Key factors behind labour market marginalization of young immigrants: limited access to apprenticeships, state dependence or low qualifications?

Key factors behind labour market marginalization of young immigrants: limited access to apprenticeships, state dependence or low qualifications? Manuscript February 2011. Final version later published in Young, 2011, 19(2) 129 158. LIV ANNE STØREN NIFU Nordic Institute for Studies in Innovation, Research and Education, Norway ABSTRACT This paper studies the transition from school to work among youth with low or medium-low competence level. Youth who have proceeded to higher education are excluded from the analyses. The youth have all participated in Norwegian upper secondary education, but a large part has not completed their schooling. The situation of youth of non-western origin and ethnic Norwegian youth is compared. The effects of apprenticeship experience, qualifications and state dependence on the labour market situation are examined. State dependence is understood as the effect of unemployment experience at an early career stage on future labour market opportunities. The ethnic minority youth do not have the same probability of being employed as the ethnic majority youth with the same competence and grades level. But, the positive effect of apprenticeship experience was less important than what was expected. Although large effects of state dependence are found, this does not explain the immigrant youth s higher rate of being without work. Keywords Apprenticeships, employment, immigrants, qualifications, youth 1

INTRODUCTION The transition from school to work is strongly influenced by the youth s qualification level (Gangle, 2002; Ryan, 2001; van Ours and Veenman, 2004). The way in which training has been carried out, for instance school-based or through apprenticeships, may also play a significant role (Ryan, 2001). The transition phase is often characterized by movements in and out of employment, job-seeking, and education/training. Moreover, youth s first experiences in the labour market may mark their further adaptation and opportunities. A transition phase characterized by movement in and out of different statuses is more typical for those youths who do not enrol in or aspire to take higher education than for those who do so. In this paper we focus only on youth who have not commenced or taken any higher education. Negative initial labour experiences among school leavers may be more important for immigrant youth than for youth with an ethnic majority background, and may constitute a significant part of an explanation of labour market marginalization of immigrant youth such as is observed in many Western countries. Research objective This paper studies the transition from school to work among youth with low or medium-low competence level. They have all attended Norwegian upper secondary education but not proceeded to post-secondary education. The competence level differs, but the highest level is completed upper secondary education. The situation of youth of non-western origin and youth of ethnic Norwegian origin is compared. The main purpose is to examine the effects on the labour market situation of a) apprenticeship experiences; b) state dependence. The latter is understood as the future effect of being in a marginalized labour market situation at an early career stage (see further elaboration below). Initial unemployment may in part determine future marginalization. To what extent do such factors affect the observed differences between youth of non-western origin and youth of ethnic Norwegian origin? Qualifications measured as grades and attained competence level will be used as control variables. UPPER SECONDARY EDUCATION IN NORWAY Upper secondary education constitutes a significant part of the national transition regime. A brief description of the Norwegian upper secondary education is thus necessary. The Norwegian youth transition regime may be described in the terms of Walther (2006) (who uses Denmark and Sweden as examples of the Nordic regime) as universalistic, 1 and which refers both to the welfare regime and the comprehensive, non-selective, school system. The participation rate in upper secondary education among the ethnic majority aged 16 18 is 92 per cent, but is lower among non-western immigrants: 74 per cent of the first-generation and 89 per cent of the second-generation non-western immigrants (Støren et al., 2007). Norwegian upper secondary education is divided into a vocational track and a general academic track. About half of all students commence in vocational programmes and the other half in general, academic programmes. Students with non-western background are overrepresented in the general track (Støren et al., 2007). General academic competence (university admission certification) is attained following the completion of three years in general programmes. The vocational students can choose between many specialized courses the second year of training, based on the broader programme they have commenced the first year. The main structure is shown in Figure 1. 2

(Figure 1 about here) For vocational students it is also possible to take the final (the third) year as a general programme in order to achieve university admission certification. In addition, a few courses within vocational programmes offer subjects the third year that lead to university admission certification. There is also the opportunity to take an additional year on a general programme after having achieved vocational competence. Vocational competence is normally attained after completing two years in school followed by a two-year in apprenticeship and after passing a final trade (craft or journeyman s) examination (the 2 + 2 model ). This is often called a dual system, referring to the (dual) contribution of training in enterprises and in school. A few vocational programmes are entirely school-based. Vocational competence is then attained after completing three years of schooling. 2 The selection to school-based training versus the 2 + 2 model is based on the type of course chosen the second year, and is normally not based on grades. Some students who follow the 2 + 2 model do not obtain an apprenticeship place. It is the employers who decide whether or not they will offer the applicant an apprenticeship. When it is impossible to provide a sufficient number of training places for those who apply for apprenticeships, the entire training may be provided in school. In such situations the county authorities are obliged to offer third year in school. The students in this group generally have poorer grades the second year of upper secondary education than those who obtain an apprenticeship (Støren et al., 2007). The final trade examination is the same as it would have been after apprenticeship training, but the proportion which fails to pass the examination is much higher than among the apprentices (Støren et al., 2007). The areas building and construction, mechanical trades and electrical trades include the largest groups of the male vocational students. Health and social care, arts and design, and the hotel- and food-processing trades include the largest groups of female vocational students. PREVIOUS RESEARCH Qualifications or discrimination? Although the relationship between educational attainment and labour market outcome is long established both empirically and theoretically (cf. the human capital theory (Becker, 1975) or the signalling and filter theories (Arrow, 1973), a number of studies show that the traditional human capital variables such as the number of years in school are not sufficient to explain differences in labour market outcome. Ethnic differences in labour market opportunities irrespective of educational level are found in many studies, for instance in Canada (Zietsma, 2007); Denmark (Skyt Nielsen et al., 2003); Flanders part of Belgium (Vanoverberghe et al., 2008), France (Frickey et al., 2006; Silberman and Fournier, 2007); Germany (Kogan, 2007); Norway (Brekke, 2007; Støren and Wiers-Jenssen, 2010); Sweden (Nekby and Özcan, 2006), and UK (Dustman et al., 2003, Battu and Sloane, 2004). However, in a Dutch study (Van Ours and Veenman, 2004) which examined job opportunities and wages among ethnic Dutch youth and second-generation immigrants, the conclusion was that ethnic background does not matter much. What did matter was the educational level. A study by Malmberg-Heimonen and Julkunen (2006) comparing immigrant and non-immigrant youth in Finland, Sweden, France and Germany, showed varying results; only in Finland did immigrant youth have a significantly lower chance of obtaining employment after longer-term unemployment than non-immigrant youth. On the other hand, in their review of research on second-generation 3

immigrants in Western Europe, and after controlling for educational level, Heath et al. (2008: 218) found that minorities from less-developed non-european origins tend to have substantially higher risks of unemployment than their respective majority groups. Although the findings differ to some extent, the overall results in studies investigating ethnic differences in the labour market indicate that theories of discrimination are necessary in order to understand differences in labour market opportunities and careers. Discrimination may take several forms, such as preference discrimination (Becker, 1971); unprejudiced discrimination (Merton, 1949) referring to situations where the employer prefers a job seeker from the majority group because of other people s prejudice; or statistical discrimination. The latter occurs when an individual is judged by the anticipated average characteristics of the group to which he or she belongs (Thurow, 1975). Research can seldom prove that discriminatory acts actually have been carried out, or what kind of discrimination that has possibly taken place. A Swedish study using experimental data 3 (Carlsson and Rooth, 2007) found that, depending on the measure used, between 9.6 and 29.4 per cent of firms studied discriminated against Arabic-sounding names. Further, their calculations indicate that discrimination was responsible for about one sixth of the nativeimmigrant unemployment gap. For that reason they conclude that language skills and other unobserved characteristics could be important in explaining the remaining gap. In addition to language qualifications, the amount of resources situated in social networks often labelled social capital (Coleman, 1988; Portes, 1998) provides a relevant explanation. The use of one s social network is an effective strategy for job seekers. In the analyses below we will examine the situation in Norway for both first- and second-generation young immigrants. We assume that the families of students who are born in Norway will possess more relevant social capital as well as a higher proficiency in the Norwegian language than the first-generation immigrants. Thus, we will take into account whether the youth are first- or second-generation immigrants. Another question is whether an unfavourable labour market situation applies to East-European immigrants to the same extent as non-european immigrants. It is probable that visible minorities (of non-european origin) are more subject to discrimination than other immigrants. This is indicated by a Swedish study (Rooth, 2002) on the labour market situation covering people who were born abroad and adopted as infants. The study by Carlsson and Rooth (2007) mentioned above also indicates this (although only the Arabic-sounding names and not Arabic appearance was visible in their study). In Canada, studies by Tran (2004) and Kunz, Milan and Schetagne (2000) have found an unfavourable situation for visible minorities when educational level is taken into account. Such findings suggest that a control be included for East-European and non-european origin. This is done in the analyses below. Type of education and institutional setting Type of education and training refers here to the distinction between entirely school-based training, and training including apprenticeship. Another distinction is between general and vocational programmes. Most students on general programmes plan take higher education (HE), but some prefer working, and some do not achieve the necessary HE entry qualifications and enter the labour market instead. Therefore, also (former) students on general programmes are of interest for our study. 4

A general finding in international studies is that in countries with an institutional link between the education and training such as an apprenticeship system (a dual system), the transition phase is smoother (Shavit and Müller, 1998; Ryan, 2001) than in countries that offer mainly general upper secondary education. Some studies, for example Wolbers (2007) and De Grip and Wolbers (2006), have shown some diverging results concerning the general positive effects on the school-to-work transition in countries in which there is a high share of upper secondary education students in vocational education or in apprenticeships. As mentioned above, most of the vocational upper secondary education in Norway is designed as a dual system. The allocation of apprenticeships is thus of large significance. If immigrants are at a disadvantage when looking for apprenticeship places, and some studies do in fact indicate this (Faist, 1993; Worbs, 2003; Helland and Støren, 2006), their later opportunities in the labour market may be harmed. Due to the fact that discrimination occurs, a dual system of vocational education may although beneficial for the majority of students reduce the labour market opportunities of immigrant youth relative to the ethnic majority youth. This point is also mentioned by Dörr and Faist (1997), referring to the fact that private employers determine access to training places. State dependence As predicted in the theory of state dependence (Heckman and Borjas, 1980; Andress, 1989) persons who experience labour market problems early in their career may have an increased risk of being unemployed at a later stage as a result of a self-enforcing process. State dependence may occur as a result (among other things) of loss of work experience, implying lack of on-the-job-training. It may also be a result of the employers sorting: firms may use unemployment records in their hiring decisions (Heckman and Borjas, 1980: 247). It may also be caused by unobserved heterogeneity: individuals differ in certain unmeasured variables that influence their probability of experiencing unemployment but are not influenced by experience of unemployment (ibid:247). One important variable on which the individuals may differ is formal qualifications. This is controlled for in the analyses below. According to the theory of state dependence we expect that those who are in a marginalized labour market position (unemployed, on labour market programs, rehabilitation/social benefits, outside labour force) at one stage will have increased risk of being in a marginalized position at later stages. The importance of state dependence may differ between the ethnic minority (immigrant) youth and majority youth. If immigrant youth have a higher unemployment rate than the ethnic majority youth when leaving school, the former may have increased risk of being unemployed (also) at a later stage because of state dependency. Thus, effects of state dependence may contribute to explaining the immigrants later disadvantage in the labour market. Since some forms of discrimination may, at least partly, have caused the initial disadvantage, both theories of discrimination and state dependence are of great relevance. In addition, one aspect of state dependence is a certain scarring effect of unemployment because of the employers sorting of job applicants. This effect may be larger for ethnic minority groups than for the ethnic majority. This has previously been shown for higher education graduates (Støren, 2004). One question is whether this also applies to persons with lower qualification levels. RESEARCH QUESTIONS 5

Below we have formulated five hypotheses based on the theories and findings sketched above: Apprenticeships Students who apply for apprenticeship and who do not obtain such a training place, may refuse an offer of a school-place. The reason can be disappointment; they are tired of school and want access to a workplace. Accordingly, they will frequently belong to the group who have not completed upper secondary education. Further, those who accept an offer of a school-place fail at the final examination more frequently than apprentices (Støren et al., 2007), and for that reason they also belong to the non-completers. The non-completers will have poorer labour market opportunities than the completers. In addition, based on the studies mentioned above, it is reasonable to assume that an apprenticeship in itself will bring about labour market-relevant social contacts. Research has also shown, as mentioned above, that non-western immigrants do not have the same access to apprenticeship as the ethnic majority. The two first hypotheses are based on these considerations and findings. H1: Apprenticeship experience will increase the probability of being employed and will compensate for poor school grades. H2: Unequal access to apprenticeship is a major cause of differences between minority and majority youth regarding their employment level. Non-Western immigrants or their families are newcomers in Norwegian society. Thus, they have more restricted labour market-relevant social network. Apprenticeships represent a good opportunity for immigrant youth to obtain such contacts, something which will increase their later chances on the labour market. Thus, the next hypothesis reads: H3: Apprenticeship experiences will be of particular importance for the employment rate among the non-western immigrants. State dependence Previous research has shown that the initial transition problems are larger among minority youth than among the ethnic majority. With reference to theories of state dependence mentioned above, it is expected that: H4: The initial transition problems (measured at T1) among the minority youth are the source for further problems at later stages (T2 and T3), and this constitutes a major cause for their lower employment level at the later stages. Further, with reference to research mentioned above it is expected that the scarring effect of previous unemployment that is the scarring aspect of state dependence is largest among the minority youth. Thus, the last hypothesis reads: H5: The effect of state dependence is larger for minority youth than for the majority youth, resulting in accumulated risk for marginalization among the minority youth. DATA 6

The data examined comprise register data for persons who embarked upon upper secondary education (for the first time) in Norway in 1999 and 2000. 4 The observation period extends to November 2005. From the data set of the entire two cohorts (98,452 persons) those who had commenced post-secondary education in 2005 or earlier (47,966 persons) are excluded. A small group of persons with Western immigrant background (0.4 per cent of the data set) are also excluded because we will compare the situation for ethnic Norwegians and persons with non-western immigrant background. The size of the data set used in the analyses is then 50,235 persons. It is important to be aware that the findings in this paper do not refer to the entire upper secondary cohorts, but only to those who have not proceeded to higher education. The reason for this delimitation of the population included in the analyses is that it gives little meaning to analyse the labour market situation for those who have proceeded to higher education and to compare it with the situation for school-leavers with lower education levels. A large part of the HE students have paid employment, but this is normally part-time employment to get extra income while studying. However, the data set that is analysed here includes persons who still are in upper secondary education although with a normal study progression they should have left the upper secondary education. The reason why they are not excluded is that many upper secondary students are delayed or are temporary drop-outs and return to upper secondary education within the observation period. Further, the tendency to return to upper secondary education varies between the ethnic majority and the minority. Thus, it is interesting to include also persons who still are in or have returned to upper secondary education in the analyses. A selection problem may arise because non-western immigrants are overrepresented among those who proceed directly to higher education after completing upper secondary education (and thus are excluded). This over-representation exists despite the fact that immigrant youth do not have better grades than the ethnic Norwegian upper secondary students (Støren, 2009). The ethnic Norwegians tend more frequently to postpone their enrolment in HE. This indicates that students with high academic ability are relatively more represented in the data set that is used among the ethnic Norwegians than among the immigrant youth. This factor adds to the fact that immigrants generally have a lower (poorer) grades level (Støren et al. 2007). However, controls are applied for grades and competence level, something which reduces this possible selection bias. Of the persons from the two cohorts analysed in this paper (i.e. had not proceeded to postsecondary education), 18 per cent had attained general academic competence, and about 30 per cent had attained a vocational competence within five years after starting in upper secondary education (total: 48 per cent). The corresponding figures if we had looked at the total upper secondary cohorts and included those who had proceeded to post-secondary education, are 52 per cent completing with general academic competence within five years, and about 18 per cent with vocational competence (total about 70 per cent). The figures illustrate that the proportion not completing the course is very high among students on the vocational courses 5 (Støren et al., 2007). The figures mentioned also imply that the persons included in the data set generally have rather low competence level: more than half of them have not completed upper secondary education. Further, when comparing this with the total cohorts, the population in the data set that is analysed is over-represented by students who have started on vocational programmes. This does not indicate a selection problem as long as we keep in mind that the results do not refer to the total cohorts, but only those who have not been enrolled into post-secondary education within the observation period. 7

Nevertheless, a selection problem could arise from the fact that there is a selection of students who start in upper secondary education, and into different types of upper secondary education. For instance, first-generation immigrants are under-represented (in total) in upper secondary education. This refers mainly to the fact that youth with rather short residence time are not enrolled (Støren et al., 2007). Then again, it is important to be aware that the results refer only to persons who have participated in upper secondary education (and to those of them who have not been enrolled into post-secondary education within the observation period). Further, students with immigrant background are, in total, under-represented on vocational programmes: girls and boys choose different programmes, and to some extent selection to the different programmes based on academic ability. Different aspirations also exist among the group concerning possible future enrolment into higher education. Most of this heterogeneity is controlled for in the analyses since controls are applied for grades, gender, immigrant background and type of programme (vocational versus general programme). However, it is not possible to take into account unobserved heterogeneity for instance in terms of language skills and the youth s aspirations. VARIABLES, METHOD AND DEFINITIONS The dependent variable We will examine the labour market status of youth at different points of time. Descriptive statistics are shown in Table 1. The first point, T1, refers to four years (and three months) after entering upper secondary education, that is in November 2003 for the 1999-cohort, and 2004 for the 2000-cohort. Conversely, T2 refers to five years after embarking on upper secondary (2004 for the 1999-cohort and 2005 for the 2000-cohort), and T3 refers to six years after starting in upper secondary education. T3 refers to the situation in November 2005 for the 1999-cohort. Data for T3 is only available for the 1999-cohort. The employment and unemployment levels depend on the inclination to return/continue in upper secondary education among those who have dropped out/are delayed, and also the extent to which the young persons have other activities. Thus, the dependent variables have several outcomes, as described below: 1) Without work. The category contains persons registered with a marginalized labour market status; that is registered as unemployed, on labour market programmes or rehabilitation measures, or on social assistance without being employed. Table A.1 in the Appendix shows the distribution of the subgroups of this category. 2) In education at upper secondary level. Some of those who are registered as students in upper secondary education are also registered as employed. Probably, most of these are apprentices. This group is included in the education category and not in the employed category. 3) Employed (only). Because apprentices and part-time students are included in the category in education (see above), this group includes persons that are employed only (i.e. not apprentices). 4) Other. The registers of Statistics Norway do not give information about this group. They may be at home, caring for children, unregistered unemployed, in military service, travelling abroad, studying on their own or having unregistered work. It is important to be aware that the percentage in the first category (without work) is much higher than the subgroup registered as unemployed. The registered unemployed constitute 35 8

40 per cent of this group. Because the total category comprises more persons than those registered unemployed, the category will be labelled Without work. The independent variables We will investigate the effects of different independent variables on different outcomes of the dependent variable. Because the dependent variable has several outcomes, multinomial logistic regression analysis is employed. The analysis shows whether the independent variables increase or decrease the probability of the different outcomes on the dependent variable. These independent variables are described below, and descriptive statistics for these variables are shown in Table A.2 in the Appendix. In addition to demographic control variables (age, gender etc.) the main independent variables refer to immigrant background, qualifications and apprenticeship experiences. Ethnic background. Ethnic Norwegians are the so-called majority group, and serves as the reference category in the regression analyses below. These are persons born in Norway with Norwegian born parents (one or both parents); alternatively they are born abroad and have Norwegian born parents. What is referred to as ethnic minority groups are (a) first-generation immigrants with a non- Western background, i.e. are born outside Norway and with both parents born abroad, and (b) second-generation non-western immigrants, that is born in Norway but where both parents are foreign born. Persons with a non-western background are born in (or their parents are born in) Asia, Africa, Latin-America or Eastern Europe. First-generation immigrants with an East-European background are, in some analyses, separated from those originating from Asia, Africa, and Latin America. The latter group is labelled non-western, else, first generation, or in the text alternatively as non-europeans. Due to the restricted number of observations all second-generation non-western immigrants are categorized together. Only a few (8 per cent) have an East-European background. Level and type of competence. Four levels of acquired formal competence (at T2) are defined: 1. Value 1 on the competence level variable: Basic course (the first year of upper secondary) is the maximum attained education level. 2. Value 2: Completion of the second year of upper secondary education is the maximum level, that is having been registered as a student during the second year of upper secondary education. 3. Value 3: Having been registered the third year (as apprentice or third year at school) at T1 or earlier, but not registered as having vocational or general academic competence at T2. 4. Value 4: Registered as having attained either vocational competence or general academic competence at T2. In addition control for whether the highest competence level was vocational or general is included, as well as whether the students were registered for a general or vocational track when they entered upper secondary education. The categories below the highest level include both delayers and drop-outs. This means persons who have not returned to upper secondary after dropping out, persons who have returned after dropping out (delayers), and others who are delayed because of failures or rechoice of study programmes. 9

Table A.2 shows that students with non-western background are overrepresented among those with low levels of educational attainment. Of the ethnic Norwegian students 49 per cent had achieved either vocational competence or general academic competence, whereas the corresponding estimate for the East-European first-generation immigrants is 37 per cent, and only 31 per cent among the first-generation non-european immigrants. Grades. The grades-variable refer to average grades on the basic course, i.e. the first year of upper secondary education, because naturally there is no information on grades at later stages among those who dropped out. A dummy-variable information on grades is included, because we miss information on grades on basic course for some students, either due to unknown practical reasons or because the students dropped out during the first school year. In Table A.2 we see that students with non-western background are overrepresented among those for whom we lack information on grades on the basic course. In addition, among those for whom we have information on grades, the average grades are better for the ethnic Norwegians than the non-western immigrants. Apprenticeship. Although most of the population in the data set (72 per cent) started on a vocational programme, only 29 per cent have been registered as apprentices during the 2001 2003 period. 6 The reasons why the percentage is not higher are several: they may not have succeeded when applying for apprenticeship, they have dropped out of education, or they have taken school-based education. The ethnic majority have far more frequently than the immigrant groups been apprentices (see Table A.2). This is evident also when taking into account that the immigrant groups more often choose the general track than the ethnic Norwegians when commencing upper secondary education. Descriptive results Table 1 displays descriptive statistics of the labour market situation at three points of time (T1, T2 and T3), by immigrant status. (Table 1 about here) Table 1 shows that the percentage without work is more or less the same at the three points of time. The percentage employed increases considerably between T1 and T3. What is decreasing is the share who are (still) in upper secondary education. The table also shows that the percentage without work is much higher among the first-generation non-western immigrants than among the ethnic Norwegians, in both cohorts and at all points of time. Furthermore, the proportion belonging to the other-group is higher among the firstgeneration immigrants than among the ethnic Norwegians. The situation for the second-generation non-western immigrants lies somewhat between. The percentage without work is lower in this group than among the first-generation immigrants, but the share belonging to the other-group is as high as among the first-generation immigrants. Among the second-generation immigrants in the 1999-cohort there is a large increase in the employment rate from T1 to T3. The situation for the second-generation is quite similar to the situation among the ethnic Norwegians, except for the high share belonging to the other-group, and except for the fact that a smaller percentage continues in, or returns to, upper secondary education. ANALYSES 10

Table 2 shows the results of the multinomial logistic regression analysis. The results are described below and illustrated in Table 3 and Figures 2 and 3. (Table 2 about here) The effects of the different independent variables are shown in four different models in Table 2. Model 1 includes only demographical variables (having children, gender, age, immigrant background) as well as type of programme when entering upper secondary education and whether living in Oslo. These variables are the base and are retained in the subsequent models. To save space, only the effects of gender, having children, and immigrant background are shown. 7 In order to show the effects of an apprenticeship, this is included in Model 2 (and retained in the following models). Control for qualifications is introduced in Model 3, and the effect of state dependence is seen in Model 4. The upper part of Table 2 examines the effects of different variables on the probability of being without work at T2 relative to being employed. In the next part the probability of staying on in upper secondary education at T2 relative to being employed is shown. The lower part shows the probability of belonging to the other-group. The reference category in Model 1 is ethnic Norwegian males on vocational programmes. For females (both immigrant and non-immigrant females) it is distinguished between having children or not. The model includes separate dummy-variables for male immigrants of different backgrounds; however, this is not the case for female immigrants. The reason is to avoid including interaction terms in the logistic model (see below) and to avoid including too many independent variables. The model is also based on preliminary analyses showing that among the females (irrespective of immigrant background) having children is the essential factor and that there is no significant difference in the risk of being without work between ethnic Norwegian and immigrant females without children. The risk of being without work is higher among immigrant males than females (if they do not have children), which is also shown in Table 2. The gender difference is probably due to the fact that the major part of the non-western female students on vocational programmes choose health and social care studies, and in this sector the demand for labour has been very strong. These work-places are in the public sector and are femaledominated. An interesting finding of Carlsson and Roth (2007) is that males responsible for recruitment discriminate to a greater extent than females. Further, Carlsson and Roth found that large firms (more than 100 employees) discriminate less frequently than smaller firms. The vast majority of male vocational students and apprenticeship seekers are exposed to maledominated firms most often of small size. 8 Model 1 shows that males with non-western origin (all the three categories) have a higher risk of being without work than the ethnic Norwegians. The effects of having been an apprentice Control for apprenticeship experience is included in Model 2, where the reference category is an ethnic Norwegian male without apprenticeship experiences. The control for apprenticeship experience refers only to males of different backgrounds. The reason is that 71 per cent of those having apprenticeship experience are males. Further, the model is specified without interaction terms 9 and it is also necessary to avoid having too many independent variables. (i.a. because of restricted number of observations in some cells). The results of Model 2 indicate that apprenticeship experience strongly reduces the risk of being without work, both among Non-Western immigrants 10 and ethnic Norwegians, however; the decrease is largest among the ethnic Norwegians. This result is opposite to what was expected: apprenticeship 11

experience does not seem to be of greater importance for immigrants than for nonimmigrants. Model 2 also shows that among males with no apprenticeship experience, there is significantly higher risk of being without work among first-generation males of East- European or non-european origin than among ethnic Norwegian males. Model 3 includes qualification related independent variables such as grades and attained competence level, which both has strong effects. (The references category in Model 3 is the same as in Model 2.) Good grades decrease the risk of being without work, also after controlling for competence level etc. When looking at the effects of these variables, one also has to take into account the second-order variables for competence level and grades. These are included because there is a curvilinear relationship between qualifications and some of the outcomes on the dependent variable. The sign of the coefficient for the second-order variable for competence level is negative. When the effects of the first- and second-order variables are added, we see that increasing competence level decreases the risk of being without work. There is a strong relationship between grades and the chance of obtaining an apprenticeship, thus, the inclusion of the qualification-related independent variables decreases the effect of having been an apprentice. This refers to both ethnic Norwegians and immigrants. The reduction of the effect of apprenticeship from Model 2 to Model 3 (i.e. after controlling for grades etc.) is so large that we find rather weak support for the expectation that apprenticeship experience compensates for poor grades. In addition, there is no extra effect for the immigrants of having apprenticeship experiences on the probability of being without work. Neither the results of Model 2 nor Model 3 confirm that having apprenticeship experience is of special importance for non-western immigrants, although apprenticeship experience decreases the probability of belonging to the othergroup among the non-western immigrants. This is shown in Table 3, which depicts estimated results of Model 3. 11 (Table 3 about here.) Even though the effects of having non-western origin on the risk of being without work tend to decrease after the inclusion of the qualification related variables in Model 3, male firstgeneration immigrants still have significantly higher risk of being without work than ethnic Norwegians. This refers to those of East-European as well as non-european origin. On the other hand, the second-generation non-western immigrants do not have significantly higher risk than the ethnic majority of being without work when controlling for competence level, grades and previous apprenticeships. Language proficiency, on which we lack information, is one factor possibly explaining some of the remaining differences between the immigrants and the ethnic majority. However, there are two reasons why lack of proficiency in Norwegian probably does not serve as a very good explanation. One is the fact that we have controlled for grades as well as formal competence level. It is probable that language proficiency is related to grades and attained competence level. The other reason concerns the length of residency in Norway among the first-generation non-western immigrants which is 12.5 years on average at T2. This is quite long, and given the fact that all the non-western immigrants included in the analyses have participated in Norwegian upper secondary education, it is reasonable to assume that the proficiency in the Norwegian language is quite good. In additional analyses we have also included control for length of residency in Norway. This is not included in Table 2. The reason for this is that this 12

control did not increase the explanatory power of the model, and further, the effect of residence time was opposite to what could have been expected. Those with the shortest duration of residency in Norway tended to have the lowest unemployment risk. 12 This may be due to a selection effect: those with short residence time who have taken Norwegian upper secondary education are probably a particularly able and motivated group. The effects of previous experiences (state dependence) The situation one year prior to T2, that is whether the person was without work at T1 or belonged to the other-group at T1, is used as an indicator of state dependence and included as control variables in Model 4 (Table 2). Whether one was in upper secondary education at T1 is also included as a control variable. In order to see whether these effects vary between the groups these control variables are specified separately for ethnic Norwegians and immigrants. (The specifications are also made in order to avoid including interaction terms in the logistic model, see endnote 9.) Due to few observations in some cells when including these variables there is made no distinction between type of immigrant background in Model 4. This implies that second-generation immigrants are categorized together with firstgeneration immigrants. All these activities (statuses) at T1 compared to being employed at T1 tend to increase the risk of being without work at T2, and this refers to both ethnic Norwegians and persons of non-western origin. The effects of these previous statuses are also the same for the two groups. The effect (at T2) of being in upper secondary education at T1 may be seen in light of the fact that those who were in upper secondary education at T1 represent a group who were delayed in their educational career. There are in particular very strong effects of being without work at T1 on the probability of being without work at T2 (cf. the coefficients 2.352 for ethnic Norwegians and 2.376 for immigrants). Being without work at T1 also increases the risk of belonging to the othergroup at T2. All the results confirm the expectation of state dependence. It was expected that state dependence would be a major cause of the immigrant youth s high unemployment level at T2, something which would result in no, or only a small, effect of being non-western if not having been without work at T1. This effect is however still significant (+0.443). Further, according to our expectations there should be a larger effect of being without work at T1 for immigrants than for ethnic Norwegians, which is not found. The reason for these findings is not manifest. A reasonable interpretation can be that it is not the effect of previous unemployment that is the major cause for the minority youth s overall greater risk of being without work at T2 relative to the ethnic Norwegians, but other reasons. One reason might be that they frequently have had temporary work-contracts at T1. Other explanations concern for example discrimination or lack of proficiency in the Norwegian language. The results of Model 4 (Table 2) are illustrated in Figure 2. (Figure 2 about here) Figure 2 displays that among those who were employed at T1, the non-western immigrants are more frequently without work at T2 (9.5 per cent) than ethnic Norwegians (6.2 per cent). This is after control for grades, competence level etc. (See the note describing reference group 13

below Figure 2 and the formula used in endnote 11.) Among those without work at T1, there is no difference between the non-western immigrants and ethnic Norwegians with regard to the estimated probability of being without work at T2 (34.4 versus 34.5 per cent). As mentioned above, this indicates that it is not an effect of previous unemployment that is the major cause for the minority youth s overall greater risk (relative to ethnic Norwegians) of being without work at T2, but other (unknown) reasons. A corresponding pattern, but even clearer, is found when analysing the situation one year later, at T3 (see Table A.3 referring only to the 1999-cohort). The effects of being without work at T1 only, or at T2 only, are more or less the same among ethnic Norwegians and those of non-western origin. However, the effect of being without work at both T1 and T2 is larger among the ethnic Norwegians than among the immigrants. Figure 3 illustrates some of the results of Table A.3. (Figure 3 about here) According to the estimates depicted in Figure 3, ethnic Norwegians who were employed at both T1 and T2 have a larger probability of being employed at T3 (85.4 per cent) than equally qualified non-western immigrants (79.5 per cent), and a lower risk of being without work (4.2 versus 7 per cent). But, among those who were without work at both T1 and T2, the situation at T3 is clearly in disfavour of ethnic Norwegians: 46.8 per cent of non-western immigrants are employed compared to 32.4 per cent of ethnic Norwegians. Further, 29.2 among the non- Western immigrants are without work, against 47.1 per cent of the ethnic Norwegians. This indicates that contrary to the expectation in Hypothesis 5 ethnic Norwegians are hit more by state dependence than the non-western immigrants. This result is probably due to selection effects that the control variables have not captured: the ethnic Norwegians with very little work experience are, probably, a socially rather marginalized group. CONCLUSIONS Support is found for Hypothesis 1: apprenticeship experience increases the probability of employment. The effect is, however, small after controlling for qualifications. The results give some, but only weak, support to a supposition of the superiority of apprenticeship system concerning transition from school to work. Further, apprenticeship experience does not seem to compensate (much) for poor grades. Referring to Hypothesis 2, it is also found that unequal access to apprenticeship constitutes a part of the source for the differences between minority and majority youth with regard to the probability of employment after leaving education and training. But contrary to our expectations in Hypothesis 2, this does not seem to be a major cause. Further, also opposite to what was expected in Hypothesis 3, we did not find that apprenticeship is of particular importance for the employment opportunities among non-western immigrants relative to the ethnic Norwegians. In Hypothesis 4 it was expected that high rate of being without work at T1 among the non- Western immigrants would be a major cause of this group s higher rate of being without work at T2, as an effect of state dependence. This expectation was not confirmed in the analyses. When controlling for the situation at T1, male immigrants still had a higher risk of being without work at T2 than equally qualified ethnic Norwegian males. 14

Very strong effects of state dependence in the youth data set as a whole were found, but these effects were equally large for immigrants and ethnic Norwegians when referring to the effects of the status at T1 on T2. Further, these effects were larger for ethnic Norwegian than for immigrants when referring to the effects on T3 of being without work at both T1 and T2. This is contrary to our expectations in Hypothesis 5. It may be caused by selection effects that our control variables have not captured. Possibly, ethnic Norwegians with low competence level and very little employment experiences represent a negatively selected group, to a higher extent than a corresponding group of non-western immigrants with different types of social risk factors such as social/health related problems, long term experience with different kinds of measures and benefits, with the result of being caught in the welfare benefit trap. Overall, the labour market disadvantage of ethnic minority youth relative to the majority is largely due to the fact that the ethnic majority generally has better grades and attained a higher competence level. This does not explain why there still are differences when controlling for qualifications. However, we did not find that non-europeans had higher risk of being without work than East-European immigrants thus the findings do not underpin the assumption that visible minorities from Asia and Africa, for example, are more exposed to discrimination than East-European immigrants. These findings provide new insight concerning the literature on youth, immigrants and the labour market in three ways: First, state dependence is an important factor behind youth unemployment, but it does not explain immigrant youth s higher unemployment level. This refers to youth with a low or medium-low qualification level. Second, apprenticeships protect against unemployment, but mainly through the effect of good grades among the apprenticeship-seekers who obtain an apprenticeship place. Third, although immigrant youth are at a disadvantage with regard to obtaining an apprenticeship place, this does not explain their higher rate of unemployment. The policy implications of the findings seem to be of two kinds. First, in the group of youth analysed here that is youth who have not proceeded to higher education many have not completed upper secondary education, and this applies to immigrant youth in particular. Low formal qualifications predict a high probability of being without work. It is a great challenge to increase the completion rates on the vocational courses in upper secondary education and among immigrant youth in particular. Secondly, the effect of state dependence, which was particularly seen among the ethnic Norwegians, indicates that measures and welfare benefits for youth outside school and work should be better targeted in order to increase their competence and provide relevant work experience. References Ai, Chunrong and Norton, Edward C. (2003) Interaction Terms in Logistic and Probit Models, Economics Letters 80: 123 29. Andress, Hans-Jürgen (1989) Recurrent unemployment The West German experience: An exploratory analysis using count data models with panel data, European Sociological Review 5(3): 275 97. Arrow, Kenneth J. (1973) Higher education as a filter, Journal of Public Economics 2(3): 193 216. Battu, Harminder and Sloane, Peter J. (2004) Over-Education and Ethnic Minorities in Britain, Manchester School 72(4): 535 59. 15