Metropolitan Economic Performance
Social Tensions and Economic Frictions

A CEPR conference on ‘Metropolitan Economic Performance’ was held in Lisbon, Portugal, on 29/31 October 1998. The conference, hosted by the Management Department of Universidade Nova de Lisboa, was organized by Pedro Telhado Pereira (Universidade Nova de Lisboa) and Klaus Zimmermann (IZA, Universität Bonn, and CEPR). The 21 papers presented covered several issues related to metropolitan economic performance, including crime, drugs and racism. Nevertheless, other topics were also addressed, including the social and economic exclusion of women, young people, the unemployed and immigrants. Return migration and the relative performance of migrants was also the subject of a number of the conference papers.

The first two papers were presented under the heading of ‘Transformation and the Cities’. Ira Gang (Rutgers University) presented ‘The Political Economy of Russian City Growth’, co-authored by Robert Stuart (Rutgers University). Drawing on recently available data, the authors investigated the patterns of population growth in medium and large Russian cities in the 1970s and 1980s, focusing on the role that might have been played by the controls peculiar to the administrative command economy. These controls included formal administrative restrictions (‘sticks’), variables directly controlled by the state (‘carrots’) and other forces less directly controlled by the state (‘economic variables’). The authors’ main conclusion was that, even when other state and economic variables were included in the model specification, direct controls mattered, while restrictions on expansion did not. These results were robust with respect to different specifications, including several types of ‘carrots’ and ‘economic variables’. Other results obtained underlined the importance of broad regional differences, and the poor explanatory power of economic variables in explaining population shifts.

Francis Kramarz (CRESS-INSEE, CNRS Paris, and CEPR) suggested that since rents might have influenced immigration behaviour they should also be accounted for in the regressions. Pedro Telhado Pereira criticized the underlying assumption of price homogeneity, which might not have had a relevant influence on migration patterns. Roxane Silberman (LASMAS/CNRS, Paris) questioned the reliability of the migration data, given that some of the flows might have been illegal.

Rumiana Stoilova (Academy of Sciences, Sofia) presented ‘The City: Lights and Shadows of the Post-Totalitarian Transformation’, which adopted a sociological approach to the changes taking place in Bulgarian cities in general, and in Sofia in particular, since the early 1990s. Stoilova argued that the move from totalitarianism to a parliamentary democracy had destroyed not only the previous regime’s artificial social structures, but also the solidarity nourished by public resistance against socialism. Consequently, success and degradation were now simultaneously apparent, contributing to the disintegration and atomisation of post-totalitarian Bulgarian society. A further example of this was that, at a city level, the previous regime’s ‘equality in poverty’ had been replaced by the stratification of city districts.

Crime rates in the cities had risen, because of increasing urban poverty, the lack of efficient authorities and the inherent nature of urban environments. Statistics reflected the crime-related shortcomings of urban reforms: apart from rising crime rates, fewer crimes were being reported, fewer criminals convicted, there was growing delinquency among children, and higher complicity among underaged persons. Ethnic problems with Gypsy, Turkish and Romanian groups had also become more serious, contributing further to the process of disintegration. Most conflicts occurred between Gypsies and native Bulgarians. There were, nevertheless, some positive aspects to city life: mostly, these had to do with better job opportunities, when compared to the countryside. But entrepreneurship, reflected in the growing numbers of new small businesses, was also more pervasive in the cities.

Ira Gang enquired about the prevalence of crime within the different ethnic groups. Jeff Frank (Royal Holloway College, London) and Peder Pederson (Aarhus Universitet) suggested the need for further research on income inequalities across ethnic groups and the contribution of reform to those inequalities.

The succeeding two papers addressed different aspects of social exclusion. In ‘Unemployment and Crime: New Answers to an Old Question’, Kerry Papps and Rainer Winklemann (University of Canterbury, New Zealand, and CEPR) – who presented the paper – argued that establishing whether or not there was a causal link from crime to unemployment was a task with relevant policy implications. If such a link existed, it should be taken into account when performing cost-benefit analyses of policies with the potential to reduce unemployment. Analysing panel data for New Zealand’s 16 regions, the authors regressed crime rates on unemployment for the period 1984–96. A pooled regression had indicated a significantly positive elasticity of crime rates with respect to unemployment rates, but a graphical analysis of the variables across time suggested that the preliminary model had been misspecified and that its results were not valid. The authors then estimated a two-way fixed-effects model, the results of which showed no evidence of a link between unemployment and crime. Since period effects were found to be significant when accounting for crime rates, they adopted a random-region, fixed time-effects model, which produced similar results, except in the two categories of drugs and other anti-social offences, and abuse-of-property offences.

An augmented model used to investigate the role of two other variables – the clearance rate and income per capita – showed that the former variable had an ambiguous impact overall, but with positive or negative effects for specific crime categories. The ambiguity of this result was tentatively explained by delays in the formation of beliefs and by the endogeneity of the clearance-rate variable. The coefficient of income, however, was unambiguously negative, suggesting that the effect of an increase in legal income-earning opportunities outweighed that of increased illegal opportunities. Finally, to overcome problems of endogeneity and lagged perceptions, the authors considered the previous year’s clearance rate, which provided more evidence for a negative effect of the deterrence rate. Their conclusions were thus that unemployment could not explain changes in the overall crime rate, and that policy-makers may have more success in fighting crime by attempting to manipulate the deterrence rate and the average household income, than by dealing with unemployment.

Jan van Ours (CentER, Tilburg University, and CEPR), argued that the authors should not completely dismiss an unemployment effect on overall crime since the similarity of patterns across regions decreased the power of the regressions. Francis Kramarz considered that different regressions should be run for each region. Jeff Frank suggested using either lagged values of unemployment or long-term unemployment instead of overall unemployment.

‘The Effect of Neighbourhood Characteristics on the Labour Supply of Welfare Recipients’, was written by Gerard van den Berg (Free University Amsterdam, Tinbergen Institute, and CEPR), Bas van der Klaaw (Free University Amsterdam, Tinbergen Institute) and Jan van Ours (CentER, Tilburg University, and CEPR), and was presented by van Ours. Their paper investigated the individual transition rate from welfare to work using 1994–6 data on Rotterdam and focusing on whether the individual’s neighbourhood was a relevant variable. Assuming that demand-side conditions were not relevant at the neighbourhood level (since commuting costs were low), the authors concentrated on supply-side differences. They also concentrated on ‘welfare recipients’, namely those aged 18 years and older, who were legally allowed to stay in the Netherlands, had no other income but had an obligation to search for a job. Their data identified three different subsamples of recipients: Dutch job-losers, non-Dutch job-losers and Dutch school-leavers. Three neighbourhood effects were considered: local labour-market effects (when each neighbourhood acted independently of the others); spillover effects (when individuals adopted the behaviour of their neighbours); and selection effects (when individuals chose a neighbourhood with similar job-search patterns to their own).

The estimation results, based on a mixed proportional hazard model, indicated that the overall individual transition rate decreased as the duration of welfare collection increased. Furthermore, neighbourhood effects on the overall unemployment rate were relevant for the Dutch recipients (both job-losers and school-leavers), but did not matter for the non-Dutch job-losers. This phenomenon was particularly acute for younger Dutch welfare recipients. Martin Klinthäll (Lunds Universitet) and Rainer Winklemann suggested that administrative neighbourhood divisions might be somewhat arbitrary, thus weakening the results. Andrea Ichino (European University Institute, IGIER, Università Bocconi, Milano, and CEPR) stressed the importance of controlling for the length of time during which immigrants have lived in a particular neighbourhood.

The session on ‘Return Migration’ was opened by Martin Klinthäll (Lunds Universitet) with ‘Patterns of Return Migration from Sweden 1970–93’. In a wider project, the author had been analysing the economic impact of immigration and the associated problems of social integration and exclusion. This paper paid particular attention to a comparison of the patterns of return migration to Germany and Greece, with the specific aim of testing the savings-target hypothesis, according to which immigrants see themselves as temporary migrants, planning to return to their home country when they have accomplished some savings target. The hazard of return migration, therefore, is expected to depend positively on age, income, unemployment in the host country, absence of children and being single. Conversely, it is supposed to depend negatively on unemployment in the home country, with an ambiguous effect from changing relative wages.

Klinthäll drew on data from the Swedish Longitudinal Immigrant database, which accounts for several occurrences in immigrants’ lives (although it might suffer from an under-registration problem). His sample included only men, aged between 16 and 65, born in either Germany or Greece, who had immigrated to Sweden after 1967. The hazard of return migration was estimated via the Cox proportional hazards model. Some of the results clashed with the model’s predictions: for example, the older age groups did not show a significantly higher risk of return migration than younger cohorts. The author estimated different models for German and Greek immigrants, but again the savings-target hypothesis could not be confirmed, although for different reasons for each nationality. While income induced a U-shaped risk of return for German men, it had a negative effect for Greek men. With regard to age, the Greek group conformed to the hypothesis under scrutiny, while the German return risk decreased with age. A suggested explanation for the Greek immigrants’ results pointed to the possibility of some ‘exclusion effect’, whereby successful immigrants, who obtained a high income, did not wish to return, while those who got stuck with a low income preferred to return to Greece. Thomas Bauer (IZA, Universität Bonn, and CEPR) suggested excluding underaged people from the regressions, since these might be the children of those immigrants who take the return decision. Most people in the younger cohorts did not return on account of their own economic status, thus distorting the results.

‘Is there a Wage Premium for Returning Irish Migrants?’ was the paper presented by Alan Barrett (Economic and Social Research Institute, Dublin, and CEPR), whose  co-author  was  Philip  O’Connell. The paper set out to characterize the educational profile of returning migrants, who currently outnumbered the traditional emigrants, and to investigate whether returnees received an above-average wage, after controlling for the standard relevant variables. Data from the Irish Labour Force Survey showed that return migrants (and non-Irish immigrants) had higher levels of education

than the resident population. This pattern was unsurprising, since the proportion of skilled emigrants during the 1980s was higher than that for the overall population. Another possible contributory factor might have been the rise in earnings dispersion, which was likely to increase the return to more educated individuals. A policy-related outcome of these findings was that concerns about ‘skill shortages’ (and ‘brain drain’) were less relevant, in that the migration mechanism itself was dealing with them.

The issue of the return migrants’ wage premium was addressed by analysing answers to questionnaires sent out to a random sample of 1992 graduates. Some of these individuals were return migrants, having been abroad for six months or more since graduation. Estimations performed – without accounting for self-selection – indicated a significantly positive wage premium for returnees. The premium was large for men, but was not statistically significant for women, although the length of time away was a variable more relevant for women than for men. These results were understandable given that, on average, men stayed abroad longer than women.

Andrea Ichino wondered whether some Irish emigrants had received further education while abroad, thus decreasing the scope for a ‘brain drain’. Francis Kramarz thought that threshold levels should be accounted for when estimating the returnees’ wage premium. He also suggested drawing on the British Labour Force survey for further information on Irish migrants in the United Kingdom.

A second panel on ‘Social Exclusion’ was opened by Jacques Silber (Bar-Ilan University, Ramat Gan) who presented ‘On Inequality in the Quality of Life in Israel: New Immigrants Versus Old-Timers’, a paper written
together with Nira Yacouel. The authors attempted to devise separate measurements for the standard of living, the quality of life, and the efficiency in transforming the former into the latter, among Israeli individuals in 1992–3. Their approach drew on the distinction between resources (standards of living) and functionings (quality of life), and also on the concept of a ‘distance function’, namely the amount by which a given quantity vector should be divided so that a given utility level was achieved). 

The authors used data obtained from a time-survey to build a vector of resources (including information about households, such as number of cars or TV sets), a vector of functionings (with information on – say – the number of recreational activities or the individual’s satisfaction with her health), and general information on the individual’s characteristics. The regressions performed indicated that there was no significant relation between the standard-of-living and the quality-of-living indices, while the opposite held true for the relation between the transformation-efficiency index and both the standard- and the quality-of-living indices. The results also showed that there was more inequality between the individuals in terms of their standard of living than in terms of both their quality of life and the efficiency of their transformation.

Francis Kramarz remarked on the need to control for household size and to consider the characteristics of products, as reflected in their different degrees of substitutability. Christian Dustmann (University College London and CEPR) claimed that some of the survey questions were ambiguous in the sense that it was difficult to tell whether a specific answer meant more or less welfare. Andrea Ichino felt there was a need for more information on the individuals’ characteristics, especially their education, and suggested that some of the questions be dropped from subsequent analyses

Christian Dustmann’s (University College London and CEPR) own paper was entitled ‘Attitudes to Ethnic Minorities, Ethnic Context and Locational Decisions’, and was co-authored by Ian Preston. Their point of departure was that negative attitudes towards minorities may be affected by the ethnic composition of the locality in which individuals live. Although racially intolerant people were unlikely to choose to live in areas with large ethnic populations, the latter were also unlikely to be keen on areas where they might expect to experience racial intolerance. Therefore, empirical results on the impact of ethnic composition on attitudes, which did not account for such phenomena, were likely to be biased downwards.

The authors tested their prediction with 1980s data for England, drawn from the British Social Attitudes survey. This dataset provided information on a range of attitudes towards minorities (both directly and indirectly reported), which was then collapsed into binary indicators. There was also extensive socio-economic information on respondents, including education, income, age and labour-market status. The estimation strategy was based on an instrumental-variables estimator, whereby it was assumed that individuals were constrained to choose their neighbourhood within their initial district.

The results of a ‘standard’ estimation indicated, inter alia, that manual workers tended to have more hostile attitudes, that being unemployed had positive effects on attitudes, and that individuals educated beyond age 18 had more favourable attitudes. The scope of ethnic concentration yielded an (insignificantly) negative effect on attitudes. The picture changed, however, with a second estimation that accounted for potential location biases: the effect of ethnic concentration on attitudes became clearly more negative. Both results were interpreted by the authors as evidence that high concentrations of ethnic minorities in England were associated with more hostile attitudes and that there was a considerable downward bias in estimations which regressed attitude variables straightforwardly on ethnic concentration indicators.

Rainer Winklemann pointed out that in a long-term equilibrium location issues would not matter and thus there would be no such bias. Francis Kramarz suggested that the estimation should account for the likely relative ease with which richer individuals could travel and move to a different district or country. Andrea Ichino proposed that survey non-respondents should be considered as either anti- or pro-immigrant so as to set lower and upper boundaries for the results. He also suggested that there should be some testing for non-linearities in the ethnic concentration effect, in the sense that there might be a threshold when previously unimportant concentration levels became relevant and hostility increased. Martin Klinthäll thought that more weight should be attached to the public housing variable, since it might constrain those individuals living in such places not to move, regardless of the ethnic mix in their ward and their attitudes towards ethnic minorities.

The first paper in a session on ‘Discrimination’ was ‘Estimating Labour Market Discrimination with Selectivity Corrected Wage Equations: Methodological Considerations and an Illustration from Israel’, presented by Shoshana Neuman (Bar-Ilan University, Ramat Gan, and CEPR) and written with Ronald Oaxaca. The authors put forward several methods for decomposing wages when there is sample-selection or selectivity bias, a phenomenon that might arise at the stage of joining the employed labour force and when a specific occupation is chosen. Three new terms arise when selectivity is accounted for in a wage inequality analysis. The first measures the effects of, say, gender differences in the parameters of the selectivity equation on the wage differential. The second accounts for the effects of gender differences in the variables that determine employment in the area under comparison. And the third term captures the effects of gender differences in the wage response to the probability of employment in such an area.

The interpretation of these terms along the lines of discrimination, endowments or selectivity is somewhat ambiguous, which led the authors to define four different types of wage decompositions: considering the first term as discrimination and the following two as endowments; considering the first and third terms as discrimination and the middle one as endowments; regarding the first term as discrimination, the second as an endowments-related measure, and the last as a measure of selectivity; and deeming all three coefficients as accounting for selectivity, with the discrimination and endowments components treated according to the standard Oaxaca approach.

The authors illustrated these different methodologies with an analysis of wage discrimination among Israeli professionals due to both ethnicity (Easterners or Westerners) and gender. They drew on census data covering earnings, human-capital variables, socio-economic attributes and labour-market characteristics of individuals. First, they estimated entrance equations (which provided a correction for the selectivity bias). Second, they estimated Mincer-type wage equations, using one set for the standard Oaxaca decomposition, and correcting the second for the selectivity bias (using the Heckman procedure). Overall, gender wage differentials were found to be larger than ethnic wage differentials. Among several other results, the authors found that three out of the five decompositions used indicated the existence of some favouritism towards Eastern men, although their wages were lower, on average, than those of Western men. Andrea Ichino questioned the adequacy of the particular instrumental variable used in the estimation and pointed out that the results were bound to be very sensitive to such a variable.

In ‘Glass Ceilings or Sticky Floors?’ Alison Booth, Marco Francesconi and the presenter, Jeff Frank (Royal Holloway College, London), accounted for the empirical result that women in Britian are as likely as men to be promoted, but receive lower pay rises upon promotion. (The British Household Panel Data indicated that, upon promotion, men received 20.4% pay rises, but women only 9.8%.) The authors built a three-period model which reproduced this result. Its underlying features were that firms induced workers to invest in human capital by committing to a promotion rule and to a minimum post-promotion wage. Furthermore, firms had the option to match a higher offer from another firm to a promoted worker. Discrimination was brought in by assuming that firms regarded women as less productive than men, although objectively this was not true. Under these circumstances, whereas the incentive to acquire human capital worked similarly for both sexes leading to similar promotion rates, discrimination made it less likely that firms would match outside wage offers to women. Thus women ended up receiving lower pay rises. The model’s predictions were tested empirically. The most important result was that the gender gap of the promotion rate became insignificant when the occupation variable was accounted for, whether or not proxies for the worker’s effort were included, thus supporting the theoretical conclusions.

Jan van Ours remarked that the estimation results were bound to be sensitive to the definition of a ‘promotion’, since a change of occupation might easily be mistaken for promotion. Francis Kramarz suggested that job complexity increased as people moved up the hierarchy, which might explain the empirical results, if men were relatively more numerous in higher ranks.

The third session on ‘Social Exclusion’ was opened by Adrian Ziderman (Bar-Ilan University, Ramat Gan), who presented ‘Vocational Education in Israel: Wage Effects of Vocational Education, Occupation, and the VocEd-Occupation Match’, which was written jointly with Shoshana Neuman (Bar-Ilan University, Ramat Gan, and CEPR). The aim of the paper was to verify the authors’ claim (based on an earlier study of an Israeli sample) that the wage advantage of vocational-school graduates over others working in related occupations stemmed from work in occupations related to their vocational studies, or from employment in a well-paid occupation, and was not the direct result of the training received (as Hotchkiss had suggested for a US sample). The wage effects of vocational education (‘VocEd’) were estimated using Hotchkiss’ specification, i.e. including a variable representing vocational-school attenders currently employed in training-related occupations, among others. The estimates showed that VocEd alone did not confer higher earnings on Israeli workers, and that VocEd completers employed in a training-related occupation did earn more than other groups, even when a matched-occupation variable was included.

The authors explained the differences between their and Hotchkiss’ results by pointing to his ‘problematic’ use of the ‘wage of the first job after leaving high school’ as the dependent variable. In their view, this led to results that were unduly pessimistic with regard to the labour-market outcomes of vocational schooling in the United States. Christian Dustmann stressed that there was a problem of selectivity relating to those individuals who were matched to the jobs for which they were trained. Francis Kramarz argued that the authors’ approach to determining the training-related occupation was not consistent.

Peter Jensen (Aarhus Universitet) presented ‘Labour Market Integration of Immigrants in Denmark’, which was a joint work with Peder Pedersen (Aarhus Universitet), M. Rosholm and N. Smith. In enquiring whether immigrants in Denmark were integrated or marginalized in the labour market, the authors considered several specific issues: transitions in the labour market between the states of employment, unemployment and withdrawal from the labour force; the length of time spent on welfare or income-support benefits; differences between first- and second-generation immigrants; differences between immigrants and refugees; and issues of discrimination, lack of qualifications, educational attainment and intergenerational transmission. The period of analysis was 1984–96. A competing-risks model was used for studying the durations of unemployment and welfare benefits, and a decomposition analysis was employed to compare the immigrants with the rest of the Danish population.

Several preliminary results were obtained: (1) the duration of unemployment was longer for the first than for the second generation of immigrants, and longer for all immigrants than for native Danes; (2) duration dependence was negative; (3) immigrants and the Danes were alike in terms of destinations; (4) the duration of subsequent employment was longer for Danes than for immigrants, and longer for second than for first generation immigrants; and (5) education was more important for immigrants. Christian Dustmann remarked that the decomposition technique was meaningful only if the characteristics were comparable, which was not the case with education for natives and immigrants. Francis Kramarz and Jan van Ours disagreed over whether there was some sort of unobserved heterogeneity present, and van Ours remarked on the limited nature of the investigation given that the authors were talking about reduced form models.

Andrea Ichino (European University Institute, IGIER, Università Bocconi, Milano, and CEPR) opened the session on ‘Unemployment and Employment’ with a paper entitled ‘How Painful is Unemployment? Consumption and Job Losses in Four European Countries’, which was written with Samuel Bentolila (CEMFI, Madrid, and CEPR). The authors analysed the relationship between unemployment and consumption in Italy, Germany, Spain and Great Britain with a view to exploring how this relationship could be affected by different social and institutional frameworks. In particular, they focused on the role of extended family networks in unemployment. Their preliminary results indicated first, that consumption losses associated with unemployment were higher in Germany than in Italy, and higher in Italy than in Spain; and second, that Britain was the only country in which unemployment of a male household head was associated with significant reductions in food expenditures. The findings supported the hypothesis that extended family networks provided a fundamental source of insurance against unemployment in southern Europe. The authors acknowledged, however, that their analysis suffered from a potentially serious problem of endogeneity of the unemployment indicators with respect to consumption decisions, and that this made it very difficult to interpret the observed associations between unemployment and consumption losses as causal effects.

Klaus Zimmermann questioned whether savings rates behaved differently in Germany and in Italy, but remarked further that there were some differences between replacement rates in Germany and Britain. Jan van Ours said that the relationship between unemployment and consumption differed in a boom and in a depression. Consequently, differences between the two countries, especially in terms of the duration of unemployment, may be ascribable to the business-cycle phase. He also criticized the authors’ apparent lack of attention to issues concerning unemployment duration. Francis Kramarz suggested using the expected duration of unemployment as a control variable.

‘Self-employment and Windfall Gains in Britain: Evidence from Panel Data’ was presented by Mark Taylor (Institute for Social and Economic Research, University of Essex). Taylor set out to investigate the effects of unanticipated windfall gains on entry to self-employment, survival in self-employment and – given survival – the subsequent growth of enterprise income. He developed a framework in which income and the utility derived from self-employment were dependent upon tastes, preferences and effort expended in the business. This allowed individuals with varying preferences and wealth to choose to devote less time to work in favour of increased leisure, or to concentrate on areas of their businesses that provided more job satisfaction at the cost of a smaller financial return. The receipt of a windfall in this framework could result in falls in self-employment income, or even transitions out of self-employment.

Multivariate analysis led to the conclusion that the amount and type of payment received were important predictors of becoming self-employed. Redundancy payments increased the probability of entering self-employment, but job-related bonus payments had the opposite effect. This suggested that losing a job, together with the associated lump-sum compensation payment, provided a catalyst for self-employment, while individuals in jobs that rewarded individual performances were less likely to start a business. The size of the windfall also had significant effects on the income of the self-employed, suggesting that the growth of enterprises was constrained by a lack of capital. Clearly, however, receipt of a payment of a particular magnitude enabled the self-employed to maximize their utility elsewhere.

Jan van Ours suggested that the age at which a windfall was received might be relevant for the response. He also recommended that the possibilities of non-linearities in responses be looked at more carefully. For Christian Dustmann, the intensity of participation was also related in some way to the windfall gain.

Peder Pedersen (Aarhus Universitet) opened the extended conference sessions on ‘Performance of Migrants’ by reverting to the issue of return migration – a question that lay at the very frontier of research on international migration. Drawing on the results of the recent publication, ‘Scandinavians without Borders – Skill Migration and the European Integration Process’, of which he had been editor, Pedersen referred to the analysis of return migration by nationals of Norway, Denmark and Sweden who had emigrated in 1981. The high proportion of Danish returnees – about 65% were back within five years – indicated that most of the emigration was of a temporary nature. Norway presented a very similar return rate, but for Sweden the share was only slightly above 45%. A crucial factor at work was the influence of skills, since the return rates differed for different education levels. Although the cumulative return rates were not monotonically related to the level of education in any of the three national groups – at least for the whole period from 1981–90 – they were lower for the lowest educational level, when considering a migration period longer than five years.

Pedersen also presented the main points of the joint paper ‘Declining Employment Assimilation of Immigrants in Sweden: Observed or Unobserved Characteristics?’, written by Pieter Bevelander (Lund University) and Helena Skyt Nielsen (Aarhus Universitet and Centre for Labour Market and Social Research), neither of whom was able to be present. The paper sought to throw light on the reasons for the big decline in employment assimilation of immigrants in Sweden between 1970 and 1990 by analysing the determinants of the probability of full-time employment for immigrants. Apart from growing differences in formal qualifications between native Swedes and immigrants, two factors of major potential relevance here were changes in the composition of the immigrant population, and a change in the ‘character’ of the Swedish economy. The former were reflected in a rise in ‘family reunion immigration’ and in flows of refugees from non-European countries from the mid-1970s onwards, in place of the labour flows from European countries that previously had characterized immigration into Sweden. At the same time, the economy had undergone a structural transformation, with working processes becoming more information- and communication-intense. The labour-market consequences were an increase in the demand for highly skilled workers, and a rise in the importance of informal skills, such as cultural-specific proficiency and language skills.

The authors had estimated a logit model for the probability of obtaining full-time employment and had decomposed the differences into explained (differences in qualifications) and unexplained parts. The most striking finding was that low qualifications, in terms of observed human capital, did not explain much of the difference between the employment rates of Swedes and immigrants in 1990. The main part was unexplained. Although some evidence of discrimination was found, this could not be the only explanation because for all groups, including the more culturally similar, there was a significant unexplained component. The structural-change hypothesis offered an alternative explanation.

Roxane Silberman (LASMAS-Institut du Longitudinal, CNRS, Paris) presented ‘Educational Attainment and Unemployment for Immigrants’ Children in France: An Investigation of the Discrimination Hypothesis’, which was written with Irène Fournier. Since unemployment, especially structural unemployment, can lead to problems with integration of immigrants, the authors’ objective was to describe more precisely the rates and determinants of unemployment for immigrants’ children. Numerous studies had stressed the increasing role of educational attainment levels as unemployment grows. Some recent studies pointed to social background as an explanation for the unfavourable academic achievements of immigrants’ children. The authors therefore employed a classic two-phase model of status attainment to examine the determinants of educational level and subsequently used these to explore the labour-market position. Although there was no specific discriminatory mechanism against immigrants’ children in the educational system, the possibility that nationality might be a factor underlying inequality in the level of education attained, as well as in the position regarding unemployment, needed to be considered. Being a highly synthetic variable, however, national origin could hide several effects and required careful interpretation. Silberman also referred to more recent work about the importance of networks for successful job search.

The main results suggested that the traditional variables of social origin and parental achievement were important factors explaining the gross differences observed between native French and immigrants’ children. In terms of labour-market position, the observed difference between the level of education of French and immigrant workers was an important explanatory factor. The results, however, did not support the thesis of selective and intentional discrimination linked to national origin. Klaus Zimmermann questioned the alleged effects of networks on firms. Francis Kramarz suggested the Enquête Emploi as a possible source of additional information on wages.

The paper entitled ‘Portuguese Migrants in the German Labour Market: Performance and Self-selection’ was written jointly by Thomas Bauer (IZA, Universität Bonn, and CEPR), Pedro Pereira (Aarhus Universitet), Michael Vogler and Klaus Zimmermann (IZA, Universität Bonn, and CEPR), and was presented by Pedro Pereira. The paper’s results confirmed the effectiveness of the German guestworker system – an active recruitment policy that began in the 1950s, but ceased after the first oil shock in 1973 – the purpose of which had been to meet the excess demand for unskilled blue-collar workers. Since recruitment of guestworkers was oriented towards the needs of German firms, the workers had been selected on the basis of qualifications. A demand for particular types of workers, however, engenders a self-selecting supply response, with the consequence that the individuals who decide to migrate might not be those in whom the receiving country otherwise would be interested. The authors sought to study this self-selection problem by analysing the characteristics of Portuguese guestworkers. In addition, the migrants’ performance was compared not only with the natives’ performance but also with that of non-migrating Portuguese workers. To this end, matched micro data from the sending and receiving regions was used.

The comparisons revealed that Portuguese guestworkers had a lower level of education than Germans and those who stayed in Portugal. Compared with the Portuguese non-migrants, however, a higher proportion of the migrants had vocational training. Estimates of earnings equations showed that non-migrating Portuguese workers would have received higher wages, if they had migrated to Germany, than those who actually did migrate. The reason for this was that non-migrant Portuguese workers had better characteristics, in the sense that they had higher education levels and that, for each educational level, remuneration in Portugal was higher than in Germany. Conversely, those who did migrate would have earned less than their German counterparts if they had remained in Portugal.

Rainer Winklemann wondered why the returns to education in Germany were so low. If hours of work were controlled for, might this provide the basis for explaining the higher Portuguese earnings? Andrea Ichino thought that it made no sense to control for the returns to education for the three different occupational statuses considered in the paper. Pedro Pereira replied that it is known that the returns to education came not only through wages but also from the occupation.

Thomas Bauer (IZA, Universität Bonn, and CEPR) presented ‘Occupational Mobility of Ethnic Migrants’, a joint paper with Klaus Zimmermann (IZA, Universität Bonn, and CEPR). The authors examined the determinants of the probability of a post-migration change in occupational status, and its variability with length of residency in Germany. Of particular interest was the experience of ethnically German immigrants, and the possibility of downward mobility in occupational status. Drawing on the ‘immigration sample’ of the German Socio-economic Panel, the authors analysed the inflows in 1994–6, dividing the sample into three sub-groups: two kinds of ethnic Germans – the Übersiedler (from the former German Democratic Republic) and the Aussiedler (from elsewhere in Eastern Europe) – and non-Germans.

The authors’ hypothesis was that the ethnic Germans, having obtained their human capital in socialist economies, would experience some downward occupational mobility in the West German market economy. Their rationale was that there existed a problem of international transferability of human capital, with higher levels of education likely to be less transferable, and more educated migrants therefore experienced a larger downward adjustment in their relative labour-market position. As these migrants also had stronger incentives to invest in the host country’s country-specific human capital, however, they could be expected to regain their former labour-market position more rapidly.

Two different models of occupational mobility were estimated. The first – a standard binomial probit model – accounted for the possibility that a change in occupational status occurred after migration; the second – an ordered probit model – considered several possibilities: the individual was not working, experienced downward mobility, experienced no change or moved upwards. The results showed that a representative Übersiedler had a lower probability of a change in occupational status than a similar Aussiedler, whose probability was slightly lower than that for a similar foreigner. The downward-mobility results showed the same pattern. Overall, the results confirmed the human capital transferability hypothesis. Interestingly, although migrants with a university degree regained their original occupational status after 14 years of residence in Germany, migrants with only primary-school qualifications needed 28 years.

Eric Zwint (SELAPO, Universität München, and IZA, Universität Bonn) presented ‘Panel Analysis of Wages and Unemployment of Ethnic Germans’. The aim of this paper was to analyse the earnings and unemployment of ethnic German immigrants who had moved to Germany shortly before and after the fall of the Iron Curtain. Zwint noted that there had been a change in the dominant countries of origin of these Aussiedler migrants from Poland or Romania to the territories of the former USSR. Migrants from the former USSR, however, exhibited poor language skills and had a significantly higher risk of being unemployed. Thus ethnic German immigrants from Eastern Europe did not seem to have been in an advantageous position relative to other immigrants – indeed, their integration into the labour market had proved rather problematical.

Francis Kramarz was critical of the use of a random-effects probit estimation because, with panel data, there would be a correlation between the mean of the independent variable and the error term, and a consequent loss of efficiency. Andrea Ichino, however, took issue with Kramarz’s view. Maria Baganha (Universidade de Coimbra) enquired whether language might be a major factor in explaining differences in migration experiences between Polish and Romanian migrants. 

Kirk Scott (Lunds Universitet) presented ‘Labour Market Entrance and Income Assimilation: An Analysis of Longitudinal Data from Sweden, 1970–1994’, written jointly with Tommy Bengtsson (Lunds Universitet). The objective of the paper was to analyse why post-1970 immigrants to Sweden had assimilated less well economically into Swedish society than earlier cohorts. The evidence was that, while controlling for education, age and country of origin, immigrants increasingly had fared worse in terms of income and employment. The authors found the explanation in the structural changes in the Swedish economy, which had led to increasing emphasis on interpersonal skills as a primary factor in securing employment.

They estimated the effects of structural change on different nationalities and individual characteristic categories by controlling for the effects of cyclical changes in labour demand and through the use of the vacancy/unemployment ratio. The results revealed that the structural variable ‘relative machine prices’ – which was a measure of demand for skilled versus unskilled labour – had differing impacts on different nationalities, at least initially. These impacts supported the hypothesis that immigrants from countries that are culturally and historically ‘closer’ to Sweden (e.g. Norwegians) should have less difficulty than those from a greater distance (e.g. Greeks and Poles). The findings also showed that higher educational levels lessened the impact of structural change. Thus the paper concluded that it is not merely changes in the supply of immigrants, but also changes in the structural demand for labour, that affect immigrants’ prospects in the destination country.

Rainer Winklemann suggested undertaking the analysis by industry. Francis Kramarz supported the idea of industrial decomposition to see whether prices had changed, and proposed the use of international, instead of Swedish, prices because these would be less endogenous and more appropriate given that Sweden is an open economy.

The final paper was ‘The Duration Until First Investment in Post-Migration Education’ by Dan-Olof Rooth (Lunds Universitet). This study identified the determinants of the time that elapsed before adult immigrants into Sweden between 1987 and 1991 were able to receive post-migration education. Two different measures of duration were used: the time until first enrolment in a Swedish university, and the time until first enrolment in a Swedish secondary-primary educational institution. A distinction was also made between refugee immigrants and others receiving a permanent visa.

The results showed that immigrants from different origins differed in terms of their pre-immigration education levels, but the processes determining which individuals invested first in a Swedish university education and which in a secondary-primary education were quite similar. The higher the level of pre-immigration education the larger the positive effect on the inflow to education. Age was also relevant, especially for the oldest immigrants, who appeared to be deterred from going into education. The 1991 increase in the level of Swedish unemployment increased the inflow to secondary-primary education, and decreased the inflow to university, implying a desire to invest in Swedish-specific human capital.