paid little in relation to their responsibilities. Although the Registration Act directed the municipalities to pay their clerks a "just" sum for registration work, the payments ranged downwards from forty dollars to a low of two dollars in 1871. Thus many clerks gave low priority to the registration work. A further problem was that municipal registrars only submitted their returns annually. Thus a county registrar received registrations up to a year after they had been filed and often could not trace errors or obtain missing information for them. Thomas S. Shenston expressed the helplessness of local registrars under the 1869 system.3 Few vital events in his Brant County jurisdiction were reported, and information obtained under the Registration Act was "very defective and really of very little use to the few individuals who comply with its provision." By way of illustration he observed that ministers are the most intelligent and best educated class in the community, and the County of Brant is the smallest and most compact county in the Province. Yet notwithstanding this, a number of their annual marriage returns do not reach me til long after the time appointed by the Act. Some to my knowledge have never been returned at all - some came by Tost/ written on small slips of paper and very defective, unaccompanied by any

33 The Evolution of Ontario Civil Registration letter ... How easy for anyone to send me such a return by way of a joke, or purposefully incorrect. Is it on such testimony as some of these returns furnished that the legality of Marriages and the legitimacy of children are to be proven?

Arguing that the compulsory centralized system was "worse than nothing," Shenston urged its replacement by a decentralized, voluntary system administered by county clerks and financed through user fees (for registrations, searches, and certificates). Since such a system would serve only legal purposes, it could emphasize safeguards against fraudulent registration. Thus Shenston suggested a registration form for deaths that required the name of the attending physician, names of persons present at the time of death, name of undertaker, location of burial plot, name and residence of the officiating minister, the names, residences, and occupations of two bearers, names of siblings living and dead, and the signatures of a physician, two bearers, and a witness. The 1875 Registration Act. The provincial government chose to strengthen the existing system, not replace it. The 1875 act, which took effect on i January 1876, required householders to register a death prior to interment (rather than within ten days), in return for which they were to be issued a certificate of death.4 If, despite the above clause, a clergyman buried a body without seeing a death certificate, he was obliged to report the death within ten days. Municipalities now were to pay their clerks ten cents per registration (versus the old "just" sum), and the clerks were to forward their returns semi-annually, not annually as before. The act abolished the position of county registrar, thereby placing municipal registrars in direct communication with the provincial office and trimming $3,750 from the Branch's budget. The attending medical practitioner now was to submit the death certificate to the registrar of the municipality in which the death occurred, not his own municipality of residence; this removed the possibility that the householder's report of a death and the physician's death certificate would go to different local officials. The 1875 act softened the provision for penalties, however. Where more than one person was required to submit a return (e.g., the head of a family and the attending physician for a death) and a return was "made by any one of such persons," then "the other of such persons shall not be liable to any penalty in respect of his default." As long as registration remained a general responsibility of the provincial secretary, whose department was charged with keeping all correspondence and records of government, central authorities

34 Facts of Life

were unlikely to give the registration laws sufficient attention to make them work.5 Thus another important feature of the 1875 act was its provision for the appointment of a provincial inspector of vital statistics. Hugh S. Crewe was inspector for the 1876-91 period, while also serving as chief clerk in the Branch office.6 During 1876 Crewe visited the offices of about 160 Division Registrars, and personally called upon a number of clergymen, medical men, undertakers, sextons, and overseers of cemeteries, and asked their co-operation in securing the most correct returns possible. In several cases he was compelled to order the prosecution of defaulters, but usually, when the intimation got abroad that legal proceedings would be taken in all cases of neglect, it seemed sufficient to induce those in default to comply at once with the requirements of the Act... among the Division Registrars visited, he found a number of them who looked upon the Registration Act as useless, and simply entered any registrations that were offered to them ... on the other hand, a number of Division Registrars displayed a great deal of energy to get all the registrations possible, as well as to impress upon the inhabitants ... the value of correct registrations.7

Death returns for 1876, the first year under the new act, increased over those for 1875 by 95 per cent, almost entirely because of improved returns from cities, and birth and marriage registrations increased by 48 and 15 per cent respectively. In his review of the decade 1871-80 Inspector Crewe estimated that the completeness of the registrations had increased from 75 to 90 per cent for marriages, from 50 to 70 per cent for births, and from 32 to 60 per cent for deaths. He also found that the quality of the returns had improved and that the city returns were complete enough to serve as the basis for provincial estimates. Here he judged completeness by comparing Ontario's reported vital rates with English vital rates that he believed were applicable to Ontario (e.g., a death rate of 18 per 1,000 population, the English rate if returns from industrial cities were excluded). Despite the improvement, officials found much to criticize. Returns from the townships commonly fell short of the vital events estimated for them. The 1891 birth returns for Ottawa were down so sharply that Dr Peter H. Bryce, the province's first deputy registrar-general (1891-1904), excluded them from the statistics for cities in his annual report. Colonel R.B. Hamilton, the second inspector of vital statistics (1892-1905), believed that extreme variation in reported municipal vital rates was symptomatic of incomplete returns. Thus he was distressed that the 1893 birth rates calculated for 115 municipalities ranged from 46 to o, while reported death rates ranged from 26 to 2. To his disgust, a "thriving town in Eastern Ontario" registered 11

35 The Evolution of Ontario Civil Registration

deaths in 1891, although 44 persons had been buried in its two cemeteries.8 Officials offered various explanations for the defective state of the returns. As Inspector Crewe ventured in his report for 1876, urban returns were better than rural returns in part because urban dwellers lived nearer to a municipal registrar. Thus they could register an event more conveniently, and for the same reason urban registrars could keep a closer eye on vital events. In his report for 1892 Inspector Hamilton observed that only the Division Registrar whose heart is in his work ... could overcome ... apparently insurmountable difficulties. To do this he no doubt has to make strenuous efforts ... interview indifferent parent, negligent doctor, and careless minister ... keep both eyes and ears open to learn of the latest arrival ... scan all local newspapers, and has to look in at the undertaker's to learn what quarter came the latest order for funeral trappings - in fact, has to keep himself posted on the movements of all the residents of his municipality, some of whom may reside fifteen or twenty miles from his own home. Then, after all these efforts, what reward does he reap for his pains? ... outside the cities of the Province, the average amount received by Division Registrars during the year is less than ten dollars.

In 1894 as for years past, a clerk's decision to prosecute violators of the Registration Act meant "almost as much as his position was worth," yet the 1875 act did not mandate the provincial inspector to initiate prosecutions. Since they were not required to keep permanent local copies of registrations, clerks generally kept haphazard records, and the inspector had difficulty checking their work. In many communities, Inspector Hamilton concluded for 1894, the Registration Act was a "dead letter" - its provision for penalties had been so long ignored that they were "as much to be feared as the stories which so often have been retailed to frighten us in our childhood's days." Under the 1875 act registration was biased against certain population groups as well as incomplete. As noted above, rural vital events were exceptionally underreported. In his report for 1876 Inspector Crewe speculated that native Ontarians were less likely to register vital events than the British-born, who had become accustomed to civil registration in the old country. Above all, the deaths of young children were "the least likely to come under the notice of the registrar, and the registering of them ... [was] the most likely to be neglected by parents and friends."9 As well as being incomplete and statistically biased, returns under the 1875 act were flawed by missing and defective data. Of the 1876

36 Facts of Life

returns, 3.6 per cent left out age, and cause of death was omitted for one in seventeen, in some cases because no physician was present but largely because physicians were indifferent to civil registration. Even where reported, nineteenth-century age data were suspect and hard to verify, especially for deaths of older persons, many of whom were not Canadian-born and whose births pre-dated civil registration systems.10 Ambiguous entries for cause of death also gave cause for complaint, in some cases because death occurred without a physician present. As the 1879 report noted, doctors frequently faced "inadequate observation ... or symptoms which were indistinct, complicated, and only explicable by autopsy." Even so, reports for the 18705 and i88os criticized entries such as "convulsions," "debility," "haemorrhage," "old age," "dropsy," and "infantile debility." In 1878 "old age," "infantile debility," and "convulsions" accounted for nearly a fifth of reported causes of death and ranked second, third, and fifth as leading causes of death (with 10.2, 6.5, and 2.0 per cent respectively). The popular "old age" entry, the report suggested, was the convenient response of indifferent physicians, though it also resulted from a natural inclination to attribute death to "natural, irremedial decay," lest uncharitable persons suspect that want of medical skill was at fault. Another suggestion, in the 1882 report, was that where old people died without a physician present, "old age" was the best explanation available to the persons reporting their deaths. In retrospect, the returns may have been better than officials realized. Crude birth and death rates were declining during the nineteenth century in most Western nations, with the result that the registrar-general's estimates of Ontario vital rates - derived from English experience around 1870 - were increasingly higher than was appropriate. In their interpretation of the returns, officials also did not allow for population aging, which was effecting a disproportionate decline in mortality for younger age groups and eventually declines in marriage and birth rates.11 Furthermore, differences between rural and urban reporting of vital events were less substantial than appeared at first glance. Extreme community variation in vital rates resulted from small population numbers as well as incomplete registration.12 As the Branch noted for Ottawa in 1886 and more generally for Ontario in 1890, the proliferation of hospitals in urban locations caused rural dwellers to add to urban death totals.13 In fact, during the twelve years preceding the founding of Ingersoll's hospital in 1909, 26 per cent of the 1,176 deaths registered in the town were those of residents of nearby hamlets and townships.14 Finally, quite apart from the influence of

37 The Evolution of Ontario Civil Registration hospitals, urban mortality returns were inflated by the removal of elderly (high mortality risk) farm couples to urban places upon retirement. Perhaps officials also judged harshly the quality of the returns. Remarks on the Ingersoll registrations showed that area residents sometimes did not know particulars such as age, nationality, and date of death, especially where the decedents were transients or elderly persons. By the late i88os official complaints of physicians' entries for cause of death may have arisen as much from the novel microbiological understanding of disease as from "deficient" physician reporting.15 Even during the 19805 Ingersoll doctors occasionally attributed death to "old age," in some cases because autopsy did not reveal a specific cause.16 The Ontario Public-Health Movement. A key to the further improvement of Ontario's civil registration - a strong public-health movement - did not exist in the 18705. The Ontario Public Health Act of 1873 merely reaffirmed pre-Confederation statutes (1835, 1849), which provided for the appointment of temporary provincial and local health boards to deal with emergencies such as the outbreak of epidemic disease.17 Acting under the 1849 Municipal Act, which permitted municipal by-laws on public health, some municipalities organized permanent public-health committees that became styled "local health boards," but these had no legal standing.18 During the i88os, however, a continuance of Ontario's urban and industrial growth brought public-health issues to government attention. As the chairman of the Ontario Health Board later reported, sewage- and water-disposal problems exceeded the financial resources of many local governments.19 Thus in 1882 a second Ontario Public Health Act established a permanent central Board of Health and gave it advisory powers over local health boards. The act also provided for a provincial secretary of health. Dr Peter H. Bryce was so appointed, and his work included inducing municipalities to appoint local health boards, even though these were not officially recognized. Under the act medical men largely replaced laymen as public-health leaders. Through to 1925, all three men serving as provincial health secretary were physicians, as were the first six men named to the seven-member provincial Health Board. Throughout Ontario, moreover, physicians acting as "local correspondents" reported weekly to the provincial board on local morbidity, thereby enabling the board to publish weekly reports for the province.20 In 1884 a third Public Health Act made local health boards compulsory, in the process giving them legal standing, and placed them

38 Facts of Life

under the supervision of the provincial board, which hitherto had only advisory powers. The 1884 act also enabled local boards to appoint medical health officers and sanitary officers, and it authorized the provincial board to appoint local boards where municipal authorities failed to act.21 Many municipalities (mostly townships) subsequently did not appoint local boards, while others appointed one but denied it adequate financial support.22 This resistance probably explains why government changed the position of secretary from part time to full time in 1884.23 Ontario's public-health legislation had direct implications for civil registration. As the 1882 act stated, the provincial Health Board was "especially [to] study the vital statistics of the Province, and ... make ... use of the collected records of death and sickness among the people." The act also directed the secretary of the provincial board to assist "in preparing the annual report of the Registrar-General." Following on this development, government named Bryce to the new position of deputy registrar-general in 1891, while retaining him as secretary of the provincial Health Board. As well as strengthening the ties between public health and civil registration, this new parttime post marked a step in the development of a permanent, separate bureaucracy for vital statistics.24 Then, in 1892, the Branch replaced Crewe, a part-time inspector, with a full-time inspector, Col. R.B. Hamilton.25 A Failed Dominion Government Initiative, 1882-91. In 1882, in response to representations from the Canadian Medical Association and various provincial governments, the federal minister of agriculture introduced a scheme to collect mortality statistics from eleven cities (the provincial capitals and cities with more than 25,000 population).26 The minister obtained the data through medical officers of local health boards, where these had been appointed. With an annual budget of $10,000, the ministry provided the forms and contributed to the expense of the data collection to a maximum of $400 per city. By the end of the decade the system embraced twenty-six cities, sixteen of them in Ontario. During the 1885-91 period, however, the national system reported fewer deaths than provincial registration in five of the nine cities for which comparable data are available, including 2,932 fewer deaths for Toronto.27 Thus in 1891 a federal government statistician found the data "so incomplete as to be practically useless, besides involving considerable expenditure of public money," and the national system was terminated.28 Henceforth Canada reported deaths for census years in its census enumerations, but even this practice ended after

39 The Evolution of Ontario Civil Registration 1911, when the "returns were so bad that they were quietly dropped overboard."29 The 1896 Registration Act. A third Registration Act in 1896 forbade clergy, undertakers, and keepers of cemeteries to bury a body without first obtaining a death certificate. It also obliged cemetery keepers to submit semi-annual returns of burials to the registrar "of the division in which the burial ground is situate."30 Second, it required physicians to report births they had attended. It also instructed the physician to submit a death certificate to the registrar in whose municipality the death had occurred, not to the registrar for the physician's own municipality of residence. Third, clergymen were to report marriages they had celebrated within thirty days from the date of marriages, not within ninety days, as under the 1869 and 1875 acts. Fourth, the inspector of vital statistics, not the local registrar, was now responsible for initiating legal proceedings against persons "neglecting or refusing to make the required reports." Penalties now applied to any non-compliance with the law, not just wilful non-compliance. As before, however, a physician was exempt from penalty if a family member had made a return of the vital event in question. Fifth, the act directed the municipal registrars to retain the original registrations and send copies to the provincial office (a change in recordkeeping), and it raised their fee from ten to twenty cents per registration. Finally, the registrar-general henceforth was to be a member of Cabinet, but not necessarily the provincial secretary. The last change caught up with developments. Under an 1888 act to establish the Department of Agriculture, the minister of agriculture was registrar-general for the 1888-92 period, and the provincial treasurer assumed the position during the 1893-97 period.31 The 1896 act took legal effect on i July and was implemented on i August. Thus the immediate benefits of the act are evident from a comparison of the returns for 1895 and 1897, respectively the last full year under the 1875 act and the first full year under the new legislation. Assuming that population increased at a uniform annual rate for the decade 1891-1901, it increased by only 0.006 per cent during this two-year period.32 Yet mortality registrations were up by 23 per cent, and birth and marriage registrations were up by 14 and 9 per cent respectively. A comparison of age-specific mortality data for 1891 and 1900 also encouraged officials.33 As shown in Table 2.1, the aging of the Ontario population was causing an increase in reported deaths for the older age groups and a decrease for younger age groups. Despite the trend, the number of infant death registrations increased sharply, and the number for the one-to-four age group

40 Facts of Life Table 2.1 A Comparison of Age-Specific Death Totals, 1891 and 1900

Age Group

Death Registrations 1891

Death Registrations 1900

% Change

Infant 1-4 5-9 10-14 15-19 20-24 25-29 30-39 40-49 50-59 60-69 70-79 80 +

4,206 1,988 1,540 722 740 876 911 1,423 1,270 1,386 1,843 2,214 2,224

7,163 1,989 803 563 941 1,257 1,083 1,908 1,819 1,931 2,949 3,825 3,099

+ 70 0 -52 -22 + 27 + 43 + 18 + 33 + 93 + 40 + 60 + 72 + 39

held constant. To Dr Bryce this indicated an improved reporting of infant and early childhood deaths under the 1896 act. By 1898 Inspector Hamilton believed that returns from older, settled parts of Ontario were 85 per cent complete for deaths and 75 per cent complete for births. In 1902 Dr Bryce judged that the returns were largely complete except for births and that province's low reported death rate evidenced a salubrious environment rather than incomplete registration.34 In this regard, officials underestimated population increase during the period 1901-10, and so calculated vital rates that were higher - implying more complete returns - than the actual population warranted.33 The 1908 and 1919 Vital Statistics Acts. The two later Vital Statistics acts changed little in the 1896 legislation. The 1908 act made the attendant liable to penalty for not reporting a birth regardless of whether a parent had registered it. It also required municipal registrars to submit their returns quarterly, not semi-annually as under the 1896 act. The 1919 act was passed to prepare Ontario for entry into a national registration area. Coming into force i January 1920, it required a physician to report a birth within forty-eight hours and the cause of death within twenty-four hours; in both instances the specific interval replaced "forthwith."36 Second, it required municipal registrars to submit monthly returns to the central office, not quarterly returns as under the 1908 act. It also required keepers of

41 The Evolution of Ontario Civil Registration

cemeteries to submit monthly records of burials to the registrar of the municipality in which the burial ground was situate. THE COVERAGE OF P R O V I N C I A L REGISTRATIONS,

1896-1920

Death Registrations. An acid test of completeness of registered deaths was whether their numbers exceeded those of censusreported deaths. Enumerated deaths were always incomplete - by as much as 20 per cent, according to the Dominion statistician.37 Census materials were subject to errors by enumerators, who failed to ask questions pertaining to vital events or recorded responses incorrectly. Enumeration also missed deaths of persons who lived alone or whose households had dissolved after their deaths - in both situations, no respondent was left to report the death. Finally, the lack of concurrence between calendar and census years (enumeration day generally was in early April) and the retrospective nature of the questions in the enumerations ("Were there any deaths in the household during the twelve months preceding enumeration day?") generated timereference errors by respondents, who might think of the census year as longer or shorter than a calendar year. Time-reference errors caused more underenumeration than overenumeration. In general, the more distant the time of death was from enumeration day, the less likely it would be reported. As shown in Table 2.2, monthly ratios of census deaths (reported at one time in the year) to registered deaths (reported throughout the year) indicate a temporal stacking of census-reported deaths for the three months preceding enumeration day. Despite its serious defects, census enumeration consistently reported more deaths for Ontario than the province's civil registration for corresponding time periods. As shown in Table 2.3, registered deaths were 10 per cent fewer in number than the enumerated deaths for the 1900-01 census year, by which time the Ontario deputy registrar-general considered death registrations to be complete. Registered deaths appear to have exceeded the number of enumerated deaths for the first time in 1911, making that year the earliest for which one can assume reasonably complete Ontario mortality statistics. Birth Registrations. As evidenced by delayed registrations of births, provincial birth records remained incomplete into the 19205. With the advent of old age pensions, automobile driver's licences, and various requirements of the welfare state, proof of age and naturalization

42 Facts of Life Table 2.2 Number of Enumerated Deaths per 100 Registered Deaths in Ontario, 1880-81, 1890-91, 1900-01 Month

1880-81

1890-91

1900-01

April (1880, 1890, 1900)

101.7 101.0 100.5 105.6 108.2 107.1 112.1 114.8 110.8 123.0 124.2 133.6

105.0 92.4 81.6 89.6 103.1 102.7 97.8 97.3 116.1 124.9 131.6 134.9

105.4 105.7 105.0 104.4 106.2 102.4 102.6 108.5 105.4 110.6 110.6 115.9

May June July August September October November December January February March

Note: The 1871, 1881, 1891, and 1901 census years ended respectively on 2, 4, 5, and 1 April. Thus the registration data are for the first quarter of the year in which the census year ended and the last three-quarters of the year preceding.

Table 2.3 Census Deaths, Registered Deaths, and Registered Deaths as a Percentage of Census Deaths

Year

Census Deaths

Registered Deaths

Percentage of Census Deaths

1871 1881 1891 1901

18,063 22,937 24, 21 12 33,272

8,043' 20,219 22,793 29,942-1

45 78 94 90

1 Annual mean for 1870-71. 2 Stillbirths added to correspond with registration data. 3 Stillbirths excluded to correspond with census data.

became increasingly necessary for Canadians. Each year, therefore, thousands of Ontario-born persons whose births were not registered applied for delayed registration. Where accompanied by evidence deemed satisfactory (e.g., declaration of a parent, baptismal certificate), registration was granted. The influence of pension requirements on these data is impressive. Filed by year of birth, delayed registrations increased in number for each successive birth year until 1924, after which the numbers dropped abruptly. The explanation is that persons born in 1924 or earlier had turned 65 - the age for pension eligibility - by 1986 (the

43 The Evolution of Ontario Civil Registration last registration year whose records were accessible to the writer), whereas persons born in later years had not. The province holds some 200,000 delayed registrations for the 1924 and older birth groups, and each year pushes the total higher.38 Yet even this statistic understates the number of persons whose births are missed by civil registration. As elaborated in chapter 4, many such persons die before life-course events can prompt them to apply for delayed registration. Marriage Registrations. From the outset, civil registration coverage was apparently more complete for marriages than for births and deaths. By official estimate the marriage returns were 75 per cent complete in 1871, 90 per cent complete in 1880, and nearly complete by 1890. New evidence in the late 18905 supported this judgment. By amendment in 1897, the Marriage Act required issuers of marriage licences to report the particulars to the registrar-general's office on cards provided for the purpose. Branch clerks then tried to match each card to a registration. By corresponding with local registrars and clergymen about unmatched cards, officials learned that some of the marriages had been called off or held in a different municipality than the one stated. In only 980 cases for 1899 - about 6 per cent of the year's reported marriage total - did officials discover than the clergymen had not registered the marriages, and by 1901 this statistic was down to 2.4 per cent. Also satisfying was that Ontario averaged only two divorces per year during the 1884-93 period; thus, the remarriage of divorced persons did not inflate the reported marriage rate, as was the case in certain states of the United States. As Inspector Hamilton discovered in 1892, however, a cross-border marriage traffic was distorting reported marriage rates. In Ottawa the traffic was two-way across the Quebec border, which caused Ottawa's reported rate to fluctuate erratically. A more serious problem in 1894 was a one-way influx of American couples to get married in Ontario border towns - especially Windsor, which was next to "Detroit, a large city, famous for its divorces." During the 1894-1903 period Windsor's reported marriage rate was between eight and eleven times that for the province. In 1903 both parties were American in 933 (78 per cent) of Windsor's 1,193 marriages, and 80 of the Americans were divorced persons. By 1912 Windsor's reported marriage rate was fifteen times the provincial rate, and 92.3 per cent of its marriages were between Americans. Dr Charles A. Hodgetts, the deputy registrar-general, was appalled at the "wholesale traffic" in Essex County - "a blot on the good name of the Province." At Sandwich in 1905 one clergyman performed 297

44

Facts of Life

of the town's 318 marriages, and three members of his family acted as witnesses 163, 111, and 75 times respectively. Many Essex County licence issuers, Hodgetts claimed in 1908, were "in commercial life, often jewellers, who benefit from the sale of rings." For Hodgetts the problem was moral, not just statistical. The marriage law was intended to screen out the mentally or physically unfit, the immature, and the unscrupulous from marriage, and safeguard the family, virtuous womanhood, and the vigour of the British race.39 From this vantage point the traffic in the "marrying county" was "repugnant to that sense of propriety which is supposed to mark the social qualities of the Canadian people." Ontario's lax 1875 Marriage Act was the source of the problem. To get a marriage licence under the 1874 Marriage Act, one of the parties had to swear an affidavit stating that one of the parties had resided in the county where the marriage was to take place for fifteen days prior to issuance of the licence. The 1875 act, however, provided a "loophole" - where neither of the parties met the residence requirement, the affidavit need merely state that the reason for solemnizing the marriage in such a place was "not to evade due publicity or for any other improper purpose."40 Having shown that "abuses of The Marriage Act were occurring," Hodgett's successor as deputy registrar-general, Dr John S.W. McCullough, claimed credit for an amendment to the Marriage Act that closed the "loophole."41 After 15 May 1913, where neither party met the residence requirement, the applicants had to produce issues of newspapers showing that "notice of the intended marriage ... has been published once a week for three successive weeks immediately preceding the application for the licence." The amendment was effective. Between 1912 and 1914 Windsor's marriage total plunged by 81 per cent, from 3,429 to 622, and the Niagara Falls total fell 65 per cent, from 839 to 291. ONTARIO IN THE NATIONAL R E G I S T R A T I O N A R E A , 1920-50

The 1918 Canada Statistics Act turned Canada's Census and Statistics Office (established 1905) into the Dominion Bureau of Statistics, headed by the Dominion statistician. The bureau's mandate was to centralize, consolidate, and integrate the "purely statistical work" of the different Dominion government departments and provincial governments.42 Vital statistics, then, was one of many potential interests in the Dominion Bureau of Statistics, not the sole concern, as in the provincial Registrar-General's Branch. At the same time it was an

45 The Evolution of Ontario Civil Registration

early candidate for the bureau's attention. In 1918 vital statistics from the various provincial systems were not available for national purposes. One province [New Brunswick] did not maintain provincial records of births and deaths. In the others the legislation and methods differed widely; for instance the international classification of deaths was not universally followed, and each province had its own scheme for the collection, compilation, and publication of the results. Of the twenty-four items which the death certificate usually calls for, two provinces omitted sixteen, another fifteen, and another thirteen. Standards of administration, i.e., the degree of enforcement of registration, differed between the provinces and sometimes within the same province from year to year. Thus whilst it was possible to assemble certain totals, the latter were incomplete and otherwise inadequate to the purposes of a national system of vital statistics.43

Following consultation with the provinces, the Union of Canadian Municipalities, and special-interest groups in 1918, the bureau drafted a model provincial Vital Statistics Act and model registration forms for births, marriages, and deaths; these defined the minimum standard for a proposed national registration area. In 1919 Ontario amended its Vital Statistics Act to conform to the model. In 1920 the national registration area was launched, with all provinces except Quebec participating.44 In 1923 the bureau issued Vital Statistics, 1921, the first of its annual reports on vital statistics. Finally, Quebec joined the national registration area in 1926.45 Under the 1918 agreement the participating provinces sent certified copies (or loaned the originals) of their completed registrations to the bureau, where clerks recorded the information on punch cards for compilation on its Hollerith punch-card machines.46 The punched cards were exclusively for statistical work; unlike the provincial offices, the bureau had no direct dealings with the public. The bureau offered important services in return. First, at no cost, it supplied the provinces with standard forms for registering vital events. On request, questions would be added to the forms issued to a particular province. In this regard the bureau's national standard was a minimum only, and the provinces retained the "right of individual treatment ... [and] independent compilation and publication."47 Second, the bureau prepared the statistical tables that a province requested for its annual report.48 This work was convenient for the bureau, which had the punch-card machines, prepared the requisite cards for its own annual report, and compiled the national statistics

46

Facts of Life

by working upwards from small-area and provincial subtotals. By contrast, it had been a burden for the Ontario office. Using manual tabulation, provincial staff had taken six to eight months to complete the province's statistical tables for iQoS.49 Third, some time after alphabetical tabulating machines became commercially available in 1934, the bureau began to issue the provinces "routine monthly listings of current births, marriages and deaths to facilitate their searches of records."50 This service presumably ceased to have value for Ontario in 1945, when the provincial office acquired its own punch-card equipment. Uses of Vital Statistics, 1920-50: Public Health. Due to changes in Canadian health problems, vital statistics became less salient in public-health statistics. As a bureau official explained in 1954, High mortality from infective and contagious diseases in youth has given place to higher incidence of chronic and degenerative illness in later life. Prolonged life expectancy and the emergence of crippling disabling conditions has posed new problems in medical care, preventive medicine, and hospital administration, which, in turn, have prompted the extension and development of morbidity and institutional statistics to a point of equal importance with traditional vital statistics [i.e., mortality statistics for deaths by cause]. In all aspects of health statistics, increased knowledge of the purely medical characteristics of disease has permitted diversion of effort towards statistics of the social and economic attributes of illness and hospitalization.51 In Ontario, provincial government administrative changes in 1931 reflected the loosening of ties between vital statistics and public health. Until then the deputy registrars-general - Peter H. Bryce (1891-1904), Charles A. Hodgetts (1905-09), John W.S. McCullough (1910-25), and William J. Bell (1926-30) - had been medical doctors and also Ontario's chief public-health officials. In 1925, moreover, the Branch had been transferred to the newly created Department of Health. In 1931, however, the Branch returned to the Provincial Secretary's Department, and the deputy registrar-general henceforth was not a medical man. Bell's successor, Francis V. Johns (1931-45), was the assistant provincial secretary, and Johns' successor was Ralph B. Wallace (1946-66), a Branch clerk with nineteen years seniority. Uses of Vital Statistics, 1920-50: Social Programs. Towards the end of the Second World War vital statistics became increasingly useful for government's administration of programs. Statistics about the

47 The Evolution of Ontario Civil Registration

incidence of births and deaths at different ages, for example, helped government to predict the fiscal burden of programs such as family allowances and old age pensions. Similarly, population trends calculated from vital and migration statistics shed light on economic trends. Changes in marriage and birth rates, for example, are of considerable importance in forecasting the probable level of construction activity, with its important bearing on other economic activities. An increase in the number of marriages is likely to mean that more homes will be needed, and an increase in the birth rate indicates that attendance at primary schools will increase in five or six years' time and at high schools in thirteen or fourteen years, and that the demand for other facilities will tend to rise as time goes on. The effect of a decrease in the marriage or birth rate will obviously have the opposite implications ... Other factors, of course have to be taken into account when estimating the effects of these changes. For example, the incidence of marriages on the demand for houses is apt to be less during a depression because the couple may live for a time with one of their families.52

Thus, in addition to its routine work in vital statistics, the bureau undertook special projects to meet emergent needs for social programs.53 For example, the Canadian government wished to verify applications for family allowances, which were scheduled for introduction on i January 1945. Accordingly, between 1944 and 1948 the bureau constructed a punch-card-based alphabetical index of births occurring within the provinces since i January 1925, a massive task requiring one hundred extra clerks. A second need was to link records of death to records of birth so as to provide an "automatic death clearance" for the administration of social programs. Early in 1945, for example, the deputy minister of welfare pointed out "how particularly important it is that a record of all the deaths occurring within the first six months of this year should be sent to us ... in time to prevent us from unwittingly making payments to parents who quite improperly made application in February, March, or April in respect to children who have since died."54 Accordingly, between 1945 and 1950 the bureau prepared a punch-card-based alphabetical index of deaths occurring in Canada since i January 1925. The birth and death indexes in turn supported a grander project, the construction of a national register of population. Modelled on England's proposal for a peacetime national register, the idea was to assign each person a registration number at birth and then to add records of the person's subsequent vital events. Thus the register

48 Facts of Life

would link records of a person's different vital events, which was not the case with the previous "segmented treatment for facts of birth, marriage, death, adoption, and divorce." Standardization in Vital Statistics, 1920-50. Despite the Dominionprovincial agreement on model registration forms and the model Vital Statistics Act, the national registration area started as a loose federation of provincial systems, with considerable provincial variation in statistical practice. As discussed in chapter 5, for example, each province followed its own definitions for live birth, stillbirth, and infant death, in no small part because standard definitions were lacking at the international level. To advance the work of standardization, bureau officials collaborated with the provinces and international organizations - the League of Nations until 1939, and the United Nations after the Second World War. Beginning in 1929 bureau and provincial officials co-operated through meetings of the Vital Statistics Section of the Canadian Public Health Association. To the same end, the bureau organized Dominion-provincial conferences on vital statistics in 1918 (two), 1943, 1944, and 1947; and annual meetings of the Vital Statistics Council for Canada, an organization created in 1945 to facilitate co-operation on technical matters.55 During the 1920-50 period the bureau and the provinces resolved several issues that receive detailed discussion in later chapters. These included standard definitions and forms for the reporting of live births, stillbirths, and infant deaths; a new standard form for the death certificate; and standard definitions of residence, to allow the reallocation of registrations from municipality of occurrence (the basis for collection) to municipality of residence for the person having the vital event (the information desired for calculation of vital rates). The bureau responded primarily to international developments when orchestrating a national solution for the first two problems. By contrast, its solution to the residence issue answered primarily to pressure from health boards of the larger cities; this issue mattered considerably for local-area statistics but little for statistics at the provincial and national levels.

ONTARIO'S CIVIL REGISTRATION AND V I T A L S T A T I S T I C S AT M I D - C E N T U R Y Ontario's population was 4.6 million in 1951, up from 2.2 million in 1921 and 1.6 million in 1871. In the course of this 184 per cent increase in population, the number of municipal registrars had risen 62 per cent: from 588 in 1870 to 932 in 1920 and 954 in 1950.56

49 The Evolution of Ontario Civil Registration

Ontario's annual number of registered vital events had increased by 354 per cent: from 43,887 in 1871 to 133,055 in 1920 and 199,294 in 1950. Over time, moreover, people had made increasing use of the registrations. Whereas the central office had only 833 requests for certificates of vital events in 1900, it had responded to 114,521 requests in 1926. To handle the work, the Branch staff had increased to 156 for the period 1945-49, up from 44 for the 1920-24 period and 5 for the period 1870-74. In addition, the Branch had acquired its own punch-card equipment in 1945 and a full-time deputy registrargeneral, Ralph B. Wallace, in 1946. Information about the Dominion Bureau of Statistics is surprisingly hard to come by, except for 1952, when the bureau published a description of its activities. In that year a statistician and staff of 41 manned the Vital Statistics Section. In addition, 135 key-punch clerks and machine operators in the Mechanical Tabulation Division compiled the published vital statistics for the bureau and the provinces.57 In any case, the great challenges in civil registration were largely over by mid-century. At the provincial level the coverage and quality of Ontario civil registrations were no longer serious problems. At the national level the issues of standardization were largely resolved, and the basis for Canada's national population register was in place. At the same time, change in the uses of vital statistics had pushed the Vital Statistics Section to the sidelines of statistical research. "The Vital Statistics Section," the Dominion statistician observed in 1952, "operates basically as a service organization in the sense that the statistics are published for the most part in primary form. Although rates are computed and published there is little present involvement in any sort of intensive statistical research. Health administrators and other agencies using vital statistics material carry out further detailed analyses or interpretations which they may require and often are aided in doing so by officers of the Division."58 As in the Ontario Registrar-General's Branch, in other words, the work had become administrative routine, performed largely by machines and single women, under male supervision.59 In several ways 1950 marks the end of an era in the history of Ontario civil registration and vital statistics. Accordingly, the book turns now to the examination of special topics in the pre-1950 period. Chapters 3 to 7 provide empirical illustrations of how aspects of definition influence the statistics and suggest general strategies for responding. In the process they extend the theme of social construction to particular statistical areas.

3 Death in Ingersoll, 1880-1972: A Case-Study Approach to the Revision of Defective Mortality Statistics WITH KEVIN MCQUILLAN

Ontario mortality registrations have major deficiencies for statistical purposes. First, they give incomplete coverage of deaths occurring in a locality, especially for nineteenth-century years. Second, as long as the coverage is incomplete, it is likely to be biased - that is, exceptionally incomplete for certain population groups (e.g., defined by age, sex, or social class). Third, increasingly into the twentieth century, deaths occurring in a locality differ from deaths involving its residents. Clearly the mortality statistics that officials compiled from the registrations cannot be taken at face value. Yet how can one measure the deficiencies in order to correct for them? This chapter explores a case-study solution to these empirical problems, using evidence for the town of Ingersoll and its two contiguous townships. Within its limits, the case study is an important genre in the literature. On the one hand, it is tricky to generalize case-study findings. Each locality differs somewhat from all other localities the "typical community" is a discarded fiction. Furthermore, casestudy research findings may be idiosyncratic due to random fluctuations in small numbers. As well, the findings from different case studies often are not readily comparable. Commonly the source materials vary, as do the scholar's methodological decisions. On the other hand, the case study permits a more in-depth analysis than a general study provides. Using published summary statistics and standard estimation techniques, for example, demographers can only describe long-term mortality trends by age and sex for large populations.1 In a case study, by contrast, one can use individual-

5i Death in Ingersoll, 1880-1972

level data and record linkage to elaborate details behind the trends, such as social-class differences and patterns of death by cause. Furthermore, case studies elucidate local variation in mortality experience. During the nineteenth century more than now, changes in living standards, social-class formation, and other factors influencing mortality were local processes, and uneven economic development often enhanced differences among localities. Thus the sources of mortality decline, manifested in changing patterns of mortality by cause of death, varied greatly among countries and localities within countries.2 Case studies capture this local particularism better than time-series data for large populations. Accordingly, this chapter shows the feasibility and limits of the case-study approach to Ontario historical mortality. First, it introduces the study area. Second, it describes the available source materials, including death registrations, and shows how to combine information from different sources to obtain the most complete record possible for locally occurring deaths. Third, it assesses the completeness of the mortality file for locally occurring deaths and how these deaths correspond to deaths of local residents. Fourth, it presents mortality trends calculated from the data for the period 1881-1972 and compares the Ingersoll trends (as calculated) to the provincial trends (as estimated). Here the plausibility of the calculated local trends is a further test of the adequacy of the study data. Finally, it discusses the implications of the Ingersoll research findings for the case-study approach. THE STUDY AREA

As shown in Figure 3.1, the study area comprised the Oxford County town of Ingersoll and its contiguous townships, North Oxford and West Oxford. Population in these municipalities increased from 4,462 in 1852 to 8,205 in 1901 and 12,402 in 1971. Ingersoll, incorporated as a village in 1852 and a town in 1865, held 27 per cent of the studyarea population in 1852, 56 per cent in 1901, and 63 per cent in 1971. Based on wheat and lumber exports in the 18505, the local economy shifted to livestock and dairy exports in the i86os, while the town economy blended local service industries with industrial production for local and non-local markets. The opening of the Great Western Railway in 1853 sparked the early development of the town, but its location between London, southwestern Ontario's regional metropolis thirty-four kilometres to the west, and Woodstock, the county seat seventeen kilometres to the east, imposed upper limits on the town's economic and population growth.

52

Facts of Life

Figure 3.1 The Study Area: Ingersoll, North Oxford Township, and West Oxford Township

53 Death in Ingersoll, 1880-1972

The three municipalities were not a community in a functional sense. Though Ingersoll was at the centre of the two townships, the eastern ends of the townships abutted Woodstock, the county seat. Thus Woodstock was the principal urban centre for persons in eastern extremities of the townships. Ingersoll, by contrast, was the principal urban centre for certain rural populations to the north, south, and west of the study area. Ingersoll is eminently suitable for a case study. Its population is large enough to support statistical analysis and yet small enough to be manageable. Compared with larger urban places such as London, its locally occurring deaths also are more easily related to population, since fewer non-residents are involved. The inclusion of the two townships captures corporate Ingersoll's urban fringe and allows for the influence of town-township interaction on deaths reported for the town. As shown below, growing numbers of township residents died in the town hospital and town nursing homes. For the town alone, these deaths are "non-resident." With the townships included in the study area, however, the deaths are "resident" and can be related to the local population. At the same time, the inclusion of the townships in the study area limits the usefulness of published statistics on residence. Beginning in 1924, the numbers of "non-resident" deaths in cities and towns were published by the Dominion Bureau of Statistics in Vital Statistics, its printed annual report. Subsequently the bureau prepared special reports on mortality by place of residence for the periods 1930-32 and 1935-36. For each city, town, and remaining parts of a county, these documents give totals for locally occurring deaths, locally occurring deaths of non-residents, deaths of residents occurring outside the locality, and all deaths of local residents. Beginning in 1944, the bureau included this information routinely in Vital Statistics. Unfortunately, these data are not reported separately for the townships, and the statistics for Ingersoll, as noted above, class deaths of township residents as non-resident. ASSEMBLING THE DATA

Problems with evidence bedevil any study of Ontario mortality. Lacking a dominant state church, the province has no counterpart to the comprehensive parish records (registers of baptisms, marriages, and burials) that are the basis for studies of Quebec Roman Catholic and various European populations. To document provincial mortality, therefore, one must rely primarily on civil registration data, which date from 1869.

54

Facts of Life

Provincial death registrations offer important advantages for the study of mortality; unlike parish records, for example, they report medically certified cause of death. Yet they miss many nineteenthcentury deaths and only gradually become complete. Thus calculations from early registration data mislead; they appear to show rising mortality levels but really show improved registration. The early data also are biased. As shown below, for example, they disproportionately underreport deaths in unskilled, working-class households, thereby masking class differences in mortality, an important feature of nineteenth-century populations. Similarly, although appearing to show that mortality rates are higher for urban than for rural populations, the data really may show a better reporting of vital events in urban places, the reporting of rural vital events in urban places for convenience, and the emergence of hospitals as central sites for births and deaths. The local case study permits a partial solution to problems described above. Where death registrations for a given locality are incomplete, other documentary sources report local deaths that provincial registration misses. Thus, combining reports of deaths from all sources (using record linkage to eliminate duplicate cases) provides a more complete record of local deaths than do death registrations alone. Using various methods, one may then estimate how completely the additional sources report the missing deaths or, more specifically, when data from all sources become reliable. Investigation turns up six documentary sources with information on Ingersoll-area deaths. The principal source, provincial death registrations, dates from i July 1869. Other sources include the burial register of the Ingersoll Rural Cemetery Company (founded in 1864), Ingersoll newspapers (1854-1919), the 1871 manuscript census mortality schedule, the Roman Catholic burial register (1850-), and the Church of England burial register (1838-). The various sources, however, report deaths for different populations. Provincial death registrations report deaths occurring in the municipality, which include deaths of non-residents and miss deaths of residents that occurred elsewhere. In other words, death registrations report deaths for a population that differs somewhat from the municipal population. By contrast, the Roman Catholic and Church of England burial registers and the 1871 census mortality schedule state a decedent's residence but not the location of death. The Ingersoll Rural Cemetery Company burial register lists all persons buried in the cemetery, regardless of where they died or had been resident. Newspaper death notices and obituaries report deaths that the editor deemed to be of local interest and also deaths of persons whose

55 Death in Ingersoll, 1880-1972

relatives paid for a notice (a twenty-five cent cost in the Ingersoll Chronicle during the 18903). As both the burial register and newspapers report where a death occurred, one can select the locally occurring deaths from these sources and use them in combination with death registrations. Ideally one would select deaths by decedents' municipality of residence so that these could be related to the local population, which is represented by enumerated census populations and defined largely by residence. The nature of death registrations, the principal documentary source, however, requires selection by municipality of occurrence. Thus data for the case study (the Ingersoll mortality file) include cases from death registrations, the cemetery register, and newspapers, but not the 1871 census mortality schedule and the Roman Catholic and Church of England parish records. Death registrations provide the starting-point for the Ingersoll mortality file. These include all deaths registered for the community from i July 1869 until 1942, and also for three-year clusters around later census years (1950-52, 1960-62, 1970-72). Expedience justifies collecting data for later years for three-year clusters only. In the mid19408 the registrar-general dropped the practice of filing death registrations by locality. From this point on, therefore, the assembling of death records for the study area was more time-consuming than for earlier years. In 1975 North Oxford and West Oxford townships disappeared in a restructuring of Oxford County; thus the study period does not extend to 1980-82. The second source on local deaths is the Ingersoll Rural Cemetery Company burial register. The mortality file includes all deaths reported in the register as having occurred in the study area before 1934 (an arbitrary cut-off point) but that are not in the death registrations. Ingersoll newspapers, the third source, receive lower priority than the burial register because they give less information. For 1896 and earlier years the mortality file includes all newspaper-reported deaths that are stated to have occurred locally but are not reported in either of the other sources. Newspapers are not used for later years because, for reasons unknown to us, the Ingersoll newspaper reduced its coverage of vital events in 1896; by then, as elaborated below, death registrations had become virtually complete. Information missing in death registrations, such as a decedent's age or cause of death, often can be added from a second source. Sometimes different documentary sources give conflicting information about a death. Of 1,772 deaths reported in both death registrations and newspapers, reported age differs by more than one year for 6 per cent (98 cases). Of 216 deaths reported in two town

56

Facts of Life

Figure 3.2 Ingersoll-Area Mortality File: Composition by Documentary Source

newspapers during the period 1876-79, reported age again differs by more than a year for 6 per cent (13 cases). The three documentary sources hold some 18,000 reports of deaths for years covered in the study, including some 8,500 from death registrations, 6,800 from the Rural Cemetery Company register, and 2,700 from newspapers.3 Many deaths for early years are reported in more than one source; when duplicate reports are removed, some 9,000 unique cases are left. The file excludes 93 deaths registered for the study area but that really occurred elsewhere, according to marginalia on the registration forms and information on residence in the cemetery register. To conform to published statistics and interna-

57 Death in Ingersoll, 1880-1972 Table 3.1 Published Statistics for Ingersoll Deaths by Place of Occurrence and Place of Residence

Occurrence

Deaths of NonResidents

Resident Deaths Elsewhere

56 71 59 76 97 77 85

11 (11) 7( 6) 11 ( 9) 20 (16) 17 (14) 21 (20) 14(13)

3 6 1 10 12

Total

by 1930 1931 1932 1935 1936 1944 1945

7(1) 9(6)

Total

by Residence

48 70 50 66 92 63 80

Note: Figures in parentheses = deaths in public institutions. The DBS special reports assigned deaths of residents of foreign jurisdictions to the categories for residents, whereas the annual report of the registrar-general classed such deaths as non-resident. To capture the foreign resident death for 1932, the annual report statistic is used for the 1932 non-resident death total. The special reports for 1930-32 classed all deaths in the Woodstock and London asylums as resident deaths for Woodstock and London. For 1935 and later years, however, asylum decedents were classed as residents of the municipalities from which they had entered the asylum. The change in definition may have contributed to the rise in the totals for resident deaths occurring elsewhere.

tional practice, the file also excludes 380 records of stillbirths, which were identified from information on age and cause of death. The analysis excludes at the outset the data for 1870 and earlier years, which are manifestly incomplete. Figure 3.2 shows the composition of the Ingersoll mortality file by documentary source for the period 1871-1910. THE PROBLEM OF DECEDENT'S RESIDENCE

The mortality file misses deaths of residents occurring outside the study area - for example, in the Woodstock hospital (opened in 1895) or hospitals in London. Yet it includes deaths of non-residents who perished in local traffic accidents, or who died in the Ingersoll hospital (opened 1909) or a town nursing home. Thus an important concern is whether deaths in the file - that is, locally occurring deaths - closely correspond to deaths of local residents. Published statistics shed some light on differences between the two groups of deaths for the town, as noted in Table 3.1. As noted above, however, the town statistics class deaths of West Oxford and

58 Facts of Life

North Oxford township residents as non-resident. Similarly, an Ingersoll resident's death in one of the townships is classified "elsewhere." A further problem is that published data do not report age and sex distributions for "non-resident deaths" and "deaths of residents occurring elsewhere." Lastly, residence classification is unreliable until 1930, when "place of residence" information was added to registration forms. Before then classification was inferred from "length of residence at place of death" (information introduced to the forms in 1920); thus Ingersoll's published total of "non-resident" deaths for 1925 equals the number of its decedents who had resided in the town less than seven days preceding their deaths. Although information on the death registrations enables one to identify the "non-residents" who died in the study area for 1930 and later years, "residents who died elsewhere" are harder to locate. Given the impracticability of examining all provincial death registrations, the search was confined to Oxford and Middlesex counties (which include Woodstock and London) for the three-year clusters 1930-32 and 1940-42. The two-county search seems adequate for these periods. In township populations both in the study area and nearby, death usually happened at the decedent's home or in a nearby hospital. For the years considered here, 43 West Oxford and North Oxford residents died in Ingersoll (including 40 in the town hospital); 7 died in Woodstock (including 5 in the hospital); 3 died in London hospitals; and i died in another township. The two-county search may be less adequate for deaths of Ingersoll residents outside the study area, but the numbers involved are small. According to published data for the town (which start in 1944 and whose "elsewhere" includes the two townships), 16 Ingersoll residents died "elsewhere" during 1944 and 1945. Of these, 8 were located by extending the two-county search to include these years. Results of the two-county search for the 1930-32 and 1940-42 periods, reported in Table 3.2, show a close similarity between deaths occurring in the study-area (i.e., deaths in the mortality file) and deaths of study area residents. The numbers of "non-resident deaths" and "deaths of residents occurring elsewhere" are small, and the two groups of decedents are similar in number and population characteristics. One can assume that a comparable picture obtains for years before 1930, given that the trend for later years is towards smaller numbers of non-resident deaths as one moves back through time. The two-county search is impractical for later years in the file (1950-52, 1960-62, and 1970-72). Beginning in 1946, Ontario registrations are filed as received from the municipalities rather than by

59 Death in Ingersoll, 1880-1972 Table 3.2 Ingersoll Area, 1930-32, 1940-42: Deaths by Place of Residence and by Place of Occurrence Local Deaths of Residents

Non-Resident Local Deaths

Non-Local Deaths of Residents

Ingersoll Deaths by Residence

Ingersoll Deaths by Occurrence

2930-32 Number 70+ Years Infant Deaths Female Deaths

247 45% 13% 49%

18 6% 11% 28%

18 28% 6% 33%

265 44% 13% 48%

265 43% 13% 48%

1940-42 Number 70+ Years Infant Deaths Female Deaths

287 52% 10% 47%

35 34% 0% 37%

21 21% 5% 48%

308 50% 9% 46%

322 50% 9% 46%

county. Although the some 240 bound volumes of registrations for a given year are indexed by county, a major effort is required to locate the scattered returns for two particular counties. Moreover, the adequacy of the limited search is problematic for later years, when residents die outside the study area in greater numbers and probably in more dispersed geographical patterns. Partial evidence indicates that deaths in the file for the later periods are increasingly different from deaths of residents. First, non-resident decedents increase from 7 and 12 per cent of deaths in the file for the 1930-32 and 1940-42 periods to 20, 19, and 31 per cent of the file for the 1950-52, 1960-62, and 1970-72 periods. Of the nonresident totals for the last three periods, 80, 74, and 83 per cent involve residents of Oxford County or the neighbouring Middlesex County township of South Dorchester. Their removal from the file makes little or no change (o to 2 per cent) in the proportions of deaths involving infants, persons aged 70 or more, and females. Second, published data show a substantial increase in town residents who died "elsewhere," from 15 in 1940-42 to 42, 68, and 50 for the next three three-year clusters. A few of these deaths may have occurred in one of the townships, and hence within the study area. Overall, town and township residents who died outside the study area (numbers unknown) are probably comparable in number to "non-residents" dying in the study area (65, 66, and 71), but one cannot determine whether population characteristics of the two decedent groups are similar.

60

Facts of Life

To summarize, the validity of the case-study approach requires a close correspondence between deaths occurring in the study area and deaths of study-area residents. This condition is met for years before 1945, but for later years the two groups of deaths increasingly involve different persons. Thus the case-study approach is problematic for the post-1945 period. Certainly the Ingersoll mortality trends that are calculated for the later period should not be taken too literally. COMPLETENESS OF THE DATA

The mortality file is problematic for deaths occurring in the 18705. As shown above in Figure 3.2, provincial death registrations in the file miss many of the deaths reported for the community in newspapers and/or the cemetery register. At the same time, the coverage of unregistered deaths in the two other sources is unknown. After 1880, however, the death registrations include most of the deaths reported in the two other sources; they furnish 85 per cent of all cases in the file for the 1881-85 period and more than 90 per cent of the cases for later years. This suggests that death registrations are increasingly complete after 1881, which means that the same holds for the entire mortality file. To obtain a second indication of when the data become reliable, the Sekar-Deming technique is used to estimate their completeness for different time periods. Developed for use in third-world countries whose population data are incomplete, the technique is appropriate for populations whose vital events are reported independently in two documentary sources. Death registrations and newspaper reports of deaths are the two sources used here. Although independent of each other, these sources are not wholly comparable, since civil registration aims to report all local deaths, whereas newspapers report them selectively. In essence, the technique uses the relationships among three groups of deaths - deaths reported in newspapers only, death registrations only, and both sources - to predict the size of a fourth group, deaths reported in neither source.4 Following the technique, the total number of deaths is estimated through the formula ((c + NI)(C + N2))/c, where c represents deaths reported in both sources, NI represents deaths reported only in death registrations, and N2 represents deaths reported only in newspapers. Once the estimated total is known, one can calculate the percentage completeness of deaths reported from particular documentary sources or combinations thereof. Table 3.3 summarizes the results of these calculations.

61 Death in Ingersoll, 1880-1972 Table 3.3 Sekar-Deming Estimates of Completeness, Ingersoll-Area Mortality Data, 1871-75 to 1891-95

Period

Estimated Number of Deaths

1871-75 1876-80 1881-85 1886-90 1891-95

714 666 568 643 544

Completeness of Source (%) Death Registrations

Newspaper Deaths

Both Sources

All Sources

45 65 86 94 92

65 68 75 75 73

81 89 97 99 98

88 93 101 100 99

The estimates accord with the earlier indication of when death registrations become reliable. Death registrations are shown to be seriously incomplete for the 18705, but their coverage sharply improves during the i88os and is nearly complete by the 18908. The estimates also indicate that the cemetery register and newspapers pick up most of the deaths missed by civil registration, and that the mortality file is complete by 1881. By contrast, data for the 18703 appear incomplete and likely to understate actual mortality levels, thereby inflating estimates of life expectancy. However, Sekar-Deming estimates exaggerate completeness where common bias (e.g., a disproportionate underreporting of infant deaths) exists in the two sources whose records are compared.5 To test for common bias between death registrations and newspaper reports of deaths, Sekar-Deming estimates are calculated separately by documentary source for deaths of infants, persons over 65, males, and females. Judging by the results, shown in Table 3.4, the SekarDeming estimates for all deaths are problematic for the period 187180 but reliable for the period 1881-90. Specifically, a common bias is evident for the first period - the coverage in each source is exceptionally low for deaths of males. By contrast, no bias is evident in the coverage of the death registrations for the period 1881-90 - the estimate of their coverage for the different population groups approximates that for the whole population. Thus, whatever the biases in the coverage of the newspaper reports of death, the death registrations do not share them. Using occupations as the proxy for class, death registrations and newspaper reports of deaths were also examined for class bias. According to characteristics of their work tasks, occupations were assigned to one of five categories: non-manual complex, non-manual simple, manual complex, manual simple, and unclassifiable.6 The last

62 Facts of Life Table 3.4 Sekar-Deming Estimates of Completeness, Mortality Data for Selected Population Groups in the Ingersoll Area, 1871-90 Completeness of source (%) LI ji iftiwim

Population Groups

Number of Deaths

Death Registrations

Newspapers

All Sources

Sex 1871-80: Males 1871-80: Females 1881-90: Males 1881-90: Females

662 656 602 608

44 (55) 56 (55) 90 (90) 91 (90)

55 67 75 76

98 92 98 100

Infant Deaths 1871-80 1881-90

257 239

51 (55) 88 (90)

62 60

94 96

Age 65 + 1871-80 1881-90

246 304

58 (55) 90 (90)

74 79

94 96

Note: Estimates in parentheses are for the whole population.

category included occupations such as "gentleman" and cases without information on occupation. The test for class bias is whether death registrations underreport deaths in households headed by persons with simple (unskilled) manual occupations. The answer requires information on the proportion of the Ingersoll population living in such households and the proportion of registered deaths reported for them. If death registrations are unbiased, these households normally would account for more than their proportionate share of deaths registered for Ingersoll. The expectation here is that household populations in the category experienced higher mortality rates than the general population, in consequence of lower nutrition and living standards. Census enumerations probably disproportionately underreport persons in the category; this gives a second reason for expecting that mortality rates calculated from unbiased registration data will be higher for persons in the category than for the general population. For each of the 1871 and 1881 census populations enumerated for the study area, persons were categorized according to occupation of their household head. To determine proportions of population in the various categories, each group was divided by the total population of classifiable persons. Thus calculated, households in the unskilled

63 Death in Ingersoll, 1880-1972 Table 3.5 Ingersoll-Area Persons in the Manual-Simple Occupation Category as a Proportion of Classifiable Persons: 1871 and 1881 Enumerated Census Populations and Death Registrations for 1871-74 and 1881-84

Number All Cases Proportion

1871 Census Population

1871-74 Decedents

1881 Census Population

1881-84 Decedents

1,646 8,010 21%

36 195 19%

1,494 7,799 19%

59 272 22%

manual category accounted for 21 per cent of classifiable population in 1871 and 19 per cent in 1881.7 Death registrations were harder to classify. Many decedents were infants, females, or elderly retired persons and had no reported occupation. In such cases the occupation of spouse, father, household head, sibling, or child was used for classification. Using record linkage (to newspaper death notices, community directories, and manuscript censuses), information about decedent-related occupations was obtained for two arbitrarily selected time periods, 1871-74 and 1881-84. m this fashion 71 per cent of the decedents for the first period and 85 per cent of the decedents for the second were classified. As shown in Table 3.5, civil registration proportionately underreports deaths occurring in unskilled working-class households for the 1871-74 period. Such households hold 21 per cent of the classifiable 1871 census population but supply only 19 per cent of the registered deaths. By the 1881-84 period, however, the bias observed for the first period is not evident, or at least is sharply attenuated. The households in question provide 22 per cent of the registered deaths, though holding only 19 per cent of the classifiable 1881 census population. As shown in Table 3.6, age characteristics of class groups in the two census populations are similar. This removes the possibility that unusual age characteristics of persons in the unskilled manual category (e.g., a disproportionately younger population) caused their underrepresentation in mortality data for the 1871-74 period and that a disappearance of the unusual characteristics made the category's representation more proportionate in death registrations for the 188184 period. Thus the class bias in death registrations and newspaper reports of deaths is real for the 1871-74 period, as is its attenuation in death

64

Facts of Life

Table 3.6 1871 and 1881 Enumerated Census Populations for Ingersoll Area: Age Composition by Occupational Categories 1871

0-14 % Non-manual Complex 38 33 Non-manual Simple 41 Manual Complex 44 Manual Simple Unclassifiable 20 31 No Occupation 40 Total

1881

60+ %

Mean Age

Median Age

0-14 60+ % %

4 4 6 6 22 10 6

24 24 23 23 36 27 24

22 23 18 17 33 21 19

34 32 35 39 24 24 34

4 5 7 6 24 13 8

Mean Age

Median Age

25 25 26 24 36 30 26

21 22 21 20 30 22 21

Note: 1871 N = 8,675; no data = 1. 1881 N = 8,658; no data = 7.

registrations by the 1881-84 period. This leads to two important conclusions. First, death registrations and the whole mortality file may be more incomplete for the 18705 than is indicated by SekarDeming estimates. Secondly, the attenuation or disappearance of bias in the death registrations confirms that these records give more complete coverage by the second time period. To summarize, three different measures indicate that the mortality file is complete for deaths occurring in the study area after 1880, but not for deaths occurring earlier. To press further the evaluation of the data, the chapter turns now to consideration of the trends calculated from them. TRENDS CALCULATED INGERSOLL DATA

FROM THE

The crude death rates for the study area are the simplest measure of its mortality. To allow for random fluctuations in deaths for the small Ingersoll population, Figure 3.3 presents these rates for overlapping twenty-year intervals. In each calculation the numerator is the average annual death total for the interval, and the denominator is the census population at the mid-point. The trend line shows a gradual decline, which accords with the known decline in provincial mortality levels for the study period. The adjusted trend line reflects a correction for a 9 per cent undercount of deaths (the conservative Sekar-Deming estimate) in the 1871-90 period.

65 Death in Ingersoll, 1880-1972

Figure 3.3 Crude and Adjusted Crude Death Rates for the Ingersoll Area

Death rates calculated separately for different age and sex groups furnish a more refined measure of mortality change. This is because mortality is age-related, and the age composition of population changes. For example, persons aged 60 years or more moved from 6 per cent of the Ingersoll population in 1871 to 13 per cent of the 1971 population; similarly persons aged 15 years or less moved from 40 to 30 per cent of the population. The calculation of rates for population groups requires information on age-sex distributions in census populations, which was obtained from manuscript census data for 1871, 1881, and 1891. The manuscript data for later censuses are closed to research, and the published data commonly do not give information at the municipal level. Thus the age-sex distribution published for the South Oxford census district, which contained the community, was assumed for the 1901 study area population. The 1911 and 1921 census publications report

66 Facts of Life

age-sex distributions for Ingersoll but not the townships. Thus the age-sex distributions published for the rural sector of the South Oxford census district were assumed for the township populations. The censuses 1931 through to 1961 lump the infant and i to 4 age groups into a single category, and the 1951 and 1961 censuses similarly reduce the number of categories for persons in the 25 to 64 age range; in these cases one must estimate how the population in the broad categories was distributed among the smaller categories used in the analysis. Inaccuracies in census data affect the computation of mortality rates. Particularly for infants and children, nineteenth-century censuses commonly underenumerate population, thereby inflating mortality rates as calculated.8 They also show "age-heaping," a tendency of respondents unsure of their age to answer in round numbers. This leads to artificially large numbers of people at ages such as 40, 50, and 60. Where age reporting is of poor quality, ages of elderly persons commonly are overstated, which inflates the number of persons in the older age categories and deflates death rates calculated for them. Census data for the community are used without adjustment, however, because one cannot distinguish reliably between systematic errors and random variability in the age distribution of a small population. Age-specific death rates, reported in Table 3.7, are calculated similarly to crude death rates. For each age group the average number of deaths occurring over a twenty-year period is divided by the number of persons in that category at the mid-point of the interval. Asterisks indicate where Ingersoll rates differ by 15 per cent or more from estimated provincial rates.9 The infant death rates present a special problem. On the one hand these rates have unusual importance in mortality analysis. A Statistics Canada study for the 1921-81 period, for example, credits "the reduction in infant mortality ... for more than 40% of the increase in life expectancy at birth for males and more than 25% in the case of females."10 On the other hand, the data available to calculate the rates are problematic. Ideally, the rates in Table 3.7 would state the number of infant deaths per 1,000 live births in the study area. Birth statistics, however, are unreliable for the denominator. The registration of local births is seriously incomplete for early parts of the study period, and locally registered births differ from births involving resident mothers towards the end of the study period. Thus census totals for infants rather than birth statistics are used for calculation, even though the censuses probably underreport infants.

67 Death in Ingersoll, 1880-1972 Table 3.7 Age-Specific Death Rates for the Ingersoll Area, 1871-90 to 1950-72

Age Males 0 1-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45^9 50-54 55-59 60-64 65-69 70 + Females 0 1-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-19 50-54 55-59 60-64 65-69 70 +

187190

18811900

18911910

190120

191130

192140

193052

194062

295072

130*

173 14* 4 2 4 7 6 6 8* 8* 10* 13* 20* 18* 36* 97*

117*

118 8* 3 2 3 4 4 5 6 5* 8* 10* 17* 23* 32* 105

107** 6* 2 2 3 3 4 5 5 6 6* 9* 18 22* 40 79*

94** 6** 3** 2 2 1* 3 4 4 5 9 9 20** 25 41 81*

76** 3 1 0* 2 1* 1* 1* 4 3* 6 7* 16 22 40 77*

36 1 1 0* 1 2 1 0* 3 3 6 9 15 28 36 91

18* 1 1 0 1 1 1 1 2 3 3* 9 13* 25 37 91

111*

120 10* 3 2 3 6 6 6 9 9 8

112**

74 4* 2 1 2 3 5 5 7 5 9 9 16 23 34 93

73** 3 1 0* 2 2 3 4 4 5 7 8 12 22 26* 103

58** 3** 0* 0* 0*

26 1 0 0 0

3** 2 4 5 7 8* 23** 29 85

2** 2 2 3* 7 8 18 27 81

15* 1 0 0 0 1 1 1 1 1* 3 6 9 14 23 76

16* 6 3 5 5 9 7 11 10* 10* 11* 17* 20* 27* 93* 95* 20 7 4 5 8 9 9 0 8 11 10 21 25 33 82

18 4 4 4 7 6 7 13**

9 10 14 15* 25* 55** 83*

11* 3 2 2 6 3 6 6* 3* 8* 11* 12* 22* 35* 101*

18**

14* 23* 39 89*

8 3 2 2 4 6 6 9 8 8 15 16 28 28* 83*

1 1*

1 1

* 15 + % lower than the estimated provincial rate. ** 15 + % higher.

The Ingersoll mortality file may be incomplete for infant deaths, the numerator data for calculation of the rates. More than deaths at later ages, an infant death may go unreported in all three data sources. To die, moreover, an infant must first be born live. Thus, as discussed in chapter 5, the reported number of infant deaths partly

68

Facts of Life

Table 3.8 Expectation of Life at Specified Ages, Ingersoll Area and Ontario, 1881-1900 to 1950-72

18911910

190120

191130

192140

293052

194062

195072

Ingersoll Males 0 50 20 45 11 65

57 48 13

57 49 14

58 48 12

60 48 12

64 50 11

69 52 14

70 52 13

Ontario 0 20 65

49 42 11

55 45 11

58 46 12

61 49 13

65 50 12

67 51 13

68 51 13

Ingersoll Females 50 0 42 20 9 65

54 45 12

58 48 16

62 48 14

66 52 16

69 54 16

73 55 15

75 56 16

Ontario 0 20 65

52 44 12

58 46 12

62 49 13

64 50 13

68 52 14

72 55 15

75 57 16

Age

18811900

Males 46 41 11

Females 49 43 11

reflects how physicians or other persons reporting the births distinguished between live birth and stillbirth (which is unknown). The validity of the reported totals in turn depends on whether we, the analysts, would accept the contemporary distinction (if we knew it).11 The infant-mortality issue clearly requires considerable new research using primary data and cannot be resolved here. Nevertheless, the Table 3.7 rates are plausible in the context of the literature, and it would take a major error to have an important effect on the other chapter findings. More generally, the Table 3.7 rates for the early decades show the expected "U-shaped" pattern of mortality by age, in which risk of death is high in infancy and early childhood, declines rapidly through later childhood and adolescence before beginning a gradual ascent to high levels in older age. Death rates for all age groups declined over time, but the largest declines were centred on infancy and early childhood. The age-specific death rates presented in Table 3.7 were used to calculate period life tables for the Ingersoll area population. Table 3.8 reports the estimates of expectation of life at ages o, 20, and 65, and also the corresponding estimates for Ontario.12 Over the entire study

69 Death in Ingersoll, 1880-1972

period, life expectancy at birth for Ingersoll males increased from 50 to 70, while the provincial figure rose from 46 to 68. The data on life expectancy at age 65 confirm the well-established view that length of life past age 65 has changed little through time. Most of the gain in expectation of life at birth came from reductions in the risk of death in the earliest years of life. Males gained less than females and in a different rhythm. Whereas male advances in life expectancy were clustered at the beginning and near the end of the study period, female advances were distributed evenly through time. The Ingersoll data show excess female mortality at the outset of the study period, followed by excess male mortality after the 1901-20 interval (where excess refers to the amount by which mortality levels for one sex exceeded those for the other). Although Ingersoll's mortality experience (as calculated) is broadly similar to provincial experience (as estimated), certain differences are noteworthy. For example, the adult male mortality for Ingersoll before the 1911-30 period is markedly below the provincial level (and also the levels for Ingersoll and Ontario females). Similarly, as shown in Figure 3.4, Ingersoll's excess female mortality at the beginning of the study period contrasts with the consistent excess male mortality in the provincial pattern. Although one cannot rule out an underreporting of male deaths in the study data as a possibility (e.g., the mortality file may underreport male deaths before the 1911-30 interval, with a particularly severe underreporting for the 1891-1910 interval), this explanation seems unlikely. First, various statistical tests fail to identify sex bias in the mortality file for the 1871-96 period. Second, the hypothetical severe underreporting of male deaths for the 1891-1910 interval works against the findings (described above) that the mortality file is more complete for these years than for the 1881-1900 interval.13 Third, the provincial estimates may exaggerate real sex differences between the community and provincial patterns. Standard estimation techniques consistently fail to predict actual sex differences in mortality and understate how these changed through time. As recently shown, they overstate excess male mortality in high mortality conditions (which works against a finding of excess female mortality) and understate the male excess in low mortality conditions.14 Finally, the play of small numbers may influence the rates calculated for one or both sexes in the study population. CONCLUSION

This chapter attempts a case-study approach to the revision of defective mortality data. First, it shows how one can use information in

7O

Facts of Life

Figure 3.4 Female Advantage in Life Expectancy at Birth, Ingersoll and Ontario

other sources (the cemetery register and newspaper death notices) to identify locally occurring deaths that civil registration misses. Second, it recommends four measures of the completeness of area death registrations and the entire mortality file: i) registration deaths as a proportion of all deaths in the file; ii) the Sekar-Deming estimates; iii) the attenuation of social-class bias in the registration coverage, as inferred from comparison of manuscript census and registration data for occupation; and iv) the credibility of trends calculated from the data, as measured by similarity with trends estimated for the province. Because the findings for one measure are ambiguous, the best test for completeness is whether the findings for the four different measures are in accord. Third, it uses information in the registrations, published statistics, and a "two-county search" to probe the differences between locally occurring deaths and deaths of local residents. In the process it shows how a careful selection of study-

71 Death in Ingersoll, 1880-1972

area boundaries can minimize the problem; in this case the inclusion of the two townships captured most of the non-residents who died in the town. In general the study findings show that the case study is feasible and a partial solution for the deficiencies of mortality registration. On the plus side the study data appear to be adequate for identifying local mortality trends for the 1881-1945 period. On the negative side the data are too incomplete for use before 1881, and the residence problem makes their value problematic after 1945. Furthermore, infant death rates pose a special problem, which is only partly dealt with here, and we cannot clearly answer whether the study data underreport adult male deaths for early intervals within the 18811945 period. Finally, especially for groups within the population, random fluctuations due to small numbers may mask the real trends. Given that the case study is feasible for the 1881-1945 period, it has substantial potential. As calculated, the Ingersoll trends are for the most part unexceptional; indeed, their chief use here is to help validate the case-study approach by showing their accord with trends estimated for the province. Certain findings, however, hold additional interest. In particular, the Ingersoll pattern for sex differences in mortality challenges conventional wisdom and invites further investigation. More generally, differences between the Ingersoll and provincial mortality patterns signal how case studies can elaborate local variation in Ontario mortality experience. Additional work with the Ingersoll data could elaborate local conditions that influenced the trends. The analysis of mortality by cause of death, for example, might show the relative importance of different environmental influences, such as food, water, and air contamination, birthing and weaning traditions, and nutrition levels. Similarly, a class-specific analysis could shed light on social differences in access to features of the local environment. The case study, in other words, offers a promising though partial solution to empirical problems, and it can move knowledge beyond what estimation techniques have shown. At the same time, it raises new questions and fresh ambiguities.

4 The Completeness of Birth Registrations, 1900-60

Birth registrations, the basis for published birth statistics, commonly underreport births for historical periods.1 In Ontario, for example, the compulsory civil registration of births dates from 1869, but the law was widely ignored for some years. In 1930 the eminent demographer Robert R. Kuczynski styled the 1865-95 period as a "thirty year's war against passive indifference" and the 1896-1920 period as "slow progress."2 He guessed that provincial birth registrations were only two-thirds complete for the 1875-95 period and a maximum of 85 per cent complete for any year in the 1896-1919 period. Kuczynski also judged that registration was "more complete from 1913 on" and "at least 90 per cent" complete by 1920. Two other studies found high levels of completeness through comparison of enumerated census populations with birth registrations for months corresponding to the enumeration years. In 1927 E. Stewart Macphail, chief of the Demography Division, Dominion Bureau of Statistics, estimated that Ontario birth registration was 100 per cent complete for 1921.3 In a 1931 census monograph William Richard Tracey, a bureau statistician, estimated a maximum completeness of 100 per cent for the 1927-31 period and a minimum completeness of 96 per cent for 1931.4 Like Macphail, Tracey used published (aggregate) data for his maximum estimate, but his minimum estimate was based upon tracing individuals in the census manuscript to a birth registration. Building on the earlier studies, this chapter elaborates the completeness of Ontario birth registration by analysing evidence from

73 The Completeness of Birth Registrations delayed registrations of birth. First, it places Ontario birth registration and delayed birth registration in their historical context. Second, it analyses information about the delayed registrations to estimate actual live-birth totals for selected cohorts and the year by which the published birth statistics became reliable. Third, it examines how the coverage of birth registrations for the 1920 cohort varied by geographical region, the mother's ethnicity and social class, and the infant's sex. Because the historical coverage of birth registration differs among jurisdictions, the research findings for Ontario cannot be generalized. The documentary source and methodology, however, are pertinent to any jurisdiction whose vital events are documented primarily through civil registration. THE HISTORICAL CONTEXT

The Ontario Registration acts of 1869, 1875, and 1896 provided most of the legislative framework for birth registration in the 1900-60 study period. The 1869 act required the father (or mother, or person representing the parents) and the medical attendant to report a birth, and it provided penalties for wilful non-compliance. Although an amendment in 1869 removed the medical attendant's responsibility, the 1896 act restored it (requiring the attendant to report a birth forthwith). The 1896 act also provided penalties for any non-compliance, not just wilful non-compliance. The 1875 and 1896 acts also provided, however, that neither the physician nor the head of a family was "liable to any penalty in respect of his default" where the "other of such persons" had made a return. Despite the legal requirements, Colonel R.B. Hamilton, the provincial inspector of vital statistics, guessed that the returns for 1898 were only 80 per cent complete, and he blamed medical attendants as the principal culprits. Physicians in the city of London, for example, made "scarcely any pretence of registering births, and few of them even obtain the cards necessary for the purpose from the Division Registrar." In the city of Stratford his canvass of the physicians found "at least 90 more" than the 159 births reported. To promote better compliance with the Registration Act, the inspector placed newspaper notices, circularized physicians, and started court actions. Thus, in December 1899, he noted, "several of the leading citizens of the 'Forest City' [London] were arraigned before the magistrate there and ordered to pay a fine of one dollar and costs, and subsequently four or five physicians also found the Registration Act was not entirely a myth, as under one of its provisions they were ordered to pay a fine into the city treasury. The result

74

Facts of Life

Figure 4.1 Registered Live Births Occurring in Hospitals

in that city has also been most beneficial, and the Division Registrar is now kept busy recording entries that should have been made months ago." To aid prosecution, Hamilton urged the removal of a "weakness" in the 1896 act - the medical attendant's exemption from penalty where a parent had reported the birth. His reports for 1903 and 1904 also urged "drastic measures": the use of a "member of the Provincial detective force" to gather evidence, to be followed by prosecution "at unequal intervals." Beginning in 1905 the provincial government did not fill the position of inspector of vital statistics, and without the inspector's report, the registrar-general's annual reports were uninformative about the completeness of birth registration. Five types of circumstantial evidence, however, pointed to increasingly complete returns. First, new Vital Statistics acts tightened the requirements for medical attendants. The 1908 act made the attendant liable to penalty for not reporting a birth regardless of whether a parent had registered it, and the 1919 act required the attendant to report a birth within forty-eight hours instead of the indefinite forthwith.5 Second, as shown in Figure 4.1, the proportion of Ontario's registered live births occurring in hospitals increased dramatically, from 2 per cent in 1900 to 16 per cent in 1920, 37 per cent in 1930, 62 per cent in 1940, and 98 per cent in i960.6 Here one may surmise that

75 The Completeness of Birth Registrations hospital records of the births, required for the provincial hospital inspector, helped both medical attendants and local registrars to report them. Third, regulatory and welfare legislation enacted during the study period made birth registration increasingly useful for proof of age or citizenship. The Ontario Insurance Act of 1913, for example, made proof of the age stated in a policy a condition for payment of the benefit.7 In 1917 the Military Service Act made males 18-59 eligible for service, set up a selective draft by age category (20-34, 35-41, and 42-59) and marital status within age categories, and placed the burden of proof on the person drafted to show that he was not in the draft category to which he was assigned.8 Also in 1917 provincial statutes made women 21 years of age or more eligible for the voter's list and public office.9 In 1919 two provincial school-attendance acts increased the age of compulsory full-time attendance from 14 to 16 and prohibited the employment of a person under 17 during school hours.10 In 1920 the Ontario Mother's Allowance Act provided help to women who lacked spousal support and had two or more dependents under the age of 14 (amended to 16 in 1921).n In 1927 Ontario's Liquor Control Act repealed prohibition but forbade selling or supplying liquor to a person under 21.12 In the same year Canada's Old Age Pension Act introduced pensions for persons aged 70 or older.13 The proliferation of motor vehicles in Ontario also led to age-related regulation. Following a 1903 law that required vehicle licences, the number of licensed vehicles per 1,000 persons in Ontario increased from 2 in 1910 to 69 in 1920, 121 in 1925, and 186 in 1930 (see Figure 4-2).14 During this time a 1908 law required a commercial driver to have a chauffeur's licence and prohibited persons under 17 from driving on public roads; a 1917 law prohibited persons under 16 from driving and required chauffeur's licences for drivers aged 16 to 18; a 1922 law prohibited anyone from permitting a minor to drive or hiring a person aged 16 to 18 who did not have a chauffeur's licence; and a 1925 law required all drivers not having a chauffeur's licence to obtain an operator's licence, which they were to keep on their persons when driving and produce for a constable on demand.15 Thus, amid the growing popularity of motor vehicles, one had to be licensed to drive and meet an age requirement to be licensed. For various statutes, proof of Ontario birth registration met a citizenship requirement. The American Immigration Act of 1924, for example, allowed Canadian-born persons to apply for an immigration visa that allowed them to work in the United States for up to four months, but it also placed the burden of proof on applicants to show that they were Canadian-born and not in a restricted class.16 Other

76

Facts of Life

Figure 4.2 Licensed Motor Vehicles per 1,000 Ontario Population

statutes with citizenship requirements included the Ontario Adoption Act of 1927, which restricted issuance of an adoption order to British subjects, and the Canada Pension and Ontario Mother's Allowance acts, which limited benefits to British subjects.17 Lastly, as registration officials noted, proof of Canadian birth was required by the federal Immigration Department and all federal government departments paying benefits to "returned soldiers, athletic associations, [and] associations carrying benefit funds."18 The increase of age and citizenship requirements in Ontario caused a corresponding increase of applications for "official certificates of registration" and "letter forms certifying to registrations on file."19 As shown in Table 4.1, the number of birth certificates issued rose by 48 per cent between 1912 and 1913, probably in reaction to the 1913 Ontario Insurance Act. The registrar-general did not report the numbers of birth certificates for later years. However, the number of certificates issued for all vital events (births, marriages, and deaths) increased sharply after 1918, and in 1929 an official noted that these were "principally of birth records."20

77 The Completeness of Birth Registrations Table 4.1 Searches, Certificates Issued, and Birth Certificates Issued Year

Searches

Certificates Issued

Birth Certificates

1903 1904 1911 1912 1913 1918 1925 1926 1927 1928 1929 1930 1931

1,272 1,516 2,854 2,646 3,159 96,500* no data 114,521* 50,000* 50,000* 75,000* 57,000* 38,185*

693 810 1,219 1,311 1,750 60,000* 30,000* 56,521 42,573 41,048 43,700 27,000 18,185

307 (44%) 324 (40%) 676(56%) 689 (53%) 1,019(58%) no data no data no data no data no data no data no data no data

* Registrar-General's estimate; the statistics in parentheses show birth certificates as a proportion of all certificates. Note: The data are from Ontario, Registrar-General, Annual Reports, and Health Department, Annual Reports, 1925-30.

Table 4.2 Stated Purposes for Searches of Vital Records, 1926 Number

%

Certificates

Emigration to the United States Other Purposes

34,850 22,691

30.4 19.8

40,100 4,200 750 700 450 11,800

35.0 3.6 0.7 0.6 0.4 9.4

114,521

99.9

Letter Forms

"Insurance, athletics, etc." Mother's Allowance Adoption Federal Immigration Department Soldier's Benefits Local Police Requests and Unrecorded Total

The registrar-general's annual report for 1918 explained the record activity for that year: During the early part of the year the Military Service Act became operative in Canada, which made it necessary for practically all men from sixteen or seventeen years of age or upwards to carry certificates of birth and, in some

78 Facts of Life cases, certificates of marriage. Never before in the History of the Department has the demand been so great for documentary evidence relating to births and marriages on the part of the public. So great, indeed, was the demand that it became necessary to increase the staff by about three hundred per cent, and work twenty-four hours a day, seven days a week, in eight hour shifts. It is estimated that there were some 60,000 certificates issued during the year, and some 89,500 searches made, for which fees were received. During the same period some 7,000 searches were made and certificates issued free of charge to soldiers or their families.

During the depressed economic conditions of the 1926-30 period the "emigration of Canadian-born citizens going to the U.S.A. for the purpose of obtaining employment" accounted for 61 per cent of the certificates issued in 1926 and "two-thirds" of the certificates issued for 1928. As shown in Table 4.2 for 1926, the emigration purpose is less dominant if "letter forms certifying to registrations" also are considered.21 Thus the study period witnessed a proliferation of government age and citizenship requirements and a growing popular usage of birth registration for proof of age or citizenship. By the 19205, therefore, parents and physicians were probably more careful about registering births than had been the case for earlier years. A fourth type of evidence of improved registration, which Kuczynski used, was the birth rate calculated from the registration data.22 As shown in Figure 4.3, brief surges in the rate during 1896-97 and 1908 may have reflected short-term impacts of the 1896 and 1908 Vital Statistics acts. The gradual rise in the rate between 1899 and 1912, and the sharp rise between 1912 and 1915, apparently informed Kuczynski's judgment that the returns show "slow progress" for the 1896-1920 period and are "more complete from 1913 on." The calculated birth rates, however, may have shown change in the actual rate rather than the reporting of births. Indeed, Kuczynski implicitly interpreted the post-1915 drop in the calculated rate as a decline in the birth rate, not as evidence of less complete registration. Plainly conceding the ambiguity of his evidence, he allowed "no possibility of ascertaining the actual number of births." Fifth, early infant deaths ceased to be much of a problem for the completeness of birth registrations after 1896. These deaths were numerous. According to published statistics for 1926, for example, 8 per cent of Ontario's live-born infants died in infancy, 4 per cent died in their first month of life, and 2 per cent died in their first day of life. Parents had no incentive to register the birth in such cases for proof of age or citizenship. After 1896, however, one had to register the death to obtain a burial permit, and the registration of the death,

79 The Completeness of Birth Registrations

Figure 4.3 Registered Live Births per 1,000 Ontario Population

in turn, presumably alerted the local registrar to the birth. In the circumstances, persons who died near their day of birth probably had a better than average chance of having their birth registered. Thus Tracey linked 99 per cent of 444 Ontario infant-death registrations for 1931 to a birth registration, compared to 96 per cent for infants enumerated in the 1931 census.23 To summarize, circumstantial evidence supports Kuczynski's view that Ontario birth registration became increasingly complete during the early twentieth century. The evidence includes the tightening and vigorous enforcement of requirements in the Vital Statistics Act, the trend towards hospital births, the proliferation of government regulation with age and citizenship requirements, a growing popular use of birth registration for proof of age or citizenship, the rise in the birth rate calculated from published statistics, and the incentive to register births for early infant deaths. To test and elaborate this overview, the chapter turns now to evidence in delayed registrations of birth, a source that Kuczynski "neglected as not essential," in order "to save space."24 DELAYED REGISTRATIONS OF BIRTH: A HISTORICAL DESCRIPTION

Local registrars (municipal clerks) could not register a birth if more than a year had elapsed since the birth date. In such a case, the provincial registrar-general could grant delayed registration on receipt

8o

Facts of Life

Table 4.3 Delayed Registrations for Selected Cohorts by Period When Registration Was Granted Cohort Birth Year Period

1900

1905

1910

1915

1920

1925

1930

1935

1940

1950

1960

0 1900-04 0 0 1905-09 5 0 1910-14 0 87 128 106 1915-19 647 199 366 244 147 1920-24 119 609 355 163 681 1010 1035 1925-29 1930-34 217 256 320 662 416 224 120 166 217 302 302 355 164 170 110 1935-39 502 388 193 188 1940-14 293 369 426 689 191 191 280 297 195 125 263 239 1945-19 85 99 114 1950-54 296 263 289 265 179 97 86 83 1955-59 339 292 319 310 165 62 68 86 1960-64 400 358 356 290 174 32 65 69 1965-69 437 502 335 308 191 68 58 290 891 587 313 142 116 1970-74 81 412 832 441 212 174 118 33 1975-79 51 98 38 209 357 134 7 16 1980-84 36 48 62 20 128 16 2 5 1985-86 3886 4136 5498 5730 3918 1951 1515 1086 Total

217 176 47 88 82 34 42 68 48 21 823

139 107 50 76 58 39 17 4 490

339 152 87 86 19 0 683

Note: Figures in boldface are referenced in discussion.

of appropriate evidence and at his discretion. Between 1915 and 1986 his office issued delayed registrations for 3,886 births in the 1900 cohort. His annual report for 1900, however, was issued on 31 December 1901, well before he had issued any of the delayed registrations. In this fashion, the published birth statistics for 1900 (and all other years) excluded delayed registrations and were manifestly incomplete. Each birth group gathers delayed registrations throughout its lifecourse as certain of its members request birth certificates to prove age or citizenship and discover that their births are not registered. Table 4.3 shows the number of delayed registrations for selected cohorts according to period of issuance. Table 4.4 converts the Table 4.3 subtotals into cohort age categories that correspond to the periods of issuance (with values referenced in discussion shown in boldface). Both historical and life-course events influence cohort accumulation rates. As Table 4.4 data show, for example, the numbers of delayed registrations issued to cohorts 1900 through 1920 increase as each group approaches 65, the age of pension eligibility. In Table 4.3 the numbers for the 1900, 1905, and 1910 cohorts increase for the 1925-29 period. Here probable influences include the new American

81 The Completeness of Birth Registrations Table 4.4 Delayed Registrations for Selected Cohorts by Age Category Age

Category

0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40^4 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-86 Total

Cohort Birth Year

1900 1905

1910 1915 1920 1925 1930

0 5 106 147 0 244 0 0 128 355 87 366 609 0 416 662 647 199 1035 355 320 302 502 119 1010 302 689 681 256 195 217 217 426 297 179 280 166 369 265 165 289 293 191 310 174 191 263 319 290 191 296 292 336 308 313 587 339 358 335 441 891 832 357 400 502 437 290 412 209 128* 81 38 20* 58 16 33 16* 7 5* 2 3886 4136 5498 5730 3918

163 224 164 388 125 99 83 86 69 142 212 134 62*

1951

120 170 193 263 114 86 68 65 116 174 98 48*

1515

1935 110 188 239 85 97 62 32 68 118 51 36*

1086

1940 1950 1960 217 176 47 88 82 34 42 68 48 21*

139 107 50 76 58 39 17 4*

339 152 87 86 19 0*

823

490

683

First two years of category only. Figures in boldface are referenced in discussion.

immigration requirements (1924) and the introduction of automobile driver's licences (1925) and repeal of prohibition (1927), each with provisions excluding minors. Similarly, wartime legislation to mobilize manpower in 1917 and 1940 caused a rise in the numbers for cohorts whose males were in or approaching the ages of military service.25 After noting the huge increase in the number of searches and certificates requested following the implementation of the Military Service Act in 1918, the registrar-general continued: Such huge amount of searching revealed the fact that in the earlier days registration of vital statistics was not nearly so well observed as it is at present. A large number of births and marriages was found not to be registered, and, in all cases where it was possible, the Department gave the privilege of registering, and in this way some 5,000 births were registered. A larger number would have been registered had the Department not been obliged to refuse many the privilege because there was no person available who was qualified to bring sufficient evidence to make the registration. Because historical events exerted an age-selective influence, cohorts differed in their accumulations of delayed registrations of birth. As

82

Facts of Life

Table 4.3 data show, the Military Service Act increased the number of delayed registrations issued to the 1900 cohort, aged 18 in 1918, but did not affect the number for the 1905 cohort, then aged 13. Similarly, the 1940 conscription legislation increased the numbers for cohorts 1905 through 1925, whose ages that year ranged from 15 to 35, but did not affect the numbers for cohorts older or younger. Given such differences, certain cohorts more than others receive large numbers of delayed registrations as they pass through the younger age categories, when comparatively few of their members have died. Relative to the others, therefore, these cohorts produce fewer applicants as they encounter critical life-course events at later ages. Similarly, younger cohorts are likely to have more complete birth registration than the older cohorts, which gives them smaller pools of potential applicants for delayed registration. THE REQUIREMENTS

FOR DELAYED

REGISTRATION

Although the requirements for delayed registration affect the numbers issued, provincial statutes state little about them. Until 1948, for example, the legislation merely refers to "the required information in the prescribed form."26 For details one must consult the instructions that accompanied the application form for the period concerned. Under the 1931 instructions, for example, one applied for delayed registration of a birth by submitting a fee and a notarized statutory declaration of the birth by someone in a position to know about it. The declaration was to come from a parent or, if neither parent was alive, a person old enough to know about the birth. Examples of such a person, in order of preference, were i) an older sibling who could remember the birth or who, while too young to remember the birth, could remember the mother "sick in bed with a young infant with whom I grew up and always recognized as my —"; ii) an aunt or uncle present at birth, or who saw the child within three days of birth and was then informed of the child's birth date; iii) a near neighbour; and iv) a physician or nurse who attended the birth. As stated on the reverse of the form, entries in family Bibles or baptismal certificates could not substitute for the statutory declaration and effectively were discounted as evidence. The requirements became more stringent in 1945-27 Henceforth these included i) the fee; ii) a statement of birth by (in order of preference) the mother, father, a person representing the parents, the occupier of the house if he had knowledge of the birth, or the nurse in attendance; iii) a statutory declaration by the applicant or another

83 The Completeness of Birth Registrations

person; and iv) "such other evidence as may be prescribed by the regulations." The instruction sheet that accompanied the application form used in 1979 elaborates the "other evidence ... prescribed." As detailed below, the key innovation for the post-1945 period was an insistence upon documentary evidence: Note: A delayed registration of birth cannot be effected by this office unless the name, date, place of birth and parentage are clearly established by documentary evidence. The responsibility for supplying such proof rests with the person applying for the registration. Acceptable Evidence

At least one item of the following is acceptable if the record was made within four years of date of birth: 1 Baptismal Certificate or other Church Records such as a Cradle Roll, 2 Hospital Record of Birth, 3 Doctor's Office Record of Birth, 4 Newspaper Notice, 5 Insurance Policy 6 Certified Copy of any record of a Child Welfare Organization. If not available at least two items of the following: 1 A School Record, 2 Bible Record, 3 Census Records, 4 Marriage Record, 5 Baby or Birthday Books, 6 Letters or Telegrams re. birth, 7 A record of Baptism after the age of four years, 8 Other documentary evidence made at the time of birth. Note: The original copy of documentary evidence, not a photocopy, is to be submitted. Personal documents will be returned after they have served their purpose.

The difference between the pre-1945 and post-1945 requirements prompts two important observations about delayed registrations for the different cohorts. First, the insistence upon documentary evidence in the second period increased the possibility (noted above for 1918) that legitimate applicants were refused delayed registration because they could not provide enough proof. Second, it decreased the risk that delayed registration might be granted to persons actually born outside the province. Although the writer found no statistics

84

Facts of Life

Table 4.5 Proportions by Cohort of Delayed Registrations Granted by 1986 That Were Issued before 1945 Cohort

1900 1905 1910 1915 1920 1925 1930 1935 1940

1986 N = % Pre-1945 % Post-1945

3886 55 45

4136 52 48

5498 47 53

5730 46 54

3918 45 55

1951 48 52

1515 32 68

1086 27 73

823 26 74

Note: Calculated from Table 4.3 data.

for refused applications, Table 4.5 elaborates cohort exposures to the differences in risk between the two periods. For example, 68 per cent of the delayed registrations for the 1930 cohort were issued under the more stringent post-1945 requirements, as compared to only 45 per cent for the 1900 cohort. ESTIMATING ACTUAL NUMBERS OF ONTARIO LIVE BIRTHS

In a four-step procedure, this section uses information in published statistics and delayed registrations of births to estimate the numbers of live births for selected cohorts. The procedure entails three assumptions (which are relaxed later for discussion): i) that all delayed registrations issued were for bona fide applicants, namely persons whose births were not registered but had occurred in Ontario in the year for which delayed registration was granted; ii) that all bona fide applications for delayed registration were successful; and iii) that the dead, had they not died, would have applied for delayed registrations in the same proportions as the living. Step i. As shown above, Table 4.3 reports the distribution of the delayed registrations for each cohort by five-year periods of issuance. Table 4.4 then reports the subtotals for cohort age categories that correspond with the periods of issuance.28 Step 2. As data in Table 4.4 above show, the 1900 birth group accumulated delayed registrations throughout its life-course, from age 15 through age 86. As the group aged, however, a growing number of its members died and therefore did not experience historical and lifecourse pressures that prodded the survivors to apply for delayed registration. To allow for the influence of mortality on the number of delayed registrations issued to a cohort, step 2 of the procedure

85 The Completeness of Birth Registrations assumes that the dead would have received delayed registrations in the same proportions as the living had nobody in the cohort died. Only 71 per cent of the 1900 birth group, for example, were alive as the group entered the age category 40-44, during which time 293 delayed registrations were granted to group members. Fewer than 71 per cent obtained the 293 delayed registrations, however, since the cohort experienced some deaths while passing through the age category. To allow for this added group attrition, the estimate is based upon the proportion alive at the mid-point of the category.29 Thus 69.8 per cent of the cohort obtained the 293 delayed registrations. Had 100 per cent of its members been alive at age 44, the group would have received 415 delayed registrations [(293/69.8)*ioo]. Generation life tables are used to calculate the proportions alive at different ages. As such tables have not been published for Ontario, tables for Canada are used for the 1900, 1910, 1920, and 1930 cohorts.30 For each of the 1905, 1915, and 1925 cohorts, calculations are based upon the means of the values in tables for the cohorts immediately preceding and following it. Although the accuracy of the Canada generation life tables for Ontario survival rates is unknown, abridged life tables published for the period 1921-81 show little difference between Canada and Ontario in expectation of life at birth for both sexes.31 In any event, the estimates are not very sensitive to variations in overall survival. As shown in Table 4.6, the estimated incompleteness of birth registrations for the 1900 cohort changes by less than i per cent if either the 1891 (higher mortality) or 1911 (lower mortality) tables are substituted for the 1901 table. Step 3. By 1986, the last year for which the writer examined delayed registrations, the younger cohorts had yet to pass through some of the age categories that the 1900 cohort had completed. For example, the 1900 cohort was 86, an age the 1930 cohort will not reach until the year 2015. Thus to apply step 2 of the procedure to all age categories for each cohort, one must estimate the numbers of delayed registrations the younger cohorts will receive during age categories they have yet to pass through. Table 4.7 shows where such estimates are needed for the 1905 and 1930 birth groups. To predict how many delayed registrations a cohort will receive during a future age category, step 3 uses the preceding cohort's experience for the category while allowing for historical differences between the cohorts. In Table 4.7, for example, the number of delayed registrations the 1905 cohort receives for the age category 80-84 hwas

86

Facts of Life

Table 4.6 Estimates of Incompleteness in Birth Registrations for the 1900 Cohort, Assuming Different Survival Rates Life Table Selected

S(x)

D(0)

E(0)

Est. % Births Missed

1891 1901 1911

100000 100000 100000

15433 13381 11259

45.20 48.67 52.55

12.45 11.54 10.72

Difference

+ 0.91

-0.82

S(x) = number of live births for the cohort. D(0) = number dying before age 1. E(0) = expectation of life at birth.

Table 4.7 Estimates of Unregistered Births (Delayed Registrations + Births Never Registered) for the 1900, 1905, and 1930 Cohorts as of 1986 Cohort Age

Grant Period 1900 C.

Unreg. Births 1900 C.

Grant Period 1905 C.

Unreg. Births 1905 C.

Grant Period 1930 C.

Unreg. Births 1930 C.

55-59 60-64 65-69 70-74 75-79 80-84 85-86

1955-59 1960-64 1965-69 1970-74 1975-79 1980-84 1985-86

527 659 789 119 83 24 11

1960-64 1965-69 1970-74 1975-79 1980-84 1985-89 1990-91

530 786 496 158 38 16* ?

1985-89 1990-94 1995-99 2000-04 2005-09 2010-14 2015-16

60* 7 7 7 7 7 7

* Two years (1985-86) only.

not known until 1990, four years after the last year for which the writer has data. The first part of the step 3 prediction assumes (for want of contrary evidence) that the 1900 and 1905 cohorts experience the same life-course influences for the age category (e.g., relatively few survivors whose critical life-course events are largely over). For prediction one calculates the 1900 cohort's step 2 number for the age category (23.85) as a proportion of its birth registrations (45,549), which yields the statistic 0.000524. Assuming the same proportion for the 1905 cohort, it will receive .000524 of its 50,808 birth registrations, or 26.6 delayed registrations (the life-course statistic) as it passes through the age category. The second part of step 3 is to adjust the life-course statistic to allow for historical differences between the cohorts in their respective accumulations of delayed registrations. For example, if birth

87 The Completeness of Birth Registrations registration for one cohort is more complete than for the cohort preceding, then proportionately fewer of its members will require delayed registration as they encounter life-course events. Alternatively, if historical events cause proportionately more members in the preceding cohort to obtain delayed registration in the younger age categories, when relatively few cohort members have died, then proportionately more members in the second cohort will seek delayed registration as they encounter critical life-course events in later years. To allow for historical differences between two cohorts, the step 3 method keys on the net difference at entry into the age category for which prediction is required. For each cohort one calculates the delayed registrations for all previous age categories as a proportion of its birth registrations. The proportion for the second cohort, which requires the prediction, then is divided by the proportion for the cohort preceding to obtain the history statistic. The history statistic multiplied by the life-course statistic yields the complete step 3 prediction for the age category. To return to the example above, the life-course part of the prediction was that the 1905 cohort would receive 26.6 delayed registrations as it passed through the age category 80-84. As the 1905 cohort reached age 80, however, its delayed registrations as a proportion of its birth registrations were equal to only 0.91 of the corresponding figure for the 1900 cohort. Thus the completed step 3 prediction is that the 1905 cohort will receive 0.91 of 26.6, or 24.12 delayed registrations of birth as it passes through the age category 80-84. The accuracy of the predictions for future years (1987 + ) is problematic. As shown in Table 4.8, the step 3 method predicts poorly the known numbers of delayed registrations for pre-1986 age categories of the 1905, 1910, and 1915 cohorts. As shown in Table 4.9, however, the step 3 predictions involve small numbers and cause little change in the cohort birth estimates. Thus the writer accepts the predicted values for the numbers of delayed registrations the cohorts receive after 1986. These values in turn serve to estimate the numbers that the cohorts would have received from 100 per cent of their members (see step 2 above). Step 4. The percentage incompleteness in cohort birth registrations is obtained by dividing the estimate (obtained through steps 1-3) by the number of unregistered births (the estimate minus the number of registrations). In Figure 4.4, the trend line shows continuous increase in the completeness of birth registrations between 1910 and 1930. Contrary to expectation, however, it shows that registration was more complete in 1900 and 1905 than in 1910.

88 Facts of Life Table 4.8 Accuracy of Step 3 Estimates for Predicting Known Values for Previous Age Categories Age Category

2905 Delayed Regs.

% Prediction Error

1910 Delayed Regs.

% Prediction Error

1915 Delayed Regs.

% Prediction Error

40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79

191 263 292 358 502 290 81 16

+ 79 - 20 + 14 + 5 - 12 + 68 - 24 + 120

289 319 356 335 891 412 38

- 12 + 12 + 10 + 42 - 28 - 5 + 189

310 290 308 587 832 209

+ 3 + 22 + 26 - 39 + 20 + 115

Table 4.9 Cases Added in Step 3 as Percentages of Birth Registrations and Estimated Birth Totals

Stage 3 Estimates, N: % Birth Registrations % Estimate

1905

2920

2925

2920

2925

2930

45 0.1 0.1

101 0.2 0.2

162 0.2 0.2

309 0.4 0.4

427 0.6 0.6

332 0.5 0.5

The statistics for 1900 and 1905 are not plausible and require adjustment. First, they go against Kuczynski's impression of "slow progress" for the 1896-1920 period. Second, on the basis of circumstantial evidence - the provincial inspector's vigorous effort to enforce the Vital Statistics Act, the tightening of the requirements in the act in 1908 and 1919, and the small increase in the proportion of births occurring in hospitals - one might expect a slight increase in the completeness of birth registration between 1900 and 1910. Third, an implicit assumption in steps 2 and 3 - that all persons whose births were missed by civil registration eventually would have applied for delayed registration, had nobody in the cohort died - is problematic for the older birth groups. Notwithstanding the effects of the Military Service Act on the 1900 cohort, age-related government regulation possibly had less influence on life courses for the older cohorts than for those of the younger. The step 4 adjustment to the figures for 1900 and 1905 makes two assumptions. One is that the cohort with the highest estimated incompleteness (1910) is the first for which the step 3 estimates are

89 The Completeness of Birth Registrations

Figure 4.4 Estimated Incompleteness in Live Birth Registrations

reliable. The other is that the completeness of the returns rose more slowly before 1910 than after, in keeping with Kuczynski's description of the data ("slow progress" for the 1896-1920 period, with an accelerated improvement in the returns beginning about 1913). Thus the writer arbitrarily assumes a i per cent increase for each of the intervals 1900-05 and 1905-10, which is about half the 2.1 per cent increase estimated for the period 1910-15. Figure 4.4 above shows the effect of the step 4 revisions, and Table 4.10 shows the final estimates of cohort birth numbers. The step 4 estimates are higher than Kuczynski's posited 85 per cent maximum completeness for any year before 1920. Here two assumptions of the estimation procedure may inflate the estimates. The first - that all persons whose births were missed by civil registration would eventually have applied for delayed registration, had nobody died - may be tenuous for more than the 1900 and 1905 cohorts. The second -

90 Facts of Life Table 4.10 Ontario Live Births: Published Totals (Registrations) and Step 4 Estimates

Registrations Delayed Regs. Never Reg. (Est.) Estimated Births % Unregistered % Completeness

2900

2905

2910

2925

2920

2925

2930

45,549 3,886 2,600 52,035 14.2 85.8

50,808 4,136 2,591 57,535 13.2 86.8

54,755 5,518 1,184 61,457 12.2 87.8

67,032 5,730 1,067 73,829 10.1 89.9

72,511 3,918

70,122 1,951

71,263 1,515

715

594

438

77,144

72,667

73,216

6.4

3.6

2.7

93.6

96.4

97.3

that all persons missed by civil registration could have obtained delayed registration of their births - also is problematic. As noted above for 1918, many applications were refused for want of acceptable evidence, and the standards of evidence became more rigorous in 1945. A possible offsetting influence was the issuance of delayed registration to persons actually born outside the jurisdiction. Even so, the actual completeness of the birth returns probably is less than the step 4 estimates indicate. For the 19205 the step 4 estimates support Kuczynski's judgment that registrations for the period were "at least 90 per cent" complete, but are lower than the Macphail estimate for 1921 and the Tracey maximum estimate for the 1927-31 period, each of which places the completeness at 100 per cent. Each of these estimates is calculated from published (aggregate) data and entails comparison between a census population for infants and an expected census population computed from statistics for births and deaths. As the Table 4.10 data for delayed registrations effectively demonstrate, Tracey's comparison-method estimates are high and therefore of little help in evaluating the step 4 estimate. However, the step 4 estimate for 1930 (97.3 per cent) is similar to Tracey's more reliable minimum estimate for the 1931 census enumeration year (96 per cent completeness). For this estimate Tracey drew a systematic sample of 5,763 infants from the 1931 manuscript census for Ontario. He then matched 89 per cent of his sample population to birth registrations for months corresponding to the enumeration year, in each case limiting his search to birth registrations for the infant's (census) county of residence. After experimentation he adjusted the total for matched cases to 92 per cent to allow for births that were registered outside the census county of residence. Lastly, he considered other influences that deflated the total for matched cases: the frequent occurrence of American immigrants' reporting a Canadian birthplace for children "whose age indicates that they were born previous to the date of migration"; the difficulty of tracing

91 The Completeness of Birth Registrations illegitimate children and children adopted "subsequent to registration and before the census"; the misspelling of names by census enumerators; and "incomplete searches by clerks seeking to match the transcripts."32 To allow for these influences, he "put the deficiency of birth registrations at not over half the percentage unmatched"; for Ontario, one-half of the adjusted total for unmatched cases (8 per cent) yields the estimate of 4 per cent incompleteness in birth registrations for the enumeration year. BIASED INCOMPLETENESS IN LIVE-BIRTH REGISTRATIONS

Did chance alone determine which births Ontario's civil registration missed, or were systematic influences at work? To elucidate bias in Ontario birth registration, the following section compares information in delayed registrations for the 1920 cohort with corresponding information about the cohort's registered live births. The information about the 3,918 delayed registrations comes from a random sample; information about the 72,511 live births comes from published statistics or, where these are lacking, from a random sample of birth registrations. Each random sample holds 384 cases; assuming maximum variability in the distribution for any attribute, the sample size allows a confidence interval of o.io and a confidence level of 0.05 (i.e., for any attribute, one can be 95 per cent confident that the sample distribution is within 5 per cent of the population distribution).33 Geographical Bias. If no geographical bias existed in birth registrations, then published birth statistics and delayed registrations would have similar geographical distributions. To test for this possibility, the writer compared actual distributions in the delayed-registrations sample with the distributions predicted on the assumption that the distributions for registered births and delayed registrations were identical. As shown in Table 4.11, no statistically significant difference obtained between the predicted and actual distributions for the geographical aggregates of cities, towns, and counties (exclusive of towns and cities). Apparently the completeness of birth registrations did not differ between rural and urban populations. However, the northern Ontario region accounted for a disproportionately high number of delayed registrations, a sign that its birth registrations were exceptionally incomplete. Unlike cities in the aggregate, Toronto had a disproportionately high number of delayed registrations.34 This did not necessarily indicate an exceptional incompleteness in city birth registrations, how-

92

Facts of Life

Table 4.11 Predicted1 and Actual Numbers of Delayed Registrations of Births in the Sample Rural-Urban

Predicted Number

Cities 184 Towns 30 Counties (Rural) 170 Chi square = 0.314663 DF = 3 Not statistically significant at the 0.5 level Region

Predicted Number

94 Eastern Ontario Southwest Ontario 153 Northern Ontario 49 Toronto- York County 88 Chi square = 16.72463 DF = T Statistically significant at the 0. 1 level

Actual Number

Percentage Difference

185 26 173

0 -13

Actual Number

Percentage Difference

77 122 88 97

-16 -31

Selected Municipalities

Predicted Number

Actual Number

Toronto (City) Sudbury (County) Fort William (City) Windsor (City) Temiskaming (County) Renfrew (County) Bruce (County) Ottawa (City) York (County) Hamilton (City) Carleton (County) St. Catherines (City) Muskoka (County)

72 5 7 6 6 5 4 18 16 17 5 4 3

92 24 11 12 12 11 9 4 5 12 1 1 0

+ 2

+ 39 + 9

1 From the distributions shown in published statistics for live births.

ever. As discussed in chapter 7, Toronto registrations were unusually prone to "boundary confusion" errors on the part of persons reporting the births. Especially before 1932, when local officials began to use detailed street maps to make corrections, some delayed registrations issued for Toronto probably had occurred in urban portions of York County townships, just outside the city limits.35 This might explain why 30 per cent of the delayed registrations for Toronto were issued before 1932, compared to 21 per cent for the entire provincial sample (see Figure 4.5 for the contrasting Toronto and

93 The Completeness of Birth Registrations

Figure 4.5 1920 Cohort: Proportions of Delayed Registrations by Age Categories

Ontario trends for period of issuance). It might also explain why York County received fewer than its expected number of delayed registrations (see Table 4.11 above). Sex Bias. Females account for 55 per cent of the cases in the delayedregistrations sample but only 49 per cent of the published total for live births. The difference is suggestive but inconclusive about sex bias in birth registration. Sample error may inflate (or deflate) the proportion of females in the delayed registrations. Furthermore, some of the female surplus in the delayed registrations comes from the greater longevity of females: more females than males survive to experience life-course events in older age categories that prompt applications for delayed registration. Yet the female surplus exists despite a male bias in certain of the life-course events (e.g., obtaining an automobile driver's licence, responding to wartime military service regulations) that prompted

94

Facts of Life

cohort members to apply for delayed registration. Moreover, evidence on sex ratios suggests an anti-female bias in Ontario birth registration until the 19305. During the 1880-1940 period the actual ratio of male to female live births probably rises irregularly due to sex-selective reductions in prenatal mortality.36 Figure 4.6 shows this pattern for England and Wales, whose birth registration was quite complete.37 The Ontario ratios calculated from birth registrations for the period 1980-1920, however, show a declining pattern and are notably higher than the English-Welsh ratios for the same years. The likely explanation is that biased incomplete registration masks a rising trend until the 19205, when more complete registration attenuates the bias. For the same reason the Ontario ratios begin to mimic the EnglishWelsh ratios during the 19205 and closely match them during the 19305. Social-Class Bias. Father's occupation is the only available proxy for the social-class status of a newborn's family. Both birth registrations and delayed registrations report this information, although for illegitimate births they usually report only the mother's occupation, if she has one. To test for social-class differences, occupations are categorized in a three-stage process. First they are separated into farm and non-farm categories; non-farm occupations then are placed in the categories of manual and non-manual; lastly, occupations in each of these categories are classified as simple or complex, according to the complexities of the work task involved.38 Table 4.12 summarizes results of the classification. The manual simple category, and to a lesser extent the non-manual simple category, may be taken as crude indicators of low social standing. From this standpoint the Table 4.12 data for delayed registrations show a slight overrepresentation of lower-class parents and therefore a slight bias against them in birth registration. Although suggestive, the evidence of class bias is inconclusive. Occupation is a notoriously ambiguous indicator of social class; the difference between the two distributions is small, and the use of two samples doubles the potential for distortion through sample error. Ethnic Bias. Although birth registrations report "race" and place of birth for both parents, the forms used for delayed registration solicit this information only for the 1945-59 period, which provides just 49 cases in the sample. The number is small, and its ethnic composition may differ from that for cases involving earlier or later years of issuance for delayed registration.

95 The Completeness of Birth Registrations

Figure 4.6 Ratio of Male to Female Registered Births, Ontario and England, 1881^85 to 1956-60

Although limited as evidence, the partial data suggest two important ethnic biases in Ontario birth registrations for 1920. The first is a disproportionate underreporting of births for continental European ethnic groups; these account for just 6 per cent of cases in the birth registration sample but 35 per cent (17 cases) of the 49 delayed registrations. The second is an unusually complete reporting of French Canadian births; franco-Ontarians account for 13 per cent of cases in the birth-registration sample but received only 8 per cent (4 cases) of the 49 delayed registrations. The finding for French Canadians accords with viewpoints expressed in the registrar-general's annual reports for 1898 and 1910. His report for 1910 also gives a partial explanation: In most of the counties settled by the French people of Ontario, a great many of the births are registered by the priests of the parishes. The people being

96 Facts of Life Table 4.12 Proportions of Fathers' Occupations in Different Occupational Categories, for 1920: Birth Registrations and Delayed Registrations %

Nonmanual Complex

Sample Birth Registrations1 Delayed Registrations2 Difference

Nonmanual Simple

Manual Complex

Manual Simple

Farm

11

11

32

19

26

7 -4

15 +4

29 -3

26 +7

21 D

1 N = 376. 2 N = 325.

Roman Catholics generally, are anxious to have their children baptized as soon after birth as possible, and in this manner the priests get their names and register them, although there is no provision in the Vital Statistics Act giving them this power, yet, if these clergymen did not make the registrations, very many, indeed most of them, would go unregistered, as it is impossible for the Division Registrar to know where births take place.

Here the priests were continuing a practice begun in Quebec, where civil registration law required clergy to compile lists of births, marriages, and deaths from information in church parish registers. The parish registers, in turn, provided nearly complete coverage of vital events for French Canadian and other Roman Catholic populations.39 CONCLUSION

In 1905 Ontario's inspector of vital statistics recognized that provincial birth statistics were unreliable. Similarly Kuczynski's 1930 monograph described birth statistics for English-speaking Canada up to 1920 as "utterly inadequate." Although useful, these early views were general, largely intuitive, and gave little historical context. By using evidence from delayed registrations, this chapter strengthens the empirical basis for such judgments and gives more precise information about the incompleteness and biases of birth registration and when it becomes reliable. Further, it uses circumstantial evidence the secular hospitalization of births, proliferation of government age and citizenship requirements, and tightening of the legislation and enforcement for civil registration - to elaborate the historical context. The principal finding is that Ontario birth registration moved from 86 to 97 per cent completeness between 1900 and 1930. This accords

97 The Completeness of Birth Registrations

with Kuczynski's estimate of "at least 90 per cent" completeness for the 19205 and also with Tracey's minimum estimate for 1931. At the same time it describes higher levels of completeness than Kuczynski's 85 per cent maximum for years before 1920. Here the study estimates may be high due to assumptions made in the estimation technique. A second finding is that Ontario birth registration was biased against certain population groups as long as it remained incomplete. As shown by evidence for the 1920 cohort, birth registration was unusually complete for French Canadians but disproportionately incomplete for northern Ontario, females, and continental European ethnic groups.

5 Getting Born: How Definition Influenced Statistics for Infant Deaths and Stillbirths, 1926-51

During the late 19203 Ontario reported a lower infant death rate than Quebec and a higher rate for stillbirths. In part, however, the provincial differences in the reported rates reflected physician, legal, and civil registration definitions for live birth and stillbirth. Some births that Ontario physicians would have considered stillborn, for example, were registered as live in Quebec so that the province's Roman Catholic physicians could baptize them; in the process the Quebec physicians added to their province's reported totals for infant deaths in the first day of life. More generally, the international statistics lacked standard definitions and were not comparable. In 1932 Canada adopted the 1925 League of Nations definition of stillbirth as its standard definition for statistical purposes. The League definition clashed with Ontario's legal definition, however, and for some time the Ontario registrations did not report the information needed to apply it. After the League definition failed to catch on internationally, Canada modified its standard definition in 1943 and 1950. Thus the definitional influence acted on published statistics in two ways: the national standard definition changed over time; and the application of the 1932 standard definition changed over time and differed by province. To elaborate the definitional influence on infant death and stillbirth statistics for the 1920-50 period, this chapter introduces the pertinent demographic terms and describes the trends for various reported infant death rates. Second, it discusses contemporary interest in statistics for infant and foetal death. Third, it describes early

99 Getting Born

background: the statutes, the content of the registration forms, and the statistics through to 1930. Fourth, it discusses issues of definition that emerged during the 19305 and how Canadian civil registration responded to them. Fifth, it analyses how application of Canada's statistical definition affected the reported infant death and stillbirth rates for Ontario and Quebec. In the process, it sheds light on the slow standardization of Canadian and international statistical practice. DEFINITIONS AND TRENDS IN THE REPORTED DEATH RATES

Live birth means infant life after complete delivery where infant life, unlike foetal life, exists independently of the mother. Infant death is the death of a live-born child under one year of age (an arbitrary end-point that suits time measurement in Western societies). Until 1943 Canada's definition required also that the foetus be viable (sufficiently developed to support infant life). Foetal death includes stillbirths and abortions (miscarriage). Stillbirth refers to the absence of life after complete birth of a viable foetus. Where the delivered foetus is non-viable, the birth event is classed as abortion; abortion occurs spontaneously (naturally) or is induced through human intervention. Neo-natal infant deaths occurred within twenty-eight days ("one month") of birth. Whereas post-neo-natal deaths arose primarily from environmental influences (manifested in acute infectious and gastrointestinal diseases), neo-natal deaths and stillbirths tended to have congenital and maternal causes.1 During the study period publichealth programs had greater success against environmental causes than against congenital/maternal causes. Thus, with a 50 per cent fall in Ontario's reported infant death rate between 1930 and 1949 (from 74 to 37), the neo-natal proportion of the deaths rose from 52 to 62 per cent. Similarly, the trend for the neo-natal death rate sometimes differed from the trend for all infant deaths. In 1937, for example, Quebec reported a sharp rise in its infant death rate (from 83 to 100) but a drop in its neo-natal rate (from 40 to 39). The conventional infant mortality rate reports the number of infant deaths occurring in a given year per 1,000 live births in the same year. Similarly, the conventional neo-natal death rate and stillbirth rate state the number of these events occurring in a given year per 1,000 live births occurring in the same year. Some infant deaths in a given year are for infants born during the previous year. Thus one must calculate an adjusted rate to match the deaths to the live births for

ioo Facts of Life

their respective birth years.2 This study reports the conventional rates only. First, its concern is not the rates but the numerator data used to calculate them. Second, the rates used as evidence are for neo-natal deaths and deaths of infants in the first day of life, which do not require adjustment.3 As shown in Table 5.1, Ontario reported declining trends in the rates for stillbirths, neo-natal deaths in the first day of life, all neonatal deaths, and infant deaths. Relative to Quebec, Ontario reported lower rates for neo-natal and infant deaths. Second, its reported stillbirth rate was higher until 1940 but lower thereafter.4 As discussed below, the trend lines for the reported neo-natal and early neo-natal death rates shed light on how definition influenced the numerator data (the chapter subject). One must interpret the trend lines cautiously, however, because they also reflect change in i) the mortality risks for different categories of risk, such as the mother's age, parity (number of delivered pregnancies), and marital status; ii) the distributions of mothers among the risk-factor categories; and iii) the registration coverage for live births, the denominator data.5 CONTEMPORARY INTEREST IN STATISTICS FOR INFANT DEATH AND STILLBIRTH

During the early twentieth century, government officials attached considerable weight to statistics for births, infant deaths, and stillbirths. On the one hand they regarded Canada as an underpopulated "young" country. On the other they were nativists, preferring that population grow primarily through the natural increase of Canada's "British" stock, not the immigration of "foreigners." From this perspective the apparent declining trend for the province's birth rate was lamentable. "The French and German settlers," Ontario's deputy registrar-general noted in 1909, "contribute more largely to the birthrate than any other class of our people ... In this respect they set a good example to our more purely Canadian people. It has of late become the fashion to have small families, or none at all."6 Second, the provincial rates for infant and foetal deaths represented an appalling haemorrhage of population.7 The haemorrhage appalled because it seemed largely unnecessary. Encouraged by the declining trend in Ontario's reported infant death rate after 1900, officials believed that most infant deaths were preventable through some combination of better pre-natal care for mothers; better feeding and care of infants; the extension of hospital

ioi Getting Born Table 5.1 Ontario and Quebec Reported Rates for Stillbirths, Deaths in the First Day of Life, Neo-natal Deaths, and All Infant Deaths, 1926-50 All Infant Death

All Neo-natal

First Day Neo-natal

Stillbirth Year

Ont.

P.Q.

Ont.

P.Q.

Ont.

P.Q.

Ont.

P.Q.

1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960

42 41 41 40 38 38 36 34 34 34 33 32 31 31 30 29 27 25 24 23 21 20 19 19 18 17 17 15 15 15 14 13 13 13 12

24 25 28 28 29 29 29 31 29 31 31 31 30 30 30 30 31 27 28 28 26 26 24 25 24 23 22 20 21 19 19 18 17 16 17

17 16 16 16 15 15 14 13 12 14 13 13 13 12 11 12 10 13 13 13 12 11 11 10 10 9 9 9 9 9 9 10 9 9 9

24 20 21 20 19 17 13 14 13 11 12 12 11 12 11 12 12 12 11 10 10 11 10 10 13 11 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

43 41 39 42 39 37 36 34 31 34 31 31 29 28 26 27 25 27 27 27 26 24 24 23 23 21 21 19 18 18 17 18 18 17 17

59 54 53 52 51 49 43 43 42 39 40 39 38 36 35 36 35 34 34 32 30 31 29 27 30 26 28 26 23 23 25 25 23 21 19

78 71 71 76 74 70 62 60 57 56 55 55 49 46 43 46 40 42 43 41 37 36 35 37 35 31 31 28 26 26 25 25 25 24 24

142 129 124 121 120 113 94 95 97 92 83 100 83 78 70 76 70 67 68 62 55 57 54 52 51 48 50 45 40 38 41 38 36 33 30

and clinical services; better and less crowded housing; social assistance to mothers to offset "the expense attending parturition"; and immunization programs, the isolation of infected persons, and sanitary measures to reduce infant morbidity from communicable and gastro-intestinal disease. Officials appreciated that neo-natal deaths

io2 Facts of Life

presented special difficulty but believed that even these were open to reduction through careful pre-natal and obstetrical care. Thus the "infant soldier," a McGill University professor insisted, "meets his death at the opening of the battle of life because of ante-natal or natal conditions, not because of his inherent unfitness but because of our unfitness to take care of his mother and him. Take care of the mother and she will take care of the baby."8 By the 19305 officials regarded the infant mortality rate as "the most sensitive index we possess to the general health conditions of any community or nation" and a "measuring rod" for estimating the "effectiveness of our efforts in the field of child health."9 They were also broadening their interest in infant mortality to include foetal mortality (stillbirths and abortions). Foetal deaths, they noted, were intimately related to neo-natal deaths in terms of the "causative factors involved."10 "Indeed," a 1946 study concluded, the dividing line between the infant and stillbirth death rates was "more a matter of definition, and the judgment by the physician in each case, than of a fundamental difference in their nature."11 Given that stillbirths alone constituted "about 30 per cent of the total loss of life under one year of age," foetal deaths were part of the "annual loss to our child population." At the same time the reported stillbirth rate showed a downward trend, and a 1936 study guessed that half of all foetal deaths were preventable through "proper antenatal care and improved obstetrical practice in treating the complications of childbirth." Thus, like "the expectation of life at birth and the rate of infant mortality," the 1946 study opined, the stillbirth rate was "a statistical index of a people's health." Even so, infant and foetal death statistics were not the benign, selfinterpreting "hard facts" of professional lore. To begin with, physicians probably reported infant deaths more reliably over time, prodded by the rising salience of infant mortality as a public-health issue. Second, officials had a vested interest in the statistics. By documenting high infant mortality while claiming that it was preventable, they made a case for programs to deal with it; similarly, by documenting a declining trend for infant deaths, they established that their interventions worked. Third, the statistics touched on the relative efficacy and politics of particular programs. Statistical arguments for enhanced medical services, for example, served the vested interests of provincial physicians, whereas statistical arguments for social assistance to mothers did not. Finally, by serving as evidence for any government interventions, infant death statistics meant coercion for population. Not all Ontarians regarded infant mortality as a preventable problem that justified infringements on civil liberties.

103 Getting Born Table 5.2 Ontario Stillbirths Registered as Births and as Deaths Year

Births

Deaths

Year

Births

Deaths

1914 1915 1916 1917 1918 1919 1920

1,469 2,246 2,055 1,425 2,198 2,091 2,495

2,745 2,807 2,518 2,486 1,339 2,463 2,868

1921 1922 1923 1924 1925 1926 1927

3,234 3,115 3,028 2,594 2,780 2,812 2,758

3,046 2,946 2,960 2,648 2,774 2,802 2,754

Indeed, supporters of the eugenics movement regarded traditional high rates for infant and foetal death as nature's culling of substandard genetic stock.12 THE ONTARIO

S T A T I S T I C S B E F O R E 1930

Ontario's 1869 Registration Act required the registration of an infant death, and its 1896 act prohibited the burial of a dead infant without a death certificate and burial permit. Although neither statute referred to stillbirths, the 1909 version of the birth registration form directed local officials to register a stillbirth twice, once as a birth and once as a death. Through to 1912, Ontario's published statistics for births and deaths included stillbirths. Thereafter, to conform to international practice, the registrar-general's annual reports reported stillbirths separately, while excluding them from other tables.13 The 1919 Vital Statistics Act made the double registration of a stillbirth mandatory and directed that the birth registration should state "stillborn" rather than a child's first name. In the town of Ingersoll, at least, the registrar ignored the latter requirement. For compilation, the physician's entry of "stillbirth" for the cause of death identified a stillbirth on a death registration. In birth registrations through to 1911, officials relied on entries of "stillborn" in the column for "Remarks." In 1912 a revised version of the form required the informant to "state if Twin, Triplet, Illegitimate or Stillbirth." In 1920 another revised version of the form introduced the separate question: "Was the child born live?" Before 1921, as shown in Table 5.2, Ontario registered more stillbirths as deaths than as births. Thereafter the two counts were similar, which the registrar-general interpreted as a sign of more complete coverage in the registrations. On average the birth registration count was slightly higher after 1920. Thus Dominion officials

104 Facts of Life

chose it for published tables, and the registrar-general did not even report the death count after 1927. By 1930 the province's counts of infant deaths and stillbirths appeared to be complete. Adequate statutory requirements were in place, and public-health interest in infant and foetal deaths gave physicians and the public a strong incentive to register these events. When officials confronted issues of definition, however, they discovered that the apparent completeness was an illusion.

CANADA'S RESPONSE TO ISSUES OF D E F I N I T I O N , 1930-50 International statistics for stillbirth were not comparable during the 19205. The Dutch, French, and Belgian totals included live-born children who died "before the day of the declaration which must be made within three days following birth." By contrast, England and the United States Census Bureau restricted the term to children who, after complete birth, showed "no sign of life," such as "breathing, action of the heart, pulsations in the cord, or movement of the voluntary muscles." The definitions in thirty-one of the American states also required that the foetus be "viable," as defined by a specified minimum utero-gestational age; four states treated breathing as the only acceptable evidence of life; and one state, Ohio, left the stillbirth classification entirely "to the appreciation of the attending physician."14 Before 1932 Canada's statistical definition required a minimum gestational age of six months (twenty-six weeks). The provinces, not the Dominion Bureau of Statistics, administered civil registration, however, and each province had its own requirements for gestational age and evidence of life (see Table 5-3).15 The extent of disagreement among the provincial definitions is unclear. To Dr Eugene Gagnon, the Quebec Health Department statistician for the city of Montreal, for example, the British Columbia requirement for gestational age, "twenty-eight weeks," meant "6.5 months"; by contrast, the requirement in Ontario, Prince Edward Island, and New Brunswick - "seven months" - was "a full seven months," or thirty to thirty-one weeks. As in the 1925 League of Nations definition, however, "seven months" could have meant "seven lunar months," or twenty-eight weeks, the same as the British Columbia requirement. In 1925 the Health Committee of the League of Nations proposed the following international standard definition for stillborn: "a birth of a foetus after twenty-eight weeks' pregnancy in which pulmonary respiration does not occur; such a foetus may die: (a) before,

105 Getting Born Table 5.3 Canada, 1930: Provincial Registrars' Answers to Questions Concerning Gestational Age and Criteria for Life Province

Gestational Age (min.)

Criterion for Life

PEI Nova Scotia New Brunswick Quebec Ontario Manitoba

7 months No requirement 7 months 6 months (26 weeks) 7 months 5 months

Saskatchewan Alberta British Columbia

6 months 24 weeks 28 weeks

No rule Heart action, breathing Respiration Respiration No rule, doctor's discretion Heart action or breathing, movement of voluntary muscles Respiration Action of heart or lungs Respiration

Note: The questions read: (a) At what age is a foetus considered as a life or a stillbirth? (b) What is considered as evidence of life in a newborn foetus or child?

(b) during, or (c) after birth, but before it has breathed." The definition entailed two distinct propositions: i) a minimum period of uterogestation, below which a delivered foetus was pre-viable and aborted, not stillborn; and ii) breathing as the only acceptable evidence of infant life. Both propositions met resistance on scientific, legal, and religious grounds. Scientific Considerations. The Vital Statistics Section of the American Public Health Association argued against the twenty-eight-week utero-gestation period as the criterion for foetal viability. Some live births occurred before twenty-eight weeks' gestation, and in "one case of record ... a child born in the i8th week was alive at the age of 9 years." In the circumstances, registration should cover all foetuses that had advanced to eighteen weeks (or 120 days or the "fifth month"), the proximate point of "quickening," or foetal movement within the womb. The American committee also preferred "any sign of life" to the breathing requirement in the League definition. In some cases the physician's evidence did not show clearly whether the child had breathed or only attempted to breathe. In other cases respiration was artificially induced up to one-half hour after birth in a child whose heart was beating. Was such a child born dead, the committee asked, and then "raised from the dead?" By contrast, Dr Gagnon of Montreal endorsed the League definition.16 He conceded that the "borderline between viability and non-

106 Facts of Life

viability is not easy to draw, as some foetuses of six months of uterine gestation have acquired a greater development than others at full term pregnancy." Nevertheless, he continued, "it is generally agreed that before the end of six months [twenty-six weeks] a foetus cannot be born viable [and] there is now a tendency among physiologists and obstetricians to extend that limit to the end of twenty-eight weeks or 196 days." Similarly, breathing (spontaneous respiration) was "the first vital act made by the child" and the appropriate criterion for distinguishing infant life from foetal life. Legal Considerations. Because civil registration served administrative and legal as well as statistical purposes, legal definitions of live birth were an issue. In France and Quebec, "where the Napoleonic code [was] the foundation of civil law, ... a child [could not] inherit or have benefit of a will or a donation unless it [was] born live and viable." Based on case law, the Quebec legal criteria for life included respiration after complete birth and a minimum gestational age of "six months" (twenty-six weeks), as opposed to the twenty-eightweek requirement in the League definition. Other Canadian provinces, England, and the United States used the English common-law criterion for live birth - "any sign of life," regardless of gestational age. Thus the gestational-age requirement in the League definition clashed with the legal definition in all provinces, and the breathing requirement clashed with the legal definition in all provinces except Quebec. Religious/Cultural Considerations. Ontario reported a lower infant death rate than-Quebec, yet a higher rate for stillbirths. In part, Dr Gagnon argued, the differences reflected the greater inclination of Quebec's Roman Catholic physicians to declare a foetus live-born, thereby adding to the statistics for live-born infants who died during their first day of life. As he explained: It is a Roman Catholic doctrine that baptism is a necessary condition of the admittance of a human being into heaven. In consequence of that doctrine the right and duty to minister baptism are not reserved exclusively to ministers of the church, but are also granted to any member of the church. As a matter of fact, the attending physician of Roman Catholic faith is often called to minister baptism to a new-born child, whenever there is any possibility that the child be still alive. In case of doubt as to the child being alive or dead, he must give it the benefit of the doubt, and pronounce it alive ... [and] will declare it alive on the death certificate, while his protestant

107 Getting Born confreres will pronounce it a stillbirth ... Thence the inaccuracy of stillbirth statistics based only on the declaration of the attending physician, and the necessity of a statistical definition of the word stillbirth applicable to all births, irrespective of religious feelings or creeds.

Effectively, the English common-law "any sign of life" understanding of live birth suited Quebec Roman Catholic religious needs better than the breathing requirement in Quebec's legal definition. The Statistical Implications of the League Definition. Based on calculations from Boston hospital data, the 1927 American Public Health Association committee estimated that the breathing requirement would make a reported infant death total 2.5 per cent lower than if the "any sign of life" criterion were used. When the breathing requirement was applied to Quebec's statistics in 1932, however, its reported infant death total dropped by about 4.3 per cent. Perhaps the American estimate was low. Alternatively, Quebec's Roman Catholic doctrinal bias in favour of live-birth declaration was more extreme than the American committee's "any sign of life" understanding of live birth. According to Canadian estimate, the 28-week utero-gestation requirement removed about 5 per cent of the infant deaths from its published total. Some foetuses under the minimum age were born live, and a high proportion of these died around the time of birth. The League definition, however, classed these cases as abortions, thereby removing them from the statistics for infant death. Canada's Statistical Definition. In 1932, following discussion with the provinces, the Dominion bureau of Statistics adopted the League definition for Canadian statistics.17 By 1938, however, Canadian officials appreciated that no other nation had followed Canada's lead. Thus in 1943 the bureau dropped the gestational-age requirement from its definition, and in 1950 it replaced the "breathing" requirement with "any sign of life." With these changes Canada anticipated the standard definition for live birth that the Expert Committee on Health Statistics of the World Health Organization proposed in 1950.18 The WHO committee also recommended that all foetal deaths be registered and that the terms "stillbirth" and "abortion" be discontinued unless "essential for internal use within a nation." Canada retained the terms, however, and continued to register stillbirths (defined by a minimum of twenty-eight weeks, utero-gestation and the absence of "any sign of life") but not abortions.19

io8 Facts of Life THE A P P L I C A T I O N OF C A N A D A ' S STATISTICAL DEFINITION IN ONTARIO AND QUEBEC, 1932-5!

By agreement, the Dominion Bureau of Statistics supplied the provinces with standard forms for registering infant deaths and births; received copies of each province's completed registrations; and compiled each province's published statistics from the copies. After adopting the League definition for stillbirth in 1932, however, the bureau and the provinces did not revise the standard forms to include questions about gestational age and signs of life.20 Thus the sole information about Ingersoll's eleven reported stillbirths in 1932, for example, was that each foetus was not "born live" (the birth registration information); and "stillbirth" was the cause of death (the death registration information). However, Ontario death registrations sometimes reported the gestational age for infant deaths from prematurity (e.g., an entry such as "premature 8 mos."). During the period 1930-33, for example, Ingersoll physicians registered fifteen infant deaths from prematurity and stated the gestational age for ten of them. Prematurity in turn was Ontario's leading reported cause of infant death. In 1932 it accounted for 30 per cent of the infant deaths, 49 per cent of the neo-natal deaths, and 67 per cent of the infant deaths in the first day of life. Second, regulations issued under Ontario's 1931 Public Hospitals Act provided the provincial health minister with "a special return ... of all infant [and foetal] deaths occurring in a public hospital, resulting directly or indirectly from pregnancy."21 These returns gave "replies to the questions: 'Was the child full term?' and 'If not, state degree of prematurity.'" The 1933 hospital returns, for example, reported 1,070 foetal deaths, stated the gestational age for 1,045 °f them, and reported an age under the twenty-eight-week limit for 93 of them.22 Judging by the evidence for the years 1934-37, tne hospital returns gave the provincial health minister a comprehensive second coverage of infant deaths and stillbirths in public hospitals.23 Between 1931 and 1951, moreover, the proportion of registered infant deaths occurring in public hospitals rose from 42 to 82 per cent, and the proportion for registered stillbirths rose from (about) 45 to 93 per cent. To summarize, the Ontario registration forms did not ask about signs of life or gestational age, the information that bureau officials required to class by the official definition. The province's death registrations did report the gestational age for some infant deaths from a reported cause of prematurity, and Ontario public hospital returns

109 Getting Born

reported the gestational age for nearly half the province's reported infant deaths and stillbirths. Information from the hospital returns could not influence the statistics, however, unless provincial officials made it available to the bureau clerks who did the counting (for example, by noting it on the registrations before sending copies of them to Ottawa). In the circumstances, Ontario's 1932 statistics for infant deaths and stillbirths primarily reflected physician classifications, which were problematic in terms of the statistical definition. First, physicians could not class by the breathing requirement without violating the Ontario Vital Statistics Act. Section 28 (4) of the 1919 act stated the English common-law requirement: "No child which shows any evidence of life after birth shall be registered as a stillborn."24 A 1943 amendment added the statistical definition ("a child born after twenty-eight weeks of pregnancy and in whom breathing does not occur after complete birth shall be deemed stillborn"), but the act retained the "any evidence of life" provision in Section 28 (4). Second, "the [physician] error in estimating period of gestation [was] a sizeable one."25 Third, as late as 1935 "the word 'stillbirth' even as now defined is confusing to practitioners. For example, the foetus may not be expelled till the seventh month, yet death may have occurred [before the seventh month] in utero and the foetus may be expelled in a macerated state. It is evident that there is wide variation in the minds of practising physicians as to what constitutes a stillbirth."26 Fourth, physician awareness of the 1932 official definition was questionable. The 1932 revised edition of the Physician's Pocket Reference Book," a pamphlet published by the bureau, stated the importance of two facts, gestational age and whether the child breathed after complete birth, for a proper classing of stillbirths. As Dr Gagnon noted, however: if these two questions are not specifically asked on the [standard death certificate] form used, how often will these details be given? Stillbirths are not of very common occurrence in the average physician's practice. The physician who is called to fill the death certificate for a new-born baby may have read very carefully the pocket reference book, or he may not have read it at all. Supposing he understood it a month ago and fully understood the meaning of the paragraph relating to stillbirths, he is very likely, unless ... gifted with an uncommon memory, to have forgotten all about it or, if he still remembers something of it, he will be at a loss to remember where he read it ... [Gagnon noted] with great pleasure this paragraph on stillbirth included in the pocket reference book, but while preparing this report, and desiring to quote the text of the booklet, he had some difficulty in finding

no

Facts of Life

the page on which it appears. This illustrates to some extent what we may expect from the ordinary practitioner who is called on to state that the child was "stillborn" or not, according to our definition.27

Quebec was the only province for which bureau officials could apply Canada's official statistical definition. In 1927 Quebec introduced a special form to register embryos and foetuses, and the form included questions about gestational age and signs of life. Then in 1932, in co-operation with the bureau, Quebec introduced a modified version of its special form to register stillbirths and infant deaths in the first day of life.28 Consequently the Ontario-Quebec differences in reported infant death and stillbirth rates partly reflected different applications of the 1932 standard definition (see Appendix A for discussion of the evidence). First, the breathing requirement was a dead letter for Ontario; for Quebec, by contrast, it subtracted 4.2 from the reported infant death rate and added 4.2 to the reported stillbirth rate. Second, bureau officials ignored the gestational-age requirement when compiling the infant death statistics for both provinces. They lacked comprehensive information about gestational age for the Ontario registrations. They had the information for the Quebec registrations but did not use it.29 Third, the application of the gestational-age requirement removed 4.2 from Quebec's reported stillbirth rate but only 1.1 from the rate reported for Ontario. Here the Quebec special-form registrations reported comprehensive information for the requirement. The questions on Ontario's standard-form registrations did not ask about gestational age; thus provincial officials must have added the gestational age for some registrations, using the information from public hospitals. Although Ontario's lack of comprehensive information for gestational age inflated its reported stillbirth total, the province's statistical tradition possibly deflated the total. Ontario's pre-1932 requirement for a judgment of foetal viability had been twenty-eight weeks (assuming seven lunar months), compared to twenty-six weeks in Quebec's definition and Canada's pre-1932 definition. It is problematic, however, whether physician reporting practices had followed the province's pre-1932 standard definition. The Development of a National Standard Form for Stillbirth Registration. Canadian officials saw several advantages in Quebec's special-form registration for stillbirths and infants in the first day of life. By reporting the gestational age and signs of life, the special form allowed a uniform application of the national statistical definition.

in Getting Born Yet it also reported the information for alternative definitions to suit legal or denominational religious needs. Finally, compared to the standard death-certificate form, the special form could obtain more satisfactory information about causes of stillbirth. The death certificate for each of Ingersoll's registered stillbirths during the 1928-33 period, for example, stated "stillbirth" (the fatal outcome) as the cause of death, whereas the issue was what had caused the foetal death. Discussion of the content of the national standard special form delayed its introduction until the 1939-41 period. Stillbirths and early neo-natal deaths, for example, had maternal as well as foetal causes. Could the special form ask questions about both and yet also show the causal sequence and the "underlying cause," for classification in published statistics? How could officials resolve such questions in the absence of a satisfactory nosology (taxonomy) for stillbirth causes - an international problem? A subcommittee of the Vital Statistics Section of the Canadian Public Health Association undertook the work of drafting a national standard form for stillbirth registration. In 1938 the draft form was agreed upon and given clinical trials, and all provinces introduced it over the next three years.30 The death-certificate portion of the final draft was identical to that of the standard death registration form. An important consideration for officials was the improved reporting of cause for stillbirths on the standard form, following the bureau's revision of its questions relating to cause of death in 1935.31 On i January 1940 Quebec introduced a modified version of the national standard form that continued the province's past practices and reflected its distinctive needs.32 First, the Quebec form retained the question about signs of life. The penultimate draft of the national form had included the question, but it was one of several dropped because they were "not essential to the immediate objective."33 Second, Quebec was "the only province which has incorporated in its new stillbirth certificate the suggestion made by the [subcommittee that births occurring to the mother before 28 weeks gestation as well as those after this period should be recorded." Third, Quebec was the only province to use the special form for "a child having lived less than 24 hours" as well as stillbirths. In essence, Quebec's version of the national special form was the same as its earlier provincial special form; thus its introduction made no difference to the statistics reported for the province. Nor did Ontario's introduction of the special form on i July 1940 affect its statistics. Between 1939 (the last full year under dual registration) and 1941 (the first full year under special-form registration) the stillbirth rate dipped slightly, but in the context of a declining

112 Facts of Life

trend, and the death rate for infants in the first day of life did not change.34 The explanation was threefold. First, the breathing requirement in the statistical definition remained a dead letter - the national special form did not ask about signs of life. Second, Ontario did not use the form for infant deaths in the first day of life. Third, the province began to obtain comprehensive information about gestational age before its introduction of the special form. About 1935 the Dominion Bureau of Statistics had introduced a question about gestational age to the standard birth registration form. It may have been 1938 or so, however, before the use of the new forms by local registrars became universal.35 Canada's Abandonment of the League of Nations Definition in 1943 and 1950. In 1943 Canada began to include live-born foetuses in its published statistics, regardless of gestational age. Officials expected the change to increase the reported totals for infant deaths by about 5 per cent,36 and this seems to have happened. In the context of declining trends, Ontario's reported rates for neo-natal deaths, infant deaths in the first day of life, and infant deaths from prematurity in the first day of life increased by two, three, and three respectively. Similar increases obtained in the 1943 neo-natal infant death rates for British Columbia, Saskatchewan, Manitoba, New Brunswick, and Nova Scotia, and in the 1944 rates for Prince Edward Island and Alberta - perhaps the definitional change was implemented after a one-year time-lag for these two provinces. The Quebec rates for neonatal deaths and infant deaths in the first day of life, by contrast, continued a declining trend after 1942. This reinforces the impression, based on evidence for 1932, that the Quebec statistics had never followed the so-called Canadian practice of excluding such cases. In 1950 Canada substituted "any sign of life" for the breathing requirement in its statistical definition. As shown above in Table 5.1, this had no discernible effect on Ontario's reported rates for stillbirths and infant deaths in the first day of life. This was to be expected because Ontario statistics had never followed the requirement. The Quebec statistics, by contrast, had followed the breathing requirement. Thus Canada's abandonment of the breathing requirement should have increased Quebec's reported rate for infant deaths in the first day of life and decreased its reported stillbirth rate. To some extent, that is what happened. Although the 1950 drop in Quebec's reported stillbirth rate was small (-0.5), its reported rate for infant deaths in the first day of life rose by an unprecedented 2.9, after some years of a declining trend.37

ii3 Getting Born CONCLUSION

Definition influenced published statistics for infant deaths and stillbirths. In 1932 a restrictive statistical definition for signs of life (breathing) and foetal viability (twenty-eight weeks gestational age) potentially made infant death totals about 7.5 to 9.3 per cent lower than would have obtained under the more general 1950 definition (no gestational-age requirement and any sign of life). Similarly the increase in the gestational-age requirement for foetal viability in 1932 (from twenty-six to twenty-eight weeks) potentially lowered reported stillbirth totals by about 15 per cent (the actual effect in Quebec).38 The 1932 statistical definition, however, was applied unevenly between Ontario and Quebec. Quebec's special-form registrations reported information about gestational age and signs of life for stillbirths and deaths of infants in the first day of life. Ontario's standard forms did not. Thus bureau officials could apply the 1932 definition comprehensively for Quebec, although they chose to ignore the gestational-age requirement for that province's infant death statistics. By contrast, bureau officials could apply the 1932 definition only partially for Ontario, based on information about gestational age in the returns from public hospitals. In 1935 a change in the birth registration form allowed a fuller application of the gestational-age requirement in the definition. Partly for this reason, Ontario's introduction of the national special form for stillbirth registration in 1940 had no discernible effect on provincial totals for infant deaths and stillbirths. Unlike Quebec's 1932 and 1940 special forms, however, the Ontario form did not ask about signs of life and was not used for deaths of infants in the first day of life. Thus the breathing requirement in the official definition was a dead letter. Physicians could not observe it without violating the Ontario Vital Statistics Act and legal definition, and the registrations did not give officials the information needed to transcend the physician classifications. Canada had adopted the League of Nations standard definition in 1932 to help make provincial and international statistics for stillbirths comparable. A decade later, however, the Quebec and Ontario statistics still differed with respect to the breathing and gestational-age requirements in the standard definition. At the same time, Canada was the only nation to have adopted the League definition in its entirety, and its statistical practices continued to differ from those of other countries. In the circumstances, Canada moved towards the new international standard definition for live birth that the World

ii4 Facts of Life

Health Organization was to recommend in 1950. In 1943 the Dominion Bureau of Statistics began to include all live births in its statistics, regardless of gestational age, and in 1950 it replaced the breathing requirement with "any sign of life." The influence of the 1932 statistical definition on the Ontario statistics can be summarized briefly. First, the gradual enforcement of the twenty-eight-week gestational-age requirement between 1932 and 1935 made the provincial totals 5 per cent lower than would have obtained under the 1943 rules, which did not have the requirement. Second, the breathing requirement in the official definition made no difference because it was a dead letter. Third, the increase in the gestational-age requirement from twenty-six to twenty-eight weeks slightly reduced the province's reported stillbirth totals, apparently by excluding some of the stillbirths reported for public hospitals. The definitional influence on differences between the Ontario and Quebec statistics changed over time and is tricky to discern. Based on the evidence, the reported difference between the Quebec and Ontario infant mortality rates was from 1.4 to 3.0 less under the 1932 statistical definition than under the 1931 provincial statistical practices. On the one hand, Quebec applied the breathing requirement during the 1932-49 period, while Ontario did not. In 1932 this removed 4.2 from the difference between the reported provincial rates. On the other hand, Ontario excluded deaths of live-born, "nonviable" foetuses from its infant death totals until 1943, whereas the Quebec totals included them. These contrasting practices added to the difference by 1.2 in 1932 and a minimum of 2.8 in 1942.39 When did Ontario civil registration obtain a reliable coverage for infant deaths and stillbirths? What was the real trend for its infant death rate during the study period? What was the actual difference between the Quebec and Ontario infant mortality rates? How much did the declining trend for infant mortality contribute to the increase for expectation of life at birth? Such questions held considerable significance for contemporaries, who treated the infant death rate, expectation of life at birth, and even the stillbirth rate as measures of the quality of life and the efficacy of public-health programs. Yet the answers depended partly on the definitions used, and these in turn changed over time, varied between and within provinces, and were socially constructed in local, provincial, national, and international historical contexts. Finally, the definition issue illustrates the slow integration of Canada's national civil registration system. During the 19205 standard definitions and registration forms were lacking at the international level, let alone within Canada. Moreover, each move towards

ii5 Getting Born standardization in Canadian civil registration required discussion and agreement between the Dominion Bureau of Statistics and the provinces, usually with the mediation of the Vital Statistics Section of the Canadian Public Health Association. Even then, provincial and physician compliance with the nationally agreed definitions varied. In statistics as in culture, Quebec especially persisted as a "distinct society" within the Canadian confederation.

6 Fatal Pregnancies in Ontario, 1920-35: A Study of the Nature of Statistics for Deaths by Cause

Published statistics for deaths by cause raise several points of interpretation. What did physicians understand by the causes they reported on the death certificates? How reliable and uniform were their diagnostic skills? Did contemporary issues make them less likely to report certain causes than others? How did the design of the death-certificate form and the administrative practices of officials influence the information that physicians provided? How did officials use the information in the death certificates to compile their published statistics? During the 1900-50 period, for example, international published statistics were organized around the concept of an underlying cause: a single disease that initiated the sequence of morbid conditions leading to death. Thus, even where a death certificate reported two or more causes, only one could prevail for classification. That being so, how did officials make their selection? How should one interpret the published statistics in view of their neglect of the reported causes that were not the underlying cause? Two recent multiple-cause approaches elaborate the nature and limitations of the underlying-cause statistics.1 In the first, the extended underlying-cause model, cause-combination categories group the underlying cause with associated or contributory causes that are believed to have modified its behaviour (e.g., the "arteriosclerotic heartdisease" category groups "arteriosclerosis" and "heart disease"). In the second, the pattern-of-failure model, by contrast, the causecombination categories group all reported causes of death but make no assumption about causal sequence.

H7 Fatal Pregnancies, 1920-35

The total-mentions approach offers a another "basic way" to convert "multiple-cause information... into numerator data for mortality analysis." As the name suggests, it involves counting the number of times a disease is mentioned on a death certificate, whether as the underlying cause or as a contributory cause. Because many death certificates report multiple causes, the number of mentions exceeds the number of deaths if mentions for two or more diseases are counted. The approach makes no assumptions about the relative importance of different causes, and it risks counting some deaths for which a disease is present but incidental. Especially where the physician's identification of the underlying cause is problematic, however, it shows the potential incidence of a disease as a primary or contributory cause of death. This chapter explores the nature of Ontario's published statistics for deaths by cause, using the statistics for pregnancy-associated deaths as an example. First, it introduces pertinent definitions and outlines the scope of the inquiry. Second, it discusses how contemporary interest in maternal deaths influenced the statistics. Third, it describes how officials compiled published statistics for maternal death during the 1920-35 period. Fourth, the chapter provides a detailed investigation of the Ontario statistics for the 1920-35 period. To this end it uses evidence in provincial death registrations and published statistics to obtain a type of total-mentions count of maternal deaths for the years 1920 and 1930; it then compares these counts with total-mentions counts reported in contemporary investigations and the underlying-cause counts in the published statistics. Finally, it discusses the quality and coverage of the published statistics for years after 1935. DEFINITIONS AND

SCOPE OF THE STUDY

Maternal death includes deaths caused either directly or indirectly by pregnancy but excludes deaths for which pregnancy is incidental (e.g., a traffic fatality). Direct maternal deaths result directly from the experiences of pregnancy, labour, and the puerperium. Haemorrhage, infection, and toxaemia are three classic causes of these deaths. (See Appendix C for definitions of selected medical terms.) In the indirect cases, pregnancy contributes to death from a non-puerperal cause. For example, death can result from a pre-existing chronic disease (e.g., heart condition, tuberculosis) that would not have proved fatal without the additional stress of pregnancy. Pregnancy also lowers a woman's cell-mediated immunity to acute infectious disease (e.g., influenza, typhoid), thus rendering her more apt to catch the disease and less likely to recover from it.2

n8 Facts of Life

The chapter definition for maternal death follows the principal literature definition. Some obstetrician authors, however, limit the definition to deaths that are preventable through improvements in obstetric practice; thus they exclude deaths from genetic or social causes (e.g., molar pregnancies, ectopic gestation, criminal abortion). Others limit the definition to deaths associated with childbirth; thus they too exclude criminal abortion deaths, which arise from attempts to avoid childbearing. By contrast, the chapter definition includes these deaths because they arise from pregnancy. The maternal mortality rate, the standard measure of the maternal mortality risk, states the number of maternal deaths per 1,000 live births. Where civil registration data serve for its calculation, scholars conventionally obtain the numerator data from published totals for the puerperal-state class of deaths. Although the puerperium is the recovery period following labour, the term in published statistics also includes pregnancy and labour. The maternal mortality rate is a crude measure of woman's risk of death from pregnancy. Whereas the denominator allows only one pregnancy outcome (live birth), the numerator allows four (live birth, stillbirth, abortion, and maternal death with no delivery). Induced abortions present a special problem; each induced abortion death adds one to the numerator but zero to the denominator; at the same time, each induced abortion, fatal or not, subtracts one from the denominator by eliminating a prospective live birth. The maternal mortality rate also reflects "compositional" influences.3 Because a mother's mortality risk increases with age, for example, a decline in the proportion of mothers in the older age groups reduces the death total without any change in the age-specific risks. Thus one must control for the influence of risk-factor distributions (age, parity, and marital status) on the trend for the maternal mortality rate to obtain the trend for the maternal mortality risk. The Study Period. The years 1920-35 approximate the last stage for high maternal mortality in Ontario. Moreover, the information solicited for maternal deaths is consistent during the period and also is transitional between that for the periods preceding and following. To launch Canada's national system of civil registration in 1920, the Dominion Bureau of Statistics issued new forms for registering deaths and births. The 1920 form for births introduced questions about the mother's age and parity (number of delivered pregnancies). As shown in Appendix B and discussed below, the 1920 form for deaths also held new questions, but solicited less information than the form that was to replace it in 1935.

ii9 Fatal Pregnancies, 1920-35

The Study Years. The years 1920 and 1930 were broadly similar for maternal mortality, but had interesting differences. Maternal deaths were numerous in both years: Ontario's published totals for deaths in the puerperal-state class (489 and 440 respectively) were the highest in provincial history. Moreover, the form for registering deaths, and therefore the information solicited, was identical for the two years. However, epidemic influenza caused many indirect maternal deaths during February and March of 1920, whereas no comparable influence acted during 1930.4 And physicians probably reported puerperal causes more reliably in 1930 than in 1920, when they had been less aware of maternal mortality as a public-health problem.5 MATERNAL MORTALITY AS A CONTEMPORARY ISSUE

The maternal mortality rate for Ontario has declined dramatically since the study period. The province reported 558.3 maternal deaths per 100,000 live births for the period 1920-35. The comparable 198185 statistic was 4.5, or 124 times less - a reflection of new drug therapies (sulphanilamides, then antibiotics) and blood-transfusion techniques introduced after 1937, improvements in ante-natal and post-natal care, improved detection and management of problem cases, and change in the age-parity distributions for mothers.6 Contemporaries also regarded their maternal mortality rate as high.7 By the 19205 Canada's reported maternal mortality rate had changed little since the nineteenth century, in contrast to the "improvement in the infant death-rate and every other death rate and the gradual tendency towards the lengthening of human life."8 As Ontario and Dominion health department investigations documented, moreover, 20 to 25 per cent of the province's pregnancyrelated deaths were not reported as such in published vital statistics.9 Noting the lower maternal death rates reported for certain countries, public-health officials believed that up to 50 per cent of Canada's maternal deaths were preventable through better pre-natal care and physician training in obstetrics. The thread of maternal neglect also linked maternal deaths to high stillbirth and neo-natal infant death rates. Thus Canada's maternal deaths were a "national disgrace," "preventable maternal wastage," an unacceptable "cost paid by our mothers for the renewal of the race," and a discredit to the medical profession. In her 1928 report for the Dominion Health Department inquiry, Dr Helen MacMurchy described the "long procession of 1,532 Canadian mothers going to their graves in the 59th year of

12O

Facts of Life

Confederation ... These mothers left behind them 5,073 motherless children. They died young. The youngest was 15 years old, and the oldest was only 48. Their sun has gone down while it was day. They did not live out half their years. Their loss was the greatest bereavement that their home and country suffered in that year. If these homes had been in Denmark instead of Canada, a thousand of the mothers would not have died."10 Nevertheless, culture and group interests, not just mortality trends, made maternal mortality a major public-health issue during the 19205 and 1930S.11 First, the high maternal death rates worked against an important goal of pro-natalist, nativist officials - that anglo-Canadian women should have large families to increase population. Second, the high rates called into question the legitimacy of the control of midwifery by male physicians. Unlike Canada, for example, Holland relied primarily on female midwives and home deliveries, yet it reported much lower rates for maternal death. How then were Canadian physicians to sustain the myth that normal pregnancy was a pathological condition, requiring the attendance of a physician and, increasingly, hospitalization? Thus both pro-natalist government officials and physicians came to view maternal mortality as an urgent problem for investigation. Their search for solutions in turn exposed competing interests among family physicians, hospital obstetricians, public-health officials, and nurses. Public-health programs, for example, were potentially a cutting edge of "state medicine" and a threat to physician autonomy. Clearly interpretation of maternal mortality statistics should allow for historical influences. Following the quickening of public interest in maternal mortality, for example, even contemporaries wondered whether an improved reporting of puerperal causes on the death certificates by physicians masked a declining trend for the maternal death rate.12 As MacMurchy observed, the rate was higher in "those provinces and municipalities where most interest has been taken, where most enquiry has been made and where the grave importance of maternal mortality has been realized." MacMurchy herself was a propagandist for maternal- and child-welfare programs; perhaps this predisposed her to find more maternal deaths than Canada's published vital statistics had reported.13 The widespread stigmatization and concealment of induced abortion undoubtedly caused physicians to underreport it as a cause of death. During the 19308, however, physicians probably reported abortion deaths more fully than before. At this time abortion received international attention as a social problem, in part because pronatalists linked an apparent rise in the incidence of abortion to a

121 Fatal Pregnancies, 1920-35 declining trend in the birth rate. Partly too, by reporting abortion deaths more fully, physicians could establish social causes, rather than physician incompetence, as the source of some maternal deaths. Both pro-natalists and medical men sought to lower the maternal death rate primarily through improvements in medical services, perhaps with mothers' allowances to offset the costs of these services during parturition.14 By contrast, both groups were cool to a birthcontrol approach (e.g., to lower the proportion of mothers in ageparity categories carrying above-average risk). This may explain why civil registrations reported the mother's parity for births (reflecting the pro-natalist interest in family size) but not for deaths (reflecting a disinclination to discover the maternal mortality risk associated with repeated childbearing). THE NATURE OF THE P U B L I S H E D STATISTICS

This section considers six aspects of the published statistics: i) the sources and format for published statistics for the puerperal-state class of deaths; ii) the role of Manuals of the International List of Causes of Death in classification; iii) the concept of the "underlying cause" of death; iv) the status of the physician's choice for "primary cause"; v) Canada's "rules for choice" of cause for classification; and vi) published tables for "pregnancy-associated" deaths. The Sources and Format. Ontario's published mortality statistics reported deaths by classes, categories, and subcategories of cause. As shown in Table 6.1, the puerperal-state class of deaths included seven categories in 1920 and also nine subcategories in 1930. The greater detail for 1930 reflected a secular trend towards more detailed classification, and also a growing interest in maternal mortality as a public-health problem. Manuals of the International List. Canadian classification followed the International List of Causes of Death. The list, originally established by an international commission in 1893, was revised periodically (1900, 1909, 1920, 1929, 1938, 1948, 1955, 1965, 1975), and each revision changed how pregnancy-related deaths were reported. To interpret the published statistics for 1920 and 1930, one must consult respectively the manuals for the second (1909) and third (1920) revisions. Both revisions placed criminal abortion deaths, for example, in the homicide by other means category of the external causes class, not in the puerperal-state class. Similarly, within the puerperal-state class, they

122

Facts of Life

Table 6.1 Deaths Reported for the Puerperal-State Class for 1920 and 1930 Code

1920 Annual Report

134 135 136 137 138 139

Accidents of Pregnancy Puerperal Haemorrhage Other Accidents of Labour Puerperal Septicaemia Puerperal Albuminuria and Convulsions Puerperal Phlegmasia Alba Dolens, Puerperal Embolism, Sudden Death in Puerperium Following Childbirth (not otherwise specified)

140

76 45 51 156 123 21 17 489

Total Code

1930 Annual Report

143

Accidents of Pregnancy 143a Abortion 143b Ectopic Gestation 143c Others under this title 143d Abortion, Self-induced Puerperal Haemorrhage Other Accidents of Labour 145a Caesarean Section 145b Difficult Labour 145c Other surgical operations 145d Uncontrolled Vomiting 145e Ruptured Uterus in Labour Puerperal Septicaemia Puerperal Phlegmasia Alba Dolens, Puerperal Embolism, Sudden Death in Puerperium Puerperal Albuminuria and Convulsions Following Childbirth (not otherwise specified)

144 145

146 147 148 149 Total

28 3 17 7 1 39 55 16 12 13 10 4 153 42 110 13 440

Source: Ontario, Registrar-General, Annual Reports, 1920, 1930.

assigned deaths from abortion with sepsis to the category for puerperal septicaemia, not the subcategory for abortion.15 The Underlying Cause of Death. This organizing concept for Canadian and international classification developed at the turn of the century, when the microbiological understanding of disease was displacing the humoral view (see discussion in the Prologue). In addition, infant deaths and epidemic infectious diseases were then the primary public-health problems. In this context officials generally could abstract death to a single cause that had occurred simultaneously with death.16

123 Fatal Pregnancies, 1920-35 During the study period infant mortality and infectious-disease epidemics were receding in importance, and chronic degenerative diseases were becoming more common. Death increasingly involved two or more diseases that operated over time, interacted with host characteristics such as aging, and did not occur in obvious sequence. Thus physicians increasingly reported multiple causes on the death certificate. Related influences to the same end included a growth of physician belief in the importance of contributory causes, a secular increase of medical knowledge for diagnosis, a shift to more detailed categories under new revisions of the International List, and the introduction of more detailed questions on the 1935 form for the medical certificate (see Appendix B). By the 19205, therefore, officials appreciated that much of the information on the death certificate was lost through underlying-cause classification. The introduction of the table for "pregnancy-associated deaths" in the bureau's published statistics for 1922 and the bureau's special report on contributory causes of death for 1926 were signs of this concern. Similarly, Dr MacMurchy's classic 1928 study of maternal mortality anticipated the total-mentions approach; for this reason her statistics for all causes exceeded the number of cases in her file.17 The Status of the Physician's Choice for Primary Cause. The physician sometimes did not enter the puerperal cause on the death certificate. As W.J. Bell lamented for Ontario in 1921, "Some physicians prefer to report ... almost any other cause rather than puerperal sepsis or one of the other causes of maternal mortality." A township clerk had called his attention to one "B.K.E, aged 36 years and mother of two children [who] gave birth to a stillborn infant on December 8, 1920. On January 3, 1921, she died after great suffering. The cause of death given on the doctor's certificate was: Primary: 'Sepsis and Phlebitis'[;] Contributory: 'Pulmonary Thrombosis.'"18 The physician might have omitted the puerperal cause out of delicacy (e.g., deaths involving an unmarried woman or illegal abortion), ignorance of the woman's case history (e.g., he had not attended the woman before her death), indifference to vital statistics, or desire to conceal malpractice. Officials noted, however, that the certificate form was ambiguous about the information required. Consequently, physicians often stated the immediate or terminal condition for primary cause while ignoring the underlying cause or entering it as contributory.™ In the circumstances, provincial registration officials used all the information on the death certificate, not the physician's choice for

124 Facts of Life Table 6.2 Percentage of International Assignments to the Puerperal-State Class in 1931 for 1,073 United States Deaths Associated with Pregnancy Denmark New Zealand U.S. Australia Scotland Netherlands

99.4 92.8 92.7 92.7 92.3 91.9

Italy Canada Chile Czechoslovakia N. Ireland France

90.5 89.6 88.6 85.3 83.9 82.7

Irish Free State Sweden Estonia England/Wales Norway

81.0 80.5 79.9 78.7 76.9

primary cause, to determine the cause of death for classification.20 Where the information on the death certificate was ambiguous or incomplete, they asked local officials to obtain further details from the physicians who had reported the deaths.21 Canada's ''Rules for Choice." Through time physicians increasingly stated two or more causes on the death certificate.22 At the same time, the physician's choice for primary cause was unreliable for the underlying cause. In the circumstances, officials examined all information on the death certificate and then applied "rules for choice" to determine which stated cause prevailed for classification. Each country, however, had its own rules. Thus in 1931, when the United States Children's Bureau asked different national jurisdictions to classify 1,073 U.S. medical certificates on which pregnancy or childbirth was mentioned, the proportions assigned to the puerperal-state class (shown in Table 6.2) ranged from 77 to 99 per cent.23 The Canadian rules were "based principally on the original French manual" issued by the international commission in 1903.24 The manual provided six general guidelines, such as the selection of a "surely fatal" disease over one of "less gravity," and preference to "precise diagnosis ... over vague and indeterminate ones, such as Haemorrhage." To guide selection from specific combinations of causes, England issued a manual on Joint Causes of Death in 1912, and the United States issued similar manuals in 1914, 1925, 1933, and 1939. Without a Canadian manual on joint causes, the Dominion Bureau of Statistics "largely followed" the English rules, used the American manual in cases where the English rules seemed not to apply, and, where necessary, made special rulings in consultation with the Department of National Health.25 In the priority given to puerperal causes, however, the Canadian rules were closer to American than to English practice. As shown above in Table 6.2, Canada assigned 90 per cent of the cases in the

125 Fatal Pregnancies, 1920-35 u.s. Children's Bureau Study to the puerperal-state class, compared to 93 per cent for the United States and 79 per cent for England. Although these assignments were made under the fourth (1929) revision of the International List, which came into force in 1931, the various national priorities for puerperal causes seemed unchanged from earlier years. First, calculation from published Ontario statistics yielded similar maternal mortality rates (5.4 and 5.2) for the 1926-30 and 1931-35 periods, which preceded and followed the change of manuals. Second, differences between the rates calculated from English and Ontario statistics for the earlier 1925-30 period (respectively 4.2 and 5.8) indicated the same lower English preference for puerperal causes.26 Third, Ontario assigned 83 per cent of its pregnancyrelated deaths in 1925 to the puerperal-state class, compared to 90.9 per cent for the United States, an earlier indication of the higher American preference.27 In the only published description of the Canadian rules (see Appendix D), E.S. Macphail, chief of the Census and Vital Statistics Division of the Dominion Bureau of Statistics, noted the high order of preference given to puerperal causes.28 The rules placed puerperal fever (septicaemia), puerperal phlebitis, and puerperal diseases of the breast in the highest category of preference. They also gave high preference to other puerperal causes (puerperal haemorrhage, accidents of pregnancy or labour), although ranking these below infectious diseases (influenza, typhoid) and various general diseases (cancer, tuberculosis). A non-obstetric category was favoured, however, where "the certificate merely makes reference to pregnancy or childbirth without indicating any puerperal disease or abnormality." Where the certificate stated two or more obstetric causes, puerperal septicaemia received preference over all others. The Tables for "Pregnancy-Associated" Deaths. In its annual report for 1922, the Dominion Bureau of Statistics introduced a table for "pregnancy-associated deaths not reported as such."29 As shown in Table 6.3, the bureau's table for 1930 reported 66 pregnancyassociated deaths in Ontario, in addition to those assigned to the puerperal-state class. The table possibly missed some pregnancy-associated deaths. The Ontario table data for 1926, for example, largely matched those in a Dominion Bureau of Statistics Special Report on Contributory Causes of Death for 1926. The special report statistics in turn had been compiled through a second classification of all deaths according to contributory cause. Where more than one contributory cause was stated on the death certificate, the rules for choice had been applied for classification.

126

Facts of Life

Table 6.3 Deaths of Women Not Classified to Pregnancy or Child-Bearing but Returned as Associated Therewith, Ontario 1930 Code

Cause

007 008 010 Oil 021 031 037a 038 040 046 049e 050 063 090a 090d 090e lOOa lOla 102

Measles Scarlet Fever Diphtheria Influenza Erysipelas Tuberculosis of the Respiratory System Acute Disseminated Tuberculosis Syphilis Gonococcus Infection Cancer of the Female Genital Organs Cancer of the Genito-Urinary System Benign Tumors and Tumors Not Returned as Malignant Diseases of the Adrenals (Addison's) Valvular Disease Chronic Endocarditis Chronic Myocarditis Broncho-Pneumonia Lobar Pneumonia Pleurisy Diseases of the Pharynx and Tonsils Other Diseases of Stomach (Not Cancer) Appendicitis and Typhlitis Acute Yellow Atrophy of Liver Accidental Traumatism by Fall Homicide

109b

112 117 120 185 199

Total for All Causes

Total

1 2 1 13 2 10 1 1 1 1 1 1 1 6 2 1 1 5 1 1 1 2 3 1 6 66

Consequently, as in the primary classification, assignment to a nonobstetric class was possible where "the certificate merely makes reference to pregnancy or childbirth without indicating any puerperal disease or abnormality." The rules for choice did not affect the table data for influenza-pregnancy deaths, however. These deaths numbered seventy-seven in the table for 1926, compared to six in the special report.30 THE TOTAL-MENTIONS COUNTS OF MATERNAL DEATHS FOR 1920A N D 1 9 3 0

The Information in the Death Registrations. Ontario registered 40,440 deaths for 1920 and 37,313 deaths for 1930. Each registration stated

127 Fatal Pregnancies, 1920-35 personal information in the top section (the decedent's name, age, occupation, place of death, whether the death occurred in a hospital or institution, the date of death, marital status, and names of parents). In the bottom section - the medical certificate of death (see Appendix B) - the attending physician gave information about primary and contributory causes of death and also responded to two supplementary questions: a) "Did an operation precede death?" and b) "Was there an autopsy?" Acting alone, or on request from the Dominion Bureau of Statistics, the Ontario registrar-general's office asked municipal registrars to obtain further information concerning "death certificates of women of child-bearing age in which the cause of death suggested a puerperal condition."31 Beginning in 1925 provincial officials noted the results of their inquiries on the death certificate in red ink, and initialled their notations. On seventeen registrations for 1930 these notations showed a puerperal cause that the attending physician had not reported. Death certificates for 1920, which predate the practice of entering marginal notations, are especially likely to omit puerperal causes. For some such cases one can document the connection with pregnancy by linking the mother to an infant's birth (or death) registration, using information on her death registration (name, age, date of death, municipality of residence and street or concession address, and husband's name) and comparable information on the infant's registration (mother's name, age, date of her infant's birth, place of residence, husband's name). One can then infer maternal cause for deaths that occurred within a specified number of days of the termination of the pregnancy - the method used in European studies based on data from parish registers.32 The length of the interval is problematic; it misses maternal deaths if too short but includes non-obstetric deaths if too long. Experts differ on how to resolve the problem. The recommended interval is one calendar month for the British demographer Roger Schofield, 42 days for the International Federation of Gynecology and Obstetrics, 60 days for French demographers, 90 days for the American Medical Association (1950), and one year for the British Health Ministry (1928). This study classes a death as maternal if the death certificate reports a puerperal primary or contributory cause of death, an operation for pregnancy, or autopsy findings related to pregnancy. It also gives maternal classification to deaths that occurred within 42 days of a pregnancy termination, as established through record linkage

128 Facts of Life Table 6.4 Documentary Basis for the Identification of Deaths as Maternal

Reported Cause of Death Registrar-General's Notation Operation or Autopsy Information Link to Infant's Registration* Total

2920

%of

Cases

1920

2930 Cases

%of 1930

551 0 6 56 613

89.9 0.0 1.0 9.1 100.0

455 17 11 6 489

93.1 3.5 2.2 1.2 100.0

Documentation not from death certificate.

between the mother's death registration and an infant's birth or death registration.33 Table 6.4 summarizes the evidence used for case selection. Matching the Death Registration to Published Statistics. By linking the registration deaths to published statistics for the puerperal-state class, the writer obtained three link-status groups of maternal deaths: i) linked deaths (described in published statistics for the puerperal-state class and traced to a death registration); ii) unlinked registration deaths (identified as maternal from information on the registration, but not matching a description in published statistics for the puerperal-state class); and iii) unlinked published-statistics deaths (described in published statistics for the puerperal-state class, but not traced to a death registration). For reasons discussed below, the sums of the three group totals were treated as the total-mentions counts for the two years. Most linkage decisions were straightforward. In 1920, for example, published statistics for Haliburton County reported in the puerperal albuminuria and convulsions category one death that occurred in January and involved a married, foreign-born woman aged 40-49. A search of the county death registrations turned up a January death of a married, English-born woman aged 44 whose stated cause of death, "nephritis, high blood pressure, confined six weeks ago," fitted the published category. Clearly the two sources referred to the same death. Nevertheless, correct linkage decisions are crucial to obtaining an accurate count of maternal deaths. Each true link missed inflates the count by one (by removing a case from the linked-death group and adding a case to each of the other two groups). Similarly, each false link made deflates the count by one. Before presenting the maternal death counts, therefore, it is important to review three features of

129 Fatal Pregnancies, 1920-35

the evidence that created difficulty for linkage decisions. These are discussed briefly below and more fully in Appendix E. Variation in the Criteria for Linkage. The criteria vary between and within the two study years, according to characteristics of the published materials. The 1920 statistics provide six criteria for linkage but are less precise for Toronto than elsewhere because of the numbers involved. The 1930 criteria are the same for the province's four largest cities, but only two criteria obtain for other jurisdictions - the municipality and cause of death. Reporting Errors in Published Sources. Published sources hold obvious printing and tabulation errors. Less detectable errors, therefore, could inflate the maternal death count by causing incorrect linkage decisions (i.e., a double count of a death through assignment to both the unlinked registration and unlinked published-statistics groups). To allow for this, the writer relaxed the criteria for certain linkage decisions. Limitations of the Methodology for Record Linkage. To detect maternal deaths whose death certificates did not state an obstetric cause, the writer linked the decedent of an infant's birth or death registration. In some cases, however, the infant's birth or death may have been unregistered. In other instances the maternal death involved a nonreportable pregnancy outcome (abortion or maternal death with no delivery). Finally, searches of birth registrations were undertaken selectively and limited to the municipality in which the maternal death occurred. The examples in Table 6.5 show the limitations of the study methodology. None of the four death certificates specifies a puerperal cause, and none of the decedents can be linked to an infant's birth or death registration. Circumstantial evidence indicates, however, that officials assigned the deaths to the puerperal-state class. For each registration the stated cause of death is consistent with the category in the published source; the decedent's age and month of death matches the published description; and no other registration has information matching the published description. Presumably registration officials obtained supplementary information from the reporting physicians to connect the deaths to pregnancy. The Total-Mentions Counts. Table 6.6 presents the maternal death counts for 1920 and 1930 as the sums of the subtotals for the three link-status groups. The principal finding is that the published totals

130 Facts of Life Table 6.5 Four Registered Deaths for 1920 and Matching Descriptions in Published Statistics for the Puerperal-State Class

Documentary Source

Age Category/ Age

Month of Death

Cause of Death/Cause of Death Category

Brockville (Town) Published Statistics: Death Registration:

20-29 21

April April

Puerperal Septicaemia Septic Peritonitis

Barrie (Town) Published Statistics: Death Registration:

20-29 21

April April

Puerperal Septicaemia Septic Peritonitis

Collingwood (Town) Published Statistics: Death Registration:

30-39 37

September September

Puerperal Septicaemia Septicaemia

Niagara Falls (City) Published Statistics: Death Registration:

40-49 43

September September

Puerperal Albuminuria Uraemic Convulsions

for the puerperal-state class (489 and 440) exclude 27.3 and 16.5 per cent respectively of the total-mentions counts of maternal deaths for the two years (673 and 527). If one includes deaths reported in the published table for pregnancy-associated deaths, the proportion omitted for 1930 drops to 5 per cent. The unlinked published-statistics deaths are the most difficult to interpret. Due to printing errors in the published tables, for example, some deaths may be double counted through inclusion in both "unlinked" groups. At the same time, officials apparently classed some deaths on the; basis of information not stated on the death certificate, and the writer may interpret certain terminology more rigorously than the officials or the reporting physician did. For example, although "metria" is widely equated with "puerperal metritis," metritis sometimes arises from a non-obstetric cause; thus the writer did not class a metritis death as obstetric unless pregnancy was specified. In any case, neither of these problems affects the count: the death is still counted once, but in the unlinked published-statistics rather than the linked deaths group. In general the distorting influences on the counts are offsetting. On the one hand, the totals may be inflated by double counts and the inclusion of deaths for which pregnancy was incidental. On the

131 Fatal Pregnancies, 1920-35 Table 6.6 Maternal Deaths by Link-Group Status, 1920 and 1930 1920

%

%

Total

1930

63.7 63.7

411 52 463

78.0 9.9 87.9

184

27.3

26

4.9

60 n.data 60

8.9 8.9

29 9 38

5.5 1.7 7.2

Total Maternal Deaths

673

99.9

527

100.0

Published Statistics for Puerperal-State Class: N = % of All Maternal Deaths

489 72.7

Linked Cases (both sources) Puerperal-State Class Pregnancy- Associated Subtotal Unlinked Registrations Subtotal Unlinked Published Statistics Puerperal-State class Pregnancy-Associated Subtotal

429 n.data 429

Total

440 83.5

other hand, the counts miss maternal deaths for which the physician omitted the puerperal cause on the death certificate; the pregnancy outcome was either not registered (live birth, stillbirth) or involved a non-reportable pregnancy outcome (abortion, no delivery); and the stated causes did not prompt the registrar-general to seek further information. In the circumstances, the maternal death counts for 1920 and 1930 are plausible, and they also accord with contemporary counts for other years in the 1920-35 period. The Comparison with Contemporary Investigations. Three contemporary investigations found shortcomings in the maternal mortality statistics published for Ontario. These were W.J. Bell's study for the Ontario Health Department for 1921-22; Helen MacMurchy's national study for the Dominion Health Department for the year i July 1925 to 30 June 1926; and J.T. Phair and A.H. Seller's study for 1933, which the Ontario Health Department undertook jointly with the University of Toronto.34 Like the writer, each investigator used evidence from death registrations, birth registrations, and published statistics. In addition each investigator used physician responses to questionnaires to determine whether the deceased was pregnant at her time of death and, if so,

132 Facts of Life

whether pregnancy was a cause or contributing cause. Bell's questionnaire went to the attending physician for all female deaths in the 15-50 age group. MacMurchy's national study used the questionnaire for all deaths that the registrar-general had assigned to the puerperalstate class, plus some 2,800 additional female deaths in the 15-50 age group whose reported cause(s) might have been associated with pregnancy.35 Phair and Sellers used the questionnaire for all deaths for which the death certificate specified "pregnancy or a puerperal disease" as a cause or contributory cause. A 100 per cent response was impossible, MacMurchy noted, because in some cases the attending physician had died or left to study abroad before receiving the questionnaire, or he could not be located because he had been merely passing through the locality concerned. Nevertheless, she claimed a 92 per cent response for deaths that the registrar-general had assigned to the puerperal-state class. Phair reported a 92.5 per cent response for such deaths and a 75 per cent response overall. Bell does not report the response rate for his study, but MacMurchy places it at 95 per cent, probably a reference to deaths assigned to the puerperal-state class only.36 The Phair/Sellers maternal death count was low. First, it excluded five criminal abortion deaths that the registrar-general reported for 1933. Second, they presented their findings in June 1934, before the 1933 death registrations were complete. Under Ontario registration law, local registrars could register a vital event up to one year within its date of occurrence. To allow for late returns, the registrar-general's annual report for 1933 was submitted on 31 December 1934 and printed in 1935. Doubtless this explained why Phair and Sellers gave 334 as the published total for puerperal-state class deaths, whereas the actual published total was 346. To rectify the Phair/Sellers count (411), the writer added the five criminal abortion deaths and the twelve missed deaths for the puerperal-state class. No adjustment was possible, however, for other pregnancy-related deaths that the preliminary data missed. As shown in Table 6.7, published totals for the puerperal-state class missed between 19 and 26 per cent of the deaths found by the contemporary investigations. This compared to 17 and 27 per cent respectively for 1920 and 1930, the years in the writer's investigation. With the inclusion of table totals for pregnancy-associated deaths, the published totals still missed up to 10 per cent of special investigation counts. The shortfall for the puerperal-state class statistics alone varied with the prevalence of acute infectious disease (an important source of indirect maternal mortality). Thus epidemic

133 Fatal Pregnancies, 1920-35 Table 6.7 Ontario Maternal Deaths: Published Totals and Totals Found in Contemporary Investigations Data Source

Coverage Period

Published Puerperal State

Published PregnancyAssociated

Published Data: All Sources

Total Found

Study Data Bell Bell MacMurchy Study Data Phair/Sellers

1920 1921 1922 1925-26 1930 1933

489 (73) 387 (80) 370 (80) 368 (74) 440 (83) 346 (81)

n.d. n.d.

489387418 (90) 474 (95) 501 (95) 400 (94)

673 483 465 498 527 428

48 106 61 54

Note: Figures in parentheses are percentages of Total Found. Two of the values for "Published Pregnancy-Associated" deaths involve guesswork. The 61 cases for 1930 assumes that influenza-pregnancy deaths numbered 8, as stated in the provincial annual report, not 13, as stated in Vital Statistics. The latter statistic apparently came from an invalid summing of two statistics in the provincial report (the 8 in the table for pregnancy-associated deaths and the 5 reported for the influenza subcategory llf, which the table data should have included). The 106 cases reported for the MacMurchy coverage period is an estimate. MacMurchy's study covers the last half of 1925 and the first half of 1926, and the published statistics for pregnancyassociated deaths do not show half-yearly distributions. The estimate assumes that, for each year, the pregnancy-associated deaths for each category (e.g., tuberculosis) have the same half-year distribution as all female deaths in the category. Vital Statistics and the provincial annual report (influenza subcategory llf) give different totals for influenza-pregnancy deaths. The estimate uses the data from Vital Statistics.

influenza caused the large shortfall for 1920; a minor influenza epidemic in early 1926 added to the shortfall for MacMurchy's study period; and an unusually low incidence of infectious disease explained the small shortfall for 1930. The reportage of maternal deaths in published statistics probably improved between 1920 and 1935. The publicity on maternal mortality must have helped. As the Canadian Welfare Council also speculated, the long-term practice of returning certificates for further information doubtless "had a cumulative effect in inducing medical practitioners to mention a recent pregnancy wherever this occurred. Such cumulative effect would now tend to reduce the gap between the official maternal mortality rate and a rate which would be obtained by a complete investigation such as the Departmental Study made by Dr. MacMurchy in 1925-26."37 The Table 6.7 statistics also suggest improved coverage. For example, the completeness of "Published Data from All Sources" moved from 90 per cent in 1922 to 945 per cent for later years.

134

Facts of Life

T H E P U B L I S H E D S T A T I S T I C S A F T E R 1935

The province's reporting of maternal deaths certainly became more complete after 1935, if one includes the table totals for pregnancyassociated deaths. Some help came from a 1934 Ontario Public Health Act amendment, which enabled the provincial Health Board to require completed questionnaires from "every physician, medical officer, superintendent of a hospital, nurse, midwife or other person in charge of a maternity case in which the death of a mother takes place from causes directly or indirectly associated with pregnancy or parturition." Whereas 75 per cent of physicians had returned the Phair/Sellers questionnaires for 1933, the response rate was 93 and 96 per cent respectively for 1934 and 1935 and still higher for 1936.38 A more important influence was a new death-certificate form that the province introduced in mid-1935 (see Appendix B). Compared to the old form, the 1935 issue was far more precise about the information required for cause of death, and it asked the valuable supplementary question, "If a woman, was death associated with pregnancy?" To support the new form, the Dominion Bureau of Statistics issued a revised edition of the physician's Handbook on Death Registration and Certification in 1936, and in 1937 the Canadian Public Health Association distributed an exercise in death certification for teaching in medical schools.39 Detailed reports from seven provinces in 1936 endorsed the new death certificate as a decided improvement and indicated that fewer supplementary inquiries were necessary. In line with English experience after 1927, physician reporting errors were numerous when the new certificate was introduced, but became fewer after physicians became accustomed to it. In 1935 a Canadian Public Health Association committee found that 27 per cent of 898 Toronto certificates had been filled out incorrectly (e.g., entry of more than one cause on a line; causes listed in incorrect order), but the proportion dropped to 21 per cent for 980 Toronto certificates examined two years later.40 Despite the improvement, Ontario published statistics probably never reported all maternal deaths in a given year. Three United States studies for the 1974-81 period, for example, found that Washington state, Georgia, and New Jersey death registrations did not report the maternal cause for 25 to 50 per cent of maternal deaths.41 Second, the Ontario statistics undoubtedly harboured social bias. In the United States, for example, the physician was less likely to report multiple causes if the decedent was "black."42 Similarly, through investigation of 5,969 death certificates from the Quebec City area for 1935-36, a committee of the Canadian Public Health Association

135 Fatal Pregnancies, 1920-35 found that the physician's information was incorrect for 22.2 per cent, and deliberately so for 19.5 per cent. In some cases the physician met the request of relatives, but more often his concern was to keep the actual cause from becoming public knowledge.43 As the chief investigator explained, the problem was Canada's "open" certification: As soon as a death has occurred, the undertaker ... is called to take charge of the interment. The undertaker is responsible to the local "collector" [priest] for having the attending physician deliver the medical certificate. First, he goes to the collector and obtains from him the death form, of which the collector fills in the personal section. The undertaker then calls on the attending physician, who makes out the medical statement. From this moment the medical certificate is seen by quite a number of persons. First, it is the undertaker himself, then his assistants and sometimes his family, for the death form remains in his hands until the interment takes place. Before the burial, the death form is returned to the local collector who must keep it until his next communication with our office. As there are more than 1,800 collectors in the Province, we cannot have them sending us each death form as soon as remitted to them. Hence, each collector sends in his returns at the end of the month. Therefore a death form may remain in the office of the collector from a single day to a full month. During this time, the collector himself, his housekeeper, the caretaker of the church, as well as any person visiting the priest or the minister, may have the opportunity to see the form. In large municipalities, this may not result in a general knowledge of the cause of death, but in the smaller communities every one may come to know from what disease a certain person died. Such a situation being known to physicians, there is little wonder that, upon occasion, they feel the desire to conceal the true cause of death, and that they now demand the closed or confidential certificate. England had secured "virtual confidentiality" in 1926 by requiring the physician to deliver the certificate directly to the registrar. The Quebec solution, introduced in 1941 at the request of 92.3 per cent of the province's physicians, was the "confidential death certificate" - a registration form with a flap that the physician could fold over the death certificate and seal. The innovation worked well in Quebec. In 1942, however, the Canadian Public Health Association committee deferred consideration of its extension to other provinces until the conclusion of the war, and the question was not revived. In 1948 a new Ontario Registration Act simply made the funeral-home director responsible for gathering the required information from the relative-informant and physician and delivering it to the local registrar.

136 Facts of Life CONCLUSION One should interpret published statistics for deaths by cause in light of their historical construction. First, the information in the death certificates reflected influences such as the physician's concept of disease, the design of the death certificate, and the concealment of causes that the public stigmatized. Second, the published statistics were organized around the historical concept of an underlying cause. Thus, where the death certificate reported two or more causes, officials selected only one cause for classification. To this end they used manuals of the International List, which were revised periodically, and "rules for choice," which changed over time and varied internationally. During the study period, pregnancy caused far fewer deaths than tuberculosis, diseases of childhood and infancy, or epidemics of acute infectious diseases. Yet the puerperal deaths showed rather well the limitations of the published statistics. As shown above, provincial totals for the puerperal-state class of deaths missed 17 to 27 per cent of the total-mentions count for the province's pregnancy-related deaths, in some cases because the death certificate did not report a maternal cause and in others because the puerperal cause did not rate for classification under Canada's "rules for choice." The shortfall in the published counts probably declined towards 1935, due to improved physician reporting of puerperal causes and a decline in the prevalence of acute infectious disease. For the period 1922-51 published tables for pregnancy-associated deaths captured most of the deaths that the puerperal-state class statistics miss. Officials did not compile such tables, however, for other classes of causes.

7 Whose "Facts of Life"? The Problem of Residence in Vital Statistics, 1920-43

As Ontario civil registration improved in terms of its coverage and quality, its reporting of events by place of occurrence became problematic. With the diffusion of automobile transportation and the emergence of medical hospitals as central sites for birth and death, growing numbers of Ontarians gave birth or died outside their municipalities of residence. Thus the vital events occurring in a locality became somewhat different from the vital events involving its residents - the information required for calculation of meaningful vital rates. Officials worked gradually towards the 1911 English solution to the residence problem - the reallocation of registrations to the municipalities whose residents were having the vital events. The solution may seem straightforward, but it entailed several steps: i) the introduction of questions on the registration forms to secure reliable information about residence; ii) the routine exchange of registration copies among different provincial and national jurisdictions for vital events involving each other's residents; and iii) standard definitions of residence for inmates of public institutions (e.g., prisons, medical hospitals, asylums, tuberculosis sanatoria). Because the residence problem developed during the national era of civil registration and involved exchanges among provincial jurisdictions, it received a national solution. In 1920 the Dominion Bureau of Statistics introduced questions about residence on the registration forms. Beginning in 1924 it published statistics for non-resident events in selected jurisdictions. During the 19305 it prepared five

138

Facts of Life

special reports on births and deaths by place of residence for the period 1930-36. It introduced routine reallocation for deaths in 1935 and for births in 1936. Finally, in 1944 the bureau made reportage by residence the basis for Canada's published vital statistics.1 Henceforth the statistics for each province or municipality excluded nonresident events occurring within the jurisdiction, but included the events of residents that had happened elsewhere. To elaborate the residence problem in provincial birth and death statistics, this chapter discusses how hospitals and motor-vehicle transportation made it an issue during the quarter-century before 1944.2 Second, it examines the official response to the problem, including the information collected in the registrations and compiled for published statistics; the adoption of standard definitions of residence and how these affected the published statistics; and efforts of officials to upgrade the quality of the information reported so that it was adequate for statistical purposes. Third, it uses the evidence officials collected to probe quantitative and spatial dimensions of the residence problem. It aims, in other words, to give a general answer to the question, Whose "facts of life" were being registered? THE EMERGENCE OF THE PROBLEM

As registration officials discovered during the 19305, "nearly all transfers [from place of occurrence to place to residence] are not between provinces or states but between municipalities of the same province or state."3 The problem of residence, in other words, acted primarily on statistics at the municipal or county level. Several influences were at work, but two stand out. The first was the proliferation of motor-vehicle ownership and the elaboration of the provincial road system. As shown above in Figure 4.2, the number of licensed motor vehicles per 1,000 Ontario population rose from 24 in 1916 to 186 in 1930, before levelling off during the 19305 depression.4 The automobile, an American study notes, transformed the spatial organization of health care.5 Compared to horse and buggy transport, it delivered comfort and speed, started instantly, and could be left untended in foul weather. Relative to rail and electric radial transport, it offered greater flexibility for time of travel, routes, and destinations.6 A general effect was to tie rural and suburban populations to urban health-care resources. On the one hand, rural populations had readier access to urban physicians and hospitals. On the other, country doctors lost their local monopolies, which had been based on proximity to their clients; the upshot was a growing concentration of physicians

139 Whose Facts of Life?

in central urban places. At the same time, improved rural access to urban hospitals reduced the need to maintain rural hospitals. Finally, automobile transport facilitated movement to city suburbs. It allowed people to live in suburbs without sacrificing access to urban hospitals and physicians, and it let them commute to workplaces in the city. Thus a 1931 census monograph noted "a tremendous expansion in the residential areas of the leading cities, not generally accompanied by any increase in the municipal areas. This development is in large part a consequence of the advent of cheap and rapid transportation of the people from residence to place of business, resulting from the enormous increase in motor buses and particularly private automobiles. Large numbers of people who work in the cities are able to live in comparative quietness outside the city limits, driving themselves and their neighbours to and from work."7 The second major influence was the transformation of the hospital in Ontario society and its emergence as a central site for birth and death. During the nineteenth century, hospitals had been charity wards for the poor. During the early twentieth century they were becoming controlled environments for technological medicine that served a predominantly middle-class clientele on a fee-for-service basis.8 In the process their numbers increased, yet remained concentrated in urban locations (see Appendix G). At the same time, with the diffusion of automobile transportation, a hospital's catchment area became wider than the municipality in which it was located. Consequently, as shown in Figure 7.1 for the period 1926-50, the proportion of registered vital events occurring in "public institutions" (hospitals, asylums, sanatoria, city refuges) moved from 26 to 91 per cent for births and from 28 to 49 per cent for deaths.9 For the same reason, growing numbers of Ontarians gave birth or died outside their home municipalities. Thus in 1931 a Toronto Health Department official noted the influence of hospital locations on the vital statistics for his city: Toronto comprises one of the largest hospital centres in Canada. Of the 7,017 deaths which occurred in this city from all causes in 1930, 1,060 or 15.1 per cent were those occurring among non-residents. Non-resident births were in excess of 1,600, or over 12 per cent of the total number. On account of Toronto's excellent facilities for the isolation and treatment of acute communicable diseases, one-quarter of all deaths occurring in 1930 were those of non-residents ... Toronto's hospital facilities attract many maternity cases from outside the city. The resulting fatalities among these mothers constituted 30 per cent, or 27 of the 90 deaths which occurred from puerperal causes in 1930 ...

140 Facts of Life

Figure 7.1 Registered Live Births and Deaths in Public Institutions

On the other hand, a general hospital is situated immediately beyond the northern limits of the eastern section of the city. During 1930, 455 or 74 per cent of the total births, and 119 or 66 per cent of the total deaths which took place in this hospital occurred among residents of the city of Toronto. These events, of course, are not registered at the place of residence. In outside sanatoria, 115 Toronto residents died from tuberculosis. This is merely the number of which we have knowledge, due to the fact that the city contributes to the maintenance of its indigent residents in outside sanatoria.10 THE GATHERING OF INFORMATION ABOUT RESIDENCE

To launch the national system in 1920, the Dominion Bureau of Statistics introduced new forms for registering births and deaths and, in the process, questions about residence.11 The birth form requested the mother's "usual residence," which was the basis for the

141 Whose Facts of Life?

reallocation of births. By contrast, the death form merely asked for the decedent's "length of residence" at "place of death," "in Ontario," and "in Canada." In 1930, however, the bureau added a question about the decedent's "usual residence," the basis for the reallocation of deaths.12 The bureau made only minor changes on issuing a new form for deaths in 1935. It substituted a more precise question, "length of stay in city, town, or township where death occurred," for the old question about "length of residence at place of death," and it added a footnote to the "usual residence" question to specify that "post office address for residents in rural parts [was] not sufficient."13 To summarize, the form for registering a birth introduced a satisfactory question about residence in 1920, but the form for registering a death did not do so until 1930, when a "usual residence" question was added. Until then officials had to infer residence status from an arbitrary, unspecified interval for "length of residence at place of death," a dubious procedure, especially for hospital deaths. Even with the more precisely worded question introduced in 1935 ("length of stay in the municipality where death occurred"), the reply "sometimes refers to a former period when the decedent was a resident of the place of death. It is not unusual to find the 'length of stay' given as 'life' although another place than the place at which death occurred is stated as the residence."14 Hence the bureau ignored the information, except for non-institutional deaths with no information about "usual residence." In these cases it assumed that the decedent was a resident if the reported stay was a year or more; otherwise it returned the certificate for more information. The Dominion Bureau of Statistics, the registrars-general of the provinces, and urban municipalities with organized health departments each published summary statistics in annual reports.15 For convenience, the writer refers exclusively to Vital Statistics, the annual report of the Dominion Bureau of Statistics. As noted in chapter 2, the bureau compiled the statistics for the provinces, and its report contained most of the statistics issued in the municipal and provincial reports. For 1924 and 1925 only, Vital Statistics reported non-resident deaths for cities and towns of 1,000 or more population. Beginning in 1926 it reported non-resident births, infant deaths, and deaths for provinces, cities, and towns of 5,000 or more population; it also reported municipal non-resident subtotals for "resident in province" and "nonresident in province." Beginning in 1934 it reported non-resident stillbirths for provinces. In the meantime, to "take advantage of population data rendered available by the Census of 1931," bureau officials prepared five special reports on births and deaths by place of residence. In the process

142 Facts of Life

they learned the extent of the residence problem in birth and death statistics, found weaknesses in the registration forms for the collection of information, experimented with definitions and rulings, and prepared the ground for routine reallocation by residence. Special reports were issued for the following: i) births in Canada according to place of residence of mother, 1930-32 (1935); ii) mortality from tuberculosis in Canada according to place of residence, 193032 (1935); iii) mortality in Canada according to place of residence, 1930-32 (1934); iv) deaths in Canada classified according to residence of decedents, 1935 (2 vols., 1938); and v) births and deaths in Canada classified according to residence of decedents, 1936 (3 vols., 1939). For each province, city, town, and "rural parts of county" these studies reported the vital events by place of occurrence, the nonresident events in the occurrence total, the events of residents that had occurred "elsewhere," and the total for residents (the occurrence total, less the non-resident events, plus the resident events that had occurred elsewhere). The reports for 1930-32 also gave totals for counties inclusive of cities and towns. Thus for these years, but not for 1935-36, one can determine the proportions of municipal transfers that were within the home county. Finally, the 1935-36 special reports reported residence totals for municipal deaths by cause; for comparison, Vital Statistics gave the occurrence totals. THE I N T R O D U C T I O N OF STANDARD DEFINITIONS AND THEIR EFFECT ON PUBLISHED STATISTICS

The "usual place of abode" census definition was the model for published vital statistics. Except for medical hospitals, the census definition classed inmates of a "public institution" as residents of the municipality in which the institution was located. This affected death statistics more than birth statistics. During the 19305, only six births per year occurred in provincial tuberculosis sanatoria, and the provincial asylum inspector did not bother to report them.16 By contrast, 5 per cent of provincial deaths occurred in public institutions other than medical hospitals.17 The Special Reports. The bureau used the census definition in the special report on births for 1930-32. In the two special reports on tuberculosis deaths and all deaths for 1930-32, however, the bureau classed inmates of tuberculosis sanatoria as residents of the municipalities from which they had entered their institutions. The special reports on deaths for 1935 and births and deaths for 1936 extended this treatment to inmates of "asylums or hospitals for the insane,

143 Whose Facts of Life? Table 7.1 Non-resident Births and Deaths Reported for Ontario in Vital Statistics and the DBS Special Reports, 1930-36 Births by Occurrence: Both Sources 71,263 69,209 66,842 63,069 62,451

Nonresident Births: Vital Statistics

767 573 440

-

357

Difference from Special Report

+ 385 + 263 + 189 + 99

Year

Deaths by Occurrence: Both Sources

Nonresident Births: Vital Statistics

1930 1931 1932 1935 1936

37,313 35,705 36,469 36,317 37,571

491 403 346 369 374

Difference from Special Report

+ 242 + 213 + 188 + 203 + 195

institutions for the feebleminded, hospitals for epileptics and homes for incurables/' and also pupils dying in a boarding school.18 For other public institutions (e.g., refuges, orphanages, children's aid shelters, prisons, reformatories) the strict census definition still applied. In a second departure from the strict census definition, the special reports classed residents of foreign countries "as residents of the place in which birth or death occurred."19 Bureau publications did not discuss the rule, but its likely purpose was to compensate for the bureau's lack of reliable statistics for Canadian births and deaths occurring outside the country.20 If so, the special reports had a second, unstated characteristic: their statistics for "residents who gave birth or died elsewhere" excluded vital events that had occurred in foreign jurisdictions. The Definition in Vital Statistics. In 1944, when reportage by residence became the basis for Canada's published vital statistics, the Dominion Bureau of Statistics formally jettisoned two of its special reports rules for residence. Henceforth Vital Statistics counted: i) "births and deaths of Canadian residents occurring in the United States" in the totals for resident events; and ii) "births and deaths of United States residents ... and persons resident in other countries" in totals for non-resident events.21 Acting on these changes, Vital Statistics for 1944 introduced subtotals for events of American residents occurring in Canada and events of Canadian residents that occurred in the United States. Published materials do not state the Vital Statistics rules for earlier years, but one can infer them from statistics for non-resident events. As shown in Table 7. i for the period 1930-36, the non-resident totals are higher in Vital Statistics than in the special reports. Apparently

144 Facts of Life Table 7.2 Non-resident Vital Events per 1,000 Registered Vital Events

Year

1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941

Vital Stats. Births

Vital Stats. Deaths

10.8

13.2 11.3

8.3 6.6 6.7 5.3 5.1 5.7 4.6

10.2 10.0 --4.8 3.2

4.9 5.7 8.8

4.9 5.7

9.7 12.2 12.3 10.5

9.2 8.9 9.5 10.2 10.7

Special Report Deaths

Vital Stats. Surplus Births

Vital Stats. Surplus Deaths

5.4 4.5 3.8

6.7 5.3 4.3

+ 5.4 + 3.8 + 2.8

+ 6.5 + 6.0 + 5.2

4.6

-

4.1

4.8

+ 1.6 Births -1.1

+ 5.6 + 5.2 Deaths -4.8