Transition From School To Work : A European Perspective 9781845446406, 9780861767397

Citation preview

International Journal of Manpower

ISSN 0143-7720 Volume 23 Number 5 2002

The transition from school to work: a European perspective Guest Editor Stefan C. Wolter Paper format International Journal of Manpower includes eight issues in traditional paper format. The contents of this issue are detailed below.

Internet Online Publishing with Archive, Active Reference Linking, Emerald WIRE, Key Readings, Research Register, Institution-wide Licence, E-mail Alerting Service and Usage Statistics. Access via the Emerald Web site: http://www.emeraldinsight.com/ft See overleaf for full details of subscriber entitlements.

Access to International Journal of Manpower online 386 Editorial advisory board ___________________________ 387 Abstracts and keywords ___________________________ 388 Introduction Stefan C. Wolter ________________________________________________

390

Successful apprenticeship-to-work transitions: on the long-term change in significance of the German school-leaving certificate Felix Bu¨chel ____________________________________________________

394

The transition from apprenticeship training to work Wolfgang Franz and Volker Zimmermann ___________________________

411

School-to-work transition: apprenticeship versus vocational school in France Liliane Bonnal, Sylvie Mendes and Catherine Sofer ____________________

426

Participation in higher education: the role of cost and return expectations Charlotte Lauer _________________________________________________

443

Labour market expectations of Swiss university students Stefan C. Wolter and Andre´ Zbinden _______________________________

458

Technical/professional versus general education, labor market networks and labor market outcomes David N. Margolis and Ve´ronique Simonnet__________________________

471

Book reviews______________________________________ 493 About the authors _________________________________ 499

This issue is part of a comprehensive multiple access information service

CONTENTS

International Journal of Manpower 23,5 386

International Journal of Manpower online An advanced knowledge resource for the entire organization Access via the Emerald Web site – http://www.emeraldinsight.com/ft

Subscribers to this journal benefit from access to a fully searchable knowledge resource stretching far beyond the current volume and issue. International Journal of Manpower online is enhanced with a wealth of features to meet the need for fast, effortless and instant access to the core body of knowledge. Furthermore, this user-friendly electronic library may be networked throughout the subscribing organization to maximize the use and value of the subscription. This is augmented with advanced search facilities and ‘‘choice of access’’ through a variety of alternative journal gateways.

Emerald online access includes: Institution-wide Licence Our liberal licence allows everyone within your institution to access the journals via the Internet, making your subscription more cost-effective. It has been designed to provide a comprehensive, simple system with minimum administration. Access can be granted by IP range or username and password. ATHENS authentication is enabled.

Research Register Research Registers are Internet-based databases where you can identify inside information on research activity worldwide. Whether you are seeking information or you wish to provide details of your own research, you will find this feature invaluable.

Emerald Alert The table of contents e-mail alerting service will e-mail you the contents page of any chosen journal whenever the latest issue becomes available online. For further information please go to http://www.emeraldinsight.com/ alerts

Support Resources A comprehensive range of resources is available online that help users learn how to effectively use online information resources and that helps information professionals market resources to their users. For further information please go to http://www.emeraldinsight.com/ support

Current Issue + Journal Archive Internet access via Emerald Fulltext to information as it becomes available online and to material from past volumes. Users can choose to browse the database for relevant articles or search using the powerful search engine provided.

Active Reference Linking Direct links from the journal article references to abstracts of the most influential articles cited. Where possible, this link is to the full text of the article.

Emerald WIRE A fully searchable subject specific database, brought to you by Emerald Management Reviews, providing article abstracts from the world’s top management, computing and engineering journals. Directly relating to the journal subject area, the database is updated monthly and gives users details of how to obtain the full text of original articles.

Usage Statistics Online Journal Usage Statistics are now available. This feature allows Emerald Administrators to download their usage statistics with regard to their organization’s journal usage. Usage Statistics allow you to review the value of electronic dissemination of your journal subscriptions throughout your organization. They can also help determine the future trends for information within your organization. For further information go to http://www.emeraldinsight.com/stats

Key Readings Abstracts of articles, relating to keywords, are selected to provide readers with current awareness of interesting articles from other publications in the field. The abstracts are available online and are updated regularly upon publication of the latest issue.

Choice of Access Electronic access to this journal is available via a number of channels, enabling end users’ libraries to reach the content through their preferred delivery system. The Emerald Web site – http://www.emeraldinsight.com/ft – is the recommended means of electronic access as it provides fully searchable and high value-added access to the complete content of the journal. Refer to the next section for ‘‘how to access’’ via the Emerald Web site. Subscribers can also access and search the article content of this journal through the following services: Ebsco Online http://www.ebsco.com/online Huber E-Journals http://e-journals.hanshuber.com/english Minerva http://www.minerva.at/ OCLC Firstsearch ‘‘Electronic Collections Online’’ http://www.uk.oclc.org/oclc/menu/eco.htm RoweCom’s ‘‘Information Quest’’ http://www.informationquest.com SilverPlatter http://www.silverplatter.com SwetsBlackwell’s ‘‘SwetsNetNavigator’’ http://www.swetsnetnavigator.nl

How to access this journal through Emerald Organizations must first register for online access (instructions provided at http://www.emeraldinsight. com/register), after which the content is available to everyone within the organization’s domain. To access this journal’s content, simply log on either from the journal homepage or direct through the Emerald Web site. Emerald Customer Support Services For customer service and technical help, contact: E-mail: [email protected] Telephone: (44) 1274 785278 Fax: (44) 1274 785204

EDITORIAL ADVISORY BOARD Professor David J. Bartholomew School of Management, University of Liverpool, UK Professor Derek Bosworth Manchester School of Management, UMIST, UK Professor Martin Carnoy School of Education, Stanford University, USA Professor Peter Dawkins Melbourne Institute for Applied Economic and Social Research, Melbourne University, Australia

Professor Lord Richard Layard Centre for Economic Performance, London School of Economics, UK Professor John Mangan University of Queensland, Brisbane, Australia Professor Stephen L. Mangum Ohio State University, Ohio, USA

Professor John Fyfe W.S. Atkins plc, Epsom, UK

Professor Barrie Pettman International Management Centres, UK, and Founding Editor of International Journal of Manpower

Professor Morley Gunderson University of Toronto, Canada

Professor Abraham (Rami) Sagie Bar-Ilan University, Israel

Professor Thomas J. Hyclak Lehigh University, Bethlehem, USA

Professor David Sapsford Management School, Lancaster University, UK

Professor Susan E. Jackson Rutgers University, New Jersey, USA

Professor P.J. Sloane University of Aberdeen, Aberdeen, Scotland Professor Klaus F. Zimmerman Department of Economics, University of Bonn, Germany

Professor Harish C. Jain McMaster University, Canada Professor Geraint Johnes Management School, Lancaster University, UK

Editorial advisory board

387

International Journal of Manpower, Vol. 23 No. 5, 2002, p. 387. # MCB University Press, 0143-7720

International Journal of Manpower 23,5 388

International Journal of Manpower, Vol. 23 No. 5, 2002, Abstracts and keywords. # MCB UP Limited, 0143-7720

Successful apprenticeship-to-work transitions: on the long-term change in significance of the German schoolleaving certificate Felix Bu¨chel Keywords Labour market, Apprenticeship, Germany The quality of labor-market entry achieved by newly qualified apprentices in West Germany from 1948 to 1992 is analyzed. A bivariate probit model, using data from the BIBB/IAB employment survey, is applied to estimate simultaneously the quality of the school-toapprenticeship transition and that of the apprenticeship-to-work transition. This shows that school leavers with lower levels of general education are selected into apprenticeships with less favorable employment prospects in all analyzed time periods. However, when controlling for this selection effect, it is only in the most recent period that lower academic achievers are further penalized for the shortcomings in their general education at the apprenticeshipto-work transition. Furthermore, the crowdingout of trainees with lower levels of general education can be observed in both the less demanding and the more challenging occupational fields.

his or her decision to stay or to leave on considerations such as experimenting with several jobs.

The transition from apprenticeship training to work Wolfgang Franz and Volker Zimmermann Keywords Training, Apprenticeship, Young people, Unemployment This econometric study deals with the question as to what extent apprentices, after successfully completing their training, stay with the firm that supplied their training and, if so, how long that job tenure holds. Determinants of both decisions can be seen from both the employer’s and the employee’s viewpoint. These firms are interested in employing apprentices in order to collect the returns from their investment in their training, which frequently is associated with net costs. On the other hand, the firms dismiss apprentices if training is viewed by themselves as a screening device or if apprentices are engaged in work for which, in terms of wages, they are too expensive afterwards. The young trained worker bases

Participation in higher education: the role of cost and return expectations Charlotte Lauer Keywords Higher education, Labour market, Germany This study applies to German data a model in which the decision to attend higher education depends on the ratio of marginal cost and marginal return expected from higher education. If this ratio is below a certain threshold, the individual will choose to participate in higher education. In a simulation exercise, the impact of selected variables on this threshold, and thus on the participation probability, is quantified. The results suggests the presence of financial constraints binding participation in higher education and that the participation decision responds to some extent to return expectations in terms of labour market outcome and to financial incentives such as student support.

School-to-work transition: apprenticeship versus vocational school in France Liliane Bonnal, Sylvie Mendes and Catherine Sofer Keywords School leavers, Work, Apprenticeship, Young people, Labour market, France There has recently been a strong drive to develop apprenticeship in France, as one means of decreasing youth unemployment. Our aim in this paper is to try to measure the ‘‘pure’’ within-firm training effect on schoolto-work transition. We address the problem of the transition to the first job, using a model of simultaneous maximum likelihood estimation of several probabilities and of the parameters of the probability density function linked to the exit from unemployment. We conclude that apprentices have a distinct advantage over those who attended vocational school. This effect is stronger when we correct for the negative selection bias associated with the choice of apprenticeship.

Labour market expectations of Swiss university students Stefan C. Wolter and Andre´ Zbinden Keywords Re-training, Education, Wages Labour market expectations and especially wage expectations are important determinants for individual schooling decisions. However, research on individual expectations of students is scarce. The paper presents the Swiss results of a survey that was conducted in ten European countries. Its main findings are that point estimates of wages after graduation are close to actual wages, whereas the expectations of the wage gain in the first ten years of professional experience exceed the actual wage gains significantly. We find that rates of return to education that are calculated on the basis of individual wage and cost expectations as well as individual time preferences can be explained partially by the seniority of students, the self-perception of their academic performance and their subjective job perspectives. Technical/professional versus general education, labor market networks and labor market outcomes David N. Margolis and Ve´ronique Simonnet Keywords Networks, Education, Labour market, School leavers, France Does the choice between a general and a technical/professional education determine

the quality of the labor market network that an individual will be able to exploit throughout his or her professional life? This paper examines the hypothesis that technical and professional tracks, because they involve fewer students who are in more regular contact with each other and focus on a common, relatively narrow subject, allow students to establish more effective networks to support them in their careers. We test whether the choice of educational track has an impact on the means by which jobs are obtained and on the time to the first job of at least six months, the percentage of time spent in employment later in the career and the earnings when employed later in the career in France. Our results suggest that the educational track determines the means of obtaining a job, but conditional on the manner in which the job was obtained, the track has no additional impact on the outcome variables for the first or later jobs. However, the link between technical/professional education and job obtainment via professional networks does not hold independent of the level of education. In particular, this effect seems pertinent only for students having obtained a professional or technical baccalaure´at (relative to a general baccalaure´at) or for students having obtained a degree from a ‘‘grande e´cole’’ or engineering school (relative to graduate-level university studies).

Abstracts and keywords

389

The research register for this journal is available at http://www.emeraldinsight.com/researchregisters

International Journal of Manpower 23,5 390

International Journal of Manpower, Vol. 23 No. 5, 2002, pp. 390-393. # MCB UP Limited, 0143-7720 DOI 10.1108/01437720210436019

The current issue and full text archive of this journal is available at http://www.emeraldinsight.com/0143-7720.htm

Introduction About the Guest Editor Stefan C. Wolter is Director of the Swiss Coordination Centre for Research in Education in Aarau, Switzerland. He has studied Psychology and Economics at the University of Berne and holds a PhD in Economics. He is also Lecturer at the University of Berne and heads the Centre for Research in Economics of Education at the same university. He is research fellow of the IZA Institute for the Study of Labor in Bonn and Governing Board Member of the CERI/OECD (Centre for Educational Research and Innovation).

The transition from school to work: a European perspective In general, a person’s employment and educational biographies do not always run smoothly and in only one direction. Interruptions often occur, or the individual may begin to move in different or new directions. This generally happens when two stages of life that are governed by different systems interface. Unquestionably the most important junction is that between the educational system and the labour market, but the educational system itself may also consist of different subsystems and levels that are both separated and joined by marked interfaces. The problems that a person experiences when moving from one system to another are not a cause for concern as long as they represent isolated instances and are not systemic. However, various phenomena show that problems can accumulate at such interfaces, and that such an accumulation is a clear sign that different segments of the educational system and the labour market are not as well adapted to one another as they should be. The most conspicuous example of such transition problems is unquestionably the high rate of youth unemployment in most of the industrialized nations: on average, it is more than twice as large as the respective country’s average rate of unemployment (Cedefop, 2001). In most cases, youth unemployment primarily affects young people who are looking for their first job, i.e. precisely those who are making their initial transition from the educational system to the job market. Another disturbing factor is that youth unemployment has stayed high even in those countries that were able to sharply reduce their general rate of unemployment during the 1990s. These observations raise the question of whether the educational system and the labour market are sufficiently in tune with one another, and whether it might be possible to facilitate the transition of young people into working life. It is precisely this question that the OECD focused on in its report, comparing the functioning and efficiency of the school-to-work transition in 14 of its member states (OECD, 2000). The study came to a number of conclusions, two of which are dealt with more closely in this special issue of the International Journal of Manpower. The first point is that countries with a strong apprenticeship tradition appear in general to be more successful at dealing with the school-to-work transition. Second, if we think of the long-term integration of young people in the job market, the school-to-work transition cannot be viewed as consisting only of that brief period during which the first job is sought. Over

the longer term, young people who have managed the transition to the highest levels of the educational system, the tertiary educational level, are unquestionably better equipped for life. Even if in many countries university graduates initially have trouble finding work, long-term studies show clearly that they rhave a lower risk of being unemployed during their employable life. The OECD report shed valuable light on the systemic relationships at the macro level. In doing so, it of course based its work on published research undertaken at the micro level. But the latter is often insufficiently developed to adequately explain the phenomena observed at the macro level. In this special issue all six articles analyse the school-to-work transition on the basis of microdata. Four articles use traditional outcome data to analyse the pathways of young people into their working life and two articles also use partially or fully subjective data to enhance our understanding of the behaviour of individual students. All the articles try to shed light on the question of transition from school-towork in three continental European countries (France, Germany and Switzerland). For a long time, most European countries did not offer micro data on a regular base (like the Longitudinal Survey of Youth (NLSY) in the USA) and therefore, detailed studies on the behaviour of young people at the time of entry into the labour market were rather sporadic and limited. At least since the 1980s, many European countries have developed special youth surveys, socioeconomic panels or even cohort studies that allow a much better microeconomic and microeconometric analysis. The studies presented in this issue try to fully exploit these new databases for a better understanding of the underlying processes that shape the individual educational and work biographies of the young people of today. Except for the article by Margolis and Simonnet, all articles in this special issue were presented at the annual conference of the Swiss Society for Research in Education in 2001, at which transitions within the schooling systems and between school and the labour market was the overall theme. In parallel another collection of papers, which were also presented at the same conference, is being published in the journal Education + Training (Vol. 44 No. 4/5, 2002). In comparison to the present issue of the International Journal of Manpower, those articles deal mainly with the topic of transitions in the Swiss context and with a disciplinary focus not only on the economic aspects of the topic but also on the sociological and psychological dimensions. The research by Charlotte Lauer on the transition into higher education in Germany, is described in papers published in both issues; the article appearing in Education + Training is a shorter, non-technical version of that presented in this issue. The present special issue of the International Journal of Manpower is divided into two thematic blocks. The first block deals with the concrete schoolto-work transition, focusing thereby on the nature of the apprenticeship system. The German contributions by Felix Buechel, and by Wolfgang Franz and Volker Zimmermann, examine the apprenticeship system in the context of a society in which most young people still participate in this form of secondary

Introduction

391

International Journal of Manpower 23,5 392

education; the French contributions by Liliane Bonnal, Sylvie Mendes and Catherine Sofer and by David Margolis and Ve´onique Simonnet look at it from the viewpoint of an educational system in which only a minority of young people have an apprenticeship. However, the French studies show that, looked at from a certain viewpoint, the apprenticeship system can also produce good results even in a country whose educational tradition has developed in a completely different direction. Thus a comparison of the findings of these four studies enables us to consider whether the apprenticeship system succeeds only in a specific context, or whether at least some of its characteristics are universal in nature. This is all the more important because an ever greater number of countries have recently begun seriously considering enacting reforms of their upper secondary educational systems. As part of this, they want to increasingly offer young people apprenticeship programmes as an alternative to purely academic studies. It has to be borne in mind, of course, that an apprenticeship is not suitable for all young people, and that an evaluation of the success or failure of this form of education has to also take into account the programme’s selection procedure. This is made especially clear in the paper by Buechel, and that by Bonnal Mendes and Sofer, as these authors began by modelling the school-to-apprenticeship transition in their analysis of the school-to-work transition. The second block of papers consists of two articles that deal indirectly with the school-to-work transition by examining the interaction between the labour market and educational choice. In contrast to the first four papers, which are concerned with the upper secondary level, the focus in this block is on the tertiary educational system. Charlotte Lauer examines how socio-economic variables influence whether young people in Germany decide to participate in tertiary education. She uses real labour market data observable when the research was done to include the future returns to education and the probability of employment in her model. This procedure is entirely justified if we assume that young people have rational expectations concerning the pay-off of education. But as the paper by Stefan Wolter and Andre´ Zbinden shows, such an assumption is only partially valid – their work examines students’ expectations relating to returns to education and concludes that in light of the wage increases received in their first year of working life, students tend to overestimate the relative wage advantage gained by studying. The paper also shows that students who expect to have an especially easy time making the transition to the labour market also anticipate having a significantly higher return to education than do their fellow students, i.e. they assume that they will enjoy a more or less persistent relative advantage. Both articles show that the labour market conditions anticipated for the school-to-work transition period have an impact even earlier in time, when the individual is making the transition from one educational level to another, and that the relationship between the educational system and the labour market is thus in fact of a reciprocal nature.

Individual transitions from the educational system to the labour market can take many different forms and can only be assessed correctly if viewed in the context of a person’s overall career. This makes the empirical analysis of transitions a highly complex matter. As a result, each individual study can only contribute a single stone to the mosaic of findings that makes up the overall picture. Despite this, it is clear that educational, economic and labour market policies cannot do without this research if we want to put in place a general framework that will guarantee successful educational and working lives for our citizens. My sincerest thanks go to the Editor of the IJM, Adrian Ziderman, whose enthusiastic and professional help was indispensable in the realisation of this issue. Stefan C. Wolter References Cedefop (2001), ‘‘The transition from education to working life’’, Key Data on Vocational Training in the European Union, Cedefop Reference Series, Cedefop, Luxembourg. OECD (2000), From Initial Education to Working Life: Making Transitions Work, OECD, Paris.

Introduction

393

The research register for this journal is available at http://www.emeraldinsight.com/researchregisters

International Journal of Manpower 23,5 394

The current issue and full text archive of this journal is available at http://www.emeraldinsight.com/0143-7720.htm

Successful apprenticeshipto-work transitions On the long-term change in significance of the German school-leaving certificate Felix Bu¨chel Max Planck Institute for Human Development, Berlin, Germany Keywords Labour market, Apprenticeship, Germany Abstract The quality of labor-market entry achieved by newly qualified apprentices in West Germany from 1948 to 1992 is analyzed. A bivariate probit model, using data from the BIBB/ IAB employment survey, is applied to estimate simultaneously the quality of the school-toapprenticeship transition and that of the apprenticeship-to-work transition. This shows that school leavers with lower levels of general education are selected into apprenticeships with less favorable employment prospects in all analyzed time periods. However, when controlling for this selection effect, it is only in the most recent period that lower academic achievers are further penalized for the shortcomings in their general education at the apprenticeship-to-work transition. Furthermore, the crowding-out of trainees with lower levels of general education can be observed in both the less demanding and the more challenging occupational fields.

International Journal of Manpower, Vol. 23 No. 5, 2002, pp. 394-410. # MCB UP Limited, 0143-7720 DOI 10.1108/01437720210436028

1. Background and research questions The German dual system of vocational education has long been regarded as an exceptionally successful institution. The system affords young people leaving the general education system the opportunity to learn a trade or profession in an apprenticeship lasting several years. The term ‘‘dual’’ reflects the specific nature of this form of training – in addition to the practical skills acquired in on-the-job training, students spend one or two days a week at vocational school, where they acquire the necessary theoretical background knowledge (for an overview of the institutional organization of this system, see Franz and Soskice, 1995; Mu¨ller et al., 1998, p. 144ff). The success of this type of training program for young people is confirmed by numerous indicators, including the youth unemployment rate in international comparison (OECD, 2000). This positive evaluation of the system is validated even by more complex indicators (see Winkelmann, 1996). Indeed, the altogether very favorable image of the German ‘‘dual system’’ has prompted a number of countries – including the USA (see Harhoff and Kane, 1997; Gitter and Scheuer, 1997; Hamilton and Hamilton, 1999) – to endeavor to adopt a similar approach. Most EU countries now offer corresponding forms of apprenticeship program, albeit with differing levels of priority (see Nı´ Cheallaigh, 1995). Recently, however, a growing body of opinion in Germany has been predicting a rather bleak future for the dual system of vocational education. Two secular trends present serious problems for the institution. First, job

requirements are increasing rapidly as a result of the accelerating technological progress. This raises the question of whether instructors – particularly those working in the vocational schools – are in a position to react quickly enough and adapt the contents of the training programs to the new conditions. Second, the question arises of whether the apprentices are at all capable of satisfying the constantly increasing job requirements. This central problem is further compounded by the sustained expansion of the German education system since the late 1960s. An increasing proportion of those school leavers who, by reason of their social background, would formerly have been most likely to take up an apprenticeship, now have access to higher education. This means that, compared to the past, there is negative selection of the school leavers entering apprenticeships. As yet, the effects of these two mutually dependent and intersecting secular trends on the efficiency of the dual system – measured in terms of the employment prospects of the apprentices it produces – have only been investigated in rather rudimentary form; only very few sociological studies are available (see Handl, 1996; Konietzka, 1999). In this presentation, the employment prospects of recently qualified trainees will be operationalized in a proxy approach, based on the quality of their labormarket entry. The questions to be addressed are as follows: Have the employment prospects of newly qualified trainees deteriorated over the course of time? Does this hold especially for trainees with a poor level of general education or trainees in particular occupational groups? Furthermore, does the pattern of results hold when taking into account that disproportionate numbers of young adults with a poor educational background have always been selected into apprenticeships with poor employment prospects? Finally, are potential changes in the occupational risks of trainees with low levels of general education to be observed in the more demanding or in the less challenging occupational fields? 2. Methodology Data and case selection The empirical analyses are based on data from the BIBB/IAB employment survey. This representative multi-wave data set was gathered by the Federal Institute for Vocational Training BIBB, Berlin, in collaboration with the Federal Employment Service’s Institute for Employment Research IAB, Nuremberg. The 1991/1992 wave that is used in this study contains retrospective information on the educational and occupational careers of approximately 34,000 residents of West and East Germany (for details of this database, see Jansen and Stooß, 1993, p. 7ff. and p. 163ff.). The present analysis focuses on West Germany only. The BIBB/IAB survey was restricted to members of the economically active population[1], with non-Germans being surveyed only if they had an adequate command of the German language. As the latter kind of restriction entails obvious methodological problems, nonGerman respondents are excluded from the present analyses.

Apprenticeshipto-work transitions 395

International Journal of Manpower 23,5 396

In the present study, the focus is on all those who have completed an apprenticeship. In the interests of homogeneity, the analyses are restricted to those who served their apprenticeship in the post-war era. Accordingly, only those who gained their apprenticeship qualification between 1948 and 1992 are included in the analyses. A few individuals who served an apprenticeship after completing a higher education degree are also excluded from the analyses. Phase of transition Irrespective of the particular training system, the course of the future career path is largely determined by the quality of labor-market entry (see, e.g., Inkmann et al., 1998). Accordingly, the phase of transition from the apprenticeship to the world of work has always been of particular interest to educational and labormarket researchers (Bu¨chel, 1994; Helberger et al., 1994; Bu¨chel and Neuba¨umer, forthcoming; for an international perspective, see Shavit and Mu¨ller, 1998; Gangl, 2000). A central advantage of this approach is that the effects of age and occupational experience are essentially neutralized, making it possible for cohort or period effects to be isolated. In the data set under analysis, respondents were asked the following question: ‘‘How did you fare immediately after your apprenticeship?’’ All respondents who belonged to the economically active population ‘‘immediately’’ after completing their vocational training (employed and unemployed respondents; the latter category including those who ‘‘only had a casual job’’) were taken into consideration. Definition of a successful apprenticeship-to-work transition Those respondents who stated that they were unemployed or only had a casual job immediately after completing their apprenticeship were classified as ‘‘unsuccessful.’’ The question of whether the dual system accomplishes its goal of preparing apprentices for specific occupations was also considered: those who stated that they ‘‘found a job immediately, but not one corresponding to (their) qualifications,’’ were classified as ‘‘unsuccessful’’ in terms of their – assumed – ambition to serve an apprenticeship that would later secure them an appropriate position in which their skills were fully utilized (for further information on the concept of overeducation, see Bu¨chel, 2001). The group deemed to be ‘‘successful’’ thus contained those who stated that they found a job corresponding to their qualifications immediately after completing their apprenticeship. Periods of observation To allow for the identification of cohort or period effects, the total period of observation (1948-1992) was subdivided into three shorter periods. This division was made on the basis of the economic conditions prevailing in West Germany in each of the periods. The initial post-war period (1948-1959) was characterized by normalization and decreasing unemployment figures. This was followed by a second phase (1960-1974), with a booming economy and low unemployment. The third and final phase (1975-1992) was marked by mounting economic difficulties and growing unemployment rates; at the same time, the effects of the imposed

expansion of the education system started to become apparent. All of the following analyses were conducted for each period separately. Analysis procedure In a first descriptive step, highly aggregated frequency distributions, broken down by time period only, were computed for the respondents’ employment status immediately after completing their apprenticeship (Table I). In a second descriptive step, this information was broken down according to school-leaving certificate and gender (Table II). For the sake of clarity, data on the schoolleaving certificate was condensed into two categories: ‘‘poor education’’ (Hauptschule certificate or no qualifications at all) and ‘‘good education’’ (Realschule certificate or above). Following this, the results presented in Table II were re-examined in a multivariate approach. For each period, binary probit models were used to identify factors impacting on the quality of the apprenticeship-to-work transition (Table III). Because systematic correlations were expected between the school-leaving certificate and gender, these two central variables were controlled in the form of interaction terms. Additional control variables include occupational field of the apprenticeship, the size of the firm providing the apprenticeship training, unemployment rate at the time the newly qualified apprentice entered the labor market, and others (see tables). In a next step, account was taken of the fact that gender and schooling correlate systematically with the choice of apprenticed trade/profession (see, e.g. Ku¨hn and Zinn, 1998). Using a bivariate probit model (cf. Greene, 2000, p. 849ff.), the analysis described above was replicated, this time controlling in a second equation for the relationship between gender, schooling, and the choice of an apprenticeship with ‘‘poor’’ employment prospects (Table IV)[2]. To this end, the occupations were first ordered according to the quality of the later apprenticeship-to-work transition, and the sample then split into two parts of similar size[3]. In addition, account was taken of information already available from Table III, namely that trainees who do not serve their apprenticeship in a

Apprenticeshipto-work transitions 397

Year of completing apprenticeship 1948-1959 1960-1974 1975-1992 First employment status after completing apprenticeship Adequate position Overqualified position Unemployed Total No. of cases

89 10 1 100 2,560

89 10 1 100 4,401

86 10 4 100 5,469

Table I. Transitions from completed apprenticeship to work, by time period and Notes: Sample consists of employed people with German citizenship first employment Source: Own calculations based on data from the BIBB/IAB Employment Survey (Jansen status, in percent (West and Stooß, 1993) Germany, 1948-1992)

Table II. Transitions from completed apprenticeship to work, by time period, gender, previous schooling, and first employment status, in percent (West Germany, 1948-1992) 825 559 4,401

10 8

324 145 2,560

90 92

10 13

2,087 930

100 100

90 87

1,627 464

13 9

87 91

100 100

100 100

100 100

Source: Own calculations based on data from the BIBB/IAB Employment Survey (Jansen and Stooß, 1993)

Notes: AD: adequate position (‘‘successful transition’’ in view of previous training) OV: overqualified position UE: unemployment (‘‘unsuccessful transition’’ in view of previous training) Sample consists of employed people with German citizenship

No. of cases Hauptschule certificate or lower Men Women Realschule certificate or higher Men Women Total

10 12

90 88

OV, UE

90 89

83 81

Year of completing apprenticeship 1960-1974 First employment status after completing apprenticeship Total AD OV, UE Total AD

1,368 1,553 5,469

1,765 783

10 11

17 19

OV, UE

1975-1992

398

Schooling prior to apprenticeship Hauptschule certificate or lower (‘‘poor schooling’’) Men Women Realschule certificate or higher (‘‘good schooling’’) Men Women

AD

1948-1959

100 100

100 100

Total

International Journal of Manpower 23,5

Variable Poor schooling, male Poor schooling, female (Good schooling, male) Good schooling, female Farmer, gardener, miner (Metalworker) Electrician Textile worker Food worker Construction worker Carpenter, painter Other manufacturers Technician Clerical worker Sales clerk Health care assistant Other service occupation Firm: < 10 employees (Firm: 10-1,000 employees) Firm: > 1,000 employees No firm (external training) Public sector Duration of training (years) Unemployment rate at end of training Constant Mean of dependent variable Likelihood ratio statistic N

Year of completing apprenticeship 1948-1959 1960-1974 1975-1992 Coeff. Coeff. Coeff. –0.239*** –0.283***

0.117 0.113

–0.009 –0.084

0.184 0.412*

0.045 0.151

0.158 –0.090 –0.094 0.345** 0.289** 0.219 0.272 0.104 0.011 0.282 0.324 –0.092

0.172 0.286 –0.125 0.474*** 0.164*** 0.742 0.326 0.284*** –0.042 0.455** 0.271* –0.090

0.099 –0.028 –0.013 0.198 0.079 0.025 0.077 0.200** 0.141 0.389*** 0.063 –0.128**

0.289* –0.273** 0.095 0.090 –0.019

0.168 –0.318*** 0.060 0.138*** –0.007

0.086 –0.303*** –0.028 0.064 –0.017*

0.760***

1.113***

0.915*** 0.894 36.6 2,557

0.896 85.0 4,395

Apprenticeshipto-work transitions

–0.113 0.194

399

0.862 80.8 5,466

Notes: Significance levels: ***(p < 0.01), **(p < 0.05), *(p < 0.10) Dependent variable: 1 = ‘‘successful’’ apprenticeship-to-work transition (immediate move to adequate position); 0 = ‘‘unsuccessful’’ transition (move to overqualified position or unemployment) Sample constists of employed people with German citizenship

Table III. Determinants of a successful apprenticeship-to-work transition, by time period (binary probit Source: Own calculations based on data from the BIBB/IAB Employment Survey (Jansen models, West Germany, and Strooß, 1993) 1948-1992)

firm authorized to provide training have systematically lower chances of realizing a successful transition to the world of work. Serving an apprenticeship in an authorized firm was thus taken as an additional criterion for an ‘‘apprenticeship with good employment prospects’’[4]. In a final step in the analysis, it was investigated whether or not the pattern of results presented in Table III for the total sample in each period holds when considering only occupations with ‘‘good’’ or ‘‘bad’’ employment prospects (Table V).

Table IV. Combined determinants of first entering an apprenticeship with good employement prospects and later realizing a successful apprenticeship-to-work transition, by time period (bivariate probit models (FIML), West Germany, 1948-1992)

Equation I: (Determinants of entering an apprenticeship with ‘‘good’’ employment prospects) Poor schooling, male Poor schooling, female (Good schooling, male) Good schooling, female Constant Equation II: (Determinants of later realizing a ‘‘successful’’ apprenticeship-to-work transition) Poor schooling, male Poor schooling, female (Good schooling, male) Good schooling, female Farmer, gardener, miner (Metalworker) Electrician Textile worker Food worker Construction worker Carpenter, painter Other manufacturers Technician Clerical worker Sales clerk Health care assistant Other service occupation –0.395*** –0.746*** (. ) 0.258*** 0.212***

–0.013 –0.090 (. ) 0.047 0.131 (. ) 0.151 0.286 –0.125 0.453* 0.143 0.722** 0.306 0.264** –0.042 0.436* 0.271

–0.190** –0.393*** (. ) 0.525*** 0.069

0.123 0.126 (. ) 0.169 0.482 (. ) 0.232 –0.091 –0.095 0.417 0.362** 0.219 0.341 0.174 0.011 0.349 0.395

Year of completing apprenticeship 1960-1974 Coefficient

400

Variable

1948-1959 Coefficient

–0.260*** –0.302*** (. ) –0.102 0.114 (. ) 0.019 –0.028 –0.011 0.118 0.078 0.023 0.002 0.122 0.140 0.316* 0.061 (continued)

–0.670*** –0.626*** (. ) 0.361*** 0.082**

1975-1992 Coefficient

International Journal of Manpower 23,5

4,395

2,557

5,466

0.336 0.862 –4,888.7

5,466

1.151*** 0.051

–0.127** (. ) 0.086 –0.254** –0.027 0.064 –0.017*

1975-1992 Coefficient

Notes: Own calculations based on data from the BIBB/IAB Employment Survey (Jansen and Strooß, 1993)

Notes: Significance levels: ***(p < 0.01), **(p < 0.05), *(p < 0.10) Dependent variable of Equation I: 1 = having completed a type of apprenticeship with ‘‘good’’ employment prospects; 0 = other apprenticeships with ‘‘good’’ employment prospects: in all three periods: primary sector occupations; electricians; construction workers; technicians; clerical workers in banks, insurance agencies, and public administration; health care occupations (in all cases: if apprenticeship served in company authorized to provide apprenticeship training). Additionally for 1948-1959 period: carpenters, painters, other service occupations. Additionally for 1960-1974 period: carpenters, painters, other manufacturers Dependent variable of Equation II: 1 = ‘‘successful’’ apprenticeship-to-work transition (immediate move to adequate position); 0 = ‘‘unsuccessful’’ transition (move to overqualified position or unemployment). Sample consists of employed people with German citizenship

N

0.399 0.896 –4,201.9

0.151 0.894 –1,911.5

4,395

2,557

0.771*** –0.013

–0.090 (. ) 0.168 –0.306** 0.060 0.138*** –0.007

Dep. mean of Equation I Dep. mean of Equation II Log-L

0.874*** 0.048

–0.093 (. ) 0.289* –0.310* 0.098 0.091 –0.019

Year of completing apprenticeship 1960-1974 Coefficient

N

Constant RHO (I, II)

Firm: < 10 employees (Firm: 10-1,000 employees) Firm: > 1,000 employees No firm (external training) Public sector Duration of training (years) Unemployment rate (end of training)

Variable

1948-1959 Coefficient

Apprenticeshipto-work transitions 401

Table IV.

Table V. Determinants of realizing a successful apprenticeship-to-work transition, by time period (selected types of apprenticeships, binary probit models, West Germany, 1948-1992) 0.212

Farmer, gardener, miner (Metalworker) Electrician Textile worker Food worker Construction worker Carpenter, painter Other manufacturers Technician Clerical worker Sales clerk Health care assistant Other service occupation Firm: < 10 employees (Firm: 10-1,000 employees) Firm: > 1,000 employees No firm (external training) Public sector Duration of training (years) Unemployment rate at end of training 0.107 0.078 –0.012

0.182 0.214 0.070 (. ) 0.337

0.349 (. )

0.335* 0.151

0.053

0.010 0.050 (. ) 0.161

Poor schooling, male Poor schooling, female (Good schooling, male) Good schooling, female

Variable

402

–0.165* (. ) 0.027 –0.165 0.144 0.078 –0.024

0.23

0.209

–0.073 –0.085

(. )

0.191 0.159 (. ) 0.179

0.113 0.104 –0.055

–0.139 (. ) 0.041

0.130

0.079 –0.164 0.327 0.090 (. )

–0.135

–0.005

0.065 –0.026 (. ) –0.006

0.186 –0.061 (. ) 0.376** –0.212** –0.029 –0.010 0.029

–0.156

0.159 –0.187

(. )

0.054 –0.119 (. ) 0.158

0.028 0.158** 0.009

–0.169** (. ) 0.106

0.124

–0.158 (. )

–0.050

–0.186

–0.28

–0.333*** –0.332*** (. ) –0.171+

–0.030 –0.083 (. ) 0.063 –0.78* –0.104 –0.004 –0.034*** (continued)

0.050

0.028 –0.004

–0.131 –0.059

(. )

–0.183** –0.210** (. ) –0.006

Year of completing apprenticeship 1948-1959 1960-1974 1975-1992 Quality of apprenticeship (indicated by employment prospects): Good coefficient Poor coefficient Good coefficient Poor coefficient Good coefficient Poor coefficient

International Journal of Manpower 23,5

1,376

0.873 17.2

0.955***

2,031

0.927 13.0

1.213***

2,370

0.868 29.4

1.245***

2,492

0.891 30.2

0.946***

2,977

0.837 26.2

1.409***

Source: Own calculations based on data from the BIBB/IAB Employment Survey (Jansen and Strooß, 1993)

Notes: Significance levels: ***(p < 0.01), **(p < 0.05), *(p < 0.10) Apprenticeships with ‘‘good’’ employment prospects: see footnote range in Table IV. Apprenticeships with ‘‘poor’’ employment prospects: all others Dependent variable: 1 = ‘‘successful’’ apprenticeship-to-work transition (immediate move to adequate position); 0 = ‘‘unsuccessful’’ transition (move to over qualified position or unemployment) Sample consists of employed people with German citizenship

1,184

0.918 8.6

Mean of dependent variable Likelihood ratio statistic

N

1.012**

Constant

Variable

Year of completing apprenticeship 1948-1959 1960-1974 1975-1992 Quality of apprenticeship (indicated by employment prospects): Good coefficient Poor coefficient Good coefficient Poor coefficient Good coefficient Poor coefficient

Apprenticeshipto-work transitions 403

Table V.

International Journal of Manpower 23,5 404

3. Empirical results Descriptive results Table I contains the employment status distribution of those respondents who belonged to the economically active population immediately after completing their apprenticeship. This first highly aggregated analysis is broken down by the year the apprenticeship was completed only. The results confirm that the German dual system of vocational education has operated at a consistently high level of efficiency over an extended period of time. Newly qualified trainees in search of work were almost entirely absorbed by the labor market throughout the entire post-war era. However, some of those who realized a seamless apprenticeship-to-work transition did have to accept a position for which they were overqualified (10 percent in all periods). On the general level, it can thus be concluded that the dual system does indeed achieve its goal of enabling young adults to enter the labor market. However, the system does not appear to live up to its more specific goal of providing young adults with vocational qualifications that can be fully exploited on the labor market – not immediately, at least. In this respect, a notable proportion of the apprenticeship-to-work transitions must be regarded as ‘‘unsuccessful.’’ Although the rate of unemployment among new entrants to the labor force has risen noticeably from one to four percentage points since 1975, it is still very low. In Table II, the two categories of respondents who were ‘‘unemployed’’ or in an ‘‘overqualified position’’ immediately after completing their apprenticeship are collapsed into a single category, and compared with those who realized a ‘‘successful’’ transition to an adequate position. In addition, the results are broken down according to gender and the quality of the school-leaving certificate acquired prior to the apprenticeship. Here again, the focus is on changes to be observed over time. Even when broken down according to schooling and gender, the results bear out a remarkable stability in the quality of the apprenticeship-to-work transition throughout the entire period of investigation. In almost all categories, around 90 percent of the trainees realized an optimal transition to the world of work. A noticeable exception can be discerned in the most recent period of observation (1975-1992), however, where there is a sharp drop in the success rate of those trainees with a poor level of general education (Hauptschule certificate or no school-leaving certificate at all), falling to just over 80 percent. Gender-specific differences are negligible here. Control for apprenticed trade and other characteristics In the next step, it was tested whether this deterioration in the employment prospects of trainees with low levels of schooling holds when controlling for the apprenticed trade/profession, the structure of the firm providing the apprenticeship training, and the labor-market situation at the time the apprenticeship was completed.

The results presented in Table III substantiate the descriptive findings: It was only in the most recent period of observation that a low level of schooling began to represent a risk factor at the apprenticeship-to-work transition. Men and women were hit equally by this development. In the time from 1948 to 1975, in contrast, the quality of general schooling acquired prior to the apprenticeship did not have any effect on the probability of realizing an optimal transition. It is striking that this result holds even when controlling for the diverse apprenticeships with varying employment prospects, and for the structure of the firms providing the apprenticeship training. With respect to the control variables, it emerges that – relative to the group of metalworking occupations – the initial period of post-war reconstruction in West Germany (1948-1959) was a particularly advantageous time for newly qualified apprentices in the primary sector, construction workers, carpenters, and painters. Other than this, the chances of realizing a seamless apprenticeship-to-work transition were more or less equal across the various trades and professions. In the boom period (1960-1974), occupation-specific differences in transition prospects became more clear cut as structural change in the labor market became apparent. The prospects of those in the metalworking occupations – still a broad occupational group – began to deteriorate. Although construction workers were still in demand, as were other manufacturers and technicians, there was now an increased demand for health care assistants and for clerical workers in banks, insurance companies, and public administration. Moreover, a longer period of apprenticeship now had a positive effect. In the most recent period of investigation, consolidation of the structural change can be observed. Only respondents in the more qualified, non-manual occupations had above-average chances of realizing a smooth apprenticeship-to-work transition. Moreover, serving an apprenticeship in a small firm with less than ten employees began to have a detrimental effect. For the first time, a relationship emerged between the prevailing economic conditions and the chances of successful labor-market entry; this effect was rather weak, however. The only common feature in all three periods of investigation is that trainees who did not complete their apprenticeship in a firm authorized to provide training (i.e. who attended ‘‘external training’’ in a public financed training center) had far less auspicious prospects than their counterparts who did serve their apprenticeship in such a company. This reflects the lack of practical on-the-job training of the latter group; a deficit which prevents these apprentices from accumulating company-specific human capital. This is worthy of note, as a large proportion of those trained in the dual system are hired by the firm they trained at after completing their apprenticeship. Control for selective admission to certain apprenticeships In the next step of the analysis, account was taken of the fact that the choice of apprenticed trade is largely determined by the level of general education attained by the school leavers competing for apprenticeships. Accordingly, it can be assumed that if we omit to control for this systematic selection at the school-to-apprenticeship transition (first threshold), the schooling effect yielded

Apprenticeshipto-work transitions 405

International Journal of Manpower 23,5 406

by the analysis of the apprenticeship-to-work transition (second threshold; see Table III) will be biased. The results of a bivariate probit model estimating the quality of transitions at both thresholds simultaneously dependent on the level of schooling are presented in Table IV. The results of Equation I (determinants of a successful school-toapprenticeship transition, operationalized in terms of selection into an apprenticeship with promising employment prospects; upper portion of Table IV) show that, in all three periods under investigation, whether or not an applicant was granted access to a ‘‘good’’ apprenticeship was largely determined by the quality of his or her general school-leaving certificate. Compared with the reference group of male school leavers holding at least a Realschule certificate, candidates from the Hauptschule – independent of their gender – were at a clear disadvantage when it came to choosing apprenticeship. In all three time periods, young women with a good level of general education had the best chances of being admitted to an apprenticeship with good employment prospects. Equation II yields the main findings of this step in the analysis (determinants of a successful apprenticeship-to-work transition, operationalized as before in terms of an immediate transition to an adequate position; lower portion of Table IV). The pattern of results presented in Table III remains unchanged even after controlling for what have been shown to be very strong selection effects in the allocation of candidates to apprenticeships. The results computed for the control variables are somewhat weaker than in the original model, but the general pattern holds. In other words, in the earlier periods of investigation (1948-1974), although lower-achieving school leavers were selected into less promising apprenticeships, a ‘‘poor’’ school-leaving certificate no longer had a detrimental effect at the second threshold – the apprenticeship-to-work transition. In the most recent period of observation, in contrast, the situation for lower school achievers worsened dramatically: if school leavers with low levels of general education were able to find an apprenticeship at all, it was – as always – likely to be in an occupation with poor employment prospects. What has changed, however, is that a low level of schooling now also had an adverse effect at the second threshold, the apprenticeship-to-work transition, even when controlling for the heterogeneity of the apprenticed trades/professions. What now needs to be explored is whether the crowding-out of trainees with lower levels of education occurs primarily in their traditional occupational sector or in the sector requiring higher levels of training. Transitions from different sorts of apprenticeships The results presented in Table V show that, until 1974, the quality of the schoolleaving certificate acquired prior to the apprenticeship had no significant effect on the respondents’ chances of realizing a successful apprenticeship-to-work transition in either the less demanding apprenticed trades or the more challenging professions. In the most recent period of investigation, however, there was a marked change in the pattern of results: compared with their peers with apprenticeships in comparable occupations but higher levels of general education,

Hauptschule leavers who atypically succeeded in completing an apprenticeship in an occupation with good employment prospects had significantly lower chances of later realizing a successful apprenticeship-to-work transition. In other words, the leveling-out effect of the apprenticeship qualification apparent in the first two periods of investigation has diminished. A somewhat weaker, but nonetheless significant effect can be observed in the sector of the less demanding apprenticed trades, where a higher school-leaving certificate still pays off in the long run for those with a good level of general education who atypically decide to serve this kind of apprenticeship. Here, again, former Hauptschule students of both genders are at an equal disadvantage. Where the control variables are concerned, a fairly even picture emerges in all models. With respect to the negative influence that the harsh economic situation in the most recent period of investigation was observed to have on the chances of realizing an optimal apprenticeship-to-work transition, however, it is worth noting that this development only hit those seeking jobs in the less demanding occupations, but that the effect here was highly significant. This, too, is an important result, confirming that the career prospects of less qualified applicants are more dependent on economic factors. Comparison of the magnitude and strength of the schooling effects determined for former Hauptschule students in the models estimated for apprenticeships with ‘‘good’’ and ‘‘poor’’ employment prospects in the most recent time period shows that the crowding-out of newly qualified apprentices with low levels of schooling is more widespread in the more challenging sector than in the less demanding occupations. This might imply that within this more challenging domain (e.g. health care occupations and clerical occupations in banks, insurance companies, and public administration) job requirements in terms of vocational education and work experience are constantly rising, and that it is increasingly difficult for Hauptschule graduates to meet these demands. However, it is also conceivable that job requirements have in fact remained constant over time, but that the continuing expansion of the education system has led to increasing negative selection in the Hauptschule. The fact that this crowding-out effect has even hit apprentices with lower levels of schooling in their traditional occupational sector of less demanding jobs – albeit to a lesser extent – should be a matter of concern for educational policy-makers, as it raises the question of which functional training routes, if any, are still open to Hauptschule students. 4. Concluding remarks From the perspective of human capital theory (Mincer, 1974), in the 1948-1974 period, the components of human capital acquired in the course of German apprenticeships (vocational education and work experience) proved to be more important than those acquired in general education. From the perspective of signaling theory (Spence, 1973), it can thus be concluded that the most recently acquired or highest-level qualification (namely the apprenticeship) was of particular importance in this period. The dual system was thus able to fulfill its objective of providing occupational certificates that afforded trainees equal

Apprenticeshipto-work transitions 407

International Journal of Manpower 23,5 408

opportunities on the labor market, irrespective of the level of general education acquired prior to the apprenticeship. Since 1975, this has no longer been the case. From the perspective of human capital theory, all components of human capital now seem to be of importance. The greater importance now attached to the components acquired in general education indicates that, in an increasingly complex world of work, employees are expected to have better cognitive qualifications, and that the dual system is no longer able to compensate for deficits in the domain of general education. From the perspective of signaling theory, all educational qualifications have now become relevant. In the course of the expansion of the education system, a poor school-leaving certificate has come to be regarded as the expression of negative selection; furthermore, this unfavorable signaling effect can no longer be completely offset by gaining a ‘‘good’’ apprenticeship in the dual system of vocational education. It remains only to consider the implications of these findings for educational policy. Failure to respond to the finding that former Hauptschule students are becoming significantly less competitive on the labor market would assume that these students and their parents are aware that the increasing negative selection of Hauptschule students resulting from the expansion of the German education system incurs serious risks for those who are not capable of keeping up with the trend to higher levels of education. One possible response to these findings would be to enhance the quality of instruction in Hauptschulen. Another option would be to introduce new forms of training to the dual system, geared to Hauptschule leavers – for example, by setting up shorter or less demanding training programs in occupations with good employment prospects. Though it is clear that putting these recommendations into practice will require heavy investment in terms of both resources and time, it is safe to assume that failing to respond will, in the long run, incur far greater costs. Notes 1. Restricting the parent population to economically active respondents constitutes a systematic selection of trainess that should be taken into account when interpreting the results. 2. The distribution of the covariates is documented in Table AI in the Appendix. 3. These results are not presented in table form; information on the classification of occupations is provided in the notes to Table IV. 4. Accordingly, the cases are not distributed equally between apprenticeships with ‘‘good’’ and ‘‘poor’’ employment prospects. References Bu¨chel, F. (1994), ‘‘Overqualification at the beginning of a non-academic working career – the efficiency of the German dual system under test’’, in Horn, G.-A. and Trabold, H. (Eds), ‘‘Globalisation and structural unemployment’’, special issue of Konjunkturpolitik, Vol. 40 No. 3-4, pp. 342-68. Bu¨chel, F. (2001), ‘‘Overqualification: reasons, measurement issues and typological affinity to unemployment’’, in Descy, P. and Tessaring, M. (Eds), Training in Europe, second report on vocational training research in Europe 2000. Background Report, Vol. 2, Cedefop, Luxembourg, pp. 453-560.

Bu¨chel, F. and Neuba¨umer, R. (forthcoming), ‘‘Individuelle Berufschancen als Folge branchenspezifischer Ausbildungsstrategien’’, in Mitteilungen aus der Arbeitsmarkt- und Berufsforschung. Franz, W. and Soskice, D. (1995), ‘‘The German apprenticeship system’’, in Buttler, F., Franz, W., Schettkat, R. and Soskice, D. (Eds), Institutional Frameworks and Labor Market Performance, Routledge, Nu¨rnberg, pp. 208-xxx. Gangl, M. (2000), ‘‘Changing labour markets and early career outcomes: labour market entry in Europe over the past decade’’, MZES Working Paper No. 26, Mannheim. Gitter, R.J. and Scheuer, M. (1997), ‘‘US and German youths: unemployment and the transition from school to work’’, Monthly Labor Review, Vol. 120 No. 3, pp. 16-20. Greene, W.H. (2000), Econometric Analysis, 4th ed., Prentice-Hall, Upper Saddle River, NJ. Hamilton, S.E. and Hamilton, M.A. (1999), ‘‘Creating new pathways to adulthood by adapting German apprenticeship in the United States’’, in Heinz, W.R. (Ed.), From Education to Work: Cross-National Perspectives, Cambridge University Press, Cambridge. Handl, J. (1996), ‘‘Hat sich die berufliche Wertigkeit der Bildungsabschlu¨sse in den achtziger Jahren verringert? Eine Analyse der abha¨ngig erwerbsta¨tigen, deutschen Berufsanfa¨nger auf der Basis von Mikrozensusergebnissen’’, Ko¨lner Zeitschrift fu¨r Soziologie und Sozialpsychologie, Vol. 48 No. 2, pp. 249-73. Harhoff, D. and Kane, T.J. (1997), ‘‘Is the German apprenticeship system a panacea for the US labor market?’’, Journal of Population Economics, Vol. 10 No. 2, pp. 171-96. Helberger, C., Rendtel, U. and Schwarze, J. (1994), ‘‘Labour market entry of young people analysed by a double threshold model’’, in Zapf, W., Schupp, J. and Habich, R. (Eds), Labour Market Dynamics in Present Day Germany, Campus Velag, Frankfurt am Main, pp. 142-64. Inkmann, J., Klotz, S. and Pohlmeier, W. (1998), ‘‘Growing into work – pseudo panel data evidence on labor market entrance in Germany’’, ZEW Discussion Paper 98-47, Mannheim. Jansen, R. and Stooß, F. (1993), ‘‘Qualifikation und Erwerbssituation im geeinten Deutschland’’, ¨ berblick u¨ber die Ergebnisse der BIBB/IAB-Erhebung 1991/1992, Bielfield, Berlin Ein U and Bonn. Konietzka, D. (1999), Ausbildung und Beruf. Die Geburtsjahrga¨nge 1919-1961 auf dem Weg von der Schule in das Erwerbsleben, Westdeutscher, Vetlag, Opladen. Ku¨hn, T. and Zinn, J. (1998), ‘‘Zur Differenzierung und Reproduktion sozialer Ungleichheit im Dualen System der Berufsausbildung. Quantitative und qualitative Ergebnisse einer Verlaufsuntersuchung u¨ber junge Erwachsene in sechs Ausbildungsberufen’’, in Heinz, W.R., Dressel, W., Blaschke, D. and Engelbrech, G. (Eds), Was pra¨gt Berufsbiographien? Beitra¨ge zur Arbeitsmarkt- und Berufsforschung, Vol. 213, IAB, Nurenburg, pp. 54-88. Mincer, J. (1974), Schooling, Experience, and Earnings, National Bureau of Economic Research, New York, NY. Mu¨ller, W., Steinmann, S. and Ell, R. (1998), ‘‘Education and labour-market entry in Germany’’ in Shavit, Y. and Mu¨ller, W. (Eds), From School to Work, Oxford University Press, Oxford, pp. 143-88. Nı´ Cheallaigh, M. (1995), Apprenticeship in the EU Member States – A Comparison, Cedefop, Luxembourg. OECD (Ed.) (2000), Education at a Glance: OECD Indicators 2000 Edition, OECD, Paris. Shavit, Y. and Mu¨ller, W. (Eds) (1998), From School to Work, Oxford University Press, Oxford. Spence, M. (1973), ‘‘Job market signalling’’, Quarterly Journal of Economics, Vol. 87, pp. 355-74. Winkelmann, R. (1996), ‘‘Employment prospects and skill acquisition of apprenticeship-trained workers in Germany’’, Industrial and Labor Relations Review, Vol. 49 No. 4, pp. 658-72.

Apprenticeshipto-work transitions 409

International Journal of Manpower 23,5 410

Table AI. Unweighted means of covariates used in Equation II, Table IV

Appendix Year of completing apprenticeship 1948-1959 1960-1974 1975-1992 Poor schooling, male Poor schooling, female (Good schooling, male) Good schooling, female Farmer, gardener, miner (Metalworker) Electrician Textile worker Food worker Construction worker Carpenter, painter Other manufacturers Technician Clerical worker Sales clerk Health care assistant Other service occupation Firm: < 10 employees (Firm: 10-1,000 employees) Firm: > 1,000 employees No firm (external training) Public sector Duration of training (years) Unemployment rate at end of training N

0.635 0.181

0.474 0.211

0.322 0.142

0.056 0.039

0.127 0.021

0.284 0.021

0.044 0.051 0.045 0.078 0.096 0.038 0.017 0.169 0.127 0.021 0.024 0.402

0.076 0.026 0.033 0.036 0.044 0.028 0.031 0.225 0.160 0.043 0.037 0.299

0.069 0.014 0.035 0.036 0.042 0.030 0.027 0.259 0.133 0.090 0.044 0.289

0.073 0.055 0.064 3.035 5.950

0.087 0.076 0.099 3.068 1.142

0.107 0.079 0.139 2.916 6.741

2,557

4,395

Notes: Own calculations based on data from the BIBB/IAB Employment Survey (Jansen and Strooß, 1993)

5,466

The research register for this journal is available at http://www.emeraldinsight.com/researchregisters

The current issue and full text archive of this journal is available at http://www.emeraldinsight.com/0143-7720.htm

The transition from apprenticeship training to work Wolfgang Franz

Transition from apprenticeship training to work 411

Zentrum fu¨r Europa¨ische Wirtschaftsforschung, Mannheim, Germany, and

Volker Zimmermann Kreditanstalt fu¨r Wiederaufbau, Frankfurt, Germany Keywords Training, Apprenticeship, Young people, Unemployment Abstract This econometric study deals with the question as to what extent apprentices, after successfully completing their training, stay with the firm that supplied their training and, if so, how long that job tenure holds. Determinants of both decisions can be seen from both the employer’s and the employee’s viewpoint. These firms are interested in employing apprentices in order to collect the returns from their investment in their training, which frequently is associated with net costs. On the other hand, the firms dismiss apprentices if training is viewed by themselves as a screening device or if apprentices are engaged in work for which, in terms of wages, they are too expensive afterwards. The young trained worker bases his or her decision to stay or to leave on considerations such as experimenting with several jobs.

Introduction The increasing lack of skilled labour, especially but by far not exclusively in the IT sector, has taken centre stage in the public debate on Germany’s future developments. Schooling and vocational training are correctly identified as the most promising instruments to achieve sustainable growth rates in order to meet international competition and to improve labour market conditions. The focus of this paper is on vocational training. More specifically, we deal with job mobility after apprenticeship training in West Germany. This aspect is, of course, only one of several topics associated with apprenticeship training (Franz, 1982; Franz et al., 2000a, b; Franz and Soskice, 1995; Zimmermann, 2000). But we regard it as important for several reasons. First, difficulties during the transition from apprenticeship training to the first job are a nonnegligible source of youth unemployment. Second, leaving the firm after apprenticeship training may be voluntarily or not. The youth may prefer some experience with several types of firms in order to learn more about his or her abilities and preferences (‘‘job shopping’’). Leaving the firm may also serve as a signal for bad employment conditions associated with this firm. On the other hand, the youth may not receive a contract due to his or her poor performance. Moreover, the firm may by intention hire more apprentices than needed later, in This paper represents a reviewed version of Franz and Zimmermann (1999). The authors are solely responsible for the contents, which do not necessarily represent the opinion of the ZEW or KfW.

International Journal of Manpower, Vol. 23 No. 5, 2002, pp. 411-425. # MCB UP Limited, 0143-7720 DOI 10.1108/01437720210436037

International Journal of Manpower 23,5 412

order to select the best persons. Thirdly, in view of an increasing excess demand for skilled labour, it is interesting to examine whether the behaviour of youths and firms is changing. The following part provides a short overview of the German apprenticeship system. In 1998, about 66 percent of 16-19-year-old youths started an apprenticeship within the dual system of vocational training (Reinberg and Hummel, 2001). Although this figure is slightly declining (1990: 71 percent), apprenticeship is still the most important component of the German vocational training system. In the 1990s, 22-25 percent of the training contracts, out of all new contracts of the previous three years, were terminated prematurely (BMBF). Most of the drop-outs do not cease their efforts in vocational training but start another apprenticeship or continue education at university or in a vocational school (Alex et al., 1997). Among those apprentices who do not quit, the overwhelming number (1999: 85 percent) succeed in the final examinations. In 1998, about 65 percent of 25-34-year-olds have completed an apprenticeship (1990: 67 percent) (Reinberg and Hummel, 2001). After a successful final examination, the former apprentices have good chances to stay within the firm in which they were trained. In 1999 about 60 percent were employed by their training firm. Explanatory pattern for the duration of job tenure directly after the end of traineeship Theoretically there are two answers to the question as to whether and how long young people continue to work in the training company after the end of their traineeship, depending on whether the employer or the employee gave notice of termination. There is substantial literature on both aspects which we will evaluate briefly with regard to this specific subject (Franz, 1999). Based on a panel data set of individual firms collected by the IAB Institute for Employment Research, Table I shows the empirical relevance of the reasons for leaving the training company directly after the end of traineeship. According to the panel, young people leave the training company on their own initiative in nearly half of the companies. As many as 38 percent of the companies state that they do not require skilled workers. Only one out of ten firms reported that Reasons for withdrawal

Former trainees had other plans At present, company does not require any skilled workers Former trainees do not meet company Table I. expectations Reasons for withdrawal Company belonging to the same group takes from the training over former trainees company after the end No reasons stated of traineeship in West Source: BMBF (1998, Figure 19) Germany 1996

Percent 46 38 9 4 2

trainees left the company after the end of their traineeship because the Transition from company felt that the trainee lacked the necessary aptitude. apprenticeship When starting out with the companies, the following determinants have to training to work be considered: . With respect to the general determinants influencing demand for (qualified) labour, the expected demand for commodities, which, 413 however, does not necessarily represent an exogenous variable, and production technology play a certain role. In other words, the output development in individual sectors, be it subject to business activities or to long-term trends due to, for instance, demand shifts triggered by preference changes, as well as the capital-labour ratio of production which is determined, inter alia, by technical progress as well as by the relative labour and capital costs, can be seen as determinants. . Furthermore, when dealing with heterogeneity of labour one has to take into account that the demand for qualified labour has increased relatively to demand for low skilled labour so that the company is ceteris paribus to a greater extent interested in taking over its former trainees into regular employment. Technical progress and the increasing international interdependence of output and input markets are primarily held accountable for this structural change of labour demand in terms of quality. . Moreover, the company has to tackle the question as to whether it intends to meet the demand for skilled workers by employing external employees instead of – or in addition to – taking over its trained employees. Apart from reviewing the personal characteristics of former trainees that were hardly observable at the beginning of the traineeship or have changed (such as qualification level above or below average, morale, personal appearance), the company will nevertheless review the external skilled labour market with regard to availability and remuneration and take into account that net training costs will represent sunk costs, if a former trainee is not employed after an in-house traineeship. . Finally, one has to take into account that companies provide at the outset a high number of training places that does not correspond to the expected demand for skilled workers. This decision might well be based on rational considerations: within a personnel selection process the company can choose the most suited workers from the group of former trainees; or the net training costs are low or even negative, e.g. because trainees perform activities for which a low skilled employee would have to be employed. In (individual) cases, however, employers might heed the appeals of their associations to increase the number of training places when there is a great need for training places, or company owners might hear such requests in their personal networks and employ more trainees than necessary. This means they act in a social manner which might improve the reputation of the company. This however, does not serve as an explanation for all cases.

International Journal of Manpower 23,5 414

On the other hand there are determinants influencing the decisions of former trainees as to whether and for how long they will continue to work in the training company. The young people will analyse their options just as the employers did. When general labour demand and wage offers of other companies are increasing, it is more likely that young people will be willing to work for a different company. Furthermore, their willingness to move to another place in the region – if there are no adequate alternatives near their home – and their satisfaction with their working conditions represent significant determinants. A former trainee may also enter into a more preferred job, i.e. the training place met the trainee’s preferences only to a very limited degree, nevertheless the trainee accepted it at that point in time, because there were no other alternatives. Finally, job changes at the beginning of working life might reflect the intention to try out different activities and to determine optimum productivity. Empirical analysis Data The empirical study on the length of service after initial vocational training is performed on the basis of the most recent available cross-sectional poll conducted in cooperation between the BiBB – Federal Institute for Vocational Training – and the IAB – Institute for Employment Research – in the framework of the survey ‘‘Acquisition and application of occupational skills’’[1]. In the context of this survey, data were collected in 1991/1992 in a representative poll on initial vocational training and careers of approximately 34,000 individuals between 15 and 65 years in West and East Germany. Only a very small share of those surveyed are persons with a foreign citizenship[2]. The analysis conducted here is constrained to former trainees from West Germany, in order to avoid extraordinary effects that might arise due to the fact that the so-called Dual System, the German vocational training system which combines classroom and practical vocational training, had not been fully established in the eastern constituent states. Furthermore, among those who completed initial vocational training, only those are included who completed their traineeship in the Dual System between 1980 and 1991. The data surveyed on training in the Dual System relate to the most recent traineeship completed. In order to restrict the survey to those who completed their first traineeship, all those individuals were removed from the sample who were more than 30 years old at the end of their traineeship. Apart from that, individuals were excluded who were unable to give details on the time they spent in the training company and on the company itself. The remaining sample comprised 4,627 former trainees. Table II shows the cases left after each selection step. The length of continued work after the end of the traineeship is determined according to six categories. The poll identifies whether a former trainee left the company immediately, within a year, in a period ranging from one to under two years, and within a period ranging from two to under five years. In addition, the answers ‘‘more than five years’’ length of service and ‘‘still employed in the

training company at the time of the poll’’ were possible. Based on these possible Transition from answers, four categories for the length of continued service for the training apprenticeship company were devised accounting for censoring at the end of the observation training to work period. These categories refer to those leaving their training company immediately after completing their traineeship, within a year, within a period ranging from one year to under two years, and within a period ranging from 415 two years to under five years. A total of 192 former trainees leave the training company due to military service. Even though in reality it is possible to influence the point in time of military service to a certain degree, this process is regarded as an exogenous factor. The length of continued service of the former trainees is regarded as censored at that point in time. Based on a graduated life table (Blossfeld et al., 1986), Figure 1 shows the calculated period spent in the training company after the end of the traineeship in relation to the year of the end of traineeship for the period between 1980 and Preclusion criterion Total East Germany Traineeship not completed End of traineeship not between 1980 and 1991 Older than 30 at end of traineeship No details on job tenure No details on training company Source: BiBB/IAB survey of 1991/91, own calculations

Remaining cases 34,277 23,245 16,490 4,838 4,776 4,639 4,627

Table II. Number of individuals remaining in the sample after removal due to various selection criteria

Figure 1. Empirical survivor function for job tenure in the training firm after apprenticeship training

International Journal of Manpower 23,5 416

1991. As a rule, 400 to 450 observations were obtained per traineeship completion year. The share of employees who do not leave their training company immediately after the traineeship has increased since the beginning of the 1980s. This increase is especially strong in the second half of the 1980s. The other categories of length of continued work, also, experienced a slight increase in the continued service curves in the 1980s. In the following, some explanatory variables for the length of service underlying the study are described. Table III shows descriptive statistics of the applied variables. The economic sector to which the training company belongs allows for conclusions on various training company characteristics. First, the economic sector reflects the cost of training as well as the degree to which trainees can be entrusted with productive activities in the company already during their Variable Training firm Sector Craft Industry Civil service Trade Other sectors Firm size Less than four employees Five to nine employees 10 to 49 employees 50 to 99 employees 100 to 499 employees 500 to 999 employees 1,000 and more employees Production technology Capital-labour ratio Capital-labour ratio not available Mean

Table III. Descriptive statistics

Former apprentice Age at final examinations Gender (1 = female) General educational background No certificate Lower secondary school Secondary school Entitlement for university (of applied sciences) Other schools Macroeconomic conditions Unemployment rate (percent) Trainee-employee ratio (percent) Source: BiBB/IAB survey of 1991/91, own calculations

Percent/Mean

SD

32.8 23.3 13.8 15.5 14.7 8.9 19.8 27.8 10.6 16.1 5.6 11.2

4.6 0.193 19.9 42.1

0.257 2.06

0.1 40.3 44.0 14.9 0.7 7.7 7.7

2.99 1.71

traineeship. Traineeship costs are studied regularly for the industrial and the Transition from trade sectors (Sachversta¨ndigenkommission Kosten und Finanzierung der apprenticeship beruflichen Bildung, 1974; Falk, 1982; Noll et al., 1983). According to the most training to work recent studies, net costs of training in the craft trades are lower than those in the sectors of industry and trade. They correspond to approximately two-thirds of the net costs of training incurred by the latter sector (Bardeleben et al., 1997). 417 The fact that training costs are lower in the craft sector combined with the high volume of productive work performed by trainees during their traineeship result in no net costs in approximately one third of all training companies in the craft sector in 1991 (Bardeleben et al., 1995). Owing to this cost structure, craft businesses quite frequently train a surplus of workers, whom they do not intend to take over into regular employment after completion of training. At the same time, trainees do not regard craft businesses as attractive employers, so that it is quite frequently on their own accord that youths look for employment in other sectors after the end of their traineeship. The size of the training company also allows for conclusions on the training costs. The net training costs in larger training companies are higher. According to calculations of the BiBB – Federal Institute for Vocational Training – the net costs of training per trainee and training year in companies with over 500 employees are 6.5 times as high as in companies with less than ten employees (Bardeleben et al., 1997). The higher costs are to be attributed to the frequent deployment of full-time trainers and the fact that trainees are trained in special training facilities. The number of trainees per employee in these companies is also much lower than in small businesses. In addition, the more complex structures and processes in larger companies seem to require the formation of company-specific human capital, so that these companies are more likely to train junior skilled personnel to meet their own needs. This leads to the assumption that both the take-over rate and the length of continued service for the training company after the end of the traineeship increase with the size of the company. At the same time, large companies seem to be more attractive for former trainees than smaller businesses: companies with a larger internal labour market offer better prospects of an internal career. Very frequently, larger companies pay higher wages and salaries than smaller businesses. These are some of the considerations that lead us to assume that former trainees are more likely to stay in larger training companies than in smaller training companies. The capital-labour ratio, measured as the net capital invested per employee, describes the state of the production technology of the training company. The training companies included in the sample are subsumed under 33 industrial segments and their capital-labour ratio is considered on an annual basis. We could, for example, assume that a higher capital-labour ratio stands for more complex work processes and that skilled workers and a high company-specific human capital stock would be required for these processes. If so, a longer length of service period in the training company would be expected. However, it is also possible that a high capital-labour ratio stands for a high level of

International Journal of Manpower 23,5 418

automation, which would entail lower requirements on the skill level of the employees. If so, employment relationships are more likely to be shorter in companies with a high capital-labour ratio. There is a small number of companies for which capital-labour ratios cannot be considered. These companies were not removed from the data set but left there marked with a dummy variable indicating the fact. The educational background of the former trainees is also observed. Among the former trainees, a majority (44 percent) holds a general certificate of secondary education, while 40 percent hold a secondary school leaving certificate. A total of 15 percent hold a university or university of applied sciences entry certificate. Hardly any members of the sample had no schoolleaving certificate or certificates other than those mentioned above. The training companies are more likely to want to keep the former trainees with the higher-level leaving certificates in the company in order to augment their human capital. Against this backdrop, we would expect the length of continued work for the training company to be longer with increasing levels of leaving certificates held by the former trainees. Here, a counter effect is to be expected due to the options the former trainees have thanks to their education, since it seems as if youths have more options outside the training company the higher their level of education is. On the one hand, it seems as if former trainees with a higher level of education receive job offers from other companies more frequently, on the other hand, these youths – especially those with university or college entry certificates – have access to further education. This means that the influence of the school education level on the continued service in the training company cannot unambiguously be explained a priori. To what extent gender influences the length of continued work in the training company also cannot unambiguously be explained by theory. However, it seems as if young women interrupt their careers more readily than young men due to their family situation (Fobe and Minx, 1996). What is also known of the former trainees is their age at completion of the traineeship. Since no further personal characteristics, like for example the marital status or the number of children in their care, are known for the period under examination, these characteristics are mapped according to the age of the former trainees. Thus, reduced mobility is to be expected with increasing age due to the family situation. The average marriage age rose by approximately two years during the 1980s, i.e. among unmarried women to 25.5 and among unmarried men to almost 28 years. It is very likely that men who are older at completion of their traineeship have already rendered military or alternative military service and are therefore less likely to be forced to leave the training company due to these reasons. Furthermore, opportunities on the labour market change with increasing age. A higher age at completion of the traineeship may, if checked against the educational level, inform about the fact that the youth had started his or her traineeship late, a fact which could be attributed to difficulties in finding a training place or difficulties at school, or about the fact that this was not the

first traineeship the person underwent. This might mean that the former Transition from trainee’s opportunities on the labour market are restricted and that they are apprenticeship therefore more likely to remain in the training company. On the other hand, this training to work group might be more likely to include youths who do not live up to the expectations of the training company and whom the training company might want to release. Thus, the influence of the former trainees’ age on the length of 419 continued service for the training company is not absolutely clear from the theoretical point of view. The scope of options of the former trainee is determined, among other things, by the situation on the labour market. Therefore, the unemployment rate of each year is accounted for in the study. A rising unemployment rate will be accompanied by reduced opportunities on the labour market and thus a longer continued work in the training company. The analysis also accounts for the training rate on a macroeconomic level, i.e. the number of trainees per employee covered by social insurance. This variable allows for the examination of to what extent a changed training intensity during the period under observation can result in changes in the take-over behaviour and the length of service. Thus it can be determined whether a declining training intensity results in more trainees being taken over into regular employment after the end of their traineeship. Such an observation might inform about the fact that in this case the training companies attempt to avoid training surplus labour or that they have improved the selection criteria for trainees so as to ensure that take-over decisions can be restricted to a smaller selection of ‘‘candidates’’. A constant take-over rate combined with a lowered training intensity seems to inform about the anticipation of the future demand for skilled labour by companies, who adjust their training commitment to reduced demand for junior skilled personnel. For the period under examination, the training proportion is only available in aggregate terms. Although it would be helpful, to disaggregate it according to industry level is not possible. Figure 2 shows the development of the training rate between 1980 and 1991. With an increasing number of students finishing school, the training proportion increases up to the year 1986 and then, along with the declining number of students finishing school, the quota slumps to a measurably lower level than at the beginning of the 1980s. Regression results For the following analysis a discontinuous hazard rate model is applied (Narendranathan and Stewart, 1993; Steiner, 1997). Table IV depicts regression results and Figure 3 the graphical display of the continued service function in the training company of a reference person as well as the respective function for other persons when changing one attribute for the reference person. Table V presents the percentage change in the survivor function in the case of such variations. The reference person is defined as follows: an almost 20-year-old employee who completed training with an intermediate level of schooling

International Journal of Manpower 23,5 420

Figure 2. Trainee-employee ratio in West Germany 1980 to 1991

working in a craft business with less than five employees. Further attributes, such as the capital-labour ratio, training proportion, and unemployment rate also take on the average values of the sample. The regression coefficients in Table IV demonstrate how various attributes influence the hazard rate, i.e. the probability that former trainees will leave their training company. In order to allow for attributes influencing the takeover probability that might deviate from other factors influencing the remaining periods of employment, interaction terms of these attributes are formed with a dummy variable designating the point in time when the apprenticeship training is completed. The last two columns of Table IV show the effects of these interaction terms. Among the former trainee’s socioeconomic attributes, age effects prove to differ significantly from zero in all periods. Age is particularly important for the take-over probability. Particularly older trainees frequently leave the training company directly after the end of traineeship. A quadratic term proved to be insignificant and was dropped from the regression equation. Figure 3 reveals that the survivor function for a former trainee, who is five years older, is much lower than the curve for a former trainee who is just under 20. According to Table V the difference between the take-over rates of both groups of persons is just under 13 percent to the disadvantage of the older former trainee. Young women leave the training company significantly sooner than their male colleagues. However, if they do stay in the training company their hazard rates do not differ significantly from those of their male colleagues. The deviations from the survivor function for a one-year to five-year period presented in Table V are exclusively due to the differing take-over probabilities.

Variable

Coefficient

Constant

–3.918

–6.6

0.038 0.072

2.1 1.1

0.047 0.317

1.8 3.1

0.073

1.1

0.189

1.7

–0.015

–0.1

0.475

3.1

–0.317 –0.912 –0.209 –0.339

–3.3 –7.7 –2.2 –3.3

0.165 0.709 0.197 0.327

1.1 4.0 1.3 2.1

0.178 0.088

1.6 0.7

–0.101 0.024

–0.6 0.1

–0.116 –0.025 –0.100 –0.134 –0.284 –0.336

–1.0 –0.2 –0.7 –1.0 –1.6 –2.2

–0.231 –0.295 –0.535 –0.438 –0.835 –0.879

–1.3 –1.7 –2.6 –2.2 –3.0 –3.7

–0.025 –0.039

–2.0 –1.9

–0.006 0.230

–0.3 4.5

2.259 2.336 1.379

3.2 3.3 1.9 4,627 6,508.9

Apprentice Age Gender (male) No certificate, lower secondary, other school Entitl. for university (for appl. sciences) Training firm Sector (craft) Industry Civil service Trade Other sectors Production technology Capital-labour ratio Capital-labour ratio not available Firm size (< four employees) 5 to 9 employees 10 to 49 employees 50 to 99 employees 100 to 499 employees 500 to 999 employees 1,000 and more employees Macroeconomic conditions Unemployment rate Trainees-employees ratio Interval dummies (immediately) Within 1 year Within 1 to 2 years Within 2 to 5 years Number of cases Loglikelihood

t-value

Interaction terms hazard rate at the end of vocational training Coefficient t-value

Source: BiBB/IAB survey of 1991/91, own calculations

Similarly, school education proves to be only relevant for the take-over probability: former trainees with intermediate schooling are most frequently taken over into regular employment. The withdrawal behaviour of former trainees employed after training, however, does not differ according to their educational background. This points to a selection process after the end of traineeship that was mentioned in the last section. As the human capital of the former trainees increases, the company is more interested in keeping the young people employed. In parallel, however, young people’s options increase as well, which is reflected particularly in the low take-over rates for young people with

Transition from apprenticeship training to work 421

Table IV. Regression results

International Journal of Manpower 23,5 422

Figure 3. Survivor functions for job tenure in the training firm of former apprentices

a university (of applied sciences) entrance qualification after the end of traineeship; thus approximately 28 percent of former trainees with intermediate schooling leave the training company. Of the former trainees, 33 percent with a lower level of schooling and more than 37 percent of former trainees with a higher level of educational attainment left their training company at the end of traineeship. Compared to the reference category of the craft trades, all other economic sectors have higher survivor curves. The increased hazard rates in the craft trades apply to the take-over rate as well as to the withdrawal of former trainees who had been taken over into regular employment. This confirms the observation in the last section that firms train more apprentices than are actually needed. The amount of net training costs brings about the expected effect with regard to the size of the training company: the take-over rate increases with company size. The hazard rates in the subsequent periods, however, do not differ significantly from each other with the exception of companies with more than 1,000 employees. Major companies are thus able to retain their former trainees for a longer period of time. In doing so it is possible to amortize the investment they carried out in the trainees’ human capital. The level of the unemployment rate is a significant factor influencing the probability of continued service in the training company. Based on the observations in the last section, the level of the unemployment rate reflects the

Variable Former apprentice Age: 25 year-old Gender: female General educational background No certificate, lower secondary school Entitlement for university (of applied sciences) Training firm Sector (craft) Industry Civil service Trade Other sectors Production technology Capital-labour ratio: 10 percent higher Firm size (less than 4 employees) 5 to 9 employees 10 to 49 employees 50 to 99 employees 100 to 499 employees 500 to 999 employees 1,000 and more employees Macroeconomic conditions Unemployment rate: 10 percent higher Trainee-employee ratio: 10 percent higher

Variations of survivor function (%) After final Within one Within two Within five examinations year years years

–12.9 –11.6

–16.5 –12.9

–20.2 –14.3

–25.0 –16.1

–7.6

–9.0

–10.5

–12.4

–13.8

–13.5

–13.2

–12.9

4.0 5.3 0.3 0.3

10.1 20.0 4.3 6.6

17.1 37.9 8.8 13.6

26.8 64.8 14.9 23.6

–0.04

–0.1

–0.2

–0.3

8.7 8.1 14.8 13.6 22.7 23.9

11.2 8.7 17.1 16.6 29.2 31.6

13.9 9.2 19.6 19.8 36.5 40.3

17.6 10.0 22.9 24.3 46.8 52.5

0.7

1.1

1.5

2.0

–4.2

–3.6

–2.9

–2.1

Notes: Reference person is defined as follows: 20-year-old male former apprentice with secondary school certificate; was trained in a crafts firm with less than five employees. The variables capital-labour ratio, trainee-employee ratio and unemployment rate at sample means Source: BiBB/IAB survey of 1991/91, own calculations

former trainees’ options to find a job outside the training company. According to these calculations a 10 percent higher unemployment rate will lead to a probability increase of approximately 2 percent that they will continue to be employed in the training company five years after the end of traineeship. The aggregate training rate also proves to be a significant factor. This influence, however, is distributed unequally. The take-over probability declines with an increasing training proportion, which suggests that training exceeds demand, particularly in view of the tense situation on the labour market for young people. The survivor rate for former trainees taken over into regular

Transition from apprenticeship training to work 423

Table V. Variation of survivor function for job tenure after apprenticeship training

International Journal of Manpower 23,5 424

employment increases, however, with the training intensity of companies. This might lead to the conclusion that, when the proportion of training firms is high, young people are faced with stronger competition of a fresh generation of skilled workers which limits their possibilities on the external labour market. Summary of findings The end of traineeship is followed by considerable mobility within the company. During the period under review approximately half of the former trainees left their training company within a period of two years. The regression results suggest that reasons for these withdrawals can be found on the part of the companies as well as on the part of the young people. Furthermore, the analysis reveals that the determinants for employment and continued work of a former trainee in the training company partially differ from each other after take-over into regular employment. Contrary to continued work at a later point in time, the socioeconomic background of a trainee thus plays a vital role for take-over into regular employment. Options on the external labour market determine the trainee’s inclination not to leave the training company. Enterprises with high net training costs have longer periods of employment. This observation is in line with the hypothesis that companies attempt to amortize investments in the human capital of the young people. Furthermore, companies are willing to train more young people than are actually needed. Increased willingness to train young people, e.g. during tight labour market conditions for young people in the mid-1980s, leads to lower take-over rates. At the same time the surplus in former trainees limits the mobility of these young skilled workers in those years. Notes 1. The data were prepared and documented for the analysis by the Central Archive for Empirical Social Research (ZUMA). Neither the BiBB nor the IAB are responsible for the analysis or interpretation of the data presented in this paper. 2. It is therefore not possible to study the impact of foreign citizenship on length of job tenure. References Alex, L., Menk, A. and Schiemann, M. (1997), ‘‘Vorzeitige Lo¨sung von Ausbildungsvertra¨gen’’, Berufsbildung in Wissenschaft und Praxis, 26/1997/4, pp. 34-9. Bardeleben, R.v., Beicht, U. and Fehe´r, K. (1995), Betriebliche Kosten und Nutzen der Ausbildung. Repra¨sentative Ergebnisse aus Industrie, Handel und Handwerk, Berichte zur bernflichen, Bildung, Bonn. Bardeleben, R.v., Beicht, U. and Fehe´r, K. (1997), Was Kostet die betriebliche Bildung? Fortschreibung der Ergebnisse 1991 auf den Stand 1995, No. 210, Berischte zur beruflichen, Bildung, Bonn. Blossfeld, H.-P., Hamerle, A. and Mayer, K.U. (1986), Ereignisanalyse, Campus, Frankfurt. Bundesministierium fu¨r Bildung, Wissenschaft, Forschung und Technologie (BMBF) (Eds) (1998), Berufsbildungsbericht, BMBF.

Falk, R. (1982), ‘‘Kosten der betrieblichen Aus- und Weiterbildung. Repra¨sentative Erhebung fu¨r 1980’’, in Go¨bel, U. and Schlaffke, W. (Eds), Berichte zur Bildungspolitik 1982/83 des Instituts der deutschen Wirtschaft, Institut der Deutschen Wirtschaft Ko¨ln, Ko¨ln, pp. 63-172. Fobe, K. and Minx, B. (1996), Berufswahlprozesse im perso¨nlichen Lebenszusammenhang, Prospekt, Nu¨rnberg. Franz, W. (1982), Youth Unemployment in the Federal Republic of Germany: Theory, Empirical Results and Policy Implications, Band, Tu¨bingen. Franz, W. (1999), Arbeitsmarkto¨konomik, 4th ed., Springer, Berlin. Franz, W. and Soskice, D. (1995), ‘‘The German apprenticeship system’’, in Buttler, F., Franz, W., Schettkat, R. and Soskice, D. (Eds), Institutional Frameworks and Labor Market Performance: Comparative Views on the US and German Economies, Routledge, London, pp. 208-34. Franz, W. and Zimmermann, V. (1999), ‘‘Mobilita¨t nach der beruflichen Ausbildung. Eine empirische Studie fu¨r Westdeutschland’’, in Franz, W. (Ed.), Lohnstrukturen, Qualifikation und Mobilita¨t, Jahrbu¨cher fu¨r Nationalo¨konomie und Statistik, Vol. 219 No. 1+2, pp. 143-64. Franz, W., Steiner, V. and Zimmermann, V. (2000a), Die betriebliche Ausbildungsbereitschaft im technologischen und demographischen Wandel, Nomos, Baden-Baden. Franz, W., Inkmann, J., Pohlmeier, W. and Zimmermann, V. (2000b), ‘‘Young and out in Germany. on the youths’ chances of labor market entrance in Germany’’, in Blanchflower, D. and Freeman, R. (Eds), Youth Unemployment and Joblessness in Advanced Countries, National Bureau of Economic Research, Cambridge, MA, pp. 381-426. Narendranathan, W. and Stewart, M.B. (1993), ‘‘How does the benefit effect vary as unemployment spells lengthen?’’, Journal of Applied Econometrics, Vol. 8, pp. 361-81. Noll, I., Beicht, U., Bo¨ll, G., Malcher, W. and Wiederhold-Fritz, S. (1983), Nettokosten der betrieblichen Berufsausbildung, Bundesinstitu¨t fu¨r Berufsbildung, Berlin. Reinberg, A. and Hummel, M. (2001), ‘‘Stillstand ist Ru¨ckschritt’’, IAB Kurzbericht, 8/2001. Sachversta¨ndigenkommission Kosten und Finanzierung der beruflichen Bildung (1974), Kosten und Finanzierung der außerschulischen beruflichen Bildung, Sachverstandigenkommission Kosten und Finazierung der beruflicher Bildung, Bielefeld. Statistisches Bundesamt (various issues), Bildung im Zahlenspiegel, Fachserie 1 Reihe 4.2, Statishes Bundesamt, Berlin. Steiner, V. (1997), ‘‘Extended benefit-entitlement periods and the duration of unemployment in West Germany’’, ZEW Discussion paper, pp. 97-104. Zimmermann, V. (2000), Der Arbeitsmarkt fu¨r Jugendliche. Eine empirische Untersuchung ihres Weges in die Bescha¨ftigung, Nomos, Baden-Baden.

Transition from apprenticeship training to work 425

The research register for this journal is available at http://www.emeraldinsight.com/researchregisters

International Journal of Manpower 23,5 426

The current issue and full text archive of this journal is available at http://www.emeraldinsight.com/0143-7720.htm

School-to-work transition: apprenticeship versus vocational school in France Liliane Bonnal Gremaq, Lernea and Universite´ Toulouse, Toulouse, France

Sylvie Mendes Leo, CNRS and Faculte´ de Droit, d’Economie et de Gestion, University of Orle´ans, France, and

Catherine Sofer Team, CNRS, Universite´ Paris 1 Pantheon-Sorbonne, Maison des sciences e´conomiques, Paris, France Keywords School leavers, Work, Apprenticeship, Young people, Labour market, France Abstract There has recently been a strong drive to develop apprenticeship in France, as one means of decreasing youth unemployment. Our aim in this paper is to try to measure the ‘‘pure’’ within-firm training effect on school-to-work transition. We address the problem of the transition to the first job, using a model of simultaneous maximum likelihood estimation of several probabilities and of the parameters of the probability density function linked to the exit from unemployment. We conclude that apprentices have a distinct advantage over those who attended vocational school. This effect is stronger when we correct for the negative selection bias associated with the choice of apprenticeship.

International Journal of Manpower, Vol. 23 No. 5, 2002, pp. 426-442. # MCB UP Limited, 0143-7720 DOI 10.1108/01437720210436046

1. Introduction Does training in a firm improve skills, and subsequently be of value to firms in a way that no knowledge acquired at school can? Many policies, especially among those aimed at decreasing youth unemployment, have been recently developed in France to facilitate and increase the time spent at work in firms within schooling or training programmes. The development of apprenticeship, which has strongly increased in recent years, is one of them. The question then arises whether this specific way of learning facilitates transitions from school to work, as the German example seems to indicate, or not. The low rate of youth unemployment in Germany (at the present time, we should maybe rather speak of a relatively low rate of youth unemployment) has traditionally been attributed mainly to the development of apprenticeship in this country. But there could be other explanations, such as a demographic specificity, or institutional arrangements increasing for firms the returns from specific human capital, or more generally the benefits from youth employment, for example. Our aim in this paper is to try to measure the ‘‘pure’’ within-firm training effect by comparing the transition from school to work of apprentices with different schooling levels with that of youths who have achieved the same schooling levels, but who have followed more traditional vocational schooling programmes, i.e. have learnt mainly at school. To what extent is the way of learning an indicator of schooling? (Ryan, 2001). When

addressed, this question has generally received a positive answer (Addison and Siebert, 1994; Booth and Satchel, 1994; Sollogoub and Ulrich, 1999). We measure the ‘‘pure’’ within-firm effect by asking two questions. First, do apprentices mainly benefit from firm-specific human capital, or, put another way, are apprentices who perform well mainly those who find their first job in the firm where they carried out their apprenticeship (stayers, in what follows)? The second question relates to a possible selection bias. Young people who choose apprenticeship rather than vocational school might exhibit some characteristics exerting a direct influence upon the probability of their getting a first job rapidly. Basically, our model is a duration model for youth unemployment (Lynch, 1983, 1985; Main and Shelly, 1988; Torelli and Trivellato, 1989; Dolton et al., 1994). We introduce an original treatment by taking into account a possible selection bias in this kind of model (Fouge`re and Serandon, 1992; Dolton et al., 1994). We shall first present the French dual system for vocational training which includes apprenticeship and vocational school (section 2). The third section presents the data and some descriptive statistics. Section 4 will be devoted to estimations and results from standard and simultaneous equation models. Some concluding remarks will be developed in section 5. 2. Apprenticeship versus vocational school: Is there some specificity? In France, the CAP, corresponding to the first exit level of vocational schools (about 12 years of schooling), may be obtained either by following an apprenticeship or by attending a vocational school. While the diploma is the same, the schooling system is different in apprenticeship and vocational school. In both systems, there is not only general teaching (French, mathematics, physics, foreign languages, etc.) but also more technical teaching. However, only apprenticeship provides a mix of on-the-job experience and courses at colleges. In vocational school, training is principally provided by teachers in the form of courses, as general teaching. General teaching and vocational training are equally distributed in the timetable, with roughly 30 hours of courses a week, and are completed by a short work experience. A very important difference between apprenticeship and vocational school concerns the status of the students. In vocational school, they are ordinary students whereas apprentices are wage-earner students. In apprenticeship, the relations are tripartite: an apprentice, a firm where the apprentice carries out his apprenticeship and a college, the C.F.A. (Centre de Formation des Apprentis). These three elements are highly correlated and interactive: (1) the apprentice works in the firm during two-thirds of his/her training; (2) he/she spends ten weeks per year at the college, which provides general and technical teaching;

School-to-work transition

427

International Journal of Manpower 23,5 428

(3) teachers and employers meet regularly to exchange information about the student. For these reasons, apprenticeship and vocational school are two systems leading to the same diploma, the content of which is nevertheless very different. A comparative analysis of transition from school to work for each group can therefore be a good indicator for future education policies. 3. The data The data come from the survey from the Cereq (Centre d’Etudes et de Re´cherches sur les Qualifications) ‘‘Panel mesures jeunes’’. It is a panel of youth who quit the schooling system (including apprenticeship) in June 1989, and who were followed until December 1990. The survey gives detailed information about employment and unemployment during this period, schooling characteristics and individual characteristics. The sample size is 1,399 observations for CAP level. Tables I-III show some summary statistics. In Table I it can be seen that the percentage of males is slightly under 50 per cent (45.9 per cent), as well as the percentage of apprentices (44.5 per cent). Also note that the majority prepare a tertiary speciality (nearly 60 per cent), 63.5 per cent obtain the diploma and this percentage is slightly higher for apprentices, 64.8 per cent against 62.4 per cent. Considering now the different issues, Table II, it can be seen that 18 months after they have left school, 12.3 per cent of women are still either unemployed or not in the labour force[1] and 2 per cent of men are unemployed (most of the ‘‘inactive’’ men are doing their military service). For those who found a job, the majority occupy at first a temporary job, but a significant minority of apprentices occupy at first a permanent job, which is to a much lesser extent the case for non-apprentices. Table III shows that 34 per cent (among women) to 38 per cent (among men) of apprentices are hired by the firm in which their apprenticeship took place. When this is the case, their unemployment duration is very short: 0 to 2 months for the great majority of them. 4. Models for access to the first job Variables of particular interest for us are those relative to apprenticeship: compared to vocational school, apprenticeship might be associated with different characteristics exerting opposite effects upon the labour market: . In any case, apprenticeship will provide a higher level of specific human capital than vocational school. . Apprenticeship could provide a lower or a higher level of general human capital than vocational school. . Among apprentices, the distribution between movers and stayers might not be due purely to chance but might simply be the result of a selection process.

31.18 40.69 29.46 26.42

29.95 41.53 28.95 26.02 38.46 59.90 7.43

Speciality of the diploma Secondary Tertiary

Unemployment duration

8.48

28.01 69.75

68.82

70.05

5.76

50.78 49.29

42.52 28.35 25.55

28.51

71.49

18.71

66.20

54.56

59.90 18.77

17.44

15.06

18.74

61.53 11.53

65.12 10.96

63.47 11.22 45.89 16.15

45.17

642

43.86

757

Men

44.46

1,399

Whole sample Women

Age Father’s labour market situation: Active Unemployed and out of the labour force Mother’s labour market situation Active Unemployed Out of the labour force

Numbers of observations The individuals: Have been apprenticed Have obtained the diploma Have prepared a BEP Are men Are stayers Have not been unemployed

Sample

8.28

32.31 67.04

41.48 35.69 20.10

25.72

74.28

18.76

9.57

7.83 91.27

41.26 37.35 19.28

25.30

74.70

18.82

59.04

34.34

46.62 36.33 64.47

67.78

332

Apprentice Women

64.79

622

Sample

6.22

60.34 39.31

41.73 33.79 21.03

26.21

73.79

18.70

70.69

38.62

61.38

290

Men

6.87

43.37 54.18

41.57 23.55 30.76

33.33

66.67

18.73

56.24

62.42 20.21 45.30

777

Sample 352

61.65 21.03

62.50 18.72 69.90 30.10 43.18 28.41 29.26 42.9 55.68 5.47

425

63.06 19.53

51.06 18.74 64.24 35.76 40.24 23.29 32.00 43.76 52.94 7.76

Vocational school Women Men

School-to-work transition

429

Table I. Descriptive statistics

International Journal of Manpower 23,5 430

Table II. Current employment status at the end of the observation period

Employment status

Unemployment

Permanent Temporary Training job job programme

Inactivity

Whole sample Sample Women Men

5.72 8.85 2.02

24.23 22.59 26.17

40.82 37.52 44.71

19.94 27.61 10.90

9.29 3.43 16.20

Apprentice Sample Women Men

5.63 8.43 2.41

35.69 34.33 37.24

40.67 41.27 40.00

10.45 14.16 6.21

7.56 1.81 14.14

Vocational school Sample Women Men

5.79 9.18 5.79

15.06 13.41 15.06

40.93 34.59 40.93

27.54 38.12 27.54

10.68 4.70 10.68

Sample size Per cent among the apprentices No unemployment Unemployment duration of one month Table III. Unemployment duration of two months Apprentices hired in Unemployment duration more than two months their firm of apprenticeship (stayers) Mean unemployment duration

. .

Whole sample

Men

Women

226 36.33 71.28 6.81 9.68 9.12 0.75

112 38.62 71.43 8.04 13.39 7.14 0.59

114 34.34 75.44 6.14 7.02 11.40 0.90

Close to the latter is the ‘‘negative signalling effect’’ of being a mover. The choice between vocational school and apprenticeship might result from a self-selection process.

Our model allows us to correct for the last selection bias, linked to the choice between apprenticeship and vocational school. We can also determine if the advantage we find for apprentices in the access to a first job applies to all apprentices or applies only to stayers. But having shown that the latter holds, we cannot discriminate between the three alternative interpretations (lack of general human capital, negative selection process or negative signalling of the movers). The model consists in estimating: . the probability of choosing apprenticeship; . the probability of finding a job immediately after the end of schooling[2]; . the probability (among apprentices) of being a stayer; . an unemployment duration model. The originality of our method consists of estimating simultaneous equations, thus taking into account and correcting for possible correlations between

observed and unobserved variables included in different equations (model 2). Many interpretations can be found for these correlations. For example, some (unobservable) variables, exerting an effect upon the choice between apprenticeship and vocational school, might also have a direct influence upon either the probability of finding a job immediately or unemployment duration or both. Put another way, there might be a selection bias in the population of apprentices when considering access to the first job. Or some youth might condition their choice of apprenticeship to a high probability of being hired by the firm (and benefit from – by us – unobserved information), inducing a positive correlation between (1) and (3). Or, apprenticeship supervisors might select stayers using (unobserved) criteria also influencing (1) or (4), inducing a correlation of any sign between (1) and (3) and between (3) and (4) (negative for the former and positive for the latter if these criteria affect access to the first job positively and if apprentices perform badly on average using these criteria). Standard probit models for (1), (2), (3) and a standard duration model for (4) are obtained as a special case when these correlations are constrained to be all equal to zero (model 1). 4.1. The statistical model Let Ya ; Ye and Ym be dummy random variables. For an individual, the realisations of these variables are defined by: ( 1 if the individual chooses apprenticeship; ya ¼ 0 otherwise; 8 > < 1 if the individual finds a job immediately ye ¼ after the end of schooling; ð1Þ > : 0 otherwise; 8 > < 1 if the individual finds a job in the firm; of ym ¼

> :

apprenticeship; 0 otherwise:

Let Ya
; Ye
and Ym
be latent random variables. For an individual, the realisation of these variables is given by: 0 y
a ¼ Xa
a þ ua ; 0

y
e ¼ Xe
e þ ue ; 0

y
m ¼ Xm
m þ um ; where Xa ; Xe and Xm are vectors of regressors;
a ;
e and
m are vectors of parameters and ua ; ue and um , are realisations of the random variables Ua ; Ue and Um .

School-to-work transition

431

International Journal of Manpower 23,5 432

So equation (1) can be written:

(

ya ¼

1 if y
a  0; 0 otherwise;

( ye ¼

1 if y
e  0; 0 otherwise;

( ym ¼

1 if y
m  0; 0 otherwise:

Finally, to study the unemployment duration we consider an accelerated life model with a loglinear specification[3] given, for an individual, by: 0

yc ¼ Lntc ¼ Xc
c þ uc ; where Xc and
c are respectively vectors of regressors and parameters, uc is a random variable and tc is the observation of the random variable associated to the unemployment duration Tc . This observation can possibly be censored. We assume that the error terms ðua ; ue ; P um ; uc Þ are normally distributed with mean vector zero and covariance matrix given by: 1 0 1 ae am ac c X B ae 1 em ec c C C; ¼B @ am em 1 0 A 0 2c ac c ec c where uv 2   1; 1½ for u ¼ a; e; m; v ¼ a; e; m and u 6¼ v; c > 0. 4.2. The likelihood function We can observe different types of contribution to the likelihood function. These contributions are given by the tree shown in Figure 1. We notice that we have seven contributions. So, the likelihood function for an individual i[4] (i = 1, . . . N, where N is the sample size) is given by: X y y ð1y Þ y ð1y Þd y ð1y Þð1dÞ Li ð
a ;
e ;
m ;
c ; Þ ¼ l y1a ye ym l 2a e e l 3a e l 4a e ð1ya Þye

l5

ð1ya Þð1ye Þd

l6

ð1ya Þð1ye Þð1dÞ

l7

where d is the right censored indicator and ’ and
the pdf and cdf of a univariate standard normal distribution. So, the likelihood function is: N

X Y X L
a ;
e ;
m ;
c ; ¼ Li
a ;
e ;
m ;
c ; : i¼1

School-to-work transition

433

Figure 1. Contributions to the likelihood function

International Journal of Manpower 23,5 434

The estimations have been run separately for men and for women (Tables IV and V). We have considered the model defined in the previous section with (model 2) and without (model 1) correlation between error terms. 4.3. The results Tables IV and V show the results obtained from models 1 and 2 at the CAP level. Looking first at the choice of apprenticeship versus vocational school, it appears that a few explanatory variables have an effect on this choice. For all, the fact of not having obtained the Bepc (which is the first diploma that can be obtained in general schooling) has a positive effect on the probability of becoming an apprentice. This effect is greater for men. Generally, these young people (without Bepc) have failed in their school career and decide to pursue vocational training with few theoretical subjects. Some of them may have been forced into apprenticeship rather than having chosen it, but the data set does not give us this information. The mother’s labour market situation has no effect on the probability of choosing apprenticeship. The father’s labour market situation, however, seems to be more important for the decision. For young men, if the father is a selfemployed worker[5] or a manual worker (blue collar) the probability of choosing apprenticeship is higher. Generally, for these kinds of jobs apprenticeship is necessary. So, presumably, the young person acts like his father. This result implies that family environment might influence educational choice. For young women, we obtain the same effect as for young men concerning self-employed workers only. There is evidence of some regional differences. Vocational school is part of the standard schooling system, its supply being determined by norms (relative to the number of children) which are roughly the same throughout the whole country. Conversely, the supply of apprenticeship varies on a much larger scale among regions. Regional differences in the choice of apprenticeship can therefore be interpreted as reflecting supply effects. Models 1 and 2 give very similar results. Four variables significantly affect the probability of finding a job immediately after school. The strongest effect is that of apprenticeship: apprentices have better chances of finding a job immediately than vocational school leavers, and the impact is particularly high for young men. The effect is still reinforced for men when correlations are corrected for. In the case of men, model 2 shows a significant negative correlation between the choice of apprenticeship and the probability of an immediate job (ae < 0), indicating a negative selection bias for apprentices when considering their capacities of finding a job immediately. As expected, the parameter for apprentice (strongly) increases in model 2 relative to model 1. Apprenticeship then appears as significantly more efficient than vocational school when the objective is to find a job very rapidly. This result does not seem to hold for women.

Variables associated with: The probability of choosing apprenticeship Intercept Bepc Father’s labour market situation White collar worker Manual worker Self-employed worker Province Paris, South, Center-east West Southwest North East The probability of finding a job immediately Intercept Father’s labour market situation Unemployed and inactive Employed Speciality Secondary Tertiary With diploma Apprentice The probability of finding a job in the firm of apprenticeship (stayer) Intercept With diploma Speciality Tertiary jobs, electricity, hotel trade Building Food The pdf of the unemployment duration Intercept With diploma Speciality Tertiary jobs, hotel trade Food Electricity Building Apprentice Father’s labour market situation Manual worker Self-employed worker White collar worker Number of types of search One Two

Model 1

Model 2

–0.3398 (0.1333)*** –1.1057 (0.1385)***

–0.2929 (0.1328)*** –1.0618 (0.1396)***

Ref. 0.3227 (0.1330)*** 0.2838 (0.1756)**

0.3018 (0.1285)*** 0.2996 (0.1750) **

0.4303 0.5588 –0.1164 0.3800

Ref. (0.1294)*** (0.2441)*** (0.1647) (0.1606)***

0.1407 (0.1376)

0.3325 0.4895 –0.1722 0.4229

–0.2432 (0.1774)* 0.3161 (0.1070)***

Ref. –0.2720 (0.1126)*** 0.3388 (0.1122)*** 0.5080 (0.1140)***

–0.1569 (0.1036)* 0.3307 (0.1044)*** 1.1902 (0.2473)***

–0.1168 (0.1680) 0.2943 (0.1761)**

0.3652 (0.2746)* 0.2460 (0.1669)*

Ref. 0.0303 (0.1802) –0.6864 (0.2112)***

–0.0233 (0.1977) –0.6594 (0.2001)***

2.0739 (0.1603)*** –0.2382 (0.1162)**

2.2625 (0.1634)*** –0.0558 (0.1086)

Ref. (0.2174) (0.1407) (0.1708) (0.1399)*

Ref. 0.0589 (0.1427) –0.2118 (0.1645)* –0.1321 (0.1738) 0.1324 (0.1742)

435

(0.1277)*** (0.2345)** (0.1560) (0.1502)***

Ref. 0.2980 (0.1210)***

0.0625 –0.1494 –0.1397 0.1966

School-to-work transition

0.0145 –0.1390 –0.1469 0.5966

(0.1513) (0.1151) (0.1426) (0.2066)***

0.0540 (0.1212) –0.1470 (0.1339) –0.1555 (0.1428) 0.1251 (0.1380) (continued)

Table IV. Men at the CAP level

International Journal of Manpower 23,5 436

Table IV.

Variables associated with: Three and more c Correlation coefficients ae ac am ec Mean log-likelihood

Model 1

Model 2

–0.0848 (0.1498) 0.6254 (0.0467)***

–0.0908 (0.1184) 0.6438 (0.0448)***

– – – –

–0.6035 (0.2609)*** –0.2168 (0.2019) –0.5570 (0.2940)** 1.0333 (0.1579)*** –1.57788

–1.61202

Notes: Standard errors are given in brackets. Coefficients significance levels: * 10 per cent; ** 5 per cent and *** 1 per cent. ‘‘Secondary’’ jobs are defined by: building, electricity, food, beauty; and ‘‘tertiary’’ jobs are defined by: management, secretarial jobs, sales, etc.

For both genders, a diploma also exerts a strong positive influence in that respect, as well as, to a lesser extent (the coefficient becomes insignificant for women in model 2), having one’s father in employment (rather than unemployed or inactive). So, it appears that firms are looking for young people with diplomas (professional diplomas) and it seems that they consider apprenticeship as professional experience. For men (but not for women), having been trained to enter the tertiary sector is found to have a negative impact. It could be because this speciality (which includes secretarial, business, trade and so on) concerns more especially women. The fact that the father (for young men) and the mother (for young women) is in the labour force has a positive effect on the probability of finding a job immediately and this result stands whatever the labour market situation. It could be that the parents have some labour relations and that they can help their children more easily. We can remark that there is no regional effect on this probability (for all models, the coefficients associated to region are not statistically significant). Looking now at the probability of being a stayer (considering only apprentices), we find that a diploma again has a significant positive influence. Having been trained in the food trade sector for men, and in the hotel trade sector for women appears as significantly negative. For men, model 2 shows a significant negative correlation (am < 0) between the probability of choosing apprenticeship and that of being a stayer. This again indicates a negative selection bias against apprentices, firms mainly choosing stayers among those apprentices who do not exhibit the negative characteristics corresponding to the selection bias. Again, this result holds only for men. The parents’ labour market situation has no effect on the probability of being a stayer. The final estimation concerns the duration of unemployment.

Variables associated with: The probability of choosing apprenticeship Intercept Father’s labour market situation Manual worker Self-employed worker White collar worker Province Center-east, Southwest South West North Paris Bepc The probability of finding a job immediately Intercept Apprentice With diploma Mother’s labour market situation Unemployed, inactive Active The probability of finding a job in the firm of apprenticeship Intercept With diploma Speciality Food, secretary Hotel trade Management Beauty The pdf of the unemployment duration Intercept Father’s labour market situation Manual worker Self-employed worker White collar worker Speciality: Beauty, food, secretary Management Hotel trade With diploma Apprentice Number of types of search One Two Three and more c

Model 1

Model 2

0.3595 (0.0900)***

0.3565 (0.0915)***

Ref. 0.2755 (0.1352)** 0.1309 (0.1410)

0.2690 (0.1338)** 0.1097 (0.1099)

–0.8897 –0.2474 –1.1695 –0.5531 –0.7714

Ref. (0.1637)*** (0.1220)** (0.1570)*** (0.2141)*** (0.1172)***

–0.1889 (0.1129)** 0.3440 (0.0949)*** 0.2687 (0.0989)***

–0.8880 –0.2387 –1.1633 –0.5212 –0.7752

School-to-work transition

437

(0.1637)*** (0.1229)** (0.1586)*** (0.2256)** (0.1173)***

–0.1188 (0.1299) 0.2207 (0.1675)* 0.2649 (0.1002)***

Ref. 0.1241 (0.1016)

0.1532 (0.0874)**

0.1618 (0.2648) 0.1884 (0.1992)

–0.5169 (0.2920)** 0.2636 (0.1475)**

Ref. –0.8500 (0.3155)*** –0.1935 (0.2644) –0.1278 (0.2928)

–0.6262 (0.2324)*** –0.1118 (0.1733) –0.0684 (0.1920)

1.9614 (0.1547)***

2.2559 (0.1248)***

Ref. 0.0187 (0.0977) 0.0863 (0.1407)

0.0249 (0.1081) 0.1097 (0.1099)

–0.2625 –0.2051 –0.2011 0.3646

Ref. (0.1197)** (0.1507)* (0.0972)** (0.1078)***

–0.2309 –0.1658 –0.0656 0.3930

0.3315 0.1208 0.3185 0.7777

(0.1766)** (0.1606) (0.1520)** (0.0386)***

0.3117 0.1241 0.3210 0.7449

(0.0956 )*** (0.1200)* (0.0799) (0.1391)*** (0.1370)** (0.1213) ( 0.1161)*** (0.0311)*** (continued)

Table V. Women at the CAP level

International Journal of Manpower 23,5 438 Table V.

Variables associated with:

Model 1

Model 2

Correlation coefficients ac ec am em ae Mean log-likelihood

– – – – – –1.90988

0.0879 (0.1090) 0.9331 (0.1290)*** 0.1407 (0.2705) 2.4175 (14.1755) 0.0720 (0.1056) –1.88497

For both men and women, a diploma exerts a significant negative influence upon unemployment duration only in model 1 (in model 2, the effect is still negative but insignificant). This indicates that the positive diploma effect (upon a rapid exit from unemployment) only appears when its impact upon the other probabilities considered is not taken into account. The interpretation could be that, in model 1, a diploma in fact captures the effect of unobserved characteristics correlated to it (such as IQ, or work motivation), which are those that have the true positive effect upon exit from unemployment. Being an apprentice has a significant positive effect upon unemployment duration for both men and women: for those apprentices who do not find a job immediately, apprenticeship becomes a penalty rather than an advantage. This effect is reinforced (for men) when considering model 2 compared to model 1. Model 2 seems to dichotomise the effect of apprenticeship more strongly than model 1; a large proportion of apprentices finds a job immediately, often staying in the firm where their apprenticeship took place. Those apprentices are in a better situation than the average vocational school leaver. But another group of apprentices does not find a job immediately. It seems that, for this group, finding a job is difficult, more difficult than for the average vocational school leaver in the same situation, for the unemployment duration of the former is appreciably longer than the latter’s. Regions and specialities do not seem to have any influence upon unemployment duration in the case of men, although for women, having been trained in business or the hotel trade, facilitates employment. Again for women, we find a negative correlation between employment and job search intensity, which might simply indicate that only those who have difficulties finding a job use ‘‘official’’ means of search, and only those who have long been employed have the opportunity to try several of these means. Both for men and for women, we find a strong positive correlation between the probability of finding a job immediately and unemployment duration (ec > 0). The calculation of probabilities associated with some of the parameters allows the interpretation to be taken somewhat further.

4.4. Associated probabilities We have computed the probabilities associated to the two models (Tables VI and VII). The aim of these probabilities is to examine the effects of apprenticeship and diploma on the different studied probabilities (finding a job immediately and being a stayer). The other explanatory probabilities of the models have been taken at their mean value (namely, the probabilities are computed for apprentice and diploma, all else being equal). These calculations allow us to compare the models’ predictions and the empirical values associated to these probabilities. The predictions of the models 1 and 2 are very similar, but for apprentices the empirical values underestimate the probability of finding a job immediately (especially for men). Computing the probabilities associated with the models confirms the previous results (Table VI). Of particular interest is the comparison of the probability of finding a job immediately for apprentices and vocational school Probability immediately after the end of schooling, of being:

School-to-work transition

439

Empirical values Model 1 Model 2 Voc. Voc. Voc. Apprentice school Apprentice school Apprentice school

Men Employed Mean Without diploma With diploma Unemployed Mean Without diploma With diploma Women Employed Mean Without diploma With diploma Unemployed Mean Without diploma With diploma

70.69 62.5 75.84

62.5 58.52 64.98

82.75 76.91 85.88

66.87 59.02 71.46

81.90 75.40 85.35

70.32 62.09 75.02

29.31 37.5 24.16

37.5 41.48 35.02

17.25 23.09 14.12

33.13 40.98 28.54

18.10 24.60 14.65

29.68 37.91 24.98

59.04 45.79 65.33

51.06 45.86 54.1

66.20 59.59 69.55

52.94 45.97 56.65

67.37 60.92 70.63

54.54 47.63 58.16

40.96 54.21 34.67

48.94 54.14 45.9

33.80 40.41 30.45

47.06 54.03 43.35

32.63 39.08 29.37

45.48 52.37 41.84

Notes: In this table the compiled probabilities are conditional probabilities, i.e. they have been obtained conditionally at the state apprentice or not apprentice

Probability of finding a job immediately after:

Men

Apprenticeship Vocational school Apprenticeship and not stayer

82.75 66.87 70.91

Model 1 Women 66.2 52.94 47.91

Men 90.43 54.63 76.03

Model 2 Women 64.9 56.43 48.82

Table VI. Conditional probabilities of being immediately (un)employed

Table VII. Probability comparisons between apprenticeship and vocational school

International Journal of Manpower 23,5 440

leavers, with a large advantage to apprentices, both for men and for women. The difference between men and women is also very striking, with a much higher probability for women of being unemployed, whatever the type of schooling chosen, the worst situation for them being that of vocational school leaver. What is also measured here is the strong positive effect of obtaining one’s diploma upon the probability of immediately finding a job, with a difference of about ten points in the performance of those who did obtain the diploma and those who did not. Some new probabilities have been computed (Table VII) to shed more light upon the compared efficiency of both types of schooling and upon their specific characteristics: to compare the efficiency of apprenticeship and of vocational schooling, we computed, for both types of schooling, using models 1 and 2, the probability of finding a job immediately for an individual with the average characteristics. Model 2 corrects for the initial selection bias among apprentices and thus gives the expected result whereas the difference between the results of models 1 and 2 can be interpreted as a measure of the negative selection bias of apprentices. We also computed the probability of finding a job immediately for an average individual who has been trained as an apprentice and who is not a stayer. The idea, here, is to eliminate the specific human capital effect and to compare the performance in access to the first job of an apprentice who has not accumulated firm-specific human capital and that of a vocational school leaver. Table VII shows that, at the CAP level, apprenticeship greatly facilitates immediate access to the first job: this type of schooling performs much better in that respect than more traditional vocational school, especially for men and after correction, for them, for the selection bias in the choice of apprenticeship (90 per cent of men finding an immediate job against 55 per cent). After correction, the proportion of male apprentices finding a job immediately rises by about eight points and decreases by about 12 points for vocational school leavers, thus indicating the importance of the selection bias. The selection bias is much smaller and goes rather in the opposite direction for women. It is also of interest to compare the results of the group of apprentices who are movers and those of vocational school leavers: the former group does not benefit from firm-specific human capital[6], and is therefore, from this point of view, in the same situation as the group of vocational school leavers. As can be seen from Table VII, the results are not similar for men and for women. In the case of men, apprentices still have an advantage for finding a job immediately (model 1), an advantage which still highly increases after correction (considering model 2, 76 per cent among movers against only nearly 55 per cent among vocational school leavers find a job immediately). In the case of men, firms do not seem to consider that vocational school delivers a better quality or quantity of general human capital than apprenticeship. The opposite situation holds for women, where, although on the whole apprentices do better than vocational school leavers, movers do worse (looking at model 2, 49 per cent of

movers, against 56 per cent of vocational school leavers find a job immediately). This difference is difficult to understand, for boys and girls, while apprentices or those at vocational school, follow the same theoretical courses. But they are segregated by speciality. Possibly, the training of apprentices in women’s specialities (beauty, hairdressing, hotel trade, secretary . . . ) is more firm-specific than in men’s specialities, or the negative selection bias and/or signalling among movers is stronger for them. 5. Concluding remarks Globally, apprentices perform better, especially in the case of men. The positive effect of apprenticeship mainly results from their better performance at the very beginning of the period: they are often hired (immediately or within two months) by the firm in which they performed their apprenticeship. Concerning males, even when apprentices are ‘‘movers’’, they perform better than vocational school leavers. These results are reinforced when corrected for the negative selection bias in the choice for apprenticeship itself. Concerning women, apprentices have a much smaller advantage, which is only due to the better results of ‘‘stayers’’, and no significant selection bias appears. So the results show evidence that acquiring specific human capital pays, but also that, except for women, firms also value more the human capital acquired by apprentices than that acquired by vocational school leavers. This is true even when the returns of the firm-specific part of it are suppressed by moving from the firm where the apprenticeship took place. It is also worth noting that the variable that has the strongest positive influence upon the transition from school to the first job is the diploma, which strongly helps both apprentices and vocational school leavers to find a job sooner. The main limitation to the conclusion drawn here upon the benefits from apprenticeship concerns what happens next: we only studied the access to the first job. Other studies (Sollogoub and Ulrich, 1999) show that these benefits seem to vanish after some time on the labour market. Notes 1. We cannot control for maternity leave, but this phenomenon is quite marginal among young women aged about 18 in France. 2. As the end of schooling generally occurs at the end of June, and July and August are holiday months for most students, ‘‘immediate’’ here means within two months. 3. This modelling can be compared to a tobit model. A tobit model for the study of unemployment duration has, for instance, been used by Fouge`re and Serandon (1992), and by Dolton et al. (1994). 4. To simplify notations, we omit the individual index, i, in the contribution. 5. ‘‘Self-employed worker’’ includes the self-employed themselves but also shopkeepers, in fact very small firms in which the father is the head. 6. It could benefit from ‘‘sector-specific’’ human capital, but then vocational school students also benefit from some ‘‘specific’’ training.

School-to-work transition

441

International Journal of Manpower 23,5 442

References Addison, J.T. and Siebert, W.S. (1994), ‘‘Vocational training and the European community’’, Oxford Economic Papers, Vol. 46 No. 4. Booth, A.L. and Satchel, S.E. (1994), ‘‘Apprenticeships and job tenure’’, Oxford Economic Papers, Vol. 46 No. 4, pp. 676-95. Dolton, P.J., Makepeace, G.H. and Treble, J.G. (1994), ‘‘The youth training scheme and the schoolto-work transition’’, Oxford Economic Papers, Vol. 46 No. 4, pp. 629-57. Fouge`re, D. and Serandon, A. (1992), ‘‘La transition du syste`me e´ducatif a` l’emploi en France: le roˆle des variables scolaires et sociales’’, Revue de l’Economie Sociale, No. 27 et 28, Tome 2, pp. 89-101. Lynch, L.M. (1983), ‘‘Job search and youth unemployment’’, Oxford Economic Papers, Vol. 35, pp. 271-82. Lynch, L.M. (1985), ‘‘State dependency in youth unemployment’’, Journal of Econometrics, Vol. 28, pp. 71-84. Main, B.G.M. and Shelly, M.A. (1988), ‘‘School leavers and the search for employment’’, Oxford Economic Papers, Vol. 40, pp. 487-504. Ryan, P. (2001), ‘‘The school-to-work transition: a cross-national perspective’’, Journal of Economic Literature, Vol. XXXIX, pp. 34-92. Sollogoub, M. and Ulrich, V. (1999), ‘‘Les jeunes en apprentissage ou en lyce´e professionnel’’, Economie et Statistique, Vol. 323 No. 3, pp. 31-49. Torelli, N. and Trivellato, U. (1989), ‘‘Youth unemployment duration from the Italian labour force survey’’, European Economic Review, Vol. 33, pp. 407-15. Further reading Hashimoto, M. (1981), ‘‘Firm-specific human capital as a shared investment’’, American Economic Review, June, pp. 475-82. Kato, T. (1989), ‘‘Specific and general training in the theory of labor turnover’’, Economics Letters, Vol. 30, pp. 259-62.

The research register for this journal is available at http://www.emeraldinsight.com/researchregisters

The current issue and full text archive of this journal is available at http://www.emeraldinsight.com/0143-7720.htm

Participation in higher education

Participation in higher education

The role of cost and return expectations Charlotte Lauer

443

Centre for European Economic Research (ZEW), Mannheim, Germany Keywords Higher education, Labour market, Germany Abstract This study applies to German data a model in which the decision to attend higher education depends on the ratio of marginal cost and marginal return expected from higher education. If this ratio is below a certain threshold, the individual will choose to participate in higher education. In a simulation exercise, the impact of selected variables on this threshold, and thus on the participation probability, is quantified. The results suggests the presence of financial constraints binding participation in higher education and that the participation decision responds to some extent to return expectations in terms of labour market outcome and to financial incentives such as student support.

1. Introduction The human capital theory (Becker, 1964; Mincer, 1974) has been the basis for a very large number of empirical studies on the wage structure and on the returns to education (see for instance, the survey of Asplund and Pereira (1999) for Europe). Strangely enough, however, only a few studies deal with the basic assumption of the human capital theory: that rational individuals weigh the costs and the returns of the various educational options they face when deciding how much they want to invest in their education. The empirical evidence related to the determinants of educational achievement often focusses on the link between social origin and education (Goux and Maurin, 1998; Blossfeld, 1993; Gang and Zimmermann, 2000; Bogess, 1998; Manski et al., 1992; Lauer, 2002b, to name a few). The few studies which do investigate the role of return expectations on educational decisions generally tend to confirm the theory, even though the approaches adopted are very different and render the comparison difficult. Goux and Maurin (1999), for instance, use for France a recursive model in which individuals perfectly predict their future earnings. Kodde (1988) uses the subjective expectations of high school graduates with respect to future income, foregone earnings and unemployment. Mingat and Tan (1998), run their analysis of college enrolment rates on aggregate data. For Germany, Merz and Schimmelpfennig (1999) use individual data to examine the career choices of German high school graduates. Recent developments in economic research, however, tend to grant more importance to the role of cost considerations on educational choices. The positive correlation between family income and schooling attainment is welldocumented for the USA (e.g. Solon (1992), Hill and Duncan (1987) and more

International Journal of Manpower, Vol. 23 No. 5, 2002, pp. 443-457. # MCB UP Limited, 0143-7720 DOI 10.1108/01437720210450897

International Journal of Manpower 23,5 444

recently Acemoglu and Pischke (2001)), and has been widely interpreted as evidence of credit constraints. This, however, is contested by Cameron and Heckman (1998) and Shea (2000). Some other studies are more directly concerned with the impact of public policy on educational attendance. Schultz (1988), for instance, examines the relationship between the expansion of public schools expenditure and aggregate enrolments for about 90 countries. More recently, Hilmer (1998) examines the effect of post-secondary fees on the college attendance decision of high school graduates. This article aims at investigating the role of cost and return considerations on the participation in higher education in West Germany[1]. In particular, the study addresses the following issues: Is higher education attendance influenced by the expected labour market prospects? May public funding influence enrolments? Is there evidence of financial constraints due to family background? The article is organised as follows. After an outline of the modelling framework in section 2, section 3 presents the results of an application of the model to German data. In section 4, the results are quantified in a simulation exercise of the effects on participation in higher education of specific changes in the key cost and return variables. Section 5 concludes. 2. Methodological approach The approach chosen here consists in analysing the probability of participating in higher education at the typical age at which people intending to complete tertiary level studies should be enrolled, irrespective of their previous educational career. Thus, not the probability that an individual successfully completes a specific transition from one level to the next is examined[2], but rather the probability that the individual has successfully completed all previous transitions until the one leading to higher education. In the extent to which drop-out from university can be neglected, this approach gives information on the probability that an individual, given a certain number of characteristics, finally achieves a tertiary level degree. Focussing on individuals in age of participation in higher education in the period covered by the data (1984-1997) makes it possible to make use of the information on labour market and educational funding conditions available in this very same data. For the analysis, the model of Cameron and Heckman (1998) has been reformulated with a view to modelling explicitly the role of cost and return expectations in educational attendance decisions. Let us suppose that an individual faces k possible educational alternatives Ei,i2 {1. . .k} of increasing levels. The desired level of education is not observable, but only the actual decision of the individual[3]. The individual is assumed to opt for the educational alternative which, given his endowment, personal characteristics and any other relevant factors, maximises his utility, the latter being defined in terms of expected net returns, i.e. the difference between expected returns and expected costs of attending each of the educational tracks Ei: Maxi2fl...kg rðEi =xÞ
cðEi =xÞ

ð1Þ

where r denotes the expected return and c the expected cost associated with the attendance of educational track Ei. It is assumed that both the returns and the costs are positive and increase with the education level. The cost and return functions are assumed to have the following form: eðEi =xÞ ¼ rðEi Þ’r ðxÞ"r cðEi =xÞ ¼ cðEi Þ’c ðxÞ"c

ð2Þ

where ’r(x) is a positive function defining the effects of the observed characteristics on the expected returns to education and "r is a random variable accounting for the effect of unobserved individual heterogeneity on the expected returns. Similarly, ’c(x) is a positive function which defines the effects of the observed characteristics on the expected costs of education and "c is a positive random variable representing the impact of unobserved individual heterogeneity on the costs. Thus, the observed characteristics as well as unobserved individual heterogeneity are allowed to affect the expected returns and the expected costs in different ways. The personal shifters ’r, ’c, "r and "c are assumed not to depend on the specific education level. Without loss of generality, it is assumed that E("r) = E("c) = 1. The optimal educational decision is such that the net return is maximised, i.e. the net return associated with Ei* must be positive and at least as large as the net return at the next lower education level Ei*–1 and at the next higher education level Ei*+1: rðEi Þ’e ðxÞ"r
cðEi Þ’c ðxÞ"c > 0 rðEi Þ’r ðxÞ"r
cðEi Þ’c ðxÞ"c > rðEi
1 Þ’r ðxÞ"r
cðEi
1 Þ’c ðxÞ"c

ð3Þ

rðEi Þ’r ðxÞ"r
cðEi Þ’c ðxÞ"c  rðEiþ1 Þ’r ðxÞ"r
cðEiþ1 Þ’c ðxÞ"c Let us define: ’ðxÞ ¼

’r ðxÞ "r and " ¼ : ’c ðxÞ "c

’(x) measures the net impact of observed characteristics x and " the net effect of unobserved individual heterogeneity on the expected relation of returns to costs. Since "r > 0 and "c > 0, " > 0, ’r(x) > 0 and ’c(x) > 0, one obtains after simplification: cðEi Þ >0 rðEi Þ’ðxÞ cðEi Þ
cðEi
1 Þ 1 cðEiþ1 Þ
cðEi Þ 1  

ð4Þ

For any individual with observed characteristics x, the expected net return is positive at the optimum and the unobserved individual component is bounded

Participation in higher education 445

International Journal of Manpower 23,5 446

by the expected ratio of marginal costs to marginal returns of attending Ei* rather than Ei*–1, for the lower bound, and from attending Ei*+1 rather than Ei*, for the upper bound. Consequently, the probability for an individual to attend Ei is given by:
 cðEi Þ
cðEi
1 Þ 1 cðEiþ1 Þ
cðEi Þ 1   1.96 (resp. 2.58, 1.65), then the hypothesis that the coefficient is equal to zero is rejected at a significance level of 5 percent (resp. 1 percent, 10 percent) Dependent variable: 1 = enrolled in higher education, 0 = not enrolled

Participation in higher education 447

a

Table I. Probit estimates with robust standard errors

International Journal of Manpower 23,5 448

qualitative way, since they do not express the effect of the variables on the enrolment probability itself, but rather on the latent utility index. Some further control variables (age, gender, nationality, time trend) have been included in the model to reduce unobserved individual heterogeneity and control for sample composition effects, without being the object of the analysis per se. Descriptive statistics are given in the Appendix (Table AI). Effect of social background The expected ratio of costs to returns is likely to be influenced by the social environment individuals grew up in. Information on the economic situation of the father when the person was 15 years old may be viewed as an indicator of permanent income during childhood. As expected, the educational attainment of both parents is significantly correlated with the enrolment probability of their children. Thus, children of more highly educated parents are more likely to attain tertiary level education. This might be due to the fact that such parents value more education and are consequently more likely to encourage them to pursue further studies. The perception of the return might be higher. Moreover, highly educated parents are in a better position to help their children in their schooling duties and are more likely to have children with higher learning abilities, which reduces the cost of acquiring education and might also help to better take advantage of the qualification acquired. The occupational position of the father is also clearly related to the enrolment probability of the young persons. Having a white collar worker or a civil servant as a father instead of a blue collar worker (reference category) increases significantly the chances of being enrolled in higher education, even if parental education has been controlled for. The same holds for sons and daughters of self-employed. This may reflect long-term financial constraints which, in case of imperfect capital markets, incite children to start working at the first possible opportunity instead of continuing further studies. The GSOEP contains no direct information on parental income, but the impact of short-term financial constraints should be captured to some extent by a variable containing net other household income[6] in the year preceding the interview. Net household income has a positive effect on the enrolment probability, even though parental education and father’s occupation have been controlled for. This means that children in families facing financial difficulties have lower chances of reaching a high level of educational achievement. This points to the presence of short-term liquidity constraints binding participation in higher education. Labour market prospects Differences in the labour market outcome of education may affect educational choices since they influence the expected benefit which might be obtained from the acquisition of further education. The analysis focusses on the effects of expected outcomes in terms of wages, unemployment risk and labour force

participation, but also self-employment and public employment propensities. Note that in the context of the analysis, it is not essential whether those expectations ‘‘really correspond to what people expect to be their labour market outcome given a certain level of education, but rather whether these expectations’’ influence the perceived ratio of costs to returns and, therefore, educational attendance decisions. The estimation of these labour market expectations is based on the assumption that people observe the current labour market situation of ‘‘comparable’’ persons of the previous generation, i.e. persons with the same observed characteristics, and expect their own situation to become similar. These variables are computed through out-of-sample predictions[7], i.e. predictions for the sample we are interested in – the individuals aged 21 to 26 – are based on estimates drawn from a second sample. Thus, in a first step, the average labour market outcome (wage[8], unemployment risk, part-time employment probability, etc.) is estimated on a secondary sample of ‘‘older’’ people as a function of gender, nationality, family background, region and year[9]. The estimated coefficients are then used to predict the expected labour market outcome of individuals in the primary sample, given their personal characteristics. For the wage expectation, the procedure was slightly different, since the net present value[10] of expected life-cycle[11] income streams given personal characteristics was computed. Thus, the wage equation used to compute these predictions[12] also includes age and age squared as well as interactions between the age variables and gender in order to account for differing age-wage profiles between men and women. One further issue examined here is whether the absolute levels (e.g. the level of unemployment or the level of wages) or rather the relative returns (e.g. the unemployment risk reduction or the wage premium associated with the completion of tertiary level studies) matter. To this end, the same procedure as described above has been applied separately to university graduates and for individuals having a lower qualification. We thus obtain differentiated expected labour market outcomes in case the individual completes university education and in case he does not. The ratio of the expected outcomes in case of graduation and in case of no graduation[13] provides an indicator of the expected labour market return to higher education[14]. Labour market expectations appear to have a significant impact on enrolment decisions. The absolute wage an individual can expect to earn might affect the probability of attending a tertiary level institution in different ways. First, the prospects of earning a higher hourly wage might increase the incentive to pursue further studies in order to benefit later on from this high wage. On the other hand, a higher wage, especially among young people, implies higher opportunity costs for studying, which should decrease the incentive to pursue further studies. Therefore, the cost effect and the return effect go in opposite directions and the net effect of this variable on the expected cost to return ratio is a priori unclear. The estimation results show that the absolute level of expected wage does not influence the probability of

Participation in higher education 449

International Journal of Manpower 23,5 450

attending a tertiary level institution in a significant way. However, the expected wage return to education, i.e. the wage premium associated with the completion of higher education, which was expected to decrease the expected ratio of cost to return via the return side proves to have a significant and strong positive impact on higher education attendance. The probability of experiencing unemployment expresses the extent to which the investment in education is risky in the sense that the wage premium associated with university graduation cannot be drawn in case the person is unemployed. Whereas the absolute level of unemployment risk has a very strong impact on the enrolment probability, the reduction of the unemployment risk due to a higher education degree has a much lower impact on attendance decisions, though it is highly significant. Beyond the obvious utility of further education with a view to diminishing one’s unemployment risk in the future, one further reason for the strong effect of the unemployment risk variable might be that in times of high unemployment, especially high youth unemployment, remaining in the education system might be seen as a worthwhile alternative in the short run (high unemployment risk means lower opportunity costs for studying). Thus, the cost and the return effects go in the same direction and both contribute to lowering the enrolment threshold, i.e. the expected marginal cost-marginal return ratio, thus to favouring enrolment decision. The extent of labour force participation may also affect educational decisions, since not participating in the labour market, whatever the reason, means abstaining from reaping the benefits of education in terms of wages. Thus, differences in educational decisions across individuals, e.g. between men and women, might be due to the fact that differences in labour force attachment are anticipated at the time educational decisions are made, which modifies the perception of the return to education. And in fact, individuals with a higher risk of being employed only on a part-time basis, i.e. who face lower return expectations, appear to be significantly less likely to be enrolled in higher education. Similarly, the prospects of being non-employed appear to have a strong negative influence on higher education enrolments. However, the variables depicting relative part-time and non-employment propensities proved insignificant. The local structure of employment also seems to affect enrolments. High prospects of becoming self-employed reduce educational participation in a very strong and significant way. A possible explanation for this may be that educational credentials could act as a signal of productivity in the eyes of employers and lose relevance if one is due to become self-employed. In other words, the return to education is lower for self-employed. Finally, the higher the probability is of being employed in the public sector, e.g. because there is a tradition to work in the public sector in the family, the higher the participation is in higher education. This may be due to the fact that wages are indexed on qualification in the public sector, and thus having a higher education level

necessarily results in higher wages, and the wage return to be expected from education is highly reliable. Educational policy The estimates found for the public policy variables give an idea of the possible effectiveness of public policy in influencing enrolments in tertiary education, while controlling for the influence of other variables such as family background and return expectations. Rather surprisingly, the extent of public investment in tertiary education, measured as educational expenditure by student, proved insignificant. This could be because this is too broad a measure of the intensity of educational efforts, since the total costs of education per student arise from many sources (e.g. subjects offered, real estate prices, etc.). Against the expectations, the impact of the students-teacher ratio in the previous year proved significantly positive. This is not consistent with the interpretation of a high students-teacher ratio as an indicator of poor quality of education. An alternative interpretation could be that a high students-teacher ratio signals a high popularity of universities in the region concerned, which in turn, might be seen by potential students as an indicator for the good quality of education offered there. The regional GDP per head variable has a negative coefficient and there is no evidence of effects of demographic pressure, since the coefficient of the ratio of the population in age of being enrolled to total population proved insignificant. Public financial support to students aims at reducing the cost of education in order to increase enrolments. Public financial support of education in Germany essentially takes place within the framework of the BAfo¨G (Bundesausbildungsfo¨rderungsgesetz). Three variables were included in the model. First, the expected chance of being entitled to a BAfo¨G grant/loan was approximated by estimating, using the GSOEP data, the probability for students to receive BAfo¨G depending on family background (in particular net household income in the previous year, parental education and occupational position), nationality, year and region. Secondly, the expected BAfo¨G amount among the beneficiaries was estimated with the GSOEP data as a function of the same variables. Finally, the share of BAfo¨G which takes the form of a repayable loan is also included. Indeed, at the time BAfo¨G was introduced, it was a mere subsidy, i.e. not repayable. However, from 1974 onwards an increasingly important part of the grant had to be reimbursed and in 1983, all of the BAfo¨G had to be reimbursed. The system was reformed again and since 1990, half of the BAfo¨G amount is a grant, half is a repayable loan. The results show that the prospect of being entitled to BAfo¨G seems to have a very strong positive influence on the probability of pursuing education. Also the amount granted by BAfo¨G plays a role in higher education attendance decisions, though to a lesser extent. Thus, the higher the amount of BAfo¨G individuals can expect to get, the higher the probability is that they are enrolled in education. Conversely, the BAfo¨G loan share has a negative impact on enrolments: the larger the part of the BAfo¨G to be reimbursed, the lower the

Participation in higher education 451

International Journal of Manpower 23,5 452

probability to be enrolled in higher education. This negative coefficient is explainable by the fact that if BAfo¨G has to be reimbursed after the end of the studies, this is expected to diminish the future return to education. .

4. Simulation of changes in expected returns and educational policy The coefficient estimates indicate the direction of the effects and their significance, but provide little information on the quantitative impact on the enrolment outcome of changes in the variables. However, the estimation results can be used to simulate the effect of changes in selected variables and assess their quantitative impact on the enrolment threshold and on the enrolment probability of a person with given observed characteristics. Indeed, changes in some variables will affect the expected ratio of marginal cost to marginal return of attending university. If the changes turn out to increase the expected ratio of costs to return, this will reduce the probability of enrolment in higher education accordingly. If the changes lower the ratio of costs to returns, this will raise the probability of participation in higher education. More formally, following equation (8), the enrolment threshold for attending education level E1 rather than E0 is given by the ratio of expected marginal cost to marginal return of attending university given characteristics x, which can be recovered from the ^1 and
^ coefficients drawn from the probit estimation: Enrolment threshold ¼

cðE1 Þ
cðE0 Þ 1 1 h i ¼ exp½^1    rðE1 Þ
rðE0 Þ ’ðxÞ exp
^x

¼ exp½^ 1

^x and the attendance probabilities are given by equation (6). In Table II, the effects of a 10 percent change in the labour market return expectations and in educational policy on the expected ratio of marginal cost to

Table II. Effect of a 10 percent increase in selected explanatory variables on the participation in higher education

Enrolment threshold

Enrolment probability

Reference situation

4.14

7.78

Changes in labour market returns Expected hourly wage return Expected unemployment probability Expected part-time employment probability Expected non-employment probability

3.42 4.11 4.33 4.42

10.96 7.89 7.16 6.85

Changes in educational policy Net other household income last year Expected chance of receiving BAfo¨G Expected monthly BAfo¨G amount BAfo¨G loan share

3.97 3.60 3.78 4.25

8.38 9.99 9.17 7.40

marginal return of higher education attendance and thus on the enrolment probability itself are simulated for an individual with average characteristics. As can be seen, for an individual with average characteristics, the expected ratio of marginal cost to marginal return of enrolment in higher education amounts in the reference model to some 4.14 and the enrolment probability predicted by the model to about 7.8 percent. Simulation of changes in labour market return expectations If the expected hourly wage return, i.e. the ratio of expected life-cycle wage for holders of a tertiary level degree to expected wages in the absence of such a degree, increases by 10 percent, this lowers the expected ratio of marginal cost to marginal return of enrolment in higher education by 0.7 points for an average individual and the enrolment probability accordingly increases by more than 3 percentage points. Also a 10 percent increase in personal unemployment risk drives the higher education enrolment threshold down and causes the enrolment probability of an average person to rise accordingly. Conversely, a rise of 10 percent in the propensity of an average person to work part-time induces a rise in the expected marginal cost-marginal return ratio and thus a lower probability to be enrolled in tertiary education. However, the effect on the enrolment probability of the propensity of being completely out of work is stronger, which is consistent with intuition. Simulation of changes in educational policy Let us imagine, for instance, that parents or any other household member were given an educational allowance to compensate for the foregone earnings of the potential student, which amounts to 10 percent of net other household income of the previous year. If everything else remained unchanged, this would lower the higher education enrolment threshold, but to a limited extent. As a result, the enrolment probability would only increase slightly. If the coverage of BAfo¨G grants/loans was extended so that the expected chance of an average individual of being entitled to BAfo¨G increased by 10 percent, this would significantly lower the expected marginal cost-marginal return ratio and induce an increase in the enrolment probability by about 2.2 percentage points. An increase of 10 percent in the BAfo¨G monthly amount the average beneficiary may expect to receive also lowers the enrolment threshold and increases the enrolment probability, but to a somewhat lower extent than the extension of BAfo¨G coverage. Therefore, at the same financial costs, extending BAfo¨G coverage proves more efficient in increasing enrolments than increasing the average BAfo¨G amount granted. Finally, raising by 10 percent the proportion of BAfo¨G which has to be reimbursed induces a small increase in the expected ratio of marginal cost to marginal return, but the quantitative effect is rather small.

Participation in higher education 453

International Journal of Manpower 23,5 454

5. Conclusion Few empirical studies analyse the role of economic incentives on educational decisions, especially for Germany. This study tried to examine whether cost and return considerations have an impact on participation in higher education in West Germany. The empirical analysis is based on a model in which the decision to attend one of several educational tracks of different levels depends on the cost and return expectation of the educational alternatives, given some personal characteristics. Thus, the individual assesses the additional cost of attending higher education and the additional return of doing so, and if the ratio of the former to the latter is below a certain threshold, he or she will choose to attend higher education. The results have been quantified in a simulation exercise of the impact of changes in selected variables on the enrolment threshold and on the enrolment probability itself. The empirical analysis shows that the probability of enrolment in higher education is strongly influenced by social origin. Parental education and occupational position, in particular, are essential. However, the enrolment probability also depends on labour market prospects expectations. The absolute level of personal unemployment risk appears to be a strong incentive to participate in higher education, more than the reduction of the unemployment risk due to a higher educational degree. As far as wages are concerned, the expected return to education in terms of life-cycle wages affects significantly educational decisions, whereas the level of expected wages proves insignificant. A 10 percent change in the expected wage return to higher education was simulated and appears to reduce significantly the expected ratio of cost to return of an average person, i.e. his enrolment threshold, and raise significantly the enrolment probability. A higher risk of being employed parttime and even more of being out of work proved to reduce the utility of higher education and thus reduce the probability of being enrolled in higher education. Whereas the overall level of public expenditure for each student engaged in tertiary education did not prove to have a significant impact, there seems to be evidence that policy measures more specifically directed to potential students do have an impact. In particular, the simulation using the estimation results shows that, at the same financial costs, extending the coverage of public financial support in the form of the BAfo¨G is expected to be more efficient in increasing enrolments than increasing the amount of BAfo¨G granted. The extent of the repayable part of the financial aid, conversely, has a dampening, though limited, influence on enrolments. On the whole, the analysis suggests that cost and return aspects do affect the decision to participate in higher education in Germany. In particular, coming from a poor family background seems to reduce significantly the chances to participate in higher education. Moreover, the results suggests that the labour market outcomes individuals may expect from higher education also play a role, and that they respond to some extent to financial incentives such as policies of financial support for education in the form of BAfo¨G.

Notes 1. A shorter and non-technical version of this paper has been published in the journal Education + Training, see Lauer (2002a). 2. A problem with the transition approach is that if the completion of a high school degree is an eligibility condition for completing university studies, part of the decision to enrol in higher education has been made at previous stages of the educational career. Thus, focussing on the selected sample of high school graduates eligible for higher education may not be really informative with regard to the determinants of educational achievement. 3. Or of his parents. 4. We have at the extremes 0 = –1 and k = +1. 5. Alternative definitions of the age span have been tested and proved not to affect the results significantly. 6. Total net household income minus own net income. 7. Alternatively, one could simply take the average wages or rates of unemployment in the region or by gender. The procedure applied here is not really different, but has the advantage that it allows to differentiate the averages according to a larger number of variables and to compute personal expectations depending on a certain number of characteristics. A similar approach was adopted by Wilson et al. (2000). 8. Gross hourly wage deflated with the consumer price index, like all other variables expressed in DM amounts. 9. The choice of the explanatory variables is restricted by the necessity of using variables which are also available for the sample of young individuals. All other variables (e.g. marital status, labour market experience) are captured by the error term, which implies that individuals are assumed to adopt an average behaviour regarding these variables. 10. A real discount rate of 3 percent was used. Further tests with alternative discount rates did not appear to change the results significantly. 11. Between ages 19 and 55. 12. Owing to the loglinear functional form of the wage equation and assuming that the residuals are normally distributed, the prediction is given by exp(
^X + 12 2), where
^ is the vector of estimated coefficients, X the vector of explanatory variables and  the standard error of the prediction (cf. Greene, 1993, p. 60). 13. For unemployment, the ratio has been defined the other way round. The reduction of the unemployment risk thanks to higher education can be seen as a return to education. 14. For the sample of university graduates, the expected outcomes are estimated from age 26 and not 19 like for the less qualified. This accounts for the different lengths of studies and for the opportunity cost associated with longer studies. References Acemoglu, D. and Pischke, J.S. (2001), ‘‘Changes in the wage structure, family income and children’s education’’, European Economic Review, Vol. 45 No. 4/6, pp. 890-904. Asplund, R. and Pereira P.T. (1999), Returns to Human Capital in Europe, A Literature Review, B Series Bd. 156 , Helsinki. Becker, G. (1964), Human Capital: A Theoretical and Empirical Analysis, with Special Reference to Education, National Bureau of Economic Research, New York, NY. Blossfeld, H.P. (1993), ‘‘Changes in educational opportunities in the Federal Republic of Germany – a longitudinal study of cohorts born between 1916 and 1965’’, in Shavit and Blossfeld (Eds), Persistent Inequality – Changing Educational Attainment in Thrirteen Countries, Westview Press, Boulder, CO.

Participation in higher education 455

International Journal of Manpower 23,5 456

Bogess, S. (1998), ‘‘Family structure, economic status and educational attainment’’, Journal of Population Economics, Vol. 11, pp. 205-22. Cameron, S. and Heckman, J. (1998), ‘‘Life-cycle schooling and dynamic selection bias: model and evidence for five cohorts of American males’’, Journal of Political Economy, Vol. 106 No. 2, pp. 262-331. Gang, I. and Zimmermann, K. (2000), ‘‘Is child like parent? Educational attainment and ethnic origin’’, Journal of Human Resources, Vol. 35 No. 3, pp. 550-69. Goux, D. and Maurin, E. (1998), Social Destinations: The Role of Education and Social Origin, Insee Studies 18. Goux, D. and Maurin, E. (1999), La mobilite´ sociale et son e´volution : le roˆle des anticipations re´examine´, document de travail du CREST 9917. Greene W.H. (1993), Econometric Analysis, 2nd ed., Prentice-Hall, New York, NY. Hill, M. and Duncan, G. (1987), ‘‘Parental family income and the socioeconomic attainment of children’’, Social Sciences Resources, Vol. 16, pp. 37-73. Hilmer, M. (1998), ‘‘Post-secondary fees and the decision to attend a university or a community college’’, Journal of Public Economics, Vol. 67, pp. 329-48. Kodde, D. (1988), ‘‘Unemployment expectations and human capital formation’’, European Economic Review, Vol. 32, pp. 1645-60. Lauer, C. (2002a), ‘‘Enrolments in higher education: do economic incentives matter?’’, Education + Training, Vol. 44 No. 4/5, pp. 179-85. Lauer, C. (2002b), Family Background, Cohort and Education – A French-German Comparison, ZEW Discussion Paper 02-12. Manski, C., Sandefur, G., McLanahan, S. and Powers, D. (1992), ‘‘Alternative estimates of the effect of family structure during adolescence on high school graduation’’, Journal of the American Statistical Association, Vol. 87 No. 417. Merz, M. and Schimmelpfennig, A. (1999), Career Choices of German High School Graduates: Evidence from the GSOEP, EUI Working Papers 99/11. Mincer, J. (1974), Schooling, Experience and Earnings, National Bureau of Economic Research, New York, NY. Mingat, A. and Tan, J.P. (1998), The Mechanics of Progress in Education – Evidence from CrossCountry Data, Document de travail de l’IREDU 98/01. Schultz T. (1988), ‘‘Expansion of public school expenditure and enrollments: intercountry evidence on the effects of income, price and population growth’’, Economics of Education Review, Vol. 7 No. 2, pp. 167-83. Shea, J. (2000), ‘‘Does parents’ money matter?’’, Journal of Public Economics, Vol. 77, pp. 155-84. Solon, G. (1992), ‘‘Intergenerational income mobility in the United States’’, American Economic Review, Vol. 82, pp. 393-408. Wilson, K., Wolfe, B. and Haveman, R. (2000), ‘‘The role of expectations in adolescent schooling choices: do youths respond to economic incentives?’’, paper presented at the Conference of the International Institute of Public Finance, Sevilla, August 2000.

Appendix Variable

Mean (s.d.a)

Enrolled in higher education 0.133 Age 22.7 (1.7) Age squared 516.2 (77.4) Male 0.51 Foreign 0.11 Trend 7.3 (3.6) Trend squared 66.5 (56.1) Schooling mother 10.2 (1.7) Schooling father 11.1 (2.3) Father white collar 0.22 Father civil servant 0.10 Father self-employed 0.14 Net other hh. income last year/1,000 2.80 (2.3) Expected hourly wage (net present value) 20.9 (4.8) Expected hourly wage return (idem) 1.16 (0.11) Expected unemployment probability 0.067 (0.03) Expected relative unemployment probability 2.3 (1.8) Expected part-time employment probability 0.192 (1.8) Expected relative part-time employment probability 0.84 (0.50) Expected non-employment probability 0.183 (0.15) Expected relative non-employment probability 3.75 (4.97) Expected self-employment probability 0.086 (0.05) Expected public employment probability 0.266 (0.09) Expenditure higher education by student 17.57 (4.34) Students-teacher ratio last year 14.39 (3.06) Expected chance of receiving BAfo¨G 0.34 (0.16) Expected monthly BAfo¨G amount 498.2 (79.3) BAfo¨G loan share 74.5 (23.7) GDP per head/1,000 34.3 (5.62) Ratio of pupils/students to total population 0.17 (0.02) Note: a Standard deviation

Minimum

Maximum

0 20 400 0 0 2 4 7 7 0 0 0 0 11.3 0.80 0.011 0.9 0.001 0.01 0.017 0.63 0.020 0.054 11.71 5.97 0.00 220.7 49.4 24.0 0.06

1 25 625 1 1 14 196 18 18 1 1 1 57.9 39.9 1.48 0.245 14.4 0.558 2.67 0.449 69.58 0.417 0.606 39.75 22.00 0.78 798.9 99 62.2 0.21

Participation in higher education 457

Table AI. Summary statistics

The research register for this journal is available at http://www.emeraldinsight.com/researchregisters

International Journal of Manpower 23,5 458

The current issue and full text archive of this journal is available at http://www.emeraldinsight.com/0143-7720.htm

Labour market expectations of Swiss university students Stefan C. Wolter Swiss Coordination Centre for Research in Education (SKBF), Department of Economics and IZA, University of Berne, Bonn, Switzerland, and

Andre´ Zbinden Department of Economics, University of Berne, Berne, Switzerland Keywords Re-training, Education, Wages Abstract Labour market expectations and especially wage expectations are important determinants for individual schooling decisions. However, research on individual expectations of students is scarce. The paper presents the Swiss results of a survey that was conducted in ten European countries. Its main findings are that point estimates of wages after graduation are close to actual wages, whereas the expectations of the wage gain in the first ten years of professional experience exceed the actual wage gains significantly. We find that rates of return to education that are calculated on the basis of individual wage and cost expectations as well as individual time preferences can be explained partially by the seniority of students, the selfperception of their academic performance and their subjective job perspectives.

Introduction Studies about individual earnings or wage expectations are, although of widely recognised importance, scarce in economic literature. As a part of the European Project on the Rates of Return to Education in 15 European Countries, PURE[1], researchers from ten countries collected information on wage expectations of university students in different fields of study (see Brunello et al., 2001). The present paper goes into the details of the Swiss data gathered in this project and extends the analysis to the question of rates of return to education by constructing individual life-earnings profiles of the interviewees. A short review of the literature The literature of these kinds of empirical investigations of individual expectations was almost entirely American for a long time (with the exception of Dolton and Makepeace, 1990). Most recently a few researchers have tried to replicate the US studies in Europe. The known studies differ considerably in respect to the methods applied to find out about expectations as well as to the underlying research questions. Most of the questionnaires were written questionnaires – as in this case – with the exception of the computer-based

International Journal of Manpower, Vol. 23 No. 5, 2002, pp. 458-470. # MCB UP Limited, 0143-7720 DOI 10.1108/01437720210436064

The authors thank Rudolf Winter-Ebmer for the provision of the questionnaire as well as for the data compilation, Josef Zweimu¨ller for providing the data from the University of Zurich, Bernhard A. Weber for his assistance and Rainer Winkelmann for helpful advice. The authors also benefited from inputs at the European Conference for Educational Research in Lille and the annual conference of the Swiss Society for Educational Research. The usual disclaimer applies.

questionnaire of Dominitz and Manski (1994a,b, 1996, 1997 and also Dominitz, 1998). The older studies in particular (Smith and Powell, 1990 or Blau and Ferber, 1991) tried to find out how good students could predict the current wage level of different worker categories. The study of Betts (1996) kept the tradition, insofar as students were asked to predict the wage level of different groups of graduates and different points in time of their working life. Dominitz and Manski (1996), as well as Wolter (2000) in its replication of the US study, used different scenarios in which students were asked to predict their own (future) wage level as well as the wage level of an average person with the same characteristics (educational level, age, gender). Additionally these studies tried to elicit the individual uncertainties about wage distributions, instead of only asking for point estimates. Whereas the cited studies had to compare their survey data with actual cross-sectional wage data in order to get a picture of the ‘‘accuracy’’ of the students’ responses, the next two studies applied a different method. Webbink and Hartog (2000) used the Dutch longitudinal ‘‘Scholar’’ project in order to ask students for personal predications of starting salaries and compared them with the actual starting salaries the students got after graduation. Caravajal et al. (2000) used a small sample of students in Florida to predict salaries of graduates, which they compared with data of actual graduates of the same school participating in the same survey. Nicholson and Souleles (2001) recently published a study about income expectations of physicians and were able to show that specialty choices after graduation could be explained significantly by differences in their income expectations. Brunello et al. (2001), finally, used a written survey similar to the one of Betts (1996) but asked students about personal expectations regarding salaries for two scenarios (university degree and university entrance degree) and two points of time in their professional biography (time of graduation and ten years later). Owing to the differences in methodology and the purposes of the studies, the results are difficult to compare. We will therefore make the necessary references in the present paper. Purpose Most papers in the past have analysed very specific categories of students, in most cases students of economics or business administration. Additionally, most studies used only students from one school or university and some of the studies had rather small samples. The main purpose of this paper is to overcome some of the shortcomings of the existing papers by analysing whether there are significant differences in expectations between students of different faculties and disciplines when using a large sample of students of more than one university or school. Contrary to the existing literature, we do not limit ourselves to point estimates of wages or wage gains in a very short period of an individual work biography but extend our analysis in the second part of the present paper by attempting to construct real rates of return to education. We then analyse the determinants of differences in the expected rates of return to education.

Labour market expectations

459

International Journal of Manpower 23,5 460

The data The data collection was done in the winter term of the academic year 1999/2000 at the Universities of Berne and Zurich. These two universities have a total of some 30,000 students enrolled. In order to get comparable data the same two-page questionnaire was distributed as in all other countries of the project (see Brunello et al., 2001 for the detailed questionnaire) and adapted to the different terminology in Switzerland. The questionnaire asked for personal information (socio-economic background), information on the study behaviour, wage expectations with two different scenarios, questions related to sources of wage information, costs of study and included a little test to elicit individual time preferences. The data were collected according to the following procedure: courses in different fields of study were selected randomly and lecturers asked for permission to use their course for administering the questionnaire. Once this permission was given, the questionnaires were distributed. The rate of participation was 100 per cent (no selection problems due to non-responses) and 1,133 questionnaires were returned (54 per cent of the questionnaires came from the University of Berne). Owing to difficulties getting permission to run the survey in non-economic faculties, there is an oversampling of economics students (60.8 per cent) in the survey. In total, 16.2 per cent were law students, 8.7 per cent studied human and social sciences, 8.3 per cent medicine, 4.5 per cent technical, natural and computer sciences, the remaining students (1.5 per cent) could not be attributed to the faculties mentioned. The high proportion of economics students provoked also an oversampling of men (roughly two-thirds of the interviewees). The majority of students (84 per cent) were in the first semester of their studies (16 per cent started before 1999). The raw data were screened for outliers, missing data and inconsistent or illogical data, which resulted in a considerable reduction of the questionnaires finally analysed (depending on the use, between 25 and 40 per cent of the questionnaires had to be dropped, thus the written questionnaire resulted in a much greater waste of information than comparable surveys using computer-based survey methods). Even though all reductions were tested, none resulted in a significant change in the means of the data and all reduced the standard deviations, therefore it is fair to speak of ‘‘robust’’ data. Results Every individual had to make four predictions about salaries. The first two estimates concerned monthly entry salaries after having obtained a university entrance degree (UED) and completed university (UNI). Then they had to predict the wage rise in the first ten years of work for both educational levels, which meant monthly wages at the age of about 30 (UED) and 35 (UNI). A first descriptive analysis of the data shows that the heterogeneity of the sampled data is somewhat larger than in the Swiss study of Wolter (2000) but smaller than in the US studies of reference (Betts, 1996 and Dominitz and Manski, 1996). The standard deviation of expected salaries amounts to some 24 per cent of the mean for the starting salary after graduation from university (Wolter, 2000, 20 per cent; Betts, 1996, 28 per cent). Brunello et al. (2001) argue that the measured

heterogeneity of expectations is larger than the actual observed spread of wages in the labour market. While this is also true for the Swiss data sampled in this study, their result is not comparable with the results of Wolter (2000), who found that the test persons in his study underestimated wage inequality in Switzerland. In the latter study the real earnings distribution was compared with the subjective uncertainty about earnings (individual expectations of the wage distribution) and not to inter-individual differences in expectations. Point estimates Every test person had to give information about four data points, which can be compared with actual wage data observed on the labour market. When comparing the expectations with the actual wage data, several problems arise: . There are various possibilities to measure the deviations of expectations from actual wage data. We will present our results in the form of the mean signed error and also in the form of the absolute (mean signed) error. . Actual wage data can be collected from different sources. For all wages we use the Swiss Labour Force Survey (SLFS). This survey has the advantage of giving us wages for all ages of the workers but does not allow differentiation according to fields of study. Additionally, the number of people with a specific degree at a certain age is rather small in the SLFS and the reliability of the data can therefore be questioned. In order to have more detailed information we therefore also use the Swiss Graduate Survey (Diem, 2000) for the time of graduation from university. . Comparisons with actual wage data do not reflect the ‘‘accuracy’’ of expectations. We can only test the extent to which student expectations differ from the current wage structure. We also abstract inflation and suppose, as Manski (1993, p. 49), that students form their expectations ‘‘in the manner of practising econometricians’’ and do not consider inflation when forming their expectations. When calculating the mean signed error and comparing expectations with the SLFS data, we find that all students significantly overestimated their wages at the time of graduation from university and that only their wage expectations in the case of the UED are accurate (always compared with actual wage data). After ten years of professional experience, students tend to underestimate significantly the salaries they would obtain with a UED and still significantly overestimate the salaries they earn with a university diploma. In other words, students overestimate significantly the premium they will get for a university diploma by overestimating both the difference between the absolute wage levels of UED and UNI and the increase in wages in the first ten years of professional experience (see also Brunello et al., 2001, Tables 4 and 5). Comparing the expectations with data from the Graduate Survey 1999 (Diem, 2000), we find that at the time of graduation from university, students of economics and medicine have similar expectations to the wages observed in the

Labour market expectations

461

International Journal of Manpower 23,5 462

survey, while students of human and social sciences as well as of natural, technical and computer sciences expect wages that are significantly higher than those reported in the Graduate Survey. Law students are a particular case, since only a proportion of them really enters the labour market after graduation. Those pursuing a career as a lawyer have to complete different spells at court and in law firms. During these two years they earn, on average, less than one third of the salary their colleagues earn entering the labour market directly. The expectations reported in the survey are a mix of both cases and therefore difficult to interpret. The mean signed error naturally obliterates part of the information. The accuracy of students’ beliefs is therefore better measured with the absolute (mean signed) error (aei ), wherein overestimates and underestimates do not cancel each other out. By specification of the absolute value of the percentage wage error we applied the logarithmic error, based on the same arguments as Betts[2]. 
 exp  ðwi
wactual Þ
100  i  i ¼1
n ð1Þ aei ¼ ln   wactual i In Table I we present the absolute error; the closer the values are to one, the smaller are the deviations of expectations from actual wages. The results can only be interpreted qualitatively but clearly show that expectations are much closer to actual wages at the time of graduation and that the deviations from the current labour market data occur when predicting the increase of salaries in the first ten years of professional experience. In Table II we show a regression, as in Wolter (2000, p. 63), to check whether or not students with specific characteristics have significantly larger or smaller deviations from the actual wages observed on the labour market. Four types of variables are tested as explanatory magnitudes: (1) sociodemographic characteristics; (2) study behaviour; (3) choice of field of study; and (4) sources of information about salaries. t0 represents the absolute error of expectations at the time of graduation, t10 the absolute error after ten years of professional experience. The explanatory power of all categories of independent variables is rather low in general. As already stated in Wolter (2000), the low R2 values indicate that

Table I. Absolute mean signed error (aei )

Graduate salary UED Graduate salary UNI Salary UED after ten years of professional exp. Salary UNI after ten years of professional exp.

Full group

Men

Women

0.991 0.975 1.212 1.286

0.985 0.993 1.204 1.329

1.002 0.937 1.231 1.189

UED Independent variables Constant

UNI

t0

t10

t0

1.1058*

1.2709*

–0.1273

1.0122*

t10

Sociodemographic characteristics Women Age Father with university degree Mother with university degree In the same field of study as father In the same field of study as mother

–0.0065 0.0021 0.0499 –0.0365 –0.0162 0.1570

0.0233 –0.0017 0.0412 0.0300 0.1387* 0.1535

0.0059 0.0085 –0.0247 0.1283 0.0065 –0.4795

–0.1214* 0.0057 0.0031 –0.0060 –0.1023 0.0637

Education University of Berne (dummy) Started university before 1998 Started university in 1998 Higher than average study perf. Works as well as studies Job perspective after graduation Relative job perspective (UNI/UED)

–0.0312 –0.1439** 0.1135 0.0506 –0.0036 –0.0575** 0.0050

–0.0093 0.0475 –0.1039 –0.0194 0.0111 –0.0248 0.0129

0.0679 –0.1585 0.0089 –0.0890 0.0314 0.1951* 0.0512

–0.0734 –0.1015 –0.0464 0.0675 0.0036 0.0561 –0.0055

Field of study (economics as reference category) Social and human sciences Law Medicine Natural, technical and computer sc. Other

–0.0647 –0.0615 0.1952* –0.0035 0.0732

0.1533** 0.0512 0.0593 –0.1719 –0.0017 –0.2066 0.0911 0.0357 0.1663** –0.0349

Sources of information University career centre Friends and colleagues Daily and weekly print media University publications Specialised salary reports No information consulted

0.0038 –0.0064 –0.0141 0.0100** 0.2063** 0.1075

0.0979** –0.0616 –0.0218 0.0172 –0.0810 –0.0288

Number of observations Mean dependent variable Adjusted R2

851 1.00 0.024

773 1.20 0.018

Labour market expectations

463

–0.0618 –0.0681 0.1392 0.0188 –0.2544

–0.0237 0.0488 –0.0264 –0.0142 –0.0291 –0.0180 –0.0629 0.0094 0.3176** 0.1425 –0.0050 –0.0081 840 0.98 0.023

784 1.28 0.030

Notes: Significant variables are in italics and with asterisks standing for the 1 per cent significance (*) and 5 per cent significance (**) respectively. White heteroskedasticity-consistent standard error and covariance

deviations from actual wages can not be explained by single characteristics or group effects. Sociodemographic factors do not explain deviations from the current wage levels, except for women in one case and for students studying in the same field as their father. Women tend to be more cautious regarding the increase of salaries in the first ten years of professional experience, whereas men tend to be – compared with the actual wage level – overly optimistic. As regards educational behaviour, we find that the fact that some students work as well as study and therefore have a higher exposure to the labour

Table II. OLS regression on ‘‘absolute mean signed error’’

International Journal of Manpower 23,5 464

market has no effect on the aei . Students who started their studies before 1998 have smaller aei when expecting UED salaries but no significant difference when it comes to UNI salaries. Students who believe they have good job perspectives after graduating have smaller aei for UED salaries at the time of graduation, but tend to have higher expectations and therefore a higher aei when it comes to starting salaries after graduation from university. Differentiation according to fields of study shows that in some but not all cases students have larger aei than students of economics when it comes to predicting salary levels of UED graduates. In respect to UNI salaries, however, there are no significant differences in the aei . Furthermore, students indicating that they used specific sources of information about salaries had, contrary to what one might expect, larger deviations from the current wage level than those who did not consult such sources. In the case of specialised salary reports, students having had access to such studies had either (in the case of UNI-salaries) higher expectations or (in the case of UED-starting salaries) a higher variance in expectations, resulting in larger absolute mean signed errors. It has to be noted, however, that the largest differences do not concern the salaries of university graduates but UED-starting salaries, for which only little public information exists. Considering all four categories of independent variables, the only true surprise seems to be the non-significance of the salary-information consulting behaviour of students and of the educational background of parents. Better informed students or students with easier access to information on salaries do not seem to have expectations that are closer to the current average than those who do not possess these advantages. Rates of return to education In empirical applications of the human capital theory, simple wage differentials between a lower and a higher educational level are sometimes taken as the best proxy for rates of return to education (see e.g. Wilson et al., 2000; Lauer, 2000). We know however, that in reality true rates of return are influenced by a number of factors that could differ between students and groups of students. Individual time preferences may differ as well as direct costs of studying, the length of study or the timing of study. Therefore differences in salary expectations alone may not be an adequate substitute for rates of return expectations. We test in the following, whether differences in expected rates of return to education depend on factors other than differences in salary expectations. Although the questionnaire provided most of the needed data, one important element, namely the expected salaries beyond the first ten years of professional life, was missing. In order to calculate rates of return to education, we had to construct these years artificially. First, we assumed that between graduation and ten years later the salaries increased linearly. Second, we calculated an extended Mincerian wage equation with data from the SLFS as follows, for men and women separately:

Ln wi ¼
þ 1 Si þ 2 Exp Si þ 3 Exp2 Si þ 4 Xi þ 5 Expi þ 6 Exp2 i þ "i : ð2Þ S is a dummy for the school level, X a vector for control variables, Exp stands for experiences Exp2 for experience squared and "i for the error term. Contrary to the traditional Mincer-equation we also estimated the values of an interaction variable (Exp Si ) of school level and experience and experience squared. These two variables were then used to calculate school type specific individual earnings profiles. The individual earnings profiles had the following components: . they started either at the age of 20 for UED (t) or at the expected age of graduation (for UNI) (SUNI ) with the expected wage; and . continued for the next ten years at the expected rate of wage increase; . continued after that point with an annual increase calculated with the help of equation (2) and stopped for all individuals at the age of 65 (P) (official age of retirement in Switzerland). The individual earnings profiles were inserted in a type of cost-benefit model (see e.g. Wolter and Weber, 1999) of the following type to calculate the net present value (NPV) of a university study: NPV

UNIi

¼

P X 

WUNI
WUED

  t

1þi

t¼SUNI þ1




SUNI X



CUNI

  þ WUED t  1 þ iÞ
t :


t ð3Þ

t¼1

The direct costs of study (CUNI ) were also individual expected values taken from the questionnaire. Discounting was done at two different rates (i). In the first calculation we used a 5 per cent discount rate for all individuals, in the second calculation the individual discount rate resulting from the questionnaire. In the questionnaire students were asked whether they would be better or worse off if they received a gift of 1,020 Euros in one year than a colleague who receives 1,000 Euros now. The amount given to them in one year was subsequently augmented step by step to a final amount of 1,120 Euro. The point where students switched from ‘‘worse off’’ to ‘‘better off’’ was interpreted as the individual time preference (see Table III). The manner in which questions were asked allowed for two corner solutions in cases where students felt that they were always better off or always worse off. This leads to a potential underestimation of the mean of the reported discount rate. The mean is also significantly lower than values found in comparable studies (see e.g. Oosterbeek and van Ophem, 2000). Some students switched more than once – behaviour that is not rational – and were therefore eliminated from the data set.

Labour market expectations

465

International Journal of Manpower 23,5 466

With the help of the net present value we finally calculated a form of relative income advantage over the life time, called relative rate of return (subsequently RRE)[3]: 
NPV UNIi RRE i ¼ P  100
100: ð4Þ P
t  ðW UED Þt ð1 þ iÞ t¼1

RRE(1) was calculated on the base of the uniform discount rate of 5 per cent and RRE(2) with the individual discount rate. This relative rate of return to education is not directly comparable to similar calculations (e.g. Wolter and Weber, 1999; Weber et al., 2001) because of some limitations in the factors that were taken into account and also because the expected wages are gross wages before taxes. Therefore we limit ourselves to the analysis of determinants of differences in the expected rates of return to education. The results are presented in Table IV. The set of explanatory variables is the same used in the calculations presented in Table II, with the addition of a variable on smoking behaviour. Some researchers use smoking habits as an indicator for individual time preferences (see e.g. Festerer and Winter-Ebmer, 2000). We tested whether the inclusion of this alternative measure for time preference had an additional explanatory value besides the time preferences used in our model. According to our results the variable was never significant. The interpretation of the results in Table IV is done according to the categories of independent variables: . Sociodemographic characteristics. Apart from the age of the students, no other factor seems to have an influence on the expected size of the rates of return to education. The factor age is an artefact of the way the RRE were calculated rather than a difference in expected RREs. Older students have a shorter period to compensate for the investments, and therefore have smaller RREs. The significant negative effect therefore only proves that older students apparently do not expect higher wages that would compensate for the shorter period of earnings. Women have significantly lower wage and wage gain expectations resulting in a smaller, although not significant, expected RRE(1). At the same time Group

Table III. Time preference – individual discount rates

Economics Social and human sciences Law Medicine Natural, technical and computer sciences Women

Discount rate

t-value*

7.886 6.767 7.772 7.618 6.260 7.120

–2.005 –0.213 –0.528 –2.223 –2.511

Note: * t-value at the 5 per cent level of significance and economics students as the reference group; men are the reference group for women with a discount rate of 7.888

RRE(1) Independent variables

Coef.

Constant 13.83 Sociodemographic characteristics Women –4.29 Age –2.84* Father with university degree 4.22 Mother with university degree 6.34 In the same field of study as father 8.14 In the same field of study as mother –7.87 Education University of Berne (dummy) 9.37** Started University before 1998 –22.63* Started University in 1998 –3.95 Higher than average study perf. 12.98* Works besides studying –2.96 Job perspective after graduation 10.13* Relative job perspective (UNI/UED) 10.79* Field of study (economics as reference category) Social and human sciences 3.26 Law –25.00* Medicine –0.71 Natural, technical and computer sc. 2.95 Other 2.05 Sources of information University career centre 8.50 Friends and colleagues 0.75 Daily and weekly print media –2.08 University publications 0.51 Specialised salary reports 16.45 No information consulted 10.01 Time preference Smoking at the age of 17/18 –4.17 Number of observations 618 Mean dependent variable 37.58 Adjusted R2 0.15

RRE(2) SD 20.83

Coef. 1.987

SD 20.99

4.23 0.65 3.92 7.05 7.90 17.78

0.75 –1.61** 2.90 3.16 13.57 –17.84

3.97 0.69 4.45 6.45 8.38 15.05

4.56 6.52 6.25 3.71 3.64 3.39 2.19

9.69** –21.57* –0.27 10.35** –5.36 2.92 8.36*

4.62 7.59 7.56 4.16 4.10 2.93 2.28

11.60 6.14 8.89 9.54 7.28

–10.45 –27.02** 18.30 0.30 –10.99

8.70 11.18 9.37 10.12 10.42

8.23 3.96 3.97 5.63 11.06 5.87

7.47 1.22 –3.53 0.94 9.99 13.31**

10.06 4.06 4.16 6.70 16.71 6.70

4.44

3.92 481 25.04 0.10

467

4.03

Notes: Significant variables are in italic letters, with asterisks standing for the 1 per cent significance * and 5 per cent significance ** respectively. White heteroskedasticityconsistent standard error and covariance

.

Labour market expectations

they have significantly lower expected discount rates (see Table III), which leads to a positive sign (also non-significant) of the coefficient in the RRE(2) equation taking the individual time preferences into account. Education. Students studying at the University of Berne have higher expected RREs than students of the University of Zurich. This structural difference is difficult to explain but shows the necessity to include more than one school in a survey. Older students have

Table IV. OLS regression on ‘‘rates of return to education’’ (RRE)

International Journal of Manpower 23,5 468

.

.

significantly lower RREs than ‘‘fresh students’’ in their first semester. Students, who expect to have very good job perspectives after graduation have a higher RRE(1) but not in the case of RRE(2); therefore they must also have a higher time preference, maybe because they are more job and labour market oriented. Those expecting better job perspectives after graduation from university, relative to the perspectives after the UED, have higher RREs in both cases. Also significant is the self-evaluation of one’s academic performance. Students had to mark their personal academic performance relative to their colleagues on a scale between 1 (very good) and 6 (very poor). The average performance of all students was significantly below 3. This result itself is not surprising as it is a standard observation in psychological (e.g. Lichtenstein and Fischhoff, 1977) and economical (e.g. Schmalensee, 1976) experiments. More interestingly is the observation that those believing to be better performers also expect to have the positive financial effects associated with their academic superiority. Additionally we find that – like in other studies (e.g. Wolter, 1996) – women tend to have a more modest view of their relative performance. Their under-representation in this category could also explain partially the non-significance of the gender dummy. This would mean that the less optimistic self-perception of women is more important than the fact of their being a woman. Field of study. Apart from the field of law all other students do not differ significantly in their expectations from students of economics. The ‘‘law’’ effect was already explained in this paper and should not be over interpreted. Interesting is the observation that perhaps against intuition, students of social and human sciences have expectations that are not lower than those of economics students when controlled for other observable differences. At least two factors seem to explain this. First, compared to actual wage data, students of this field tend to overestimate their salary perspectives and secondly, even when expecting lower salaries than economists, this is compensated with significantly lower time preferences (see Table III). Sources of information. Information about salaries does not play a prominent role, with the exception of those cases where no specific sources about salaries were consulted. These students with no specific information about salaries tend to have – at least in the case of RRE(2) – significantly higher expectations than the rest.

Conclusions This study on salary expectations of students in Switzerland shows that expected wages after graduation at university are rather accurate compared to crosssectional data from different sources. Significant deviations from actual wages can be observed for law students, students of social and human sciences and students of natural, technical and computer sciences, whereas students of economics and

medicine come close to actual wages. Expectations of wage gains during the first ten years of professional experience, however, show consistently higher expected gains than actual gains, with women having the lowest degree of overestimation. Consequently the rates of return to education that were constructed on the base of the reported expectations are also higher then conventional results on the base of labour force survey data. However, significantly higher time preferences and other factors reduce somewhat the degree of overestimation. As in other studies, the found degree of heterogeneity of expectations is large and can not be explained by group specific effects or other observable differences between students. Whereas differences in point estimates of wages display the highest degree of individual – non-attributable – differences, the expected rates of return to education can be explained somewhat by perceived job prospects, the self-assessment of academic performance and the seniority of students. Women expect significantly lower wages and lower wage gains then men. These expectations are partially justified by actual wage data. There is, however no significant gender effect concerning the expected rate of return to education, which is a consequence of a significantly lower time preference of women and women having a less optimistic self-assessment of their academic performance. The observed differences, both in wage expectations and in time preferences between students of different fields of study as well as the unexplained high degree of heterogeneity of individual expectations show the need for further research. Bigger samples covering more institutions and more fields of study are needed for a better understanding of how students form their expectations about their personal future after graduation. Only with this information can we better explain and understand individual schooling decisions. Notes 1. For more information see www.etla.fi/PURE 2. ‘‘The log absolute error is used as the dependent variable since the absolute values themselves are positive, which would have rendered the normality assumption used for inference in OLS highly untenable’’ (Betts, 1996, p. 42). In the case that expected salaries = actual salaries, the log would be impossible. As this was never the case, we were able to apply the above formula. 3. Note that the relative advantage in life income that was calculated here is not directly comparable to a normal rate of return calculation but that our decision to take the life income advantage instead of the internal rate of return does not affect the calculations presented in Table IV. References Betts, J.R. (1996), ‘‘What do students know about wages? Evidence from a survey of undergraduates’’, The Journal of Human Resources, pp. 27-56. Blau, F. and Ferber, M. (1991), ‘‘Career plans and expectations of young women and men’’, Journal of Human Resources, pp. 581-607. Brunello, G., Lucifora, C. and Winter-Ebmer, R. (2001), ‘‘The wage expectations of European college students’’, Quaderni dell’Istituto di Economia dell’Impresa e del Lavoro, No. 30, Universita’ Cattolica del Sacro Cuore, Milan.

Labour market expectations

469

International Journal of Manpower 23,5 470

Caravajal, M.J., Bendana, D., Bozorgmanesh, M.A., Castillo, K., Pourmasiha, Rao, P. and Torres, J.A. (2000), ‘‘Inter-gender differentials between college student’s earnings expectations and the experience of recent graduates’’, Economics of Education Review, Vol. 19, pp. 229-43. Diem, M. (2000), Von der universita¨ren Hochschule ins Berufsleben, Absolventenbefragung 1999, Bundesamt fu¨r Statistik, Neuenburg. Dolton, P.J. and Makepeace, G.H. (1990), ‘‘The earnings of economics graduates’’, The Economic Journal, Vol. 100, pp. 237-50. Dominitz, J. (1998), ‘‘Earnings expectations, revisions, and realizations’’, The Review of Economics, pp. 374-88. Dominitz, J. and Manski, C.F. (1994a), ‘‘Eliciting student expectations of the returns to schooling’’, NBER Working Paper, No. 4936. Dominitz, J. and Manski, C.F. (1994b), ‘‘Using expectations data to study subjective income expectations’’, NBER Working Paper, No. 4937. Dominitz, J. and Manski, C.F. (1996), ‘‘Eliciting student expectations of the return to schooling’’, The Journal of Human Resources, pp. 1-26. Dominitz, J. and Manski, C.F. (1997), ‘‘Using expectations data to study subjective income expectations’’, Journal of the American Statistical Association, Vol. 92 No. 439, pp. 855-67. Festerer, J. and Winter-Ebmer, R. (2000), Smoking, Discount Rates, and Returns to Education, mimeo, University of Linz, Linz. Lauer, C. (2000), ‘‘Enrolments in higher education in West Germany’’, discussion paper, No. 00-59, ZEW, Mannheim. Lichtenstein, S. and Fischhoff, B. (1977), ‘‘Do those who know more also know more about how much they know’’, Organizational Behavior and Human Performance, pp. 159-83. Manski, C.F. (1993), ‘‘Adolescent econometricians: how do youth infer the returns to schooling?’’, in Clotfelter, C.T. and Rothschild, M. (Eds), Studies of Supply and Demand in Higher Education, The University of Chicago Press, Chicago, IL, pp. 43-60. Nicholson, S. and Souleles, N.S. (2001), ‘‘Physician income expectations and specialty choice’’, NBER Working Paper, No. 8536. Oosterbeek, H. and van Ophem, H. (2000), ‘‘Schooling choices: preferences, discount rates, and rates of return’’, Empirical Economics, Vol. 25, pp. 15-34. Schmalensee, R. (1976), ‘‘An experimental study of expectation formation’’, Econometrica, Vol. 44, pp. 17-41. Smith, H. and Powell, B. (1990), ‘‘Great expectation: variations in income expectations among college seniors’’, Sociology of Education, Vol. 63, pp. 194-207. Webbink, D. and Hartog, J. (2000), ‘‘Can students predict their starting salary? Yes!’’, Scholar Working Paper Series, WP 10/00. Weber, B.A., Wirz, A.M. and Wolter, S.C. (2001), ‘‘Switzerland’’, in Harmon, C., Walker, I. and Westergard-Nielsen, N. (Eds), Education and Earnings in Europe, A Cross Country Analysis of the Returns to Education, Edward Elgar Publishing, Cambridge, Aldershot. Wilson, K., Wolfe, B. and Haveman, R.R. (2000), ‘‘The role of expectations in adolescent schooling choices: do youths respond to economic incentives?’’, paper presented at the 15th IIPF Conference, Sevilla. Wolter, S.C. (1996), Erwartungsbildung in der Oekonomie, Haupt Verlag, Berne. Wolter, S.C. (2000), ‘‘Wage expectations: a comparison of Swiss and US students’’, Kyklos, Vol. 53 No. 1, pp. 51-69. Wolter, S.C. and Weber, B.A. (1999), ‘‘On the measurement of private rates of return on education’’, Jahrbu¨cher fu¨r Nationalo¨konomie und Statistik, Vol. 218 No. 5 & 6, pp. 605-18.

The research register for this journal is available at http://www.emeraldinsight.com/researchregisters

The current issue and full text archive of this journal is available at http://www.emeraldinsight.com/0143-7720.htm

Technical/professional versus general education, labor market networks and labor market outcomes

Technical vs general education 471

David N. Margolis CNRS, TEAM-Universite´ Paris 1 Panthe´on-Sorbonne, Paris, France, and

Ve´ronique Simonnet TEAM-Universite´ Paris 1 Panthe´on-Sorbonne, Paris, France Keywords Networks, Education, Labour market, School leavers, France Abstract Does the choice between a general and a technical/professional education determine the quality of the labor market network that an individual will be able to exploit throughout his or her professional life? This paper examines the hypothesis that technical and professional tracks, because they involve fewer students who are in more regular contact with each other and focus on a common, relatively narrow subject, allow students to establish more effective networks to support them in their careers. We test whether the choice of educational track has an impact on the means by which jobs are obtained and on the time to the first job of at least six months, the percentage of time spent in employment later in the career and the earnings when employed later in the career in France. Our results suggest that the educational track determines the means of obtaining a job, but conditional on the manner in which the job was obtained, the track has no additional impact on the outcome variables for the first or later jobs. However, the link between technical/professional education and job obtainment via professional networks does not hold independent of the level of education. In particular, this effect seems pertinent only for students having obtained a professional or technical baccalaure´at (relative to a general baccalaure´at) or for students having obtained a degree from a ‘‘grande e´cole’’ or engineering school (relative to graduate-level university studies).

1. Introduction Does the choice between a general and a technical/professional education determine the quality of the labor market network that an individual will be able to exploit throughout his or her professional life? The literature has shown that the conditions surrounding a school-leaver’s transition into the labor market are related to the educational system that he or she left. International comparisons of school systems, such as Damoiselet and Le´vy-Garboua (1999), emphasize that the implication of firms in professional training within the education system is an important determinant of the quality of an individual’s school to work transition. Simonnet and Ulrich (2000) and Bonnal et al. (2002) have shown that apprenticeships improve a student’s chance of being employed and his or her earnings when employed, relative to a professional high school education. Furthermore, Margolis et al. (2001) find results consistent with offer arrival rates that are higher for individuals who do not change sectors than for those who do, and they suggest that this may be due to

International Journal of Manpower, Vol. 23 No. 5, 2002, pp. 471-492. # MCB UP Limited, 0143-7720 DOI 10.1108/01437720210450905

International Journal of Manpower 23,5 472

sector-specific networks of varying quality, perhaps based on the track that an individual followed in school. The economic literature on the role of networks and contacts also suggests a link between network quality and labor market outcomes[1]. Montgomery (1991) develops an equilibrium model in which workers are heterogeneous in productivity and contacts provide information about the productivity of potential workers to the firm. His model implies that workers with more contacts receive higher wages and firms hiring through contacts make higher profits. Alternatively, Staiger (1990) views jobs as experience goods and models contacts as providing up front information about match quality. More recently, economists and sociologists have focused on the importance of networks in the job-finding process for school leavers. Rebick (2000) showed that more than half of all hires may be attributed to persistence in hiring by employers from faculties in Japan and that persistence appears to be related to screening of potential employees and to the assurance of supply. Mortensen and Vishwanath (1994) developed a model in which personal contacts can be important even when workers are equally productive and jobs are inspection goods. They showed that differential access to job information by direct application to employers or indirect contact through friends and relatives can induce differential outcomes in the wage earned. These results suggest that there is a link between the quality of the network of professional contacts and an individual’s transition into the labor market, and more importantly for us, that this link may be determined by choice of following a technical/professional track, as opposed to a generalist track, in school. To test this hypothesis, however, one must be able to control for the possibility that the quality of the transition (and the later career experiences) are directly affected by the educational track followed by the person. Our hypothesis is different, that the educational track primarily determines the mode of job finding, and that better professional networks tend to lead to a better school to work transition. In the extreme case, the only role that the choice of an educational track plays in an individual’s later career success arises from the quality of the networks that the educational system provides, and once this is taken into account the educational track plays no further role. In this paper, we test this hypothesis: technical and professional tracks, because they involve fewer students who are in more regular contact with each other and focus on a common, relatively narrow subject, allow students to establish more effective[2] networks to support them in their careers. Our analysis is carried out in several steps. First, we identify the variables that determine which educational track is followed by an individual and those which determine the final level of education. Given the risk of endogeneity of the educational track and level in the later steps, this first step serves essentially to instrument our key variables. In a second step, we use the estimated probability of following a general (versus a technical/professional) track and the estimated level of education to consider the means by which the individual obtained his or her first job that lasts at least six months. In theory, the higher the probability of having had a

general education, the lower should be the probability that the individual found their first stable job by means of a professional network, as individuals with a technical/professional training will have more effective networks ceteris paribus and thus there will be a higher probability that these networks will lead to an acceptable job offer. In a third step, we consider the impact of the mode of obtaining the first stable job on the time between the school leaving and hiring date. We do this in the presence of the estimated educational track and level. If the extreme version of the theory is valid (i.e. that the only role of the track is via the mode of job obtainment) and the track followed actually has an impact on the school to work transition as has been found elsewhere in the literature, this should be reflected by both the presence of a significant relation between mode of obtainment and the time to first job and by the absence of a significant relation between the educational track and the time to the first job. In a fourth step, we repeat step 2 (the analysis of the means by which a job is obtained) for the job that is available in 1997 (a minimum of five years after school leaving according to our sample selection criteria). This allows us to see whether the networks that are accessible further into the career are also dependent on the educational track followed. In addition, since the professional network model requires an individual’s contacts to be in a position to help the individual on the labor market, we may even expect to see stronger effects later in the career than at the beginning, since the probability that an individual’s contacts have themselves been employed or in a position to help is likely to increase over time. Finally, we estimate a model as in step 3, but we consider the percentage of time spent employed in 1997 and the monthly earnings on the job held in March 1997 as dependent variables. In addition, we include the estimated time to first job as a further model covariate alongside the mode of job obtainment, the educational track and the educational level. The implications of the theory for the mode of obtainment and the educational track remain the same, but the inclusion of the early career variable opens up an additional possibility for analysis. By considering the relation between the early career experiences and later career outcomes in the presence of the mode of obtainment and educational track variables, we can get a clearer picture of possible scarring effects of a poor school to work transition for later in an individual’s professional life. The structure of the paper is as follows. After discussing the data in section 2, section 3 describes our models of the determinants of educational track and educational level. Section 4 treats the means by which the individual obtained his or her first stable job, and section 5 considers the time to the first job of at least six months. Section 6 considers the means by which a job in 1997 was obtained and section 7 deals with the time spent employed and earnings in 1997. Section 8 concludes. 2. The data Our analysis is based on data from the Youth and Careers Survey (Enqueˆte jeunes et carrie`res, or EJC) undertaken by France’s National Institute for

Technical vs general education 473

International Journal of Manpower 23,5 474

Statistics and Economic Studies (INSEE) in 1997. The survey was a supplement asked of all individuals between 19 and 45 years old in the outgoing rotation group of the 1997 French Labor Force Survey (Enqueˆte emploi). It covers roughly 20,800 people, of whom 9,000 are less than 30 years old and thus participated in the ‘‘youth’’ part of the survey from which many of our variables are drawn. The EJC provides detailed information on the educational experiences of the individual, the level of education and occupation of the parents and variables describing whether the individual had health, family or sentimental problems. In addition, it provides detailed information on the individual’s career, as well as the highest level of schooling attended and the highest degree obtained (and its date), the date of the start of the first job of at least six months and the characteristics of the first stable job. The data also provide information on the means by which a job was obtained, although no information is available on methods used during job search that did not result in employment. Following Sabatier (2000), we decompose the means of obtaining a job into five categories: (1) Market-based methods, which includes spontaneous contacts with employers, sending of re´sume´s and answering help-wanted ads. (2) Personal networks, which involve contacts obtained through family and non-work related friends. (3) Institutional methods, which involve using an employment agency or unemployment insurance office. (4) Professional networks, which involve contacts obtained through people met in one’s professional life, through one’s school or training organization or through a previous employer. (5) Other methods, which include becoming self-employed and obtaining jobs through a national hiring competition[3]. The professional networks method is thus the means of obtaining a job that corresponds the closest to the theory, and on which we focus our attention. The level of schooling attended variable in the data allows us to consider four different schooling levels: (1) 0 – for individuals with a pre-high school diploma, technical/professional training (CAP or BEP) or having left high school on a general track before the final year. (2) 1 – for individuals having attended the final year of high school, either on a general or a technical/professional track. (3) 2 – for individuals having attended a university at an undergraduate level or having attended specialized technical/professional training leading to a BTS degree or a paramedical or social certification. (4) 3 – for individuals having attended a university at a graduate level or having attended a ‘‘grande e´cole’’[4] or engineering school.

In addition to the level of schooling, we created a variable to identify the educational track as general education (as opposed to technical/professional education) on the basis of the details of the educational experience. Since the distinction between tracks does not exist for the lowest levels of education, we excluded all individuals who have not reached education level 0 from our sample. These variables provide the basis for the analysis carried out in steps 1 through 3. Another advantage of the EJC is that it provides the information necessary to measure all of these variables, information on the employment situation at labor market entry and information on the current employment situation. In particular, we can measure (as in the case of the first stable job) the means by which the current job was obtained (if the individual is employed at the sample date), the percentage of time the individual spent employed during the previous year and the monthly earnings on the current job. The bringing together of all of these variables in the same data set allows us to exploit a common sample for all of the analyses, with the exception of the earnings analysis which can (obviously) only be run on employed individuals[5]. Finally, we applied several selection criteria to the construction of our data set. As already mentioned, we excluded individuals who never attended even level 0 school. In order to be guaranteed that period of labor market entry had terminated by the time of the March 1997 observation, we excluded everyone who left school after 1992. We also excluded everyone who took longer than five years to find their first stable job and those whose first stable job was found before leaving school, since (given the previous selection criterion) it would be possible that the job observed in March 1997 would not yet be the first stable job[6]. Finally, we lost an additional 10 percent of our remaining observations due to missing data on education or labor market entry variables. Table I describes the effects of each of these selection criteria on the sample size.

Technical vs general education 475

3. The choice of educational track and level Since the theoretical model that we are considering suggests that the choice of education track and level will influence future outcomes (indirectly via the quality of the professional networks and perhaps directly as well), one would expect that individuals might choose the educational track and level as a Sample selection criterion Initial data set: EJC youth subsample Left school after 1992 or did not find first job within five years of school leaving Education less than level 0 First job variables missing Analysis sample total Source: Youth and Careers Survey, authors’ calculations

Number of Observation observations lost 20,770 13,043 12,490 11,275 11,275

7,727 553 1,215

Table I. Impact of selection criteria on the sample size

International Journal of Manpower 23,5 476

function of desired future outcomes, thus rendering these variables endogenous. As a result, in step 1 of our analysis we estimate the determinants of educational track and level, making sure to include valid instruments for the analyses that follow. The results are found in Tables II and III. Our results suggest that most of our selected variables, including individual characteristics, individual context, parents’ education and parents’ occupation, are important for both the choice of a technical/professional track versus a general education and for the level of education. Young men tend to choose a technical/professional education more often than young women and to stop their schooling sooner, and young French people tend to choose a general track less often than young foreigners (results not shown). In addition to opting more often for a technical or professional education, young men tend to stop at lower levels than young women. Young people who have had sentimental problems tend to stay in school longer and opt less often for a general education. As we do not have a prior theoretical notion for the causal link between sentimental problems and most of our dependent variables, this should be considered primarily as an instrument for identification purposes only. As has been found elsewhere in the literature, children of more educated parents tend to be more educated themselves, with a monotone relation between the educational levels of both the mother and father and that of the child being present and highly significant[7]. On the other hand, the main impact of parental education on their child’s choice of educational track seems to be present for children of parents who are high school or post-high school graduates. These children tend to opt more often for a general track than for a technical or professional education. The occupation of the parents also weighs heavily on the educational outcomes of their children. Children of blue collar workers tend to stop school the earliest and those whose fathers are blue collar workers tend to choose a technical or professional education themselves[8]. Interestingly, young people whose parents (especially mothers) are blue collar workers tend to opt more for a general education. At the other end of the spectrum, children of upper level and middle managers tend to stay in school much longer and to a generalist education even more than the children of blue collar workers, although the difference is not significant when considering the mother’s occupation. 4. Means of obtaining the first stable job In our second step, we consider the role of the educational track on the means by which the first job that lasted at least six months was obtained. Is it the case that, independently of the level of education of the individual, the fact that the person followed a technical/professional track increases the likelihood that he or she will have found his or her first stable job through professional relations? To answer this question, we analyze the mode by which the first stable job was obtained via a series of probit regressions[9] that control for factors present at the time of labor market entry, the (instrumented) probability of having followed a general educational track and the (instrumented) probability of

Don’t know or deceased Secondary Technical/ professional Tertiary

0.1119** (0.0540)

–0.3307** (0.1594)

0.1087 (0.0678)

–0.3349*** (0.0256)

0.1580** (0.0795)

11,275 –6,573.81 713.52 0.000

Observations Log likelihood 2(77) P-value

0.0429 (0.0455)

0.1743*** (0.0413)

0.0866 (0.1014)

Tertiary

Technical/ professional

–0.1383*** (0.0390) 0.1229* (0.0642)

Secondary

Don’t know or deceased

Mother’s education

0.1192 (0.0459)

–0.1133 (0.0841)

Father’s education

Explanatory variables

Blue collar

Middle manager White collar

Upper manager

Artisan

Don’t know or deceased

0.2492*** (0.0608) 0.1267** (0.0582) 0.0646 (0.0509)

0.3623*** (0.0718)

0.2520*** (0.0625)

0.2255** (0.0876)

Father’s occupation

–0.0222 (0.0612) 0.2335** (0.1191) 0.1106* (0.0594) 0.0042 (0.0343) 0.1009** (0.0402)

Artisan Upper manager Middle manager White collar Blue collar

–0.1216 (0.1083)

Don’t know or deceased

Mother’s occupation

Source: Youth and Careers Survey, authors’ calculations

Notes: The model also includes controls for 25 birth years, 22 regions (plus missing), four types of nationality (plus missing) and three types of father’s nationality (plus missing). Standard errors in parentheses * indicates coefficient significant at the 10 percent level; ** at the 5 percent level; *** at the 1 percent level

Male

Family problems

Sentimental problems

Health problems

Demographic and environmental

Technical vs general education 477

Table II. Probit model of the determinants of the educational track chosen (general = 1)

Table III. Ordered probit for level of education (four possible levels) Secondary Technical/ professional Tertiary

0.3584*** (0.1293)

0.0364 (0.0613)

–0.2497*** (0.0233)

Sentimental problems

Family problems

Male

0.6036*** (0.0688)

11,275 –10,907.90 3,088.61 0.000

Observations Log likelihood 2(77) P-value

0.2391*** (0.0395)

0.4444 (0.0358)

–0.0535 (0.0958)

Tertiary

Technical/ professional

0.1790*** (0.0339) 0.6987*** (0.0558)

Secondary

Don’t know or deceased

0.3841*** (0.0399)

–0.1029 (0.0781)

Mother’s education

Blue collar

Middle manager White collar

Upper manager

Artisan

Don’t know or deceased

0.3067*** (0.0534) 0.0679 (0.0517) –0.2563 (0.0459)

0.5645*** (0.0628)

0.1692*** (0.0554)

0.0196 (0.0809)

Father’s occupation

Blue collar

Middle manager White collar

Upper manager

Artisan

Don’t know or deceased

0.2311*** (0.0518) 0.0586* (0.0312) –0.1466*** (0.0387)

0.3476*** (0.1065)

0.1784 (0.0533)

–0.1684* (0.1083)

Mother’s occupation

Source: Youth and Careers Survey, authors’ calculations

Notes: The model also includes controls for 25 birth years, 22 regions (plus missing), four types of nationality (plus missing) and three types of father’s nationality (plus missing). Standard errors in parentheses * indicates coefficient significant at the 10 percent level; ** at the 5 percent level; *** at the 1 percent level

Don’t know or deceased

0.0618 (0.0490)

Health problems

Father’s education

Explanatory variables

478

Demographic and environmental

International Journal of Manpower 23,5

having attended school up to level 1, 2 or 3. Each probit model estimates the probability of having obtained the first stable job by a particular means versus all other means of obtaining a job[10]. The results are presented in Table IV. Insofar as concerns the main hypothesis of this paper, that technical/ professional tracks lead to better professional networks and thus a higher chance of obtaining a job via such networks, the results are mixed for the first

Explanatory variable

Market

Mode of obtainment Personal Professional network network Intermediaries

Technical vs general education 479

Other

–0.0691 –0.5220 –1.2087 –0.9250 1.4586** (0.6000) (0.5942) (0.6257) (0.7766) (1.1840) Level 1 –0.8209 -4.8237* 5.6935** –3.6756 4.8036 (2.7148) (2.6426) (2.7013) (3.6262) (4.0755) Level 2 0.3060 3.8215* –3.5010* 2.0578 –3.3058 (2.0337) (1.9983) (2.0268) (2.8690) (2.7777) Level 3 1.0912 –1.7575 0.1806 –2.2174 2.8968* (1.1525) (1.1601) (1.1553) (1.6928) (1.5152) General  Level 1 –2.6848 7.1531 –11.0675* 9.8629 5.6022 (6.5209) (6.4052) (6.6173) (8.9966) (10.6235) General  Level 2 –0.9362 –8.3925** 10.3852** –4.4262 2.0708 (4.1175) (4.0136) (4.1012) (5.9736) (5.6317) General  Level 3 –2.5732 2.0333 –0.1252 3.1016 –0.7781 (2.0454) (2.0031) (2.0481) (3.0231) (2.7934) * ** *** Health problems –0.0058 0.0873 –0.1224 0.1801 –0.2032** (0.0518) (0.0504) (0.0539) (0.0632) (0.0843) Sentimental problems 0.0431 0.0103 0.0081 0.1267 –0.4935** (0.1203) (0.1170) (0.1186) (0.1302) (0.2277) Family problems 0.0138 –0.0103 –0.1391** 0.1588** 0.1018 (0.0640) (0.0640) (0.0681) (0.0773) (0.0916) ** *** Male –0.0123 2.0333 0.0807 –0.1551 0.1126** (0.0349) (2.0031) (0.0356) (0.0468) (0.0519) Observations 11,275 11,275 11,275 11,275 11,275 Log likelihood –6,613.09 –6,878.46 –6,283.66 –3,163.83 –2,542.37 2(81) 167.52 228.91 229 467.93 237.56 P value 0 0 0 0 0

General

Niveau Niveau Niveau Niveau

0 1 2 3

0.0151 0.8392 0.9065 0.5363

P(/P(mode)//track) = 0 0.9074 0.4041 0.2325 0.0582 0.0511 0.0264 0.2693 0.7193

0.1196 0.3021 0.3788 0.4861

0.4347 0.6271 0.8559 0.4540

Notes: All of the models include controls for 25 school-leaving years, 29 years of first stable job obtainment, four types of nationality (plus missing), four types of father’s nationality (plus missing), six classes for father’s occupation (plus missing) and six classes for mother’s occupation (plus missing). The educational track and level variables, as well as their interactions, are instrumented based on the models in Tables I and II. Standard errors are not corrected for instrumented regressors. Standard errors in parentheses * indicates coefficient significant at the 10 percent level, ** at the 5 percent level, *** at the 1 percent level Source: Youth and Careers Survey, authors’ calculations

Table IV. Probit of mode by which first job of at least six months was obtained

International Journal of Manpower 23,5 480

stable job. Although we can not reject the hypothesis that the educational track has no influence on the probability of having obtained one’s first stable job by professional contacts for level 0 and level 3 educated people[11], the difference in the use of professional networks by track is significant for levels 1 and 2 educated people[12]. For level 1 educated individuals it is even of the correct sign. Unfortunately, Table IV suggests that undergraduate university students tend to use professional networks more than their colleagues who attended a technical or professional school at the same level[13]. This latter result may be a side effect of a particularly effective exploitation of personal networks by individuals with technical/professional training at the level 2 education level, since level 2 students with an undergraduate university education seem to be particularly disadvantaged in terms of the effectiveness of their personal networks[14]. Insofar as concerns the other modes of obtaining one’s first stable job, Table IV suggests that market-based techniques are more often effective for young people with a level 0 general education, while personal networks are much more frequently used by people with a level 2 technical or professional education. Neither the content nor the level of education seems to have much of a significant impact on the probability of obtaining one’s first stable job in France by intermediaries, self-employment or a national competition. Given that market methods, placement agencies and self-employment or national competitions are options that are open to all young people, cases where the first stable job is found by one of these modes may signal particularly ineffective networks of labor market contacts, either professional or personal[15]. Table IV also shows that men tend to rely more often on their professional network and on self-employment or national competitions to obtain their first stable job, while women exploit employment agencies more effectively or thoroughly. As employment agencies may be a search mode of ‘‘last resort’’, this may be a sign of particular job finding difficulties that young women face, relative to young men. These results are the mirror image of what we find for individuals who have had health problems, i.e. people with health problems are more likely to have found their first stable job through a placement service than healthy individuals, and they are less likely to have successfully used professional networks, self-employment or national competitions. As a final point, it is perhaps interesting to note that individuals who have had family problems are not significantly less likely to have obtained their first stable job via personal contacts, although they are less likely to have successfully exploited their networks of professional contacts and are more likely to have found their first stable job via an employment agency. 5. Time to the first job of at least six months In our third step, we investigate the impact of the educational track, the educational level and the means by which the first stable job was obtained on the time it took after school leaving to find this job. This time is measured in months between the end of schooling and the start of the first stable job. We explicitly deduct time spent in military service (for men) from the measured

length. The educational track and level variables are instrumented as in section 4, and we also instrument the modes of obtaining the first stable job by their predicted probabilities from step 4[16]. The results of our tobit model are presented in Table V. As Table V makes clear, the educational variables have practically no significant direct impact on the time it takes to find the first stable job after school leaving (in our sample of individuals who have found such a job within five years of the end of their schooling). On the other hand, all of the modes of obtaining this first job result in a significantly longer time to first stable job relative to individuals whose first stable job was self-employment or who obtained it via a national competition. Among the different modes of obtaining the first job, the differences between professional networks, personal networks and market-based strategies are not significant. People who found their first

Technical vs general education 481

Explanatory variables Mean of obtaining the current job Market methods Personal networks Professional networks Intermediaries

39.6139*** (5.1856) 34.3565*** (4.7818) 41.4473*** (6.2959) 55.0359*** (4.6225)

Environmental and demographic

Education track and level General Level 1 Level 2 Level 3 General General General

1.0227 (4.0024) 1.4030 (16.3775) 10.3219 (12.1340) 13.6605* (7.0201)  Level 1 46.6749 (41.3841)  Level 2 –41.9184 (26.3596)  Level 3 –5.4997 (12.3973)

Observations 11,275 Log likelihood –35,894.52 2(89) 12,498.23 P-value 0.0000

Health problems Sentimental problems Family problems Male

0.3158 (0.3538) –1.3787** (0.6935) 0.3220 (0.4521) –4.5345*** (0.2533)

P(/P(mode)/track) = 0 Niveau 0 0.7983 Niveau 1 0.2153 Niveau 2 0.1478 Niveau 3 0.6700

Notes: The models include controls for 25 school-leaving years, 29 years of first stable job obtainment, four types of nationality (plus missing), four types of father’s nationality (plus missing), six classes for father’s occupation (plus missing) and six classes for mother’s occupation (plus missing). The mode of job obtainment, educational track and level variables, as well as their interactions, are instrumented based on the models in Tables I and II. Standard errors are not corrected for instrumented regressors. Standard errors in parentheses * indicates coefficient significant at the 10 percent level, ** at the 5 percent level, *** at the 1 percent level Source: Youth and Careers Survey, authors’ calculations

Table V. Probit of number of months to first job of at least six months

International Journal of Manpower 23,5 482

job via a placement agency, however, tend to take significantly longer than those who obtained their first job via either sort of network or a market-based strategy, which is unsurprising if such agencies serve as a means of last resort for obtaining a job when all else has failed. Table V also shows that men tend to find their first stable jobs faster than women, even after controlling for the fact that women find their first jobs more often via intermediaries. This is also true for young people of French nationality and young people whose fathers are of a Western European nationality (results not shown). Finally, our other instruments for this stage (school leaving year and year of first stable job) are all jointly significant, and most of them are individually significant and different from each other as well, with a clear negative relation between time to first stable job and the school leaving year and an equally clear (and stronger) positive relation between the time to first stable job and the year in which the job was obtained. 6. Means of obtaining the current job Although the theory posits a relation between the educational track and the quality of professional networks in the labor market, one must keep in mind that in order for a network to be effective, its members must be in a position to help. This is likely to have been a factor in reducing the significance of the educational track in the explanation of the use of professional networks to find one’s first stable job, as it is less likely that one’s professional contacts have had employment experiences themselves in the period right after school leaving than when we consider them five years (at least) into the career. As a result, in this section we concentrate on the role of the educational track on the use of professional networks at the sample date (March 1997), given that the sample is selected so that this date is further than five years after school leaving. The results of our series of probit estimations, controlling for the (instrumented) educational track, the (instrumented) education level and the (instrumented) time to first stable job, along with other factors that reflect the situation at the sample date, are presented in Table VI. Table VI makes clear that the distinction between educational tracks now affects only the probability that the job held at the sample date was obtained by professional networks or (slightly less clearly) employment agencies[17]. That said, a technical or professional education does not always lead to a higher probability of having found one’s job via a professional network. This is once again the case for level 1 and (at the 16 percent level) for level 3 educated people). Such a result is consistent with the theoretical model underlying this paper. Unfortunately, the relation is in the opposite direction for level 0 and level 2 educated people. The explanation of particularly useful personal networks for level 2 educated people no longer holds, although there is an indication that employment agencies have taken up this role later in the career. This is also the case for level 0 educated workers, although the difference in the probability of obtaining a job via institutional means is only significant at the 17 percent level. In a sense, this is an unsurprising result for the employment agencies which may suggest some skimming. It says that experienced[18], technically

Explanatory variable

Market

Mode of obtainment Personal Professional network network Intermediaries

Other

–1.0944 –1.2982 –0.4374 –0.4894 1.4684* (0.6695) (0.6444) (0.8059) (0.7929) (0.9503) Level 1 0.1854 –0.3233 9.9371*** –6.7676* 0.7879 (2.8281) (2.8585) (3.8013) (3.5253) (3.4023) Level 2 –0.5380 –0.4086 -5.5715* 5.2806* –1.9248 (2.1341) (2.1916) (2.8435) (2.7466) (2.3970) –4.2815*** 2.3005* Level 3 0.2819 –0.5151 3.2157** (1.2143) (1.2805) (1.6035) (1.6622) (1.3249) General  Level 1 –0.4239 –2.5586 –17.4789* 12.0849 11.5771 (7.1081) (7.0425) (8.9410) (8.6649) (8.9838) General  Level 2 1.7205 2.1256 11.5449** –9.4703* –1.3981 (4.4313) (4.4813) (5.7832) (5.4013) (4.9743) General  Level 3 –1.0099 –0.3292 –5.0880* 6.6059** –0.0058 (2.2205) (2.2846) (2.9135) (2.6768) (2.4705) Health problems –0.0817* 0.1863*** 0.1470** 0.3108*** –0.0590 (0.0445) (0.0465) (0.0582) (0.0568) (0.0549) Sentimental problems 0.0031 0.3203*** –0.2761* 0.2578** –0.2378 (0.0978) (0.1014) (0.1594) (0.1169) (0.1479) 0.0903 0.2799*** 0.0823 Family problems 0.0169 0.0885* (0.0483) (0.0518) (0.0655) (0.626) (0.9583) Male –0.0378 –0.3292 0.2145*** –0.0015 0.2432*** (0.0372) (2.2846) (0.0477) (0.0486) (0.0425) Number of months to –0.0008 0.0061 –0.0068*** 0.0070*** –0.0059*** first stable job (0.0013) (0.376) (0.0017) (0.0016) (0.0015) Observations 9,354 11,275 11,275 11,275 11,275 Log likelihood –6,191.46 –5,804.80 –3,149.11 –2,895.84 –4,373.44 2(54) 68.85 158.51 138.31 217.18 299.31 P-value 0.084 0 0 0 0

Technical vs general education

General

Niveau Niveau Niveau Niveau

0 1 2 3

0.5135 0.8957 0.7891 0.4589

P(/P(mode)//track) = 0 0.4475 0.0684 0.6405 0.0539 0.7353 0.0366 0.6871 0.1588

0.1675 0.1718 0.0696 0.0222

0.1719 0.2090 0.6258 0.5261

Notes: All of the models include controls for 25 school-leaving years, three types of nationality, three types of father’s nationality (plus missing), six classes for father’s occupation (plus missing) and six classes for mother’s occupation (plus missing). The educational track and level variables, as well as their interactions, are instrumented based on the models in Tables I and II. Standard errors are not corrected for instrumented regressors. Standard errors in parentheses * indicates coefficient significant at the 10 percent level, ** at the 5 percent level, *** at the 1 percent level Source: Youth and Careers Survey, authors’ calculations

483

Table VI. Probit of mode by which current job was obtained

International Journal of Manpower 23,5 484

trained workers who did not reach the end of high school are easier to place than the pure dropouts, and people who put their immediate post-high school years to use learning a trade or profession are easier to place than those who went to a general university whose only entry criterion is a high school diploma (baccalaure´at). In terms of the other model variables, it is interesting to note that individuals who took longer to find their first stable jobs are less likely to have found their current jobs by professional contacts, self-employment or national competitions and are more likely to have found their current jobs by intermediaries. This is consistent with our results in Table IV, which indicated that variables that tended to increase the probability that the first stable job as found by professional networks, self-employment or national competitions also tended to reduce the probability that it was found by intermediaries. Since intermediaries tended to increase the time to the first job (relative to professional networks and other means of obtaining the first stable job), this may be an indirect signal of a persistent effect of the quality of the networks on the means of job finding over time. Finally, the other variables in our models also tend to have a significant impact on the probability of obtaining the first job by at least one of the modes described. For example, health problems tend to increase the probability that the current job was obtained by networks or intermediaries, while individuals with family problems have a higher probability of having obtained their job via an employment agency while those with sentimental problems are more likely to have come about their current job by either personal contacts or intermediaries. 7. Time spent employed and earnings in 1997 The final step of our estimation is to examine the role of the educational track, the educational level, the mode of obtaining the current job and the time to the first long job (all instrumented) on the percentage of time spent employed during the previous year and on monthly earnings, each measured in March 1997. The least squares results[19] for earnings (with and without occupational controls) are presented in Tables VII and VIII and the tobit results for the number of months spent employed are presented in Table IX. Insofar as concerns the variables that describe the means by which the individual obtained the current job, none are significant in the earnings model which controls for occupation (Table VII) whereas market methods, personal networks and self-employment or national competitions are associated with significantly more months spent employed relative to those who are currently unemployed (Table IX). The size of the coefficients suggests that people who obtained their current jobs by these modes were almost always employed all 12 months of the previous year. People whose current job was found via a professional contact are also likely to have spent less of the previous year without a job than those observed without a job at the sample date, but not significantly so. On the other hand, those whose current job was found by an employment agency are likely to have spent less time during the previous year in employment than the currently unemployed or inactive, although the

Explanatory variables Means of obtaining the current job Market methods Personal networks Professional networks Intermediaries Self-employment, national competition

0.1619 (0.1590) 0.2676 (0.2026) 0.4229 (0.2792) 0.3686 (0.2481) 0.2095 (0.1535)

Education track and level General Level 1 Level 2 Level 3 General  Level 1 General  Level 2 General  Level 3

Observations Adjusted R2 F(93, 7809) P-value

0.0409 (0.1747) 0.6227 (0.7957) –0.2218 (0.5803) 0.1409 (0.3502) –1.4314 (1.8595)

Environmental, demographic and early career Health problems –0.0711*** (0.0213) Sentimental –0.0410 problems (0.0362) Family –0.0583*** problems (0.0207) Male 0.0825*** (0.0156) Number of months to first –0.0009* stable job (0.0005)

Technical vs general education 485

1.0116 (1.2216) –0.0155 (0.6141)

7,903 0.6850 185.78 0.0000

P(/P(mode)/track) = 0 Level 0 0.8147 Level 1 0.4204 Level 2 0.4247 Level 3 0.9621

Notes: The model also includes controls for 25 school-leaving years, 22 regions, log hours worked, eight industries, five firm sizes, six occupants, five types of work schedule, job seniority, public or private sector, three types of marital status, four family sizes and five classes of city size. The time to first stable job, mode of job obtainment, educational track and level variables, as well as their interactions, are instrumented based on the models in Tables II, III, V and VI. Standard errors are not corrected for instrumented regressors. Standard errors in parentheses * indicates coefficient significant at the 10 percent level, ** at the 5 percent level, *** at the 1 percent level Source: Youth and Careers Survey, authors’ calculations

difference is not significant. Still, the sign of the coefficient suggests that if there were unobserved heterogeneity in terms of employability in the population, the set of those who have passed though employment agencies to find their current jobs tend to be less employable than the stock of unemployed and inactive people at a point in time. Once again, we note that the educational track and the educational level have no significant impact on the labor market outcomes considered, in this case the percentage of time spent employed and the monthly earnings of an individual. This is somewhat surprising for the education level variables, but it should be noted that we measure the highest educational level attended, and not the highest degree obtained. Simonnet and Ulrich (2000) have shown that there is a very important diploma effect in France, and a given educational level

Table VII. Ordinary least squares model of the log of monthly earnings in the current job with control variables for occupation

International Journal of Manpower 23,5

Explanatory variables Means of obtaining the current job Market methods

486

Personal networks Professional networks Intermediaries Self-employment, national competition

0.4842*** (0.1847) 0.3481 (0.2358) 0.9426*** (0.3244) 0.4323 (0.2886) 0.7503*** (0.1782)

Education track and level General Level 1 Level 2 Level 3 General

0.2580 (0.2031) 0.8115 (0.9260) –0.2863 (0.6753) 0.8667** (0.4073)  Level 1 –0.3528 (2.1640)

Environmental, demographic and early career Health problems –0.0992*** (0.0248) Sentimental –0.0532 problems (0.0422) Family –0.0946*** problems (0.0240) Male 0.0952*** (0.0176) Number of months to first –0.0013** stable job (0.0006)

General  Level 2

0.0820 (1.4216) General  Level 3 –0.4553 (0.7148) Observations Adjusted R2 F(88, 7814) P-value

Table VIII. Ordinary least squares model of the log of monthly earnings in the current job without control variables for occupation

7,903 0.5732 121.58 0.0000

P(/P(mode)/track) = 0 Level 0 0.2041 Level 1 0.9623 Level 2 0.8247 Level 3 0.7516

Notes: The model also includes controls for 25 school-leaving years, 22 regions, log hours worked, eight industries, five firm sizes, five types of work schedule, job seniority, public or private sector, three types of marital status, four family sizes and five classes of city size. The time to first stable job, mode of job obtainment, educational track and level variables, as well as their interactions, are instrumented based on the models in Tables II, III, V and VI. Standard errors are not corrected for instrumented regressors. Standard errors in parentheses * indicates coefficient significant at the 10 percent level, ** at the 5 percent level, *** at the 1 percent level Source: Youth and Careers Survey, authors’ calculations

can cover people with a variety of different terminal degrees[20]. This may be the source of the insignificant result. Another source of the result may be that the occupational categories we have included as controls capture all of the relevant variation in wages, as there is a very strong positive relation between the level of qualification of the job and earnings (results not shown). Table VIII explores this possibility and finds that there is indeed a significant, positive effect for individuals with a level 3 education, although the coefficients for the other education variables remain insignificant. As in the case of the first stable job (Table V) and the model that includes occupational controls (Table VII), there is no significant earnings difference in Table VIII associated with the educational track beyond the indirect effect that the educational track has on the mode of job finding. Furthermore, individuals who obtained their current jobs by professional

Explanatory variables Means of obtaining the current job Market methods Personal networks Professional networks Intermediaries Self-employment, national competition

57.8956*** (13.7693) 79.7327*** (18.3869) 24.9148 (26.9817) –24.7421 (24.7125) 79.2422*** (14.5974)

Education track and level General Level 1 Level 2 Level 3 General

–10.1584 (15.6062) 5.6246 (74.6830) 1.8702 (53.6218) –18.4778 (32.5723)  Level 1 –24.8848 (165.7653)

Environmental, demographic and early career Health problems Sentimental problems Family problems Male Number of months to first stable job

5.4905** (2.2149) 1.6213 (3.7318) 6.2024*** (2.1405) 8.9901*** (1.3663)

Technical vs general education 487

–0.0339 (0.0456)

General  Level 2

21.7441 (107.6463) General  Level 3 27.6729 (54.0295) Observations Log likelihood 2(70) P-value

11,275 –10,059.9310 1,324.8 0.0000

P(/P(mode)/track) = 0 Level 0 0.5151 Level 1 0.8196 Level 2 0.9209 Level 3 0.7080

Notes: The model also includes controls for 25 school-leaving years, 21 regions, three types of marital status, four family sizes and five classes of city size. The time to first stable job, mode of job obtainment, educational track and level variables, as well as their interactions, are instrumented based on the models in Tables I, II, IV and V. Standard errors are not corrected for instrumented regressors. Standard errors in parentheses * indicates coefficient significant at the 10 percent level, ** at the 5 percent level, *** at the 1 percent level Source: Youth and Careers Survey, authors’ calculations

networks are the highest paid, followed by those who obtained their jobs through national hiring competitions or by becoming self-employed and by those who used market methods[21]. Married people tend to earn more and spend more time employed than people of any other marital status, and men tend to earn more and spend more time in employment than women. People who have had health or family problems, although they tend to work more also tend to earn less while working. Finally, earnings tend to decline slightly with the time it took an individual to find his or her first stable job (between 0.13 and 0.09 percent less per additional month), although the effect is only marginally significant. This may reflect some persistence in individual heterogeneity, as was also (weakly) suggested by the results concerning the use of intermediaries.

Table IX. Probit of number of months worked during the previous year, 1997

International Journal of Manpower 23,5 488

8. Conclusion This paper has attempted to systematically address the question of whether the type of education (technical/professional or general) determines the network of professional contacts that a person has available when he or she enters the labor market. We have found that the education track does tend to influence the means by which jobs are obtained, both at the start of the career and later on in the career, and that conditional on the means of obtaining a job the educational track has no further influence on our chosen labor market outcomes (time to first stable job, percentage of time spent employed over the previous year and monthly earnings, the latter two measured at least five years after school leaving). Although this is encouraging for the model, the effects of the educational track are not uniform across education levels, with the most theory-consistent results being obtained for individuals who attended the final year of high school and those who went on to graduate-level studies. Of course, care should be taken in interpreting our results. We have shown that there are clear differences between the types of students that pursue a professional or technical, as opposed to a general, education. As we can only measure the means by which a job was obtained (and not all of the different modes used during the job search process), it may simply be the case that students on the general track do not have access to professional networks, at least at the beginning of their careers, and thus we should not expect to see this mode of obtaining a job as being particularly successful for these people. However, as time goes on the networks of both types of individuals grow with their previous employment experience, and thus one should see the differences in the modes of obtaining a job that appear when comparing Tables IV and VI. Yet there may be an alternative explanation, related to unobserved individual heterogeneity. It may the case that the unobserved characteristics that make an individual more likely to follow a professional or technical track are also those that make them less able to successfully exploit alternative means of finding a job, such as personal networks, national job competitions or market-based methods. These characteristics also affect the labor market outcomes that we measure, and the ‘‘mode of job obtainment’’ variables absorb more of the omitted variable bias than the educational track variables in outcome models. This hypothesis is, however, not testable without multiple observations on the outcome variable, and since our data do not provide this information, we unfortunately can not clearly exclude this possibility. Notes 1. We do not discuss here the research in sociology on the role although an extensive literature has developed at least since Granovetter (1974). 2. Throughout this paper we pay particular attention to using opposed to ‘‘efficient’’. By ‘‘effective’’, we mean successful in

of contacts for careers, the pioneering work of the term ‘‘effective’’, as finding a job, whereas

3. 4.

5. 6. 7. 8. 9.

10.

11. 12.

13. 14.

15.

‘‘efficient’’ implies a conversion of job search inputs (means employed, hours spent searching, number of CVs mailed out, etc.) into an output (a job). Since we are only able to observe the mode of job search that leads to the first acceptable job (an outcome of the search process) and no inputs, we thus can not evaluate the efficiency by which a given mode of job search converts inputs into acceptable jobs, but only which mode ‘‘effectively’’ generated the accepted job offer. All regular jobs in the French public sector are accessible only through national hiring competitions. A ‘‘grande e´cole’’ is a specialized school that only accepts students with a high school diploma who have passed an entrance exam. The entrance exam is usually preceded by two to three years of preparatory courses, which implies that students attending a grande e´cole have typically been finished with high school for as long as students attending university at the graduate (2e`me cycle (licence, maıˆtrise) and 3e`me cycle (DEA/DESS, the`se)) level. Unfortunately, we have no information that allows us to identify individuals who attended preparatory courses but did not follow up with either a grande e´cole or university studies. Were such data available, these individuals would be classed in education level 2. Of course, all individuals in the sample can participate in the selection equation that accompanies the earnings model. Those who started their first stable job immediately on school leaving are not excluded from the sample, however. The reference category for parents’ education is primary. The reference category for parents’ occupation is farmer. We also estimated a multinomial logit model for the different modes of job obtainment, which yielded qualitatively similar results. As we would like our analyses to be comparable with those undertaken in section 6, and since individuals are not necessarily employed at the sample date in 1997, we would have been required to add an additional outcome, ‘‘no successful mode’’, to the multinomial logit. As the standard multinomial logit specification does not account for correlations between residuals in the latent models, the primary impact of choosing a set of probit estimations over a single multinomial logit estimation lies in the different function form chosen for the perturbation. As we have no prior information concerning the correct functional form, we chose the set of probit estimations to maintain a comparable methodology throughout the paper. Since our sample is selected on the basis of having obtained the first stable job within five years of school leaving and since we excluded everyone for whom this variable was missing, the sum of the estimated probabilities should be (and is) 1. In section 6, however, some individuals may not be employed at the date of the survey, in which case the sum of the estimated probabilities will be less than 1. The P-values for the corresponding tests are shown at the bottom of Table IV. The education interaction terms in these and the following models are constructed as the product of the predicted probability of following a general track and the predicted probability of having attended the given level of schooling. The two models were not estimated jointly, and thus all covariance between level and track is assumed absent. It should be noted that the coefficient is only marginally significant, although the P-value of the corresponding hypothesis test is better than for high school graduates. Once again, we would like to point out that the term ‘‘effectiveness’’ refers to a search method that is observed to have led to employment. As we are unable to measure inputs to the search process, or even other methods that are attempted but do not produce the first acceptable job offers, our data do not allow us to consider the efficiency of different search methods. It should be noted that market methods are likely to be the most sensitive to differences in individual motivation, although the same might also be said for self-employment. If such a sort of unobserved individual heterogeneity is indeed present, then market methods need not be less likely than networks to lead to the first stable job for some people, and thus one

Technical vs general education 489

International Journal of Manpower 23,5

16.

490 17.

18. 19.

20. 21.

should nuance the conclusion about the quality of the networks when the first job was obtained through market methods. Given the nonlinear nature of almost all of the models in this paper, the standard error correction necessary to account for the estimated nature of the instrumented regressors is not, in general, preprogrammed unless the models are explicitly estimated as systems of simultaneous nonlinear equations. For reasons of clarity, we have opted for a sequential approach as the initial instrumentation equations are not the principal source of interest for the paper. Corrected standard errors can be obtained either analytically via the procedure suggested by Murphy and Topel (1985) or by bootstrapping. Future research on this question will apply one or the other of these techniques. Note that some individuals may not be employed at the date of the survey. In this case, the sum of the estimated probabilities of having found the current job by a given mode will be less than 1, since unemployed and inactive individuals will be classed as ‘‘not this mode’’ for all of the probits. Here, the sum of the estimated probabilities is roughly 0.88. Recall that all individuals in our sample have held at least one job that lasted at least six months prior to the survey date. Tables VII and VIII present results from an ordinary least squares regression, with and without occupational control variables. Maximum likelihood estimation of more complete models with a correction for selection bias (presented in Appendix Tables AI and AII) does not allow us to reject the hypothesis of the absence of selection bias, although it does consume degrees of freedom (via the selection equation that needs to be estimated) and increase standard errors unnecessarily. Evidence on the so-called ‘‘sheepskin effect’’ in the USA is provided by Hungerford and Solon (1987) and Jaeger and Page (1996). Note that we are using the intrumented probabilities of having found a job via a given mode and that these probabilities need not sum up to 1 at the sample date due to the presence of out-of-work individuals. As a result, one can include all five modes of job finding without an identification problem due to collinearity with the model intercept. This would have been the case had we used observed job finding modes instead of estimated modes.

References Bonnal, L., Mendes, S. and Sofer, C. (2002), ‘‘School-to-work transition: apprenticeship versus vocational school in France’’, International Journal of Manpower, Vol. 23 No. 5, pp. 426-42. Damoiselet, N. and Le´vy-Garboua, L. (1999), ‘‘Educational systems from an economic perspective: an international comparison’’, TSER-STT Working Paper. Granovetter, M.S. (1974), Getting a Job: A Study of Contacts and Careers, Harvard University Press, Cambridge, MA. Hungerford, T. and Solon, G. (1987), ‘‘Sheepskin effects in the returns to education’’, Review of Economics and Statistics, Vol. 69 No. 1, February, pp. 175-7. Jaeger, D.A. and Page, M.E. (1996), ‘‘Degrees matter: new evidence on sheepskin effects in the returns to education’’, Review of Economics and Statistics, Vol. 78 No. 4, November, pp. 733-40. Margolis, D.N., Simonnet, V. and Vilhuber, L. (2001), ‘‘Using early career experiences and later career outcomes to distinguish between models of labor market behavior under institutional constraints’’, Cahiers de la MSE, se´rie blanche, No. 2001.35. Montgomery, J.D. (1991), ‘‘Social networks and labor market outcomes’’, American Economic Review, Vol. 81 No. 5, pp. 1407-18. Mortensen, D.T. and Vishwanath, T. (1994), ‘‘Personal contacts and earnings. It is who you know!’’, Labour Economics, Vol. 1 No. 2, pp. 187-201. Murphy, K.M. and Topel, R.H. (1985), ‘‘Estimation and inference in two-step econometric models’’, Journal of Business & Economic Statistics, Vol. 3 No. 4, October, pp. 370-9.

Rebick, M.E. (2000), ‘‘The importance of networks in the market for university graduates in Japan: a longitudinal analysis of hiring patterns’’, Oxford Economic Papers, Vol. 52 No. 3, pp. 471-96. Sabatier, M. (2000), ‘‘Modes de recherche d’emploi et transitions individuelles sur le marche´ du travail: applications microe´conome´triques au Panel Te´le´phonique du Cereq 1989-1993’’, doctoral dissertation of the Universite´ d’Auvergne. Simonnet, V. and Ulrich, V. (2000), ‘‘La formation professionnelle et l’insertion sur le marche´ du travail: une analyse multicrite`res’’, Economie et Statistique, No. 337-8. Staiger, D. (1990), ‘‘The effect of connections on the wages and mobility of young workers’’, MIT mimeo, Cambridge, MA.

Technical vs general education 491

Appendix

Explanatory variables Means of obtaining the current job Market methods Personal networks Professional networks Intermediaries Self-employment, national competition

Education track and level –0.0840 (0.1419) –0.0992 (0.1503)

General Level 1

0.0047 Level 2 (0.2263) –0.3194** Level 3 (0.1525) –0.0019 General  Level 1 (0.1343) General  Level 2 General  Level 3

Observations 10,558 Log likelihood –6,714.0220 2(85) 17,087.61 P-value 0.0000 P (no selection bias) 0.3762

Environmental, demographic and early career

–0.1222 Male 0.1081*** (0.1643) (0.0135) 0.3635 Number of (0.7584) months to –0.0011** first stable job (0.0005) –0.0669 (0.5641) –0.0481 (0.3263) –0.8559 (1.8245) 0.8491 (1.2008) 0.3008 (0.5891) P(/P(mode)/track) = 0 Level 0 0.4570 Level 1 0.5646 Level 2 0.5732 Level 3 0.7321

Notes: The models also includes controls for 25 school-leaving years, 21 regions, log hours worked, seven industries, six firm sizes, six occupations, five types of work schedule, job seniority, public or private sector and five classes of city size. The time to first stable job, mode of job obtainment, educational track and level variables, as well as their interactions, are instrumented based on the models in Tables II, III, V and VI. The selection equation contains the same variables as Table IX. Standard errors are not corrected for instrumented regressors. Standard errors in parentheses * indicates coefficient significant at the 10 percent level, ** at the 5 percent level, *** at the 1 percent level Source: Youth and Careers Survey, authors’ calculations

Table AI. Selection-corrected model of the log of monthly earnings in the current job with control variables for occupation

International Journal of Manpower 23,5

Explanatory variables Means of obtaining the current job Market methods

492

Personal networks Professional networks Intermediaries Self-employment, national competition

Education track and level 0.1263 (0.1679) –0.1268 (0.1779)

General Level 1

0.3635 Level 2 (0.2663) –0.5563*** Level 3 (0.1800) –0.4137*** General (0.1568) General General

Observations 10,558 Log likelihood –7,937.0650 2(84) 10,379.41 P-value 0.0000 P (no selection bias) 0.1831

Table AII. Selection-corrected model of the log of monthly earnings in the current job without control variables for occupation

Environmental, demographic and early career

0.0104 Male 0.1355*** (0.1920) (0.0151) 0.4848 Number of (0.8861) months to –0.0016*** first stable job (0.0006) –0.1006 (0.6591) 0.6591* (0.3811)  Level 1 0.6380 (2.1320)  Level 2 –0.1883 (1.4026)  Level 3 0.0349 (0.6882) P(/P(mode)/track) = 0 Level 0 0.9566 Level 1 0.7438 Level 2 0.9061 Level 3 0.9680

Notes: The model also includes controls for 25 school-leaving years, 21 regions, log hours worked, seven industries, six firm sizes, five types of work schedule, job seniority, public or private sector and five classes of city size. The time to first stable job, mode of job obtainment, educational track and level variables, as well as their interactions, are instrumented based on the models in Tables II, III, V and VI. The selection equation contains the same variables as Table IX. Standard errors are not corrected for instrumented regressors. Standard errors in parentheses * indicates coefficient significant at the 10 percent level, ** at the 5 percent level, *** at the 1 percent level Source: Youth and Careers Survey, authors’ calculations

Book reviews 130 Years of Catching up with the West: A Comparative Perspective on Hungarian Industry, Science and Technology Policy-making since Industrialization Peter S. Biegelbauer Ashgate Aldershot, Brookfield, VT, Singapore, Sydney 2000 xiv + 250 pp. ISBN 1 84014 930 2 Since the fall of the Berlin wall, libraries’ bookshelves dedicated to East Central Europe studies have been overflowing. The early 1990s was the period of ‘‘transitology’’ when social sciences and literature focused on the East-, Centralor Central-East European political, economic and social tranformation. The best of these analyses have looked at two key elements of this process; the element of rupture and change and the elements of continuity, tradition and inheritance. The theoretical approach used by the author of ‘‘path-dependency’’ has helped economists and sociologists give a differentiated overview of the East Central European changes and their social outcome embedded in the complex and varied histories of these countries’ industrialization and/or modernization efforts (Ferge, 2000; Chavance and Magnin, 1996; Stark and Grabher, 1997). After a decade – researchers have once more begun to utilize this approach (Western authors with more skill than the local ones!) to enlarge the historical as well as comparative perspective of the analysis of the recent East Central European transformations. These specific experiences can be utilized for deepening our knowledge on the dynamics of social change. Biegelbauer’s book is an excellent result of such an effort. Though the book seems to avoid using the concepts of ‘‘modernization’’ or ‘‘development’’, both difficult to operationalize because of their shifting content, the author instead puts technical and scientific innovations at the centre of the analysis. This leaves the possibility for the reader to judge the character of changes as ‘‘development’’, ‘‘modernization’’ or simply – as the author does – ‘‘catching up with the West’’. In my reading, Biegelbauer is presenting a ‘‘modernization story’’ traversing through two world wars and revolutions, through the rise and fall of European powers – at the West and at the East – since the late nineteenth century. His work documents the different periods of Hungarian industrialization, political system changes and economic developments in the past 130 years, shows the political, social, economic and cultural motivations of the several transitions of this period and identifies the international and local social actors of modernization efforts. At the same time it puts emphasis on the rapid and radically changing international scene and regional context of the economic

Book reviews

493

International Journal of Manpower, Vol. 23 No. 5, 2002, pp. 493-498. # MCB UP Limited, 0143-7720

International Journal of Manpower 23,5 494

and social development for such a small country as Hungary. The availability of changing patterns, methods and resources – which are utilized in the course of industrialization, as well as the specificities of Hungarian society and policymaking to adapt them, are the issues that enlarge the analysis into a comparative perspective. It is the systems theory of Niklas Luhmann which forms the author’s theoretical basis for his approach. This helps him provide a more general interpretation of the Hungarian case and to draw conclusions on the ‘‘intertwined nature of the different forms of innovations, in science, technology, economy and politics’’ (p. 8). Technological innovations, the course of industrialization and the development of the science and research institutions of Hungary are well documented and presented in the changing political and social context of four historical periods of this country’s past 130 years. These are distinguished by three transitions according to Biegelbauer. Being aware of the fundamental differences between absolute- and enlightened monarchical, fascist, pluralistdemocratic, Soviet-style, realsocialist-reformist and the transforming democratic political and social systems of this country, the author demonstrates that this science and technology policy, as a social sub-system, had its relative autonomy. This may explain both the continuity of its development in changing national and international political contexts and ruptures during seemingly stable political regimes. Within this framework, the author shows how Hungary’s S&T system was established, more than seven decades since the establishment of the Dual Austro-Hungarian Monarch in 1867 until the Second World War. Not ‘‘only’’ did political systems change during this period, but with the fall of the Monarchy and the rise of nation states in the region Hungary lost huge territories, industrial capacities, raw materials, higher education and research institutes. This halted its dynamic development and fostered stagnation after the First World War. The book at the same time presents convincing data about the role of state support and about the – somewhat weakening – Austrian and the growing German influence in industrialization, in the education and research system and in technology transfer, all of which helped establish the country’s higher education system, its academic and industrial research institutions and the acceptance of technological innovations. The chapters on the ‘‘three transitions’’ since the Second World War follow the same logic of analysis which present the interdependence of Hungary’s political and economic history, international environment and its S+T system. The role of such changing paradigms like ‘‘science push’’ and ‘‘demand pull’’ is connected to shifts from a centrally planned and controlled industrialization model of the 1950s to decentralized and market oriented patterns of development from the 1960s until the late 1980s. The author links these transitions to changes in the origin and in the availability of innovations, ideas and notions, and development patterns. He then goes on to describes how the Hungarian S&T system went through its ‘‘first transition’’, the Soviet, then through other ‘‘transitions’’ since the 1970s and 1980s which incorporated US

and German S&T concepts and innovations. At the same time he calls attention to the originality of the Hungarian S&T system which introduced as early as the 1970s institutional and financial solutions that combined state and marketbased ressources in research and development, several years before such Western paradigm changes (pp. 85-7). The sixth chapter of the book gives a wide overview of how the different initiatives of the 1990s were made to renew the Hungarian S&T system, higher education, R&D organizations and technology transfer mechanisms in accordance with the new political institutions and market structures at the time. Also mentioned were the sometimes controversial effects of speeding up foreign investments in frame of the country’s ‘‘catching up’’ efforts. The systematic collection of historical data on research expenditure and on its composition, on the structure of R&D insitutions and on their personnel, on productivity, on sectoral shifts within the European and the Hungarian economy, on the international mobility of students and academics, on international economic and scientific exchange processes, is an extermely valuable contribution of the author to further research not only in the field of S&T systems, but also in terms of economic history and economic sociology. The underlying ‘‘catching up’’ efforts behind and together with political and S&T policy moves of the East Central European countries on the one hand, and the ways and means of the economic, scientific and technological influences of the West have been realized on the other hand, resulting together in a diversity which enriched the dominant European development pattern (that of the ‘‘West’’). This diversity within the actual dominant model is demonstrated by the comparative data of the seventh chapter describing the specificities of the S&T systems of the East Central European countries in the context of their other social and economic sub-systems and international relations. However the past ten years are too short to come to a more definitive conclusion concerning these differences, though they permit the author to show the different degree of centralisation and varied presence of corporatist institutions in the S&T systems of these countries as well as the different division of S&D functions between universities and research organisations (especially in the case of Austria, Slovenia and Hungary). The arguments of the last chapters show that the success of the latest transformation of Hungary in these 130 years was and remains dependent on the whole of its S&T system: ‘‘. . . to build economic and political structures and achieve growth and living standards comparable to the economically most highly developed countries . . .’’ as Biegelbauer defines success (p. 214) ‘‘. . . will be a direct function of the ability of government to facilitate an interaction between a variety of social interests . . . unthinkable without an S&T system producing ideas for the solution of problems of society, in its economic, political and social subsystems’’. Drawing lessons from the historical and comparative analysis he is arguing for functioning communication structures for Hungarian public policies, for better coordination of economic and S&T policies, for facilitating knowledge transfer, for maintaining and reinforcing the country’s

Book reviews

495

International Journal of Manpower 23,5 496

excellent education system and – last but not least – putting accent on R&D activities while promoting foreign investments. The author has successfully integrated into his study all the important Hungarian research experiences of the past 20 years in the field of S&T systems, and the works of such experts as Bala´zs, Darvas, Inzelt, Mosoni-Fried, Nyı´ri, Pe´teri, Pungor, Tama´s, Va´mos are basic elements of his book. The abundance of this Hungarian literature underlies one of Biegelbauer’s findings about the strength of self-reflection and the importance of underlying ideas and notions behind S&T policies. It also indicates the contribution of Hungarian social science analysis in the past decades to innovations in ‘‘catching up with the West’’. Let’s hope that – despite several immediate problems and serious social consequences of the radical transformation – these efforts of the different sub-systems, among them science and technology, will result in a better society to live in. Agnes Simonyi Institute of Sociology, Eo¨tvo¨s Lora´nd University, Budapest, Hungary References Chavance, B. and Magnin, O. (1996), ‘‘L’e´me´rgence d’e´conomies mixtes ’de´pendentes du chemin’ ´ l’Est du nouveau?, dans l’Europe centrale post-socialiste’’, in Delorme, R. (Ed.), A L’Harmattan, Paris, pp. 115-53. Ferge, Zs. (2000), ‘‘In defence of messy or multi-principle conracts’’, European Journal of Social Security, Vol. 2/1, pp. 7-33. Stark D. and Grabher, G. (Eds) (1997), Restructuring Networks in Post-socialism: Legacies, Linkages, Localities, Oxford University Press, Oxford.

Employment and Citizenship in Britain and France Edited by John Edwards and Jean-Paul Re´vauger Ashgate Aldershot 2000 xiii + 282 pp. ISBN 0 7546 1294 5 hardback At first sight, this seems to be a report on studies of labour market flexibility in Britain and France. In reality, it never addresses the topic. The book consists of papers presented at a conference in Aix-en-Provence in September 1998, the second in a series for a ‘‘policy transferability programme’’. Apparently the first conference, in 1997, did address employment flexibility, and generated a good deal of consensus on the nature and value of increased flexibility while suggesting different effects in Britain and France. Unfortunately, I could find no reference to the report on this important precursor to the present volume anywhere in the book, not even in the editor’s introduction. There is no evidence that the contributors to this second volume had access to it, as

essential background information. Moreover, there is no central theme or focus to the essays in this book, as so often happens with conference proceedings. The authors of the 13 essays are a disparate mixture of economists, political scientists, sociologists and social policy specialists. Added to that are the very different approaches of the French and English contributors, although, paradoxically, some of the French contributors are specialists in British society and culture. Most of the essays report political and public debates, or else explore the authors’ personal views on issues. Some of the chapters address moral arguments around employment flexibility, and the associated insecurity – at this point we are a long way from the social scientific study of policies to support employment flexibility. The editors say the purpose of the conference was to examine cultural, social, economic and linguistic factors which might facilitate the transfer of policy ideas and policies between Britain and France. However they insist that there was no interest in comparative analyses of policies between the two countries, and very few papers are explicitly comparative – apart from the occasional aside. The editors also conclude that national perspectives on employment flexibility are so different in the two countries that common policies are ruled out anyway. An alternative view is that in 1997 the European Union finally accepted the arguments in favour of greater labour market flexibility in Europe long advanced by British governments, so that ‘‘flexibility’’ was finally adopted as a positive strategy at the 1997 European Employment Summit. What is certain is that Britain is ahead of the game in this policy area, and should be examined for both positive and negative lessons by more dirigiste European polities. Whatever the reason, the book is a failure. The 13 essays are too disparate in focus and presentation, too poorly informed on current developments in employment flexibility and on employees’ responses to the new contractual arrangements, and too little concerned with the nature of the policy-making culture to offer any useful addition to social scientific knowledge on policies to support employment flexibility. Academics can no longer rely on their intellectual status alone for Luddite criticisms of new developments and social change to carry weight. Policies designed around ‘‘standard’’ full-time permanent jobs will of course be ill-adjusted to other types of career. It requires imagination to see how policies can be revised to accommodate the more complex work histories that are now emerging. Pensions are an obvious example, but the contributors here simply reiterate the well-known problems. Part-time work is another example. France recently introduced legislation to legalise and promote part-time jobs. Despite the vehement opposition of feminists, social scientists and trade unionists, this produced an increase in part-time working; and although France has a higher percentage of involuntary part-time workers than in other EU countries, the great majority of people working part-time in France do so voluntarily, despite repeated claims to the contrary (Blossfeld and Hakim, 1997, pp. 37, 140-42). Conflicting perspectives on part-time work in Britain and France might have made an excellent topic for

Book reviews

497

International Journal of Manpower 23,5 498

this book, but it is never addressed, even though part-time work is by far the most important type of employment flexibility, with the fastest growth rates. Pension problems also derive from the popularity of part-time work among women: they work part-time or not at all while their children are young then later expect a full pension anyway. However in Holland women receive a full pension irrespective of their work history – thus overcoming the problem at a stroke. None of the contributors ever mentions this (old) policy which facilitates and supports the expansion of flexible employment. The most interesting chapter is a case study of people working in the arts, culture and entertainment industry, where short-term and insecure employment, and self-employment, have long been the norm. This might have prompted a discussion of the differences between occupational labour markets and careers (where self-employment is common) versus internal labour markets and employer-based careers, and how national policies support these two systems – but again, the paper goes nowhere. It does, however, provide some interesting information on the rapid growth of this industry in Britain and France, suggesting that the growth of employment flexibility may be specific to industries where it already occurs and is accepted. Perhaps the greatest weakness of this volume is that it displays no awareness of the rich research-based literature on all forms of flexible employment, covering both employer and worker perspectives – as illustrated by Purcell’s (2000) excellent collection. It is no longer necessary to speculate on workers’ reactions, there are numerous studies available. But even the most informed essays in this book seem unaware of relevant research and use superficial data. For example, Hancock and McCreadie’s chapter provides a useful review of pension provision in Britain, with analyses of General Household Survey data. However, even they are unaware of the research on trends in job tenure which show little change over the past 20-30 years in Britain. Similarly they seem to be unaware of two major national surveys of retirement decisions and retirement incomes in Britain, in 1988 and 1994, which have undoubtedly informed policy thinking on pensions. I do not see how academics can discuss social and cultural factors affecting reactions to policy unless they have first taken account of the factual evidence that underpins policy development. Catherine Hakim London School of Economics, London, UK References Blossfeld, H.-P. and Hakim, C. (Eds) (1997), Between Equalization and Marginalization: Women Working Part-Time in Europe and the USA, Oxford University Press, Oxford. Purcell, K. (Ed.) (2000), Changing Boundaries in Employment, Bristol Academic Press, Bristol.

About the authors Liliane Bonnal Liliane Bonnal is Associated Professor of Economics at Toulouse University and member of the research laboratory Gremaq (CNRS) and Leerna (Inra). Her main research topic is applied econometrics. E-mail: [email protected]

About the authors

499

Felix Bu¨chel Felix Bu¨chel is a tenured Senior Research Scientist at the Max Planck Institute for Human Development, Berlin, Germany. Bu¨chel is affiliated as a Senior Lecturer to the Department of Economics at the Technical University of Berlin, where he worked as an Assistant Professor from 1992-1998 and received his post-doctoral degree (Habilitation) in Economics. Other affiliations include the DIW Berlin and IZA Bonn. His main research interests are in the fields of economics of education, labour economics and social policy research. E-mail: [email protected] Wolfgang Franz Wolfgang Franz is President of the Center for European Economic Analysis (ZEW) in Mannheim and Professor of Economics at the University of Mannheim. He is a member of various scientific councils, including the Scientific Advisory Board of the Federal Ministry of Economic Affairs. Wolfgang Franz studied at the University of Mannheim and was Research Fellow at Harvard University and the National Bureau of Economic Research (NBER). He is author of numerous books and articles on labour economics and empirical methods in economics. E-mail: [email protected] Charlotte Lauer After graduating from the Ecole Supe´rieure des Sciences Economiques et Commerciales (ESSEC) in Paris, Charlotte Lauer completed a Master of European Economics at the University of Saarland in Germany. In 1998, she joined the Center of European Economics (ZEW) in Mannheim, where she still works as a researcher on issues related to education and employment. She is currently completing a PhD at the University of Mannheim under the supervision of Professor W. Franz. E-mail: [email protected] David N. Margolis David N. Margolis is a ‘‘Charge´ de Recherche’’ of the Centre National de Recherche Scientifique (CNRS), assigned to TEAM, University of Paris 1 Panthe´on-Sorbonne. He is also ‘‘Chercheur Associe´’’, of the Centre de Recherche en Economie et Statistique (CREST), Paris, Research Fellow of the Institute for the Study of Labor (IZA), Bonn, Executive Comittee Member of the European Association of Labour Economists (EALE) and Co-Editor of the Revue E´conomique. Between 1994 and 1997 he was Associate Professor at the University of Montre´al. E-mail: [email protected]

International Journal of Manpower, Vol. 23 No. 5, 2002, pp. 499-500. # MCB UP Limited, 0143-7720

International Journal of Manpower 23,5 500

Sylvie Mendes Sylvie Mendes is a Phd student at the University of Orle´ans and member of the research laboratory LEO (CNRS). Her main research topics are education systems and labour economics. E-mail: [email protected] Ve´ronique Simonnet Ve´ronique Simonnet is lecturer (Maıˆtre de Confe´rences) at the University of Paris 1 Panthe´on-Sorbonne. She is a researcher in the TEAM (The´orie et Applications en Microe´conomie et Macroe´conomie) laboratory of the same university in the research field of human resources and social policies. E-mail: Ve´[email protected] Catherine Sofer Catherine Sofer is Professor of Economics at the University Paris 1 Panthe´onSorbonne and member of the research laboratory TEAM (CNRS et Universite´ Paris1). Her main research topics are labour economics and economics of the family. Her most recent articles were publised in journals like Journal of Labor Economics or Journal of Population Economics. She is also an Associate Editor of the Review of Economics of the Household. E-mail: [email protected] Stefan C. Wolter Stefan C. Wolter is Director of the Swiss Coordination Centre for Research in Education in Aarau, Switzerland. He has studied Psychology and Economics at the University of Berne and holds a PhD in Economics. He is also lecturer at the University of Berne and heads the Centre for Research in Economics of Education at the same university. He is Research Fellow of the IZA Institute for the Study of Labor in Bonn and Governing Board Member of the CERI/OECD (Centre for Educational Research and Innovation). E-mail: [email protected] Andre´ Zbinden Andre´ Zbinden is a trained teacher and studied economics at the University of Berne. He is currently working as a research assistant at the Department of Economics. E-mail: [email protected] Volker Zimmermann Volker Zimmermann studied economics at the University of Konstanz. He was Research Associate at the University of Konstanz and the Center for European Economic Research (ZEW) in Mannheim. His Phd thesis was an econometric analysis of the German labour market for youths. He works currently in the Economic Research Department of the Kreditanstalt fu¨r Wiederaufbau, the German promotional bank. E-mail: [email protected]