Discussion Paper Series

Transcription

1 Discussion Paper Series CDP No 9/6 Employment, Wae Structure, and the Economic Cycle: Differences between Immirants and Natives in Germany and the UK Christian Dustmann, Albrecht Glitz and Thorsten Voel Centre for Research and Analysis of Miration Department of Economics, University Collee London Drayton House, 3 Gordon Street, London WC1H AX

2 CReAM Discussion Paper No 9/6 Employment, Wae Structure, and the Economic Cycle: Differences between Immirants and Natives in Germany and the UK Christian Dustmann*, Albrecht Glitz* and Thorsten Voel *Centre for Research and Analysis of Miration, Department of Economics, University Collee London, Humboldt Universität zu Berlin and Centre for Research and Analysis of Miration Non-Technical Abstract Differences in the cyclical pattern of employment and waes of immirants relative to natives have larely one unnoticed in the miration literature. In this paper we show that immirants and natives react differently to the economic cycle. Based on over two decades of micro data, our investiation is for two of the larest immirant receivin countries in Europe which at the same time are characterised by different immirant populations as well as different economic cycles, Germany and the UK. Understandin the manitude, nature and possible causes of differences in responses is relevant for assessin the economic performance of immirant communities over time. We show that there are substantial differences in cyclical responses between immirants and natives. Our analysis illustrates the manitude of these differences, while distinuishin between different roups of immirants. Differences in responses may be due to differences in the skill distribution between immirant roups and natives, or differences in demand for immirants and natives of the same skills due to differential allocation of immirants and natives across industries and reions. We demonstrate that substantial differences in cyclical patterns remain, even within narrowly defined roups. Finally, we estimate a more structural factor type model that, usin reional variation in economic conditions, separates responses to economic shocks from a secular trend and allows us to obtain a summary measure for these differences within education roups. Centre for Research and Analysis of Miration Department of Economics, Drayton House, 3 Gordon Street, London WC1H AX Telephone Number: +44 () Facsimile Number: +44 ()

3 Employment, Wae Structure, and the Economic Cycle: Differences between Immirants and Natives in Germany and the UK Christian Dustmann, Albrecht Glitz (UCL, CReAM) and Thorsten Voel (Humboldt Universität zu Berlin and CReAM) 1 September 26 Abstract Differences in the cyclical pattern of employment and waes of immirants relative to natives have larely one unnoticed in the miration literature. In this paper we show that immirants and natives react differently to the economic cycle. Based on over two decades of micro data, our investiation is for two of the larest immirant receivin countries in Europe which at the same time are characterised by different immirant populations as well as different economic cycles, Germany and the UK. Understandin the manitude, nature and possible causes of differences in responses is relevant for assessin the economic performance of immirant communities over time. We show that there are substantial differences in cyclical responses between immirants and natives. Our analysis illustrates the manitude of these differences, while distinuishin between different roups of immirants. Differences in responses may be due to differences in the skill distribution between immirant roups and natives, or differences in demand for immirants and natives of the same skills due to differential allocation of immirants and natives across industries and reions. We demonstrate that substantial differences in cyclical patterns remain, even within narrowly defined roups. Finally, we estimate a more structural factor type model that, usin reional variation in economic conditions, separates responses to economic shocks from a secular trend and allows us to obtain a summary measure for these differences within education roups. Keywords: Immiration, Wae Structure, Business Cycle JEL: E32, F22, J31 1 Correspondin author: Christian Dustmann, Department of Economics, University Collee London, Drayton House, 3 Gordon Street, London, WC1H AX, UK. c.dustmann@ucl.ac.uk. Tel: Fax: This project has been supported by a rant from the Anlo-German Foundation. We are rateful to Stefan Bender and Johannes Ludsteck for support with the German data, and to Nikos Theodoropoulos for support with the UK data. All errors remain ours. 1

4 1 Introduction In this paper, we investiate the way different immirant roups respond to the economic cycle, as compared to native born workers. The possibly differential response of immirants even within skill roups, which has so far been larely inored in the economic literature, 2 has potentially important implications for immiration policy, as well as for the economic adaptation of immirant roups. Our analysis is for two of the larest economies in Europe which at the same time host lare immirant populations: Germany and the UK. Comparin these two countries is particularly interestin as they have experienced substantially different patterns of chanes in the earnins distributions of their workforces (see e.. Katz and Autor 1999) as well as different economic performance over the last decade. Furthermore, both countries have lare immirant populations, but these are very different in terms of educational backround as well as oriin composition, with the UK s immirants bein more hihly skilled than those to Germany. Also, both countries experienced larely different economic conditions over the last decade. We concentrate on unemployment and waes as economic outcomes. We show hat there are substantial differences in cyclical responses between immirants and natives in both countries. Our analysis illustrates the manitude of these differences, distinuishin between different roups of immirants. Further, we analyse some possible reasons. Differences in responses may be due to differences in the skill distribution between immirant roups and natives, or differences in demand for immirants and natives of the same skills due to different allocation of immirants and natives across industries and reions. We demonstrate that substantial differences in cyclical patterns remain, even within narrowly defined roups. We also show that developments for the relative wae position of immirants have been quite different in 2 A notable exception is recent work by Barth et al. (24, 26). 2

5 the UK and Germany, in particular over the last decade. We then estimate a more structural factor type model that, usin reional variation in economic conditions, separates responses to economic shocks from a secular trend and allows us to obtain a summary measure for these differences within education roups. This analysis confirms the larer cyclical response of unemployment for immirants, in particular those from non-oecd countries, in both countries. The structure of the paper is as follows. In the next section we provide some backround information about immiration to Germany and the UK, as well as on differences in economic outcomes for immirants in these two countries. We then discuss the data we use for our analysis, and the samples that we draw. Section 3 illustrates the economic cycle and its differences in terms of GDP rowth and unemployment rates in the UK and in Germany. Section 4 illustrates economic outcomes of different roups of immirants in both Germany and the UK over the economic cycle, and compares these to outcomes of native born workers. Section 5 investiates how much of these differences is due to differences in observable characteristics, like education and ae, and industry- and reion allocation. Section 6 estimates a model that summarises these differences in a set of parameters that allow comparison between roups and across countries. Finally, Section 7 summarises and concludes. 2 Backround and Data 2.1 Miration to Germany and the UK Both the UK and Germany experienced lare waves of immiration in the period after WWII. The first lare wave of immiration into Germany was an inflow of ethnic Germans, expelled from former German territory, totallin 12m between 1945 and 1949 (see Oezcan (24) for details). After 1955, the West-German economy experienced a stron upward swin, and immiration from Italy, Spain, Greece, Turkey, Portual and Yuoslavia in the late 195 s and early 196 s led to a rise of forein workers to 1.2m in 1965, and peaked in 1973 with 2.6m, or 3

6 12 percent of the total labour force. The stock of the forein population increased from 7. in 1961 to around 4. million in The period after 1973 was characterized by family reunification. The early 198 s saw the arrival of the first larer waves of asylum seekers. Finally, towards the end of the 198 s, and accelerated by the fall of the Berlin wall, Germany experienced a new lare immiration from the East. Ethnic German immirants (so called Aussiedler), who under Soviet rule were not allowed to move, mirated from Eastern Europe and beyond to Germany totallin 2.8 million between 1987 and 21. In 22 there were 7.3 million foreiners livin in Germany (Statistical Office of Germany), representin 8.9 percent of the total population. 3 Immiration leislation in the UK after WWII, embodied in the 1948 British Nationality Act and 195 Aliens Act, distinuished formally between Commonwealth and non- Commonwealth citizens. All Commonwealth citizens notionally enjoyed unrestricted freedom to enter the UK. In the subsequent decade, immiration reulations were proressively tihtened. The 1971 Immiration Act brouht an end to the privileed position of Commonwealth citizens, replacin the previous distinction between aliens and British subjects with one between patrials and non-patrials. The 198 s and 199 s saw continuin restrictive reforms to immiration leislation. Immiration of Commonwealth citizens was most pronounced in the two decades after the war. While the early 195 s were characterised by miration from the Caribbean, in the late 195 s a 3 Notice that the numbers for Germany refer to forein individuals, not forein-born individuals as for the UK. While official statistics in the UK define immirants accordin to their country of birth (similar to the US, Canada and Australia), the distinction in most German official statistics is accordin to nationality. This creates some problems of comparability, as forein nationals may include individuals who are born in Germany but have maintained their nationality, and exclude individuals that are born abroad, but were naturalised, such as the ethnic German immirants. 4

7 rowin number of immirants arrived from the Indian subcontinent. Later mirants arrived from Pakistan and Banladesh. Labour market shortaes in the period after the war also led to recruitment of European workers, predominantly from Southern Europe, but also from Poland. After the 1971 act, an increasin fraction of immiration was due to family unification, which remained for a time larely unrestricted. Recently, immiration has increased aain sinificantly, mainly a result of the stron British economy and, after May 24, the allowance of free movement of labour from the new accession countries. In 25 about 2.7 percent of the population in Britain mirated to the UK within the previous five years. In 22 there were 4.9 million foreiners livin in the UK, representin 8.3 percent of the total population (usin LFS data, see also Table 2). 2.2 Data and Samples Our analysis below is based on two lare lonitudinal data sets: The IABS 2 percent employment reister data for Germany, and the Labour Force Survey (LFS) for the UK. Both data sets cover approximately the same period, and are sufficiently lare to analyse minority populations. The IABS The basis for our analysis of Germany is the Employment Subsample which is made available by the Institute for Employment Research (IAB). This administrative data set comprises a 2 percent subsample of all dependent employees subject to social security contributions in Germany. This includes all wae earners and salaried employees but excludes the self-employed, civil servants and the military. In 21, 77.2 percent of all workers in the German economy were covered by social security. The data also includes all unemployed who 5

8 receive unemployment compensation. 4 The IABS does not include individuals that are out of the labour force. Because of the numerous adjustment processes in the East German labour market after German unification in 199 and the relatively small immirant population (the immirant concentration in East Germany is only about 2.5 percent in 21 compared to more than 1 percent in West Germany) we focus on West Germany throuhout, excludin Berlin. For a detailed description of the data set see Bender et al. (2). The LFS Our analysis for the UK is based on the British and the Northern Ireland Labour Force Surveys (LFS). The British LFS is a survey of private households livin in Great Britain, carried out by the Office for National Statistics (ONS) while the Northern Ireland Labour Force Survey is carried out by the Department of Finance and Personnel. Both surveys used to be conducted biennially from 1973 to 1983 and annually between 1984 and Since the sprin quarter 1992 the survey in Britain is conducted each quarter and chaned to a rotatin panel, with individuals included in five consecutive waves of the survey. In Northern Ireland the quarterly LFS was only introduced in the winter quarter of Both the British and the Northern Ireland LFS collect data on a wide rane of aspects of the labour market. Since 1984 the LFS uses the ILO definition of unemployment. For the years 1981 and 1983 the information in the LFS allows us to impute whether a person was ILO unemployed or not. Questions on earnins were not asked before the winter quarter of 1992/93 4 In 21, 74.5% of unemployed individuals in West Germany received official unemployment compensation mostly either unemployment benefits (Arbeitsloseneld) or unemployment assistance (Arbeitslosenhilfe) and are hence recorded in the IABS (Bundesaentur 24). The remainin 25.5% are in most cases unemployed individuals whose entitlement for unemployment benefits has run out and they do not qualify for the means-tested unemployment assistance. 6

9 in Great Britain. In Northern Ireland this set of questions was only included in the questionnaire in Definition of Immirants For the UK, immirant status is defined by country of birth. In contrast, official data in Germany distinuishes between forein and German citizenship rather than country of birth (followin the principle of nationality by descent). In the IABS, therefore, we only observe an individual s citizenship but neither the place of birth nor the year of entry into the country. As an individual born in Germany to forein parents does not automatically obtain German citizenship but keeps the citizenship of the parents, there is a roup of people included in our sample who were born in the country but have possibly forein citizenship. Between 1993 and 22, the share of second eneration immirants in the ae bracket which we consider for analysis is quite small, and lies between 3.5 and 7.5 percent. 5 We are well aware that these definitional problems imply that comparisons across the countries have to be made with care. Nevertheless, we also believe that these differences will not compromise the conclusions we draw from comparisons of immirants and natives in the two countries. For simplicity, we will in what follows refer to the forein sample in the German data as immirants and the German sample as natives. We will use the same terminoloy for the forein born and native born in the UK. Samples used for Analysis To account for roup differences in a parsimonious way that allows comparability across Germany and the UK, we distinuish two roups of immirants in our analysis of the two countries, those from OECD and those from non-oecd countries. We expect immirants from OECD countries to be endowed with human capital that is more suited to the requirements of the 5 Source: Tabulations provided by the Statistical Office in Germany. 7

10 host countries labour markets. Current OECD member countries are Austria, Belium, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Luxembour, Netherlands, Norway, Poland, Portual, Spain, Sweden, Switzerland, Australia, Japan, Korea, Mexico, New Zealand, United States, Canada, the Czech Republic, Hunary, Slovakia, Poland, and Turkey. As we use two different data sources for our analysis, the outcome measures used here are not fully comparable. In particular, the definition of the state of unemployment is different in the two data sets. In Appendix 1 we describe in more detail our outcome measures waes and unemployment. Composition of Immirant Populations Reflectin the different miration histories of Germany and the UK, the composition of their immirant populations differs considerably. In Table 1 and Table 2 we display the composition of the OECD and non-oecd immirant populations for both countries for the years: 1981, 1991, and 21. [Tables 1 and 2 about here] The fiures for Germany suest that in 1981, one in four immirants was from a non- OECD country; in 21, one in three immirants was from a non-oecd country. For the UK, OECD and non-oecd immirants were rouhly equal in share in 1981; like in Germany, the fraction of non-oecd immirant has relatively increased by 21, with a share of 4.6 percent in the overall population, compared to 3.2 percent for non-oecd immirants. The composition of both roups differs considerably between the two countries. While for Germany, the larest OECD roup is Turkish, it is immirants from Ireland for the UK. While the Turkish roup remains fairly constant in relative size for Germany, the Irish roup drops considerably both in percentaes as well as in total numbers (from about 6, in 1981 to 52, in 21). Immirants from India, Pakistan and Banladesh make up most of the non- 8

11 OECD roup in the UK in 21, while it is immirants from Former Yuoslavia, Asia, and Africa for Germany. Individual Characteristics In Table 3 and Table 4 we report some key characteristics for natives and immirants for the years 1981 and 21, where, as before, we distinuish between the immirant population by OECD and non-oecd oriin. For Germany (Table 3), we distinuish between three educational levels: Individuals who have no post-secondary education (low education); individuals who have post-secondary vocational trainin (intermediate education); and individuals who have a collee education (hih education). For the UK (Table 4) we areate qualifications into the same three broad classes (low, intermediate, hih). Similar to the classification in the German data, the first class refers to people without any post-secondary education. As intermediate education we code GCE A Level or equivalent, GCSE rades A*-C or equivalent and other qualifications. Hih education comprises individuals holdin a deree, for instance university deree, and other hiher education. [Tables 3 and 4 about here] Table 3 and Table 4 show that the percentae of collee raduates amon natives in Germany is far lower than in the UK. This is due to the different classification in both the German and the UK data, as a lare part of professional trainin that is offered by collees in the UK is offered by the apprenticeship system in Germany. In both countries, the percentae of collee educated in the labour force has dramatically increased amon natives. Interestin is the different educational backround of immirants compared to natives in Germany and the UK: While in Germany the percentae of collee educated in the immirant population is substantially lower than in the native born population, in the UK the percentae of those with a collee education is hiher for immirants, in both 1981 and 21, and amon OECD as well as 9

12 non-oecd immirants. Overall, immirants in Germany are considerably lower educated than in the UK. In both countries but particularly in Germany, the percentae of individuals in the lowest educational cateory is hiher amon immirants. While in 21 about 1 in 6 Germans have no post-secondary education, this is the case for more than 1 in 2 immirants, with similar percentaes for OECD and non-oecd immirants. Further below we report mean lo real waes distinuishin between males and females. For Germany, the difference in real waes in 1981 between natives and OECD immirants is about 11 percent for males, and 2 percent for females, increasin to 25 percent for males and 15 percent for females in Non-OECD immirants seem to be more disadvantaed than immirants from OECD countries. This suests a dramatic deterioration for immirants over the two decades. For the UK, wae data are only available from 1992 onwards. The data for the year 21 show that male immirants earn on averae similar waes than natives while female immirants earns on averae 16 percent more than native women. The performance of immirants from OECD countries is particularly remarkable. Men and women earn on averae 19, respectively 22 percent hiher waes than their native counterparts. Between 1981 and 21, unemployment rates for German men (women) have increased from 3.4 (4.8) percent in 1981 to 6.8 (5.8) in 21. For immirants, the unemployment rate of men (women) increased from 5. (7.4) in 1981 to 11.8 (11.1) in 21, suestin a quite dramatic increase in the unemployment ap between natives and immirants: For males the ap has increased from 47 percent to 74 percent. 7 6 ln w We compute percent differences as e 1 where ln w is the difference in lo waes between natives and immirants. 7 These unemployment rates are lower than the official unemployment rates for West Germany 4.5 (6.9) percent in 1981 and 8.6 (7.9) percent in 21 because of our sample selection; in particular, we focus on the population aed in which unemployment is relatively low. 1

13 For the UK, unemployment has one down over the period in overall terms. As for Germany there is a substantial difference in the unemployment ap between immirants and natives in 21: for males the overall difference is 83 percent compared to only about 35 percent in For the UK, it is the non-oecd immirants that are mainly responsible for this difference. In the bottom row of Table 3 and Table 4 we present the normalised Herfindahl index to measure reional concentration of natives and immirants. 8 This index takes on numbers between zero (individuals are perfectly equally distributed across reions) and one (complete concentration in one reion). For Germany, the index is.8 for natives, and.11 for immirants for both 1981 and 21. The hiher values for immirants are driven by their stroner concentration in North Rhine-Westphalia and Baden-Wuerttember where more than 5 percent of the immirant population lives, compared to 44.6 percent of the native population. Overall, however, immirants in Germany are not particularly concentrated in certain areas relative to the native population, and concentration has been remarkably stable between 1981 and 21. This is in stark contrast to the UK. Here the index is.1 for natives and.11 for immirants in 1981, increasin to.2 and.15 in 21 respectively. This suests a much stroner concentration for immirants, which has increased (relative to natives) over the period from 1981 to 21. Concentration is particularly stron for immirants from non-oecd countries. About 4 percent of non-oecd immirants live in London as compared to only 8 percent of the native born. To summarise it appears that over the last two decades the economic conditions of immirants relative to natives have considerably deteriorated in terms of unemployment in both Germany and the UK. In Germany, the averae waes of immirants have at the same time N =, where s i i is the share of individuals, either i 8 2 The index is defined as H s 1/ N /( 1 (1/ N) ) natives, OECD or non-oecd immirants, livin in reion i, and N is the overall number of reions. 11

14 dramatically decreased relative to natives. There is hardly any overall wae ap between immirants and natives in the UK; if at all, waes seem to be hiher for immirants. Therefore, while there seem to be some common developments in the two countries with respect to unemployment, the lare disadvantae of immirants with respect to waes is a particular feature of Germany. This could be related to the different skill structure: The tables above suest that immiration to Germany is predominantly low skilled, while it closely resembles the skill structure of natives in the UK. 3 Macroeconomic Conditions in Germany and the UK Before we discuss the way employment and waes of immirants and natives react to the economic cycle in these two countries in the next section, we briefly illustrate their macroeconomic conditions over time. In Fiure 1 we present GDP rowth and unemployment rates for Germany (left raph) and the UK (riht raph). Time series are provided by the Statistical Office and the OECD for Germany and the UK, respectively. For the years prior to reunification data for Germany refer to West Germany and from 1991 to unified Germany. The shaded areas indicate recessions in the respective economy. 9 [Fiure 1 about here] In eneral, recessions in Germany and the UK occur simultaneously. A noticeable exception was the recession of the early 199 s which hit the German economy about one year later than the UK. This was due to the hue demand for consumption and investment oods after the German re-unification. The fiures seem to indicate that both Germany and the UK experienced considerable increases in unemployment in the early-mid 198 s recession, with 9 The dates used in the construction of Fiure 1 are those published by the Economic Cycle Research Institute (ECRI) and by the Business Cycle Datin Committee of the Centre for Economic Policy Research (CEPR). 12

15 some improvement towards the end of the decade. The early 199 s recession led aain to an increase in unemployment in both countries. However, while unemployment fiures started comin down shortly after this recession in the UK, this is not the case for Germany where unemployment continued to rise throuhout the decade, with a small temporary decrease towards the end of the 199 s/early 2 s. Since the recession of the early 199 s the British economy rows at a steady pace of approximately 2.8 percent per year in real terms and the unemployment rate has continuously declined to a level of less than 5 percent in 24. In Germany, unemployment has steadily increased over the entire period, reachin about 11 percent in 24. Furthermore economic rowth was sluish with an averae annual rowth rate of about.9 percent. 4 Economic Outcomes and the Economic Cycle 4.1 Unemployment [Fiure 2 about here] In Fiure 2 we display unemployment rates of natives and immirants from OECD and non- OECD countries for Germany and the UK. For Germany, the fiures show that at the start of the 198 s unemployment rates were very similar for natives and the two roups of immirants for both males and females. The 198 s recession led to a larer increase in unemployment for immirants, but in the successive recovery phase unemployment drops slihtly faster for the two immirant roups. In the 199 s recession, unemployment rows considerably faster for immirants than it does for natives, leadin to a quite dramatic difference in unemployment between natives and both roups of immirants. Unemployment amon OECD immirants is lower throuhout than unemployment amon non-oecd immirants. Towards the end of the 199 s unemployment of immirants seems aain to drop more rapidly than unemployment of natives, but, compared to the early 198 s, there remains a lare difference between the two 13

16 immirant roups and natives. The cyclical patterns are similar for males and females althouh, interestinly, while unemployment is hiher for non-oecd men than for OECD men, it is the opposite for women: OECD women experience hiher unemployment rates than non-oecd women. The fiures suest a stron cyclical development in unemployment differences between immirants and natives. For the UK, the fiures suest differences between immirants from non-oecd countries and natives already in Unemployment for OECD immirants is quite similar to that of natives, with slihtly hiher cyclical variation. Like for Germany, the 198 s recession has a larer impact on non-oecd immirants, but unemployment drops faster for immirants in the subsequent recovery phase. In 199, differences in unemployment between immirants and natives are at the smallest level over the period between 1981 and 25. However, the early 199 recession sees unemployment of immirants risin considerably faster than of natives. In the subsequent recovery phase, unemployment oes down slihtly faster for immirants, but remains hiher for immirants. Overall, the fiures for the UK suest a similar pro-cyclical pattern in the difference in unemployment rates between natives, and the two roups of immirants. 4.2 Waes [Fiure 3 about here] In Fiure 3 we display the development of real lo waes. Aain we distinuish between males and females. Althouh there was hardly any difference in unemployment between the different roups in Germany in 198, there is a wae differential of about 11 (6) percent in favour of native men (women) relative to OECD immirants, and 12 (5) percent relative to non-oecd immirants. Durin the first recession, this differential does not increase. However, there is a dramatic increase in the wae differential from the early 199 s onwards. For non-oecd immirant men it increased from about 15 percent in 199 to 36 percent in 2 which 14

17 corresponds to an increase of the absolute wae differential by 114 percent. The increase is less dramatic for OECD immirants. For females the pattern is similar, althouh the differences between both natives and OECD immirants and natives and non-oecd immirants are smaller. For the UK we only observe waes in the LFS after This is the decade where we saw most dramatic diverence in lo waes between immirants and natives for Germany, but also a much more favourable economic development in the UK. Real waes increased steadily between 1992 and 21 with an averae wae rowth rate of 1.7 (2.5) percent per year for native men (women) (compared to only.3 (.8) percent in Germany), 2.4 (2.7) percent for OECD immirant men (women) (-.2 (.7) percent in Germany) and 1. (3.1) percent for non-oecd immirant men (women) (-1.2 (.2) percent in Germany). In both countries non-oecd immirant men have thus slower wae rowth than their native counterparts; in Germany they even experience neative wae rowth. In the UK, waes of OECD immirants are slihtly above those of natives, and non-oecd immirants. We do not observe the same deterioration in relative waes for immirants in the UK as we saw for Germany. 5 Adjustin for Composition Part of the differences in the labour market outcomes of immirants and natives which we illustrate in the previous section could be due to differences in their composition. In this section we investiate this in more detail. We analyse whether, and to what extent differences in outcomes over the business cycle can be explained by differences in skills, ae structure, industry allocation or reional allocation. We do this by sequentially conditionin out differences between natives and the two roups of immirants. In particular, we estimate the followin model: t2 it = X it α + = natives, t= t1 OECD, Non OECD y γ T + e t t it 15

18 where y it is the outcome of interest for individual i belonin to roup (natives, OECD immirants, non-oecd immirants) in period t, X it is a vector of additional controls, like education, ae etc., and e it is an error term. T t represents the interaction of the roup indicator with year dummies for each year t. The parameters γ t estimated for these interaction terms measure the averae outcome y for roup in period t, conditional on observables X it. Simple re-parameterisation allows estimatin the differences in outcomes over time relative to a reference roup. We estimate the followin model by choosin as the reference roup the native German, respectively UK born: t t = 2 it X it + t Tt + 2 α γ = OECD, t= t1 t= t1 Non OECD y γ d + e t t it d t are here year dummies for each year t. When restrictin α to zero, the estimated parameters γ t are the roup mean labour market outcomes of OECD/non-OECD immirants relative to the native population (picked up by γ t ) as we have illustrated in the fiures in the last section. By sequentially addin education and ae, reional, and, for Germany, industry controls, we eliminate differences in estimates of economic outcomes between our roups that may be due to differences in these observable characteristics. We plot the resultin estimates of γ t in the fiures below. This amounts to comparin immirants and natives who are identical in observables. In the initial estimations without controls, illustrated by the solid line, the only variable included in X it (apart from a constant term) is an indicator for the sex of the individual. In the next step, represented by the dashed line, we add ae, ae squared and interactions of our education roups and year dummies. Finally, in the last step, illustrated by the dotted line, we also include interactions of reion and year and, in the case of Germany, industry and year dummies. Unfortunately, the LFS data does not allow to condition on industry allocation since 16

19 information on industry affiliation is not available for a lare proportion of the unemployed - up to 4 percent of the observations in many years. Notice that we assume that all three roups respond in the same way to chanes in the X it, so there are no roup-specific α coefficients (althouh we allow α to vary with time by usin interactions of education, reion and industry dummies with year dummies). 5.1 Unemployment [Fiure 4 about here] In the upper panel of Fiure 4 we show the unemployment rates of OECD and non-oecd immirants relative to the unemployment rates of natives for Germany. The solid line is the unconditional differential; the dashed and dotted lines control for differences in ae and education structure, and differences in ae, education, industry and reional allocation between immirants and natives, as explained above. The fiures suest that conditionin on ae and education reduces the unemployment differential between Germans and immirants in both roups; however, there remains a lare difference, and the cyclical pattern is clearly visible. Conditionin on industry structure and reional allocation in addition does not systematically chane these differences apart from non- OECD immirants durin the period 1985 to 1995 where it tends to increase the unemployment differential and to some extent smoothen the cyclical pattern. The fiures that separate men and women look very similar to the pooled fiure and can be found in Appendix 2. In the lower panel in Fiure 4 we display the conditional and unconditional unemployment differentials for the UK. The differences between the conditional and unconditional patterns are much smaller than in Germany. That is not surprisin, as the ae and education structure of immirants in the UK resembled quite closely the native population, as we have shown above. Furthermore, althouh immirants are hihly concentrated in London, this is 17

20 not an area with particularly different unemployment rates. Overall, we see aain considerable differences between OECD and non-oecd immirants, as well as the cyclical pattern in the early 198 s and 199 s which is particularly pronounced for the roup of non-oecd immirants. 5.2 Waes [Fiure 5 about here] In Fiure 5 we display the unconditional and conditional lo wae differentials for Germany and the UK. Aain, the solid line depicts the unconditional differentials as we have already discussed in the last section. As for unemployment, we see a reduction in the wae differential between the two immirant roups and natives for Germany when we condition on ae and education, suestin that part of the differential is due to differences in the ae and education composition in the two populations. This is not surprisin as we find lare educational differences between roups in Table 3. However, there remain substantial differences, in particular for non-oecd immirants. For this roup the differential decreases further when takin account of differences in industrial and reional allocation. This suests that non-oecd immirants are particularly affected by the cycle not only because they have low educational achievements, but also because they have an unfavourable industry- and reional allocation. Conditionin on these, and until up to 199, the wae differential actually vanishes. However, after 199, controllin for education, ae, industry structure, and reional allocation can only account for around 5 percent of the widenin wae ap between natives and non-oecd immirants, still leavin a ap of more than 1 percent unexplained in 2. The ap between Germans and OECD immirants on the other hand remains at about 5 percent after also controllin for industrial and reional allocation. For separate raphs for men and women see Appendix 3. 18

21 In the lower panel of Fiure 5 we display the conditional lo wae differentials for the UK. The difference to Germany is quite strikin. Conditionin on ae and education does not affect the differential between natives and OECD immirants; it does however turn the differential between non-oecd immirants and natives neative. This is the opposite of what we find for Germany, and suests that non-oecd immirants would worsen their relative wae position in comparison to natives if they had the same ae and education structure. The sliht overall wae advantae of non-oecd immirants turns into a substantial disadvantae when keepin individual characteristics the same. Moreover, the wae differential relative to natives worsens further when we condition on reion effects: Over the period we consider, the wae disadvantae turns from bein close to zero to between 1 and 15 percent. This is due to immirants bein predominantly livin in hih-wae London. To summarise, our findins in the last section suest that for both Germany and the UK, unemployment probabilities of immirants are more sensitive to the economic cycle than those of the natives. Conditionin on individual characteristics and reional and (in the case of Germany) industry allocation reduces this differential slihtly, but the stroner pro-cyclical pattern for immirants remains. Interestin is the common pattern in both countries, despite the different skill composition of their immirant populations. For waes, differences between the two countries are partly due to the different composition of immirants and natives. The differences in averae waes in 2 between immirants from non-oecd countries and natives in Germany and the UK are similar after conditionin on composition and reional allocation: In both countries, non-oecd immirants face a substantial wae disadvantae when compared to native workers. This is very different from the unconditional differentials, where non-oecd immirants have rouhly similar averae waes in the UK than natives, while the differential is much larer in Germany, due to different composition and reional allocation of these roups in the two countries. On the other side, 19

22 before the early 199 s, the conditional wae differential was close to zero in Germany, but already neative in the UK. Over the last decade, we observe a stark diverin trend in particular for Germany, but not for the UK, where waes of immirants and natives seem to move larely in parallel. Compared to the pattern for the unemployment rate, the differential cyclical responsiveness of immirants and natives is less pronounced when lookin at their waes. 6 Differential Responses to Economic Shocks across Groups We now estimate a more structural model to summarise the evidence we have provided in the last sections, and to quantify the differential response of both different skill roups and natives and immirants within skill roups. This will also enable us to distinuish between permanent chanes in responses over time (which we capture by a time trend), and differences in responses to economic shocks. For example, the lare increase in the wae ap between immirant roups and natives that we observe in Germany since the 199 s is likely to be more a secular trend rather than a differential response to economic shocks. We implement this decomposition usin a parsimonious factor structure. The idea of our approach is similar to Hoynes (2). We utilise differences in economic shocks across reions and over time to identify the relative response of different education- and population roups to such shocks, conditional on reion effects and a time trend. Our outcome variables are unemployment and waes. This allows us to asses the manitude by which the roups react differently to economic shocks, and test whether these differences are statistically sinificant. More formally, consider the followin outcome equation: jrt y = a + b t + c f + µ + v j j j rt r jrt 2

23 where y jrt is the labour market outcome (averae lo waes or unemployment rates) of skill roup j (defined by education and sex) in reion r in time period t. The index distinuishes between different roups of immirants and natives workers. The skill specific labour market outcome is a function of a fixed roup and skill effect a j, a roup and skill specific time trend b j, a fixed reion effect µ r, and a measure of the reion specific business cycle f rt. Importantly, the common factor f rt is assumed to be identical for all skill and immirant roups. The coefficient c j then ives the responsiveness of roup (immirants or natives) with skill level j to the business cycle fluctuations in reion r at time t. To eliminate fixed roup and skill effects, we estimate the above equation in first differences: y jrt = b j + c f j rt + v jrt In our data there is no natural measure for business cycle shocks, that is, for the common factor f rt and hence for f rt. One may think of takin the unemployment rate as a measure of business cycle shocks, but in that instance left- and riht-hand side variables would be mechanically linked. In the absence of such a measure we proceed by treatin the shock as unobserved and estimate each f rt as the parameter on the interaction term of year t and reion T. 1 Denote these parameters as β. Our final estimation model is then iven by: r, rt rt jrt y = b + c β T + v (1) j j rt rt jrt The parameter of interest is c j, measurin the effect of local labour market shocks on roup in skill roup j. These c j are only identified after normalisation. We therefore set equal to one for the base roup, which we choose to be native male workers with collee c j 1 The model of this section is thus a variant of a dynamic factor model which recently have become quite popular in empirical macroeconomics. For a survey of dynamic factor models see Breitun and Eickmeier (25). 21

24 education in Germany, and native male workers with a deree in the UK. In addition, the intercept for this reference roup b j is set to zero in estimation, so the intercepts for the other roups are interpreted as the averae trend for roup with skill level j relative to the trend for the reference roup. In this model, identification is obtained by our assumption that the labour market specific shock β is identical for all roups and skill levels j. Due to our normalisation ( b = and rt j c j =1 for the reference roup) our estimates of β rt are simply the expected chane in the outcome variable of the reference roup in reion r and year t. Notice that in this model we have interactions of two coefficients, c j and β rt, so the model is non-linear in the coefficients which renders OLS inappropriate for estimation, and we use Nonlinear Least Squares to estimate it. We now return to our oriinal question. Suppose that the differential response to the economic cycle, as we have shown in the fiures above is only due to different skill compositions of the native and the immirant population, and that the cycle affects both roups (immirants and natives) equally. In this case, for a iven skill roup j, the parameter c j should be the same for immirant and native workers. For instance, for hih skilled immirant male workers the estimates should be, as for the reference roup, equal to one. This is a testable hypothesis. If these skill specific parameters are not the same across natives and mirant roups (within skill roups), then this provides evidence that business cycle shocks affect immirants differently than natives. [Tables 5 and 6 about here] Table 5 and 6 report our NLS estimation results of Equation (1) for Germany and the UK. For the latter we pool two years toether in order to sustain a sufficient number of observations 22

25 per skill and oriin roup for each of the 11 UK reions we use in our analysis. 11 We report the estimated parameters c j for the unemployment rates for each of our 18 skill roups (2 ender, 3 education, 3 nationality/oriin) in Germany and the UK in columns (1) and (2), and the respective parameters for waes in columns (3) and (4). The results for the skill specific trend coefficients b j are reported in Appendix 4 (Table 7). We report the standard errors underneath the coefficient estimates where stars (*) are used to indicate that a coefficient is statistically different from one (the parameter of the base roup) at the 5 percent level. We also test the hypothesis that responses of the two immirant roups are different from those of native workers within skill roups. Sinificant differences in estimates at the 5 percent level are marked with a (+). 6.1 Unemployment Table 5 reports results for males, and Table 6 for females. We first concentrate on males. Columns (1) and (2) report results for unemployment. For both the UK and Germany, there is a clear tendency that the lower the educational attainment, the stroner the cyclical fluctuations in the unemployment rate. For instance, for native men in Germany, the estimate increases from 1 for the reference roup with collee education to 2.41 for those with intermediate education, to 4.68 for those with low education. This suests that the unemployment response to macro shocks for the low educated is stroner by factor 4 than for the hihly educated men. The numbers for the UK are remarkably similar, with point estimates of 2.56 for the intermediate and 3.58 for the low educated men. 11 We distinuish the three constituent countries Wales, Scotland, and Northern Ireland and, in the case of Enlish reional units, we areate to the level of Government Office Reions, which are London, South East, South West, West Midlands, North West, North East, Yorkshire and the Humber, East Midlands, and East. 23

26 Within skill roups, there seems to be a hiher responsiveness of unemployment for immirants than for natives. For Germany, native men with intermediate education respond 2.41 times stroner to business cycle shocks than native German men with collee education; however, OECD immirants in the same education cateory react stroner by factor 4.19, and non-oecd immirants by factor Both estimates for immirants are sinificantly different from that of natives within the same skill roup. For the low-educated roup, OECD immirants react similar in manitude to shocks than natives (always compared to native men with hih education), with point estimates of 4.79 and 4.68 respectively. Non-OECD immirants react stroner, with a point estimate of 6.98; this estimate is aain sinificantly different from that of native men in the same skill roup. For the UK, natives and OECD immirants with medium qualifications react very similarly to shocks with estimates of 2.56 and 2.69, respectively. On the other hand, non-oecd immirants react sinificantly stroner than both of these roups, with a point estimate of For the roup of low educated workers, point estimates suest aain that both roups of immirants respond stroner than their native counterparts (4.92 respectively 4.9 vs for natives); however, estimates are not sinificantly different between roups. For women, we report results in Table 6. Estimates confirm the overall pattern that we find for men, with perhaps slihtly smaller differences across skill roups, but with immirants reactin stroner to shocks than natives. Aain, non-oecd immirants seem to be particularly sensitive to economic shocks, especially in the UK. 6.2 Waes We now turn to waes, and we report results in the last two columns of Table 5 (males) and Table 6 (females). For Germany the numbers in column (3) suest that the wae fluctuations over the business cycle are larer for the intermediate and still larer for the low skilled when 24

27 compared to the estimates of the hih skilled reference roup, with all coefficients bein sinificantly different from one. Within skill roups, thouh, we do not find statistically sinificant differences in the responsiveness of waes to economic shocks between natives and immirants. The lare diverence in the conditional wae ap between immirants and natives that we have seen in the fiures above is therefore unlikely to be due to differences in the response to shocks. Inspectin the trend coefficients which we display in Table 7 in Appendix 4, these become more neative the lower the educational attainment is, implyin a deterioration of the relative waes of low skilled workers. Furthermore, the relative downward trend is sinificantly stroner for male non-oecd immirants amon the medium and low educated as compared to natives within the same skill cateory. Results for men in the UK are displayed in the last column (4) of Table 5. There seems to be no clear differences across skill roups in the response to economic shocks. This may be due to the fact that wae data are only available since 1992, and that the British economy experienced a steady rowth over most of the last decade. Like for Germany, there is little evidence for differential response to economic shocks between immirants and natives within skill cateories. Only the estimate for non-oecd immirants with intermediate education is sinificantly hiher than its native counterpart. Other than for Germany, Table 7 in Appendix 4 do not suest lare differences in time trends across roups perhaps not surprisin after inspection of the fiures in Appendix 3. For females, results are reported in Table 6. The reference roup is still hihly educated males. For Germany, low educated women react more stronly to business cycle shocks than the reference roup but less so than equally educated males. For both countries, there is little evidence of lare differences between immirants and natives within skill roups. For Germany, only low educated non-oecd women react sinificantly differently than their native counterpart; in this case they show a lower responsiveness to economic shocks. For the UK, none 25