generation minority members, both men and women, and does not disappear with control for personal characteristics.

Transcription

1 The Curse of Inopportune Transitions: an Analysis of the Labour Market Behaviour of Migrants and Natives in the UK Draft Paper Neli Demireva Abstract: This paper focuses on the transition behaviour of migrants, second generation minority members and British-born Whites. The analysis is based on a five-quarter British LFS dataset from 1992 to We find evidence of lower labour market stability and disadvantage on the part of certain groups of migrants such as EU8, Eastern European, Middle Eastern and Turkish migrants. This penalization holds also for second generation minority members, both men and women, and does not disappear with control for personal characteristics. I. Introduction The analysis presented in this paper is based upon the panel structure of the British Labour Force Survey and explores the inequalities between immigrant and native-born workers in Britain in terms of employment trajectories. Although we know a considerable amount about such inequalities in cross-sectional perspective, there is very little research in the British context about the dynamics of employment and how such dynamics differ for immigrants and natives. The panels available in the Labour Force Survey are only one year in length (five quarters); therefore the longitudinal potential is limited. Nonetheless, given the large sample size and the availability of relevant information, this represents the best data source for such an analysis. The sample size is sufficient to differentiate among a wide range of immigrant groups from 1

2 different parts of the world and to control for some of the obvious individual-level differences between immigrant and native-born workers, such as family structure, age, and level of education. Our focus in this paper is on transitions in and out of employment. Our main contribution lies in providing some information about the little-explored with British data degree of persistence of the yearly labour market activity patterns of migrant and native men and women. Modelling transitions is an important approach that allows researchers to move the discussion from the cross-sectional examination of the incidence of disadvantage typical of workers, be it unemployment or unskilled occupations, towards a dynamic understanding of recurrent inequality. We begin, first, with an analysis of the risks of unemployment, and examine in particular whether immigrants higher risk of the transition to unemployment might be explained by the type of jobs they occupy, including controls for job characteristics such as type of occupation, industry, sector, full-time versus part-time, and temporary versus permanent contract jobs. We limit this analysis to those who are employed at a given point in time, and look then at unemployment status over the subsequent year. Second, we present an analysis of the opposite phenomenon, whether unemployed immigrants and natives have divergent chances of finding employment again within the one-year period. In both analyses, we are attentive to issues of attrition in the data, and how this might affect our conclusions. II. The Labour Market Transitions Framework Social exclusion can be successfully exposed through studying labour market transitions. A substantial body of economic literature exists that discusses the significant detrimental effects that certain transitions may have upon a worker s career path. For example, Gregory and Jukes (2001) and Arulampalam (2001) show that unemployment results in earnings some ten per cent lower than pre-unemployment earnings, which effect is indeed a persistent one (Gregory and Jukes 2001). More generally, transition studies examine the determinants of unemployment 2

3 duration and the exit rate to employment (Nickell 1979, Lancaster and Nickell 1980, Arulampalam and Stewart 1999); although recent work has focused on exits to other states as well 1. In this paper, we will follow the traditional transitions framework and outline three main labour market states: employment, unemployment and inactivity. Our modelling approach controls for a wide range of factors associated in the economic literature with the moving in and out of these transition states. We will further argue that respondents from different migrant groups have different transition rates which are of great research interest and deserve to be studied in detail. In this way, we will aim to investigate the inequality between migrants and natives encapsulated in their labour market behaviour; and contribute to both the existing transitions literature and the studies of migrants labour market performance. The economic literature does not focus specifically on differences between natives and migrants or natives and minority groups in terms of transitions. A few studies model the effects of ethnicity as part of the individual s characteristics. Thus, we know that both men and women of Black African, and Pakistani and Bangladeshi origin are significantly less likely to be in work than their White counterparts (Booth et al. 1999). Pakistani and Bangladeshi women in particular have the lowest predicted probabilities of working at the next interview if they have one or more young dependent children. Studies that look into net differences between Whites and minorities, however, present an ambivalent picture of the effect of ethnicity which is possibly due to difference in available minority sample sizes and the general focus of research 2. Barnes, Lyon and Millar (2008) using the Families and Children Study claim that ethnicity, family size and tenure influence substantially labour market transitions and the associated family income poverty and hardship status over time. Boheim and Taylor (2001) using BHPS data control for ethnicity in their transition models but do not elaborate on its effect. Gregg and Wadsworth (1996) based on the British Labour Force Survey report a significant ethnic coefficient in success of exiting unemployment using state employment agencies, which is included only in the tables in the Appendix and not discussed in the text. Clearly, for the presented studies ethnicity or the labour market outcomes of migrants have not been the major point of concern. Moreover, they may have lacked in numbers of different minority and migrant groups, therefore, in statistical power to 3

4 model thoroughly migrant s labour market choices and expose their disadvantage, which is the main focus of this paper and for which purpose an adequate sample has been used. Several indications exist, however, that the majority and the minority groups differ systematically in their transitions behaviour which pattern merits a thorough analysis. For example, the differential ability of natives and migrants to avoid unemployment and temporary employment is well-documented (Simpson et al 2007, Heath and Cheung 2006, Heath and Yu 2005, Modood et al. 1997). In addition, the existence of migration routes and sector-based vacancies naturally will preclude some migrants from occupying permanent jobs. In the case of certain groups of migrants, especially temporary seasonal workers coming outside of the EU, unemployment will result in the termination of their legal residence in the UK (DirectGovUK: Migrant Workers 2010). Other groups of migrants may be exposed to higher risks of deskilling and stagnation in precarious occupations if they lack knowledge about the functioning of the receiving country labour market (Friedberg 2000) and ways to minimise the negative effect of an unemployment spell. Even migrants from established minority communities can be exposed to detrimental labour market effects and prefer the supportive network of ethnic economies which guarantee quick transitions from unemployment and minimized risk of losing one s employment (Portes 1995, Alba and Nee 1997, Zhou 2004). Furthermore, in the mainstream labour market, workers on short-term contracts and in temporary positions (Booth et al. 2000) have been found to be less occupationally mobile, in contrast to high-skilled workers who have experienced high growth in demand (Elias 1997). Naturally, this can lead to a gap in the transition probabilities of high and low-skilled workers (Bradley et al. 2003). As the concentration of migrants in low-skilled occupations is greater than among natives (Berthoud and Blekesaune 2007), it seems natural to assume that job acquisition rates of the two groups would be correspondingly affected. To sum up, undoubtedly, the labour market paths of migrants and natives can be quite divergent, which entails that in a more dynamic approach their labour market transition probabilities should also be inspected instead of relying solely on the patterns of disadvantage observed with cross-sectional data. This study will attempt to achieve such insight using the transition models developed in the framework of economic research. 4

5 III. Immigrants and Minorities in the British Labour Market The majority of studies focusing on the performance of ethnic minorities in the British labour market test with cross-sectional data the human capital hypothesis; namely, whether ethnic minority members are still disadvantaged in terms of avoiding unemployment; obtaining similar salary levels and occupational positions as their White British counterparts with control for individual demographics, a set of labour market characteristics and most importantly their educational attainment (Chiswick 1978, Borjas 1999). It is generally believed that in comparison with the 1960s the 70s, 80s and early 90s were characterized by convergence of the economic profiles of the ethnic minority groups and the White British workforce (Robinson 1990, Ignaski and Payne 1996). Some researchers however question these optimistic findings (Modood et al. 1997); arguing that they were based on aggregate data that did not distinguish between first and second generation minority members and consequentially overstated the declining trend in ethnic minority disadvantage. This is to say that ethnic minorities in the 80s and early 90s compared to the 60s might have been taking better jobs, but they were still doing so to a lesser extent than White people with the same qualifications (Modood et al. 1997). Even if there was a positive trend in the economic profile of minority groups in Britain in the 80s, it has certainly reversed by the mid-90s. The findings of the Fourth National Survey of Ethnic Minorities carried out in 1994 showed that Caribbeans, Pakistanis and Bangladeshis suffered substantial employment and earnings disadvantage with almost two-thirds of male respondents from these groups concentrated into manual jobs. In comparison half of the Indians, and two third of Chinese men were in non-manual work (Modood et al. 1997). The good representation of Chinese and South Asians, particularly Indians, in professional, managerial and employing position at the time (the 90s) has been largely attributed to the high rate of self-employment among these groups (Clark and Drinkwater 1998, Clark and Drinkwater 2000); and is 5

6 often cited as indirect evidence of the use of entrepreneurship by minority members as a way of avoiding discrimination in the mainstream labour market and of realizing one s human capital potential (Clark and Drinkwater 2000). Recent research shows that relatively little progress in the economic standing of minority members has been registered since the 1990s. Dickens and McKnight (2008) using the Life-Time Labour Market Database established that migrants were still disadvantaged in terms of closing the pay gap with natives; although more recent migrant groups entered at a higher pay level. Berthoud and Blekesaune (2007), analyzing cross-sectional General Household Survey data from 1973 to 2003 and individual-level linked data from 1971, 1981, 1991, and 2001 Censuses, found that all ethnic minorities in Britain apart from Chinese men and women, and Black Caribbean/Black Other women had lower activity and occupational attainment prospects in comparison with the White reference group. The situation of second generation minority members is also not necessarily improving with the net disadvantage of ethnic minorities in the labour market becoming greater for men born in the UK, especially in terms of avoiding unemployment (Simpson et al. 2006, Heath and Cheung 2006, Blackaby et al. 1999). The question of who gets ahead in the British labour market becomes even more complicated if we consider the newer waves of migrants; that is to say, migrants whose prominence began to be felt at the turn of the millennium when the migrant cohort composition started dramatically to change and the share of Eastern European and Middle Eastern migrants in Britain increased rapidly. Our knowledge of the labour market patterns of these newer and less established flows of migrants is still sketchy. It is clear that there has been a noticeable decline in general of Commonwealth migrants as opposed to the rise of Central, Eastern European and Middle Eastern migrants (Berkeley et al. 2006, Haque 2002); although the post-1990 migration wave has still been dominated by the well-established Commonwealth migrants. According to the 2001 Census, the number of Eastern Europeans living in Britain is relatively low - or roughly three per cent of the population (Home Office Research and Statistics Department 2007). In some respect, we can say that the share of Old and New Commonwealth migrants diminished greatly from thirty and thirty-two per cent of the migrants in 1971 to seventeen and twenty per cent of the 6

7 migrants in 2002 (Berkeley et al. 2006), however, at the same time the visibility of the more recent migrant cohorts, or the public awareness of the whitening of the migrant flows remained low prior to the enlargement of the European Union in Historically, the European migrants arriving from countries both inside and outside the European Economic Area have never been considered problematic (Ruhs 2006, Sales 2007). Nevertheless, 2004 brought a shift in public opinion. As the share of Central and Eastern Europeans entering Britain vastly exceeded expectations and received a lot of media attention, naturally their visibility to the general public increased. Research evidence shows that scrutiny is indeed warranted as the labour market performance of these less-established White migrants has become increasingly divergent depending on their country of origin (Haque 2002, Ruhs 2006, Berkeley et al. 2006). Many of the new migrants are concentrated in temporal seasonal work (Ruhs 2006, Matthews and Ruhs 2007). Demireva (2010) established using data from the British Labour Force Survey that EU10, Eastern European, Turkish and Middle Eastern migrants were particularly disadvantaged in terms of their probabilities of being active, employed and of accessing intermediate positions which cannot be explained by lack of human capital, and their situation is not likely to improve with time. For new migrant women, inactivity rather than unemployment seemed the greatest concern. EU10 and Eastern European migrants also had very high probability of engaging in self-employment compared to British-born Whites and even old migrants. The latter trend has been recorded in other studies as well (Ruhs 2006, Clark and Drinkwater 2008). Clark and Drinkwater (2008) and Drinkwater et al. (2008) focusing on earnings equations also argued that Polish workers (and A8 workers in general) are at a particular disadvantage as they had lower rates of return to their human capital than other recent migrants, even after controlling for personal and jobrelated characteristics. The outlined literature exemplified how divergent is the labour market insertion and incorporation of the various migrant groups in Britain. This study will adopt an approach to country of birth and ethnicity distinctions that will allow us to reflect on the differences between British-born Whites, second generation minority members and migrants. Furthermore, our migrant group will include several sub-groups with which we will aim to capture the differences between countries of birth that are traditional migrant sending countries for Britain such as the New 7

8 Commonwealth, the Old Commonwealth and EU15 countries, and countries from which migration to Britain has experienced a boom in the last twenty years. Overall, our country of birth variable consists of ten categories: White British-born respondents who are our reference category in the multivariate analysis, second generation minority members born in Britain, EU15 and US migrants 3 (reference category in the multivariate analysis for only migrants), EU8, Romanians and Bulgarians, Other Eastern Europeans, Middle Eastern and Turkish migrants, New Commonwealth, Old Commonwealth 4 and Other migrants. This country of birth categorization will also help us ascertain whether the insertion of White migrants is smoother and thus indirectly test the hypothesis of easier incorporation of the less visible migrants (Sales 2007) 5 which hypothesis implies that White migrants will have an advantage over other migrant groups simply due to lack of visibility (Robinson 1990, Alba and Nee 1997, Sales 2007). Likewise, distinguishing between EU8, Romanians and Bulgarians and Other Eastern European migrants will make possible the tapping of divergence within the cohort of White migrants and will take into account the important changes which have occurred within the migrants flows since the expansion of the European Union in 2004 and Unfortunately, due to insufficient numbers we had to merge the categories of EU8 migrants and Bulgarians and Romanians in the multinomial logistic regressions looking into transitions from employment (in all only migrant models as well); however we believe that this joint category continues to capture the impact of cultural and economic phenomena such as the accession of some Central and Eastern European countries to the EU. Second generation minority members, are kept as a separate category, due to the fact that they have been raised in the receiving country and therefore accustomed to its institutional settings to a greater extent than migrants (Friedberg 2000). Their transition behaviour can be expected to follow closely that of British-born Whites although excluding them from the analysis altogether or creating a joint group of natives will be a mistake as numerous studies claim that second generation minority members are still exposed to greater labour market risks than native Whites (Simpson et al. 2006, Heath and Cheung 2006, Heath and Yu 2005). We believe that with this country of birth groupings we will be able to produce more illuminating results about migrants disadvantage and make an important contribution to the literature which has not explored the differences between migrant flows in such detail. 8

9 IV. Data and Methodology This study uses a pooled sample of the British Quarterly Labour Force Survey. Combining data over the span of several years is a common strategy in studies trying to capture the labour market performance of migrants (Heath and Cheung 2006). The Labour Force Survey in Britain follows respondents through five successive quarters every year. Around households are interviewed each quarter, and the sample consists of approximately individuals per quarter. The LFS asks direct questions on country of birth and year of arrival to the UK. Only respondents of working age 16 to 65 are included in the present sample as the focus of this study is on labour market participation. In this paper, the originally cross-sectional observations for each individual followed for five consecutive quarters have been linked together to produce longitudinal data. Unlike most econometric studies which focus on a complete sequence of labour market spells for a balanced panel of individuals (Battu et al. 2008), the panel used in this paper is unbalanced as each respondent contributes as many waves of information to the analysis as they have provided full interviews. Moreover, it is not a rotating panel. We are therefore aware that problems of attrition and non-random drop out from our sample, for example due to change of address, may bias the results 6. Nevertheless, the approach of working with only balanced panel of individuals is subjected to some serious selection biases as well especially in the case of comparing minorities and natives since balanced panels virtually disregard shorter spells which can be quite characteristic for minority individuals. In this paper, we have studied carefully the attrition effects and we will argue that in terms of the recording of respondent s activity patterns based on the LFS ilodefr 7 variable (which is our main variable of interest), British-born Whites and minority individuals exemplify similar patterns of drop-out, that is to say, similar rates of drop-out from wave to wave (see Table A.1 in Appendix 1). This similarity lends some justification to the use of the unbalanced panel and indicates that the attrition rate although in principal slightly higher among migrants than among British-born Whites, and therefore a valid concern, in practice may be less problematic 8. 9

10 Thus, our transitions and descriptive statistics sample with non-missing information on our activity variable consists of men and women, contributing and person-wave observations. 70 per cent of the sub-sample of men and 72 per cent of the subsample of women provide data from all five waves in terms of the activity variable (in the multivariate analysis, we have excluded a number of individuals who have missing data on some of the covariates). In this analysis, we do not flag job-changing as a particular spell, which some econometric studies do (Bradley et al. 2003) as they consider it an important job spell of its own. The minority and migrants job changers in our sample are however too few for any meaningful results in this direction to be derived. This paper focuses on three state transitions. The three states considered are employment, unemployment and non-participation. Nonparticipation is analysed as a separate state as our research focuses specifically on women and minorities among whom non-participation rates are even higher than among natives and among men in general (Simpson et al. 2007) 9. Furthermore, in the multivariate analysis, we model two main sets of transition: the transition from employment at time t-1 to respectively unemployment and inactivity at time t; and the transition from unemployment at time t-1 to employment and inactivity at time t. We include in the analysis different sets of covariates. The rationale behind our country groupings was already discussed in detail in the previous section. We estimate separate models for migrants and natives in an attempt to avoid the misleading results and the problem of structural zero (Upton and Cook 2004) if the variables for which only migrants provide information are present in the models including all respondents. In the models for migrants only, we include the variable years since migration. Unfortunately, the small number of migrants in our sample has allowed us to model distinctively only the transition of migrants from unemployment to employment. In the naturalization variable we distinguish between British-born and foreign-born citizens. The years since migration variable is kept as a continuous one. We look into labour market transitions distinguishing not only by country of birth of the respondents but also by men and women. Amassing the transition behaviour by gender will be a mistake as previous research has shown that the transition probabilities of women are 10

11 much different than those of men (Booth et al. 1999). Moreover, the labour market insertion of migrant men and women is also very divergent (Haque 2002, Berkeley et al. 2006, Demireva 2010). We will take account of this by performing the analysis separately for the two genders. Apart from country of birth, the multinomial models include other covariates as well. Marital status, number of children and education are predictors which have always been found to determine level of job attachment, motivation, access to opportunities in the labour market and social networks (Boheim and Taylor 2001, Montgomery 1991). The education variable has been coded after the International Standard Classification of Education (ISCED) with collapse of several categories. For example, Level 0 and Level 1 are merged to form the Pre-primary and Primary education category. Lower Secondary education category is preserved and Upper Secondary Level 3 ISCED and Post Secondary non-tertiary Level 4 ISCED are also combined into one category Upper Secondary education. Finally Level 5 and 6 of ISCED (first and second stage of Tertiary education) are combined to form the Tertiary level category. The presence of children more specifically may restrict the employment opportunities of the parents (Wanberg et al. 1999). For the transitions from employment we will provide control for industry (two categories: primary and secondary; and tertiary sector) 10, type of occupation 11, permanent-contract versus temporary-contract jobs, and full-time versus part-time employment although we will not model specifically industry or type of contract transitions as done by Haynes et al. (1999) and Booth et al. (1996). In the models looking into transitions from unemployment, we will include control for the main method of job search used in each wave as search techniques may affect differentially the chances of respondents to exit unemployment (Gregg and Wadsworth 1996, Böheim and Taylor 2001). Other important covariates include housing tenure and region of residence (that will be included in all models), claiming of benefits (a whole range of benefits are covered including disability, housing and unemployment benefits) 12 and duration of unemployment (the latter two will be included only in models analysing the transition from unemployment). Research has found that the nature of social tenure may affect disproportionately spatial mobility and labour market activity; for example local authority housing in the early 1980s has been blamed for the relative immobility of its renters and lack of desire to get jobs that will involve a longer commute (McCormick 1983, Hughes and McCormick 1981); a trend observed in more recent studies as well 13. Region of residence is often included in transition research as a spatial 11

12 dimension to the choice of respondents and the opportunities and constraints available to them (Boheim and Taylor 2001, Booth et al. 1999). Finally, all the models include standard control for year in which the survey took place. V. Some Sample Characteristics and General Patterns of Transition Behaviour Table 1 introduces the general rates associated with each of the three labour market states: employment, unemployment and inactivity among the men and women in our sample if we treat each wave as a separate cross-section. It is apparent from the proportion of women in inactivity that the latter state is much more of a problem for women than for men which justifies our approach of separating the analysis by gender. Most importantly, the observed estimates concur with those calculated from other British data sources such as the BHPS (Booth et al. 1999); or those presented in other studies using the LFS (Gregg and Wadsworth 1996) which is a positive sign for the reliability of the panel dataset used in this analysis. [Table 1 comes about here] Tables 2 and 3 report persistence of the labour market state by cross-tabulating the respondent s activity at one wave with the activity at next wave and presenting transition rates averaged over the five waves of the data. What becomes immediately obvious is that among men (Table 2), the probability of remaining at work for two consecutive interviews (Path V) is very high irrespective of country of birth. However, in the other paths divergence in the behaviour of British-born White men and migrant men can be witnessed. Path I shows that the probability of remaining in unemployment is higher among migrants, especially EU8 migrants. The pattern holds for second generation men as well conforming to the 12

13 trend of ethnic penalization 14 in terms of avoiding unemployment observed in previous studies of second generation members (Simpson et al 2007; Heath and Yu 2005). [Table 2 comes about here] Only Romanians and Bulgarians and Old Commonwealth migrant men have lower probability of remaining in unemployment. In contrast, the probability of moving from unemployment to employment is much lower among second generation men (14 per cent) and EU8 migrants (14 per cent) than among British-born men (20 per cent); and these low rates of unemployment exit pertain to Eastern European and Middle Eastern and Turkish migrant men; with Romanians and Bulgarians and Old Commonwealth migrants maintaining their advantage. Finally, it should be pointed out that with the exception of EU8 migrant men, other migrant men have lower probability of remaining in inactivity compared to British-born White men. Perhaps, the high average probability of migrants to remain in employment and lower probability to remain inactive is motivated by the existence of restrictions upon the legal residence of migrants; restrictions which encourage activity and in particular employment on the part of men. [Table 3 comes about here] Among women (Table 3), a similar pattern of the persistence of the employment state is witnessed. Nevertheless, the differences between migrant women and British-born White women are greater. First of all, women in general have higher probabilities in remaining in unemployment compared to men, however more than 60 per cent of second generation women, Romanian and Bulgarian, Eastern European, Middle-Eastern and Turkish and New Commonwealth women are likely to be in this state for two consecutive interviews. Only Old 13

14 Commonwealth women have much high probability of exiting unemployment and especially entering employment compared to British-born White women. The exit rate of unemployment to employment is very low for Eastern European and Middle Eastern and Turkish women as well as EU8 and second generation women. Inactivity seems much greater problem for all women in general although migrant women with the exception of EU8 women are less likely to lapse from unemployment in inactivity than British-born White women. To sum up, at first glance it seems that as we expected the persistence of the labour market states is different for migrants, second generation minority members and natives. Some groups are at a particularly disadvantaged position in terms of avoiding unemployment and inactivity (although the latter spell is much bigger problem for women than for men) for the duration of their presence in the panel. Such groups are EU8 migrants, Middle Eastern and Turkish and Eastern European migrants, both men and women. The situation of second generation minority members, men and women alike and of New Commonwealth women also does not appear very different which goes against the assumption that minority members born and raised in the receiving country or at least migrants from established communities in the UK will have similar labour market behaviour to British-born Whites. In addition, we see confirmed how divergent in their insertion the groups comprising the category of White migrants really are. Only Old Commonwealth, EU15 and US migrants as well as oddly enough Romanians and Bulgarians have similar transition behaviour to British-born Whites. In the next section, we will check these results further against the different sets of covariates that could influence the persistence of respondent s current labour market state, and possibly modify the effect of the predictors, migration status and country of birth, or even reinforce it. VI. Multinomial Logistic Regression Estimates 14

15 Tables 4 and 5 outline the effect of the different covariates on the probability of making the transition from unemployment as a state in time t to employment, unemployment or inactivity at t-1 for all respondents in the sample and respectively for only migrants. All the tables also report the marginal effects and the change associated with the probability of being in each transition state, calculated with the independent variables at their means. [Tables 4 and 5 come about here] Our estimates show that marriage is associated with greater probability of being employed in subsequent quarter for migrant men they being 3 percentage points more likely to be working than single men without children thus perhaps indicating their greater attachment to the labour market if they have families. However, this trend is quite reversed for all women in general and migrant women in particular who are much more likely to remain in unemployment or lapse into inactivity if they are married with or without children compared to being single without children. As established in previous research (Battu et al. 2004), owning one s home or even having a mortgage has quite a positive effect on the respondent s chances of entering employment. In fact, mortgaged ownership has a slightly greater effect in the models driven by natives, with the probability for men increasing with 3 percentage points and for women with 8 percentage points but the estimates are slightly lower for migrant men and women, the share of mortgage owners among whom must be lower as well. The positive association is nevertheless expected as mortgage owners have substantial financial commitments and this may enhance their search activity. Educational achievement also clearly matters. Booth et al. (1999) suggest that better-educated workers are more likely to receive a job offer and they may search for a job more effectively. In our sample, men and women with Upper Secondary/Below Tertiary and Tertiary education have the greatest chances of employment in subsequent quarter. In fact, the positive impact of the Upper Secondary/Below Tertiary education if anything appears to be 15

16 stronger; with the change of the probability of migrant men in that educational category to be unemployed estimated to be 4 percentage points higher compared to.002 points for the effect of Tertiary education. Going against the theoretical predictions is the insignificant impact registered for the variables years since migration and naturalization status. This shows that in the British labour market, it is not all migrants that have lower chances of successful labour market insertion but that some migrant groups are particularly penalized and their chances are not likely to improve with time. Indeed, second generation men and women have greater probability of remaining in unemployment (for women, of lapsing in inactivity as well) with marginal effect for men estimated at 6 percentage points. The same trend is noticed among Middle Eastern and Turkish migrants and New Commonwealth migrants. The latter are even disadvantaged in comparison with the other migrant groups. Overall, the models confirm the negative results outlined with the transition matrices which hold strongly despite control for a range of predictors. Only the penalization experienced by EU8 nationals and Eastern Europeans which was pronounced in the transition tables seems to be explained by the observed covariates. In addition, both the claiming of benefits and increased duration of unemployment affect negatively the respondent s probability of making the transition from unemployment; with claiming of benefits having the largest marginal change of all the independent variables at 45 percentage points for all men and slightly reduced effect for migrant men at 33 percentage points a difference which is perhaps due to lower rate of benefits eligibility among migrants. We turn next to transitions from employment. In the theoretical sections, we have hypothesized that not only migrant men and women are facing greater risks of remaining in unemployment compared to natives, but that they may be more likely also to move out of employment. Table 6 shows that almost all migrant groups, men and women in them including second generation minority men have greater odds of losing their job in a subsequent quarter. This pattern is not registered only among Old Commonwealth migrants and the now joint group of EU8, Romanian and Bulgarian migrant men who have greater odds than natives to avoid the two states of disadvantage: unemployment and inactivity. Of course, in this model, we have merged the groups of EU8, Romanian and Bulgarian migrants due to insufficient numbers which perhaps hinders the clear interpretation of the results. Even EU15 and US migrants with similar set of characteristics as British-born Whites experience 16

17 penalization in terms of their chances of remaining in employment. An inspection of the marginal effects associated with this outcome, however, suggests that the difference between migrant men and British-born Whites is not that big with only the marginal change for Middle Eastern and Turkish women for lapsing in inactivity reaching 2 percentage points. Therefore, we can say that the other predictors in the model have greater impact upon this transition state rather than migration status and country of birth. The impact of other predictors does not go out of the ordinary with married men having greater probability of preserving their job from one wave to the next. For women, however, marriage and the presence of children seems to encourage inactivity although the marginal effect for both groups does not reach one percentage point. Furthermore, full-timers, people on permanent contracts, working in the public sectors and in the highest occupational categories and those with mortgages compared to renters are the respondents enjoying greater stability in their employment status. VII. Discussion The presented results suggest that some migrant groups in the UK experience persistent disadvantage in terms of avoiding unemployment. This pattern holds for the joint category of second generation minority members as well. Although other studies have registered such trends, this is the first paper that explores the problem longitudinally. A close inspection of the marginal effects in the empirical models controlling for a range of predictors shows that the disadvantage of second generation, New Commonwealth and Middle Eastern and Turkish men to remain in unemployment or lapse in inactivity from unemployment as their state of origin is very strong; with inactivity being the more likely option. The disadvantage of Eastern Europeans and EU8 nationals compared to British-born Whites noted in the transition tables, however, seems to 17

18 disappear with control for the individual factors; thus, indicating contrary to our initial hypotheses and beliefs that some of the migrants groups with long established communities in Britain such as New Commonwealth migrants are at a greater risk of remaining out of employment. Overall, the marginal change associated with migrant status in the models focusing on transitions from employment suggests that other factors play greater role in determining the stability of the employment state although some migrant men and women continue to be more disadvantaged compared to British-born White respondents. Perhaps, the existence of different migration channels in Britain that streamline applicants according to their qualifications and industry of the job reduce the magnitude of the negative impact of migration status on the probability of retaining the acquired employment once qualifications and the characteristics of the job are controlled for. Notes: 1. Haynes et al. (1999), for example, study the inter-sectoral transitions and distinguish between employment in the same industry, employment in a different industry, unemployment and inactivity. Burdett and Taylor (1994) analyze transitions to self-employment as well and Booth et al. (1996) focus on transitions from and to employment, analyzing part-time and full-time as two separate states, in addition to unemployment and inactivity. 2. Barnes, Lyon and Millar (2008) focus on family economic transitions; Gregg and Wadsworth (1996) focus on the rate of job acquisition of the unemployed when using state employment agencies. 3. EU15 and US migrants are grouped in one category as previous analysis has shown that their labour market insertion is similar (Haque 2002, Demireva 2010) as migrants drawn from the upper end of the skill distribution. By contrast, it will be a mistake to group together Old and New Commonwealth migrants as their labour market performance is quite different (Haque 2002). 4. The term Old Commonwealth refers to Canada, Australia and New Zealand. The major sending countries of the New Commonwealth are India, Pakistan, Bangladesh, the Caribbean, Kenya and South Africa. For full list of Commonwealth states see McIntyre

19 5. It has long been argued that to be black in Britain entails that the person is simply non-white (Brown and Gay 1985) and White British people appear to discriminate alike towards Caribbeans and South Asians an assumption for which some justice has been found in the British Social Attitudes Survey (Rothon and Heath 2003) and in the most recent comprehensive field experiment in Britain. 6. Drop-out due to change of address is one of the reasons for drop-out in the LFS (Tate 2000). This could be a problem especially if migrants are more residentially mobile than natives. However, as the time interval between interviews in the LFS is only three months and the total duration of the household in the general sample is 15 months, the extent to which residential mobility in the LFS constitute a problem is questionable. The high non-response rates can be due to the fact that participation in the LFS is not mandatory (Clark and Tate 2000) and for some minorities such as Black Caribbeans this has been found problematic (Owen 1993, Lindley et al. 2004). Yet, the authors are not aware of any existing detailed information about differential response rates of migrant groups in the LFS, especially more recent migrant groups, and whether such can be expected. 7. The ILODEFR variable in the LFS captures the reported basic economic activity of the individual in the reference week. It follows the International Labour Organization ILO definition (LFSUserGuide 2010) and allows distinction between inactivity and economic activity, and employment and unemployment. 8. Jones et al. (2007: 267) provide detailed explanation of how survival rates, raw and net attrition rates could be calculated between waves as a good practice of checking the possible effect of attrition in the dataset. We argue similarly to Booth et al. (1999) that using an unbalanced panel of individuals especially as the general activity patterns of the individuals in our sample are similar to those established in previous research (Haque 2002, Clark and Drinkwater 2008). 9. It is acknowledged that other researchers have adopted a different approach. Burdett and Taylor (1994) distinguish between employment, unemployment and selfemployment, minimizing the bias of excluding inactive individuals by restricting the analysis to only those between the ages of 18 to 55. Booth et al. (1999) distinguish simply between those in paid work and those not in paid work. Bradley et al. (2003) distinguish between three types of employment: highly-skilled, skilled, unskilled. The other two states are unemployment and inactivity. We believe that our approach will allow us to best outline some of the general patterns given the number of minority individuals in each state without risking potential bias by excluding inactivity as some authors have done. 10. We distinguish only between two industries because the initial inspection of the data showed that the overall concentration of migrants in the tertiary sector is much higher for migrants than for natives. A detailed industry categorization will present the possible problem of collinearity with the occupation variable. 11. We use the LFS occupational classification distinguishing between: Managers, Professionals, Technicians and Associate Professionals, Clerks, Service workers, Skilled agricultural workers, Craft and Related Trades workers, Plant and Machine operators and Elementary occupations

20 13. Battu et al. (2008) claimed that unemployed public renters appeared much less likely to enter a distant job than private renters although employed home-owners were the ones characterized with the greatest degree of stability as one might expect. 14. The term ethnic penalty has been introduced in referral to any remaining disparity that persists in ethnic minorities chances of securing employment or higher-level jobs, or income, after taking account of measured personal characteristics such as their qualifications, human capital and the like (Heath and Yu 2005: 192). References: Alba, R. and Nee, V. (1997) Rethinking assimilation theory for a new era of immigration, International Migration Review 31: Arulampalam, W. (2001) Is unemployment really scarring? Effects of unemployment experiences on wages, Economic Journal. Arulampalam, W., Stewart, M.B., The determinants of individual unemployment durations in an era of high unemployment. Economic Journal 105,

21 Barnes, M. and Lyon, N. and Millar, J. (2008) Employment transitions And the changes in Economic circumstances Of families with children: Evidence from the families And children study, DWP, Research Report No506 Battu, H., Ma, A. and Phimister, E. (2008) Housing Tenure, Job Mobility And Unemployment In The Uk, The Economic Journal, 118 (March), Bell, B. D. (1997) The performance of Immigrants in the United Kingdom: Evidence from the GHS. Economic Journal, Berkeley, R., Khan, O. and Amikaipaker, M. (2006) What s new about immigrants in twenty-first century Britain? York: Joseph Rowntree Foundation Berthoud, R. and Blekesaune, M. (2007) Persistent employment disadvantage, Department for Work and Pensions, Research Report: 416 Blackaby, D. H., Leslie, D. G., Murphy, P. D. and O'Leary, N. C. (1999), Unemployment among Britain's ethnic minorities, Manchester School of Economic and Social Studies, 67: 1-20 Böheim, R. and Taylor, M. (2002) Job search methods, intensity and success in Britain in the 1990s, Working paper No: 0206, Johannes Kepler University of Linz Booth, A. L., GarcõÂa Serrano, C. and Jenkins, S. P. (1996). `New men and new women: is there convergence in patterns of labour market transition?' Working Paper No. 96-9, ESRC Research Centre on Micro-Social Change, University of Essex, Colchester. Booth, A.,Jenkins, S and Serrano, CG. (1999) New men and new women? A comparison of paidwork propensities from a panel data perspective, Oxford Bulletin Of Economics And Statistics, 61, 2 :

22 Booth, A.L., Francesconi, M., Frank, G., Temporary jobs: who gets them, what are they worth, and do they lead anywhere? Paper presented at EEEG, July Borjas, G. (1995) Ethnicity, neighborhoods and human capital externalities, American Economic Review 85: Borjas, G. (1999) Heaven s Door: Immigration policy and the American economy, Princeton, Princeton University Press Bradley, S., Crouchley, R and Oskrochi, R. (2003) Social exclusion and labour market transitions: a multi-state multi-spell analysis using the BHPS, Labour Economics (10): Brown, C. and Gay, P. (1985) Racial Discrimination. 17 Years after the Act. London: Policy Studies Institute Brown, C. and Gay, P. (1994) Racial Discrimination: 17 years after the Act, pp in Burstein, P. Equal employment opportunity: labour market discrimination and public policy, New York, Aldine de Gruyter Burdett, K., Taylor, M.P., Semi-markov and markov labour histories. Working Paper, vol. 27. ESRC Centre on Micro-Social Change, University of Essex, Essex. Chiswick, B. (1978) The effect of Americanization on the earnings of foreign born men, Journal of Political Economy 86 (5): Clark, K. and Drinkwater, S. (1998) Ethnicity and self-employment in Britain, Oxford Bulletin of Economics and Statistics 60: Clark, K. and Drinkwater, S. (2000) Pushed out or Pulled In? Self-Employment among Ethnic Minorities in England and Wales. Labour Economics, 7,

23 Clarke, P.S. and Tate, PF. (1999() Methodological Issues in the production and analysis of longitudinal data from the Labour Force Survey. GSS Methodology Series no17. Demireva, N. (forthcoming) New Migrants in the UK: Employment Patterns and Occupational Attainment ; accepted by The Journal of Ethnic and Migration Studies DirectGovUK: Migrant Workers 2010, last accessed Dustmann, C and Fabbri, F (2000) Language proficiency and labour market performance of immigrants in the UK. Discussion Paper No 156, IZA Bonn. Elias, P., Restructuring, reskilling and redundancy: a study of the dynamics of the UK labour market, Working Paper, vol ESRC Centre on Micro-Social Change, University of Essex. Fielding, A. J. (1995) Migration and Social Change: A Longitudinal Study of the Social Mobility of `Immigrants' in England and Wales. European Journal of Population, 11 (2): Friedberg, R. (2000) You can t take it with you? Immigrant assimilation and the portability of human capital, Journal of Labor Economics, 18(2): Gordon, M. (1964) Assimilation in American life: the role of race, religion, and national origins. New York: Oxford University Press Gregg, P. and Wadsworth, J. (1996) How effective are state employment agencies? Jobcentre use and matching in Britain, Oxford Bulletin of Economics and Statistics, 58(3),