Early career trajectories of first- and second-generation migrant graduates of professional university

ABSTRACT

This study explores how the careers of professional university graduates with a first- or second-generation non-Western migrant background evolve within the first four to eight years after graduation, as compared to their Dutch native peers. We find that in the first year after graduation, while holding constant background characteristics, both first- and second-generation migrants experienced lower employment chances, more skill mismatches and lower job satisfaction than natives. No wage differences could be observed between migrants and natives. Although the differences between first- and second-generation migrants appeared to be small in the short term, a follow-up survey four to eight years after graduation revealed evidence that second-generation – but not first-generation – migrants improved their situation overall. They maintained earnings parity with natives, and narrowed the gap in terms of job satisfaction, skill match, and to a somewhat lesser extent employment chances. Although the first generation eliminated the gap with respect to natives in terms of employment chances, they continued to show lower job satisfaction and opened up a wage gap as compared to their native Dutch peers. This suggests that early-career gaps for first-generation migrants are enduring and cannot easily be resolved.

KEYWORDS:

• Higher education
• Professional university graduates
• early career trajectories
• first- and second-generation migrants
• longitudinal analysis

1. Introduction

Over the last decades, the share of non-Western migrants in Europe has risen sharply through different waves of migration from former colonies, refugees fleeing warzones, and the permanent settlement of guest workers (OECD 2018). Several studies show that the children of these migrants perform worse in school than comparable students without migration backgrounds (Belfi et al. 2014; Gabrielli and Impicciatore 2022; Levels and Dronkers 2008; Triventi, Vlach and Pini 2022). In this paper, we investigate how migrants fare once they enter the labour market, looking at the case of Dutch professional university graduates.

While initially, non-Western migrants in Europe typically found work in manual and other types of low-skilled jobs, currently many have higher education (HE) degrees, and pursue a middle-class career (Brekke 2007; Zwysen and Longhi 2018). However, while the labour market trajectories of lower educated migrant school leavers are well-studied, relatively little is still known about the labour market trajectories of migrants who successfully managed to complete higher education in the host country. To our knowledge, only one study to date has looked into the development of early career trajectories of HE graduates with a non-Western migrant background and compared these with their non-migrant counterparts (Zwysen and Longhi 2018). This study investigated employment and wage differences in the early career of university graduates with a non-Western migrant background in the UK, between six months and three and a half years after graduation. While this study found large ethnic differences in employment in both short-and long-term, it found almost no ethnic wage differences.

It is an open question whether the labour market situation of higher educated non-Western ethnic migrants found in the UK resemble those of higher educated migrants in the Netherlands and other European countries. A country’s migration history, its education system, and its organisation of the labour market all affect whether and how ethnic penalties arise in the labour market arise (Van Tubergen and Kalmijn 2005). The Netherlands differs on a number of these characteristics from the UK. While the UK’s migrant groups mainly consist of Indian, Pakistani, and Bangladeshi migrants, the three largest migrant groups in the Netherlands are of Surinamese, Moroccan, and Turkish descent (Connor et al. 2004; Blommaert, van Tubergen, and Coenders 2012). These origin differences matter for labour market success (Van Tubergen and Kalmijn 2005). Furthermore, while research has established that migrant students in the UK typically tend to enrol in lower quality institutions (Connor et al. 2004), the quality of HE institutions in the Netherlands has been found to be very homogeneous (Klumpp, de Boer, and Vossensteyn 2014). There is however a strong tradition of vocationally oriented professional higher education in the Netherlands. There is no comparable tradition in the UK, so Zwysen and Longhi (2018) focussed on graduates of academic universities. By contrast, in the Netherlands, graduates of professional universities currently comprise almost two-thirds of the HE graduates entering the labour market each year (Allen and Belfi 2020).

The present study consequently focuses on investigating the early labour market trajectories of non-Western migrant graduates of Dutch professional universities. Aside from supplementing Zwysen and Longhi’s (2018) research, this study contributes to the broader literature on ethnic-related differences in graduate labour market careers, in at least two ways. Firstly, in order to get a more balanced view of ethnic differences in early career development, in this paper we not only focus on objective indicators of labour market success, such as employment and wages, but also on subjective success indicators, such as job satisfaction and subjective job-skill match. This is because prior research has shown that both types of career success contribute to people’s overall wellbeing (Verbruggen et al. 2015).

Secondly, this study aims to identify potential education-related barriers faced by non-Western migrant graduates that may explain potential ethnic-related labour market gaps. More specifically, this study investigates to which extent background differences with regard to human capital, parental socioeconomic background and field of study preferences can account for potential ethnic-related differences in the labour market success of HE graduates. To our knowledge, apart from Zwysen and Longhi (2018), this topic has been little studied.

The paper is organised as follows. In Section 2, we discuss the existing theoretical explanations for labour market differences between migrants and natives and between first- and second-generation migrants. Section 3 provides information on the specific migrant groups and higher education system in the Netherlands. In Section 4, we describe this study’s sample, variables, and design. The results of the descriptive and regression analyses are discussed in Section 5. The paper concludes with a discussion of its results in Section 6.

2. Literature review

2.1. Explanations for ethnic related labour market differences

In labour market research, human capital theory (Becker 1962) is by far the best-known theory in accounting for differences in labour market success. According to this theory, education and work experience are the most crucial factors for labour market success, as these factors increase the productivity of workers. However, research has shown that differences in education and work experience alone are often unable to fully explain ethnicity-related differences in career outcomes (Blommaert, van Tubergen, and Coenders 2012; Siebers and van Gastel 2015). The remaining unexplained variance in career outcomes between migrants and natives after education and experience have been accounted for has been referred to as an ethnic penalty (Heath and Cheung 2007).

Several modifications and alternatives to human capital theory have been suggested to account for this ethnic penalty. For example, it has been pointed out that human capital acquired in the country of origin by migrants may be less well transferable to, or rewarded differently in, destination countries. The argument usually revolves around assumptions about differing circumstances between origin and destination countries in terms of socio-economic circumstances and education systems (Dronkers, Levels, and de Heus 2014; Basilio, Bauer, and Kramer 2017). The fact that human capital acquired in origin countries can be less suited or less rewarded in destination countries has been used to explain differences between natives and migrants in terms of wages (Chiswick and Miller 2005), economic integration (Van Tubergen, Maas, and Flap 2004), and educational performance (Levels and Dronkers 2008).

As an alternative to human capital theory, signalling theory assumes that employers use observable signals to gauge job applicants’ productive potential, such as degrees and resumes (Spence 1973). Foreign credentials may be seen as less reliable or less valuable signals (Castagnone et al. 2015). More subtle differences within otherwise comparable qualifications may be expressed in signals such as grade point average (GPA) and extracurricular activities (such as managerial experience in a student association) (Kittelsen Røberg and Helland 2017). Migrant students generally acquire less of such specific forms of human capital during their studies than native students (Camacho and Fuligni 2015). This could account for part of the ethnic-related differences in the labour market during the school-to-work transition.

It has also been found that migrant students hold different preferences regarding fields of study than students without a migration background. For example, migrant students show a stronger preference for law and economics and less interest in technical and health care studies (Gerber and Cheung 2008; Jennissen and Oudhof 2007 Kittelsen Røberg and Helland 2017). Pásztor (2012) found that this difference in preference could be explained by a stronger preference of migrants for white-collar professions that are a certain entry ticket into the upper-middle class, as these present a clear contrast to their parents’ jobs, which are often characterised by hard physical labour, unattractive working hours and a minimal income. Jennissen and Oudhof (2007) found that migrant students’ study choices are relatively strongly driven by factors such as job security and prestige. However, in the Netherlands, such jobs are less in demand, as the labour market for school-leavers has a higher demand for highly educated personnel in jobs that involve physical tasks and unattractive working hours, such as technical and health care jobs, than in jobs in business and sales (Bakens, Fouarge, and Peeters 2018). As such, differences in field of study preferences may also account for less successful careers of migrant HE graduates.

Another frequently cited explanation for ethnic-related labour market problems has to do with the often lower parental socioeconomic status (SES) of migrants which would imply lower social and cultural capital and less information about the education system (Gabrielli, Longobardi, and Strozza 2022; Gabrielli and Impicciatore 2022). However, research has shown that effects of SES decrease as students grow older. For example, a recent literature review found a very weak relationship between parental SES and academic achievement among HE students (Rodriguez-Hernandez, Cascallar, and Kyndt 2020). As students age, other factors (such as previous academic achievement and extracurricular experience) become more important than SES. For example, Zwysen and Longhi (2018) who studied employment and earning differences in the early career of ethnic minority British graduates, found no evidence that SES played an explanatory role in these relationships.

A final explanation for ethnic career-differences is taste-based or statistical discrimination by employers. Employers may discriminate against migrants for various reasons, including a manifest or perceived dislike of migrants, or a perception of lower potential productivity (Becker 1962). Field experimental evidence strongly suggests that discrimination against ethnic minorities also plays a role in the Dutch labour market (Ramos, Thijssen, and Coenders 2019).

2.2. Explanations for labour market differences between first- and second-generation migrants

Research on career trajectories has shown that first- and second-generation migrants differ in many respects. For example, several studies have shown that first-generation migrants demonstrate considerable skill shortages in terms of literacy and numeracy in comparison to second-generation migrants, and accordingly perform worse in the labour market compared to native peers with the same education (Chiswick and Miller 1999; Levels and Dronkers 2008). Differences have also been found in country-specific knowledge such as cultural norms and customs, and this has also been found to hinder first-generation migrants’ careers (Crul and Vermeulen 2003). Furthermore, first-generation migrants’ sometimes involuntary departure from their country of origin and possible traumatic events they experienced there have also been found to disadvantage them when entering the labour market in comparison to their second-generation counterparts (Bakker, Dagevos, and Engbersen 2017).

Finally, the literature on transnationalism (cf. Basch, Glick, and Cristina 1994) suggests that first-generation migrants may be choosing different jobs, as socio-economic circumstances and labour markets in origin countries form a stronger frame of reference.¹ By contrast, second-generation migrants may have developed frames of references that are more similar to those of students without a migration background, and therefore have higher job aspirations (Heath and Cheung 2007). These findings are consistent with those of Arcarons and Muñoz-Comet (2022), who observed that first-generation migrants often suffer from a persistent concentration in vulnerable economic sectors with a precarious hold on the labour market. This underscores the importance of distinguishing between generations when investigating possible ethnic penalties in the labour market.

2.3. Implications of unfavourable labour market beginnings for later career trajectories

Although it is increasingly clear that the school-to-work transition is a process rather than a singular event (Raffe 2008), the first job still forms a crucial element determining the trajectory from school to work, as it is through this experience that one learns to transform the formal knowledge obtained in school into job-relevant outcomes (Becker 1962). As such, the more challenging the first job is, the more human capital will be accumulated within that job. This additional human capital further increases the opportunities for a good second job. In turn, a good second job will increase chances for a good third job, and so on.

It is therefore important that initial labour market experiences are taken into account when studying possible ethnic-related differences in career trajectories (Birkelund, Heggebø, and Rogstad 2017; Zwysen and Longhi 2018). Both signalling theory (Spence 1973) and career timetable theory (Lawrence 1988) predict that problems in finding an initial job may result in a so-called scarring effect, which may form a negative signal for later employers. This is because there are certain norms regarding how an individual’s career develops over time and which achievements are appropriate given one’s career stage. The further one falls behind the ‘normal’ career timetable, the more likely one is to be viewed unfavourably by prospective future employers (Verbruggen et al. 2015).

3. The Dutch context

3.1. Ethnic minority groups in the Netherlands

In the Netherlands, the inflow of non-Western migrants, took off in the 1960s and 1970s. In 2018, the Netherlands had about 17 million inhabitants, of whom 13% had a non-Western migration background (Statistics Netherlands 2019). Non-Western ethnic migrants in the Netherlands can be divided into four main groups based on origin: (1) migrants from former Dutch colonies (Surinam and the Netherlands Antilles); (2) guest workers from Mediterranean countries (e.g. Turkey and Morocco); (3) subsequent waves of migrants from these countries for purposes of family reunification; and (4) asylum seekers from Africa and the Middle East. More recently, there has been a marked increase in the prevalence of study migration to the Netherlands, which has brought in migrants from countries from further afield, like China, (Statistics Netherlands 2021). Given these different origins, there is some diversity among non-Western migrants in the Netherlands. However, almost all non-Western migrant groups share a comparatively disadvantaged socioeconomic position (Bakker, Dagevos, and Engbersen 2017; Gabrielli and Impicciatore 2022). In this regard, non-Western migrant groups are considerably different from Western migrant populations in the Netherlands, who mostly come from neighbouring countries such as Belgium and Germany (Statistics Netherlands 2019). Of the total Dutch population, 7.2% is considered a first-generation non-Western migrant, while 5.8% is a second-generation migrant (Statistics Netherlands 2019).

3.2. The Dutch higher education system

The Netherlands has a binary system of higher education, consisting of 13 m academic universities and 41 professional universities (‘Hoger Beroepsonderwijs’ in Dutch; henceforth: HBO). In general, academic universities are relatively more theoretically oriented and the professional universities more vocationally oriented. Both types of institutions offer studies in a broad and largely overlapping range of fields.² Dutch professional universities enrol around two-thirds of all higher education students in the Netherlands (Lepori and Kyvik 2010). Due to strict national accountability measures, there is relatively little difference between institutions within these two broad types of higher education in terms of quality of teaching and research (Klumpp, de Boer, and Vossensteyn 2014). Access to both types of higher education is conditional on prior qualifications. While a degree in academic upper secondary education is required for academic university enrolment, professional university enrolment is open to all students with any form of upper secondary education degree. Since non-Western migrants are strongly underrepresented in academic upper secondary education, professional university is often their chosen path towards a higher education diploma (Jennissen and Oudhof 2007).

4. Data and methods

4.1. Sample

The present study makes use of data on graduates from professional universities, who comprise almost two-thirds of the annual outflow from Dutch HE to the labour market (Allen and Belfi 2020). Partly due to their larger student volume and stronger vocational orientation, the professional universities invest more resources into monitoring the transition from study to work than professional universities, for which data are scarcer and of lower quality. We use two partly overlapping datasets for our study. For the analysis of the initial transition from HE to the labour market, we make use of data from the HBO-Monitor, a yearly graduate survey among recent graduates of professional universities. In this survey, professional university graduates are asked about their work situation one year after graduation. Around 90% of all Dutch professional universities take part in this study with full coverage of all their recent graduates with annual response rates around 40%. These data have been analysed for indications of deviations from representativeness, and this has shown the data to be strongly representative of the population as a whole (Belfi and Huijgen 2015). Key outcomes such as employment chances and income have also been found to be highly comparable to those revealed in administrative data collected by the Dutch government (Allen et al. 2019).³ For our analyses, we use data from eleven cohorts that graduated in the academic years 2007–2017, who were surveyed in the years 2008–2018 respectively.⁴

We restricted the sample to native respondents and respondents who belonged to a non-Western migrant group according to the definition of Statistics Netherlands (2019). If at least one parent was born outside the Netherlands, an individual is defined as migrant. If this is a country in Africa, Latin America or Asia, one is defined as a non-Western migrant. There are two exceptions to this general rule. First, the whole of Turkey, including the part that lies in Europe, is classified as non-Western. Second, Indonesia and Japan are classed as Western as migrants from these countries occupy a more favourable socio-economic position than migrants from other countries in Asia⁵ (for a more elaborate explanation of this definition, see Guiraudon, Phalet, and Ter Wal 2005).

If the respondents themselves were born outside the Netherlands, they are defined as having a first-generation migrant background. If they were born in the Netherlands to at least one non-Western parent, they are defined as having a second-generation migrant background. To avoid confounding the results with outcomes that are likely to be strongly influenced by differences in the life-cycle phases of graduates, we further restricted our sample to graduates who were younger than 30 when they first entered the labour market. After applying these selections, the final sample for the analyses comprised 110,635 graduates, including 3,264 first-generation non-Western migrants and 5781 second-generation non-Western migrants.

Finally, for the analysis of the graduates’ subsequent career development, we focussed on a subsample drawn from the four cohorts that graduated in 2007, 2009, 2010 and 2011, and that were first approached for participation one year after graduation, in 2008, 2010, 2011 and 2012 respectively. A follow up of all four cohorts was held in 2015, when the graduates had already been in the labour market for four to eight years. In total, 10,434 (14.5%) of the graduates who participated in the first survey could be uniquely identified in the second survey. After further restrictions in terms of age and availability of relevant labour market data were taken into account, the final sample for the career analyses comprised 5,984 graduates, including 114 first-generation non-Western migrants and198 second-generation non-Western migrants.

4.2. Variables

4.2.1. Outcome variables

Four labour market outcomes were studied: two objective and two subjective outcomes. All outcomes were measured both one year after graduation and (depending on the cohort) four to eight years after graduation. The two objective outcomes were employment status and hourly earnings. The subjective outcomes were job-skill match and job satisfaction.

Employment status: this variable was specified as a dummy. Graduates who at the time of the survey were in paid employment (including self-employment) for at least one hour per week were defined as employed.⁶ Graduates who were not in paid employment but who were actively seeking employment were defined as unemployed.

Hourly earnings: the hourly earnings of graduates were derived from reported monthly earnings in combination with hours worked. To eliminate possible estimation bias due to implausible outliers, the top and bottom 1% of the earnings distribution were excluded from the analyses. We took the natural logarithm of the remaining income measure.

Job-skill match: this outcome consisted of two separate measures of job-skill match: skill utilisation and skill shortage. First, respondents were asked to report the extent to which they felt that their skills were insufficient for their current job (skill shortage). Second, respondents were asked to report the extent to which they used their skills in their current job (skill utilisation). Both questions were rated on an ordinal five-point scale ranging from 1 (not at all) to 5 (to a high extent).

Job satisfaction: graduates were asked to report their satisfaction with their current job on an ordinal five-point scale, ranging from 1 (very dissatisfied) to 5 (very satisfied).

4.2.2. Control variables

In all analyses, the following variables were controlled for: gender (1=woman; 0=man); age (in years, linear and quadratic); cohort dummies for graduation cohorts 2008 through 2017 (with as reference year 2007); study province (reference province South Holland, with dummies for each of the 11 other Dutch provinces); living abroad at time of survey (1=yes; 0=no); highest prior education (dummies for pre-university secondary education, secondary vocational education, previous college programme or other programme, with as reference higher general secondary education), and further education directly after graduation (1=yes; 0=no).

4.2.3. Explanatory variables

To assess the explanatory role of field of study, parental SES, and human capital accumulation, the following variables were included in the analyses:

Field of study: agriculture, education, technical studies, health studies and social studies; reference category: economics.

Parental SES: operationalised as the highest level of education attained by at least one parent; dummies for high and medium SES (reference category low SES). The original question distinguished six levels: primary education and lower secondary education (classed as low SES), upper secondary education and intermediate vocational education (classed as medium SES) and professional university education and academic university education (classed as high SES). This variable was only available in the data for the second wave.

Additional human capital: expressed by GPA⁷ and extra-curricular experiences during study, operationalised by dummies indicating relevant work experience over and above the compulsory internships and administrative experience in student associations, sporting clubs and the like.

4.3. Study design

As mentioned, our analyses focus on graduates of professional universities, and do not relate to HE graduates who attended an academic university. This could potentially raise issues of selectivity. There is evidence suggesting that Dutch non-Western migrants prefer such professional universities to academic universities, which seems to be related to the stricter admission conditions for academic universities, and to a preference on the part of non-Western migrants for more applied fields of study (Pásztor 2012). To account for possible effects of selection into professional universities vs. academic universities, a Heckman Correction Model (HCM) has been applied (Heckman 1979). We estimate the HCM in two stages (cf. Wooldridge 2009). The first stage uses nationally representative school leaver data (Meng 2020) to predict the likelihood of a student to enrol in professional vs. academic university and calculate the predicted inverse Mills ratio. The variables used for this estimation were first and second-generation migration background, gender, HE enrolment age, HE enrolment year, and dummies indicating the type and level of prior education. The second stage corrects for possible the decision of enrolment by including the inverse Mills ratio (λ) as a predictor in all models. Subsequently, depending on the measurement of the outcome variables linear and (ordered) logistic regression analyses – including the inverse Mills ratio – were used to assess whether ethnic-related differences in the short-term career trajectories (one year after graduation) were present with regard to employment, income, skill shortage, skill utilisation and job satisfaction, for the total sample of 110,635 graduates.⁸

Three regression models per outcome were estimated in which native graduates served as the reference category, and differential effects for first and second-generation non-Western migrant graduates were estimated by including two additional dummies representing each group. The first of the three models was a baseline model, estimating the relation between migration status and the outcome in the short-term with no controls. In the second model, the standard control variables described in Section 4.2.2 are added to the baseline model. These control variables could potentially distort the estimated associations, but play no plausible direct role in shaping outcomes of migrants compared to those of natives. The third model was an explanatory model for the short-term outcome in question, in which three explanatory variables (mediators) were added: (1) field of study, (2) GPA, and (3) extra-curricular experiences. If the labour market careers of migrant graduates are (partly) mediated by these explanatory variables, inclusion of these variables will reduce the estimated associations.

For a limited number of cohorts and graduates we had data at our disposal collected several years after the initial graduation. This has been done for a subset of graduates from the cohorts 2007, 2009, 2010 and 2011, which were all surveyed in 2015. Since it is likely that there is some selectivity involved in determining which graduates surveyed one year after graduation were available for the follow-up survey four to eight years later, a second Heckman Correction Model (HCM) has been applied in addition to that described above, to account for selectivity in response to the follow-up survey. For the first stage, we used migration background, broad field of study, response mode for the initial survey (paper or internet), employment status, job search and training participation one year after graduation, to predict the odds that one is represented in the follow-up survey and calculate the predicted inverse Mills ratio. Subsequently, in the second stage we corrected for possible selectivity in the decision to respond by including the inverse Mills ratio (λ) as a predictor in all models used to estimate the effect of migration background on outcomes in the medium term.

To examine differences by migration background in medium-term outcomes, four models per outcome were estimated for the smaller, selective sample of graduates who completed both surveys at both time-points. The first model was equivalent to that estimated for short-term outcomes, including no controls. The second model used a subset of the standard controls used for estimating short-term outcomes, namely gender, age, graduation cohort, prior education, study region and living abroad at time of survey.⁹ To examine to what extent the medium-term outcomes were due to the possible scarring effects of the initial transition, in the third model a dynamic regression analysis was added (Honoré and Tamar 2006), in which we controlled for the corresponding outcome in the short-term. Finally, in the fourth model four mediators were added: (1) field of study, (2) parental SES, (3) GPA, and (4) extra-curricular experiences to estimate to which extent any residual ethnic-related medium-term career differences could be explained by these factors.

5. Results

5.1. Descriptive analyses

Table 1 shows descriptive evidence for the main variables of analysis by migration background one year after graduation. This table already reveals some quite marked differences between migrant and native graduates. In terms of objective labour market outcomes, the most striking difference is in the higher risk of unemployment among first- and second-generation migrants. The differences in unemployment risk between first and second-generation migrants is quite small compared to the gap with natives. Once graduates have obtained a job, there is little difference between native and migrant graduates in terms of earnings. Migrant graduates – particularly first-generation migrants – are however more likely to experience a skill shortage, less likely to utilise their skills, and are less likely to be satisfied with their current job. It is important to bear in mind that these differences in labour market outcomes could be due to differences in other characteristics of graduates, which as Table 1 reveals, are quite numerous.

Table 1. Descriptive statistics by migration background.

	Native Dutch	First-generation minority	Second-generation minority
Labour market 1 year after graduation
Unemployment	4.8%	13.0%	11.1%
Hourly wage (mean Euros)	€14.43	€14.99	€14.61
Skill shortage	6.2%	9.9%	8.8%
Skill utilisation	64.4%	54.5%	53.8%
Job satisfaction	67.9%	58.6%	57.5%
Mediators
Field of study
Agriculture	3.4%	1.4%	0.6%
Education	13.7%	5.0%	8.1%
Technical	18.2%	19.4%	13.3%
Economics	36.7%	52.6%	52.0%
Health care	13.2%	7.8%	6.9%
Social studies	14.8%	13.8%	19.0%
GPA relative to class mates	0.01	−0.06	−0.08
Extracurricular experiences
Relevant work	47.7%	46.2%	44.3%
Administrative experience	16.2%	18.7%	14.1%
Control variables
Gender: female	56.7%	56.1%	60.3%
Living abroad at time of survey	2.0%	8.7%	2.0%
Age at graduation (mean years)	23.7	26.0	24.7
Graduation year
2007	8.7%	9.8%	7.3%
2008	8.0%	8.0%	5.9%
2009	8.1%	8.8%	6.6%
2010	8.9%	9.4%	7.9%
2011	8.6%	7.7%	7.9%
2012	8.9%	8.0%	9.4%
2013	9.4%	8.6%	9.4%
2014	9.0%	9.7%	9.6%
2015	10.1%	9.2%	11.4%
2016	10.4%	9.7%	11.9%
2017	10.0%	11.1%	12.7%
Study region
North	11.8%	5.9%	3.8%
East	23.7%	14.4%	13.3%
West	39.3%	66.8%	69.3%
South	25.2%	13.0%	13.6%
Prior education
Senior general secondary education (HAVO)	53.8%	37.1%	49.6%
Pre-university education (VWO)	15.1%	11.7%	7.4%
Secondary vocational education (MBO)	26.5%	34.7%	38.3%
Previous college degree	3.0%	7.6%	2.6%
Other	1.5%	9.0%	2.1%
Further education	18.2%	18.6%	17.3%
N	101,590	3264	5781

Regarding the mediating variables, both first- and second-generation migrants show a preference for economics programmes, and the second-generation shows a preference for social studies programmes. The grades of both natives and migrant students were very close to average. Native and migrant graduates show similar extra-curricular experiences, although the second generation gained somewhat less administrative experience. Finally, as expected, Table 1 also shows small differences between natives and migrants on our control variables, with most notably migrant graduates being slightly older at time of graduation and more frequently living in the West of the Netherlands.

5.2. Multivariate analyses

5.2.1. Short-term analyses

Employment status. Table 2 shows the relationship between migration background and graduates’ employment status one year after graduation. Model 1 shows the marginal effects – the estimated percentage-point difference between graduates with a first- and second-generation non-Western migration background compared to native Dutch graduates. Both first-generation migrants (−5.5%, p < .001) and second-generation migrants (−4.6%, p < .001) are less likely to be employed one year after graduation than their native Dutch peers. Model 2 shows that the difference between native and first- and second-generation graduates is reduced by respectively 2.2 and 1.3 p.p. (percentage points) after controlling for the control variables.

Table 2. Logistic regressions on employment status (marginal effects).

	Model 1			Model 2			Model 3
	Estimate		S.E.	Estimate		S.E.	Estimate		S.E.
Migration status (ref. native Dutch)
1st generation non-western	−5.5%	***	0.3%	−3.3%	***	0.4%	−3.1%	***	0.4%
2nd generation non-western	−4.6%	***	0.2%	−3.3%	***	0.8%	−3.2%	***	0.4%
Standard controls^a	no			yes			yes
Mediators
Field of study (ref. economics)
Agriculture							−1.1%	**	0.4%
Education							2.7%	***	0.2%
Technical studies							1.2%	***	0.2%
Health studies							4.8%	***	0.3%
Social studies							1.1%	***	0.2%
GPA							1.0%	***	0.1%
Extracurricular experience
Relevant work experience							1.3%	***	0.1%
Administrative experience							0.9%	***	0.2%
Mills ratio: lambda	−0.1%	*	0.1%	2.6%	***	0.8%	1.6%	*	0.7%

Notes: ^aThe standard control variables are gender, age (linear and quadratic), graduation cohort, study province, living abroad at time of survey, prior education and further education.

*p < 0.05; **p < 0.01; ***p < 0.001.

^c N = 110,635.

Finally model 3 shows that all added mediators have significant associations with graduates’ employment prospects. Graduates of health studies programmes have particularly good employment prospects, having 4.8% (p < .001) more chance of being employed than graduates in the reference field of economics. High grades significantly enhance one’s prospects of finding work (+1.0%, p < .001) as do relevant work experience (+1.3%, p < .001) and administrative experience gained while studying (+0.9%, p < .001). Adding these mediators further explains 0.2 p.p. of the association for the first generation, and 0.1 p.p. of the impact on the estimates for the second generation.

Hourly earnings. Once graduates with a migration background obtain employment, they do not appear to suffer any penalty in terms of their earnings compared to native Dutch graduates (see Table 3). In fact, prior to the controls (Model 1) first-generation as well as second-generation migrants, have a substantial earnings advantage of respectively 2.8% (p < .001) and 0.8% (p < .05) relative to their native Dutch peers. After accounting for the covariates in Model 2 the earnings advantage disappears entirely.

Table 3. Regressions on ln(hourly earnings).

	Model 1			Model 2		Model 3
	Estimate		S.E.	Estimate	S.E.	Estimate		S.E.
Migration status (ref. native Dutch)
1st generation non-western	2.8%	***	0.5%	−0.8%	0.6%	−1.0%		0.6%
2nd generation non-western	0.8%	*	0.4%	−0.6%	0.5%	−0.7%		0.5%
Standard controls^a	no			yes		yes
Mediators
Field of study (ref. economics)
Agriculture						−1.2%	*	0.5%
Education						3.3%	***	0.3%
Technical studies						4.2%	***	0.2%
Health studies						13.1%	***	0.3%
Social studies						3.2%	***	0.2%
GPA						3.0%	***	0.1%
Extracurricular experience
Relevant work experience						1.8%	***	0.2%
Administrative experience						1.2%	***	0.2%
Mills ratio: lambda	−0.9%	***	0.1%	−0.2%	0.9%	−3.6%	***	0.9%

Notes:

^a The standard control variables are gender, age (linear and quadratic), graduation cohort, study province, living abroad at time of survey, prior education and further education.

*p < 0.05; **p < 0.01; ***p < 0.001.

^c N = 91,430.

Job-skills match. As outlined above, we use two measures to assess the match between the skills possessed and those required for the job, namely skill shortage and skill utilisation.

Skill shortage. Table 4 shows the association between migration background and the extent to which graduates report that they lack the skills required to do their job. Since the outcome is measured on a five-point scale – which can be presumed to be ordinal but with scale points that cannot necessarily be assumed to be equally spaced – we use ordered logit models to estimate these associations. The estimates in these models represent the change in the log odds of being in a higher category on the dependent variable (in this case skill shortage) that results from a one point increase in the predictor. In the case of first- and second-generation non-western migration status, the coefficient represents the change in log odds compared to natives. Model 1 shows that both first-generation (b = 0.192, p < .001) and second-generation migrants (b = 0.066, p < .05) are significantly more likely to report that they have a skill shortage than their native Dutch peers. After adding the control variables in Model 2, the log odds of being in a higher skill shortage category actually increase by 0.097, for first-generation migrants and 0.91 for second-generation migrants. This is mainly attributable to differences in age and prior education. Migrants are older on average, and more likely than natives to have entered HE via secondary vocational education. Both these attributes are associated with less skill shortages, so after controlling for them the effects of migration background become larger. The log odds are further increased after adding the mediators added in Model 3. More specifically, the log odds of being in a higher skill shortage category increase by a further 0.003 for first-generation migrants by 0.028 for second-generation migrants, indicating that differences in choice of field of study, lower GPAs, and less relevant work experience further increase the skill shortages of migrants as compared to those of natives.

Table 4. Ordered logistic regressions on skill shortage.

	Model 1			Model 2			Model 3
	Estimate		S.E.	Estimate		S.E.	Estimate		S.E.
Migration status (ref. native Dutch)
1st generation non-western	0.192	***	0.035	0.289	***	0.042	0.292	***	0.042
2nd generation non-western	0.066	*	0.026	0.157	***	0.037	0.185	***	0.037
Standard controls^a	no			yes			yes
Mediators
Field of study (ref. economics)
Agriculture							0.348	***	0.038
Education							−0.023		0.019
Technical studies							0.252	***	0.017
Health studies							0.125	***	0.019
Social studies							−0.120	***	0.018
GPA							−0.104	***	0.011
Extracurricular experience
Relevant work experience							−0.072	***	0.012
Administrative experience							−0.010		0.016
Mills ratio: lambda	0.057	***	0.005	0.247	***	0.065	0.286	***	0.065

Notes:

^a The standard control variables are gender, age (linear and quadratic), graduation cohort, study province, living abroad at time of survey, prior education and further education.

*p < 0.05; **p < 0.01; ***p < 0.001.

^c N = 104,101.

Skill utilisation. The other side of job-skills match is skill utilisation. To some extent, this may be viewed as the mirror image of skill shortages, in the sense that if one experiences a skill shortage one is presumably more likely to utilise the skills that one does possess. However, this only applies to those graduates working in a domain that is substantively related to one’s field of study. For example, a trained nurse working as a nursing assistant may fail to fully utilise his/her skills, but will probably not experience any major skill shortages. The same trained nurse working as a hotel manager may however simultaneously experience a skill shortage and fail to utilise his/her existing skills.

The importance of distinguishing these two aspects of skill match becomes clear when we look at Table 5, which shows the associations between migration background and skill utilisation. Despite experiencing a higher level of skill shortages compared to native graduates, migrant graduates are less likely to use the skills that they possess at work. This problem appears to be a bit more severe among the second-generation (b = −0.397, p < .001) than among the first-generation (b = −0.339, p < .001).

Table 5. Ordered logistic regressions on skill utilisation.

	Model 1			Model 2			Model 3
	Estimate		S.E.	Estimate		S.E.	Estimate		S.E.
Migration status (ref. native Dutch)
1st generation non-western	−0.339	***	0.034	−0.180	***	0.042	−0.074		0.042
2nd generation non-western	−0.397	***	0.026	−0.247	***	0.037	−0.176	***	0.037
Standard controls^a	no			yes			yes
Mediators
Field of study (ref. economics)
Agriculture							0.082	*	0.038
Education							1.228	***	0.020
Technical studies							0.395	***	0.017
Health studies							1.001	***	0.020
Social studies							0.147	***	0.018
GPA							0.251	***	0.011
Extracurricular experience
Relevant work experience							0.196	***	0.012
Administrative experience							0.104	***	0.016
Mills ratio: lambda	−0.060	***	0.005	0.237	***	0.064	0.044		0.065

Notes:

^a The standard control variables are gender, age (linear and quadratic), graduation cohort, study province, living abroad at time of survey, prior education and further education.

*p < 0.05; **p < 0.01; ***p < 0.001.

^c N = 104,243.

Adding the standard controls in Model 2 reduces the gap with respect to natives in the log odds of being in a higher skill utilisation category by 0.159 for first-generation migrants and by 0.150 for second-generation migrants. Additional analyses reveal that this is mainly due to age differences. The addition of the mediators in Model 3 further reduces the gap in log odds further by respectively 0.106 and 0.073. The disadvantage compared to natives remains statistically significant for the second-generation, but is no longer significant for first-generation migrants. This change is mainly attributable to the field of study choice of those with a migration background, with their preferred field of economics showing the lowest levels of utilisation.

Job satisfaction. Table 6 shows that both first- and second-generation migrants are substantially less likely to be satisfied with their current work than natives are. This negative association is somewhat stronger for the second-generation (b = −0.390, p < .001) than the first-generation (b = −0.360, p < .001), as was the case for skill utilisation. The addition of the standard controls reduces the gap with respect to natives in the log odds of moving to a higher job satisfaction category by 0.107 for first-generation migrants and by 0.073 for second-generation migrants (see Model 2). Once again, this is mainly due to the higher age of graduates with a migration background. Model 3 finally shows that the mediators explain an additional portion, namely 0.034 for first-generation migrants and 0.39 for second-generation migrants. Once again, study choice appears to be the main explanation here.

Table 6. Ordered logistic regressions on job satisfaction.

	Model 1			Model 2			Model 3
	Estimates		S.E.	Estimates		S.E.	Estimates		S.E.
Migration status (ref. native Dutch)
1st generation non-western	−0.360	***	0.034	−0.253	***	0.041	−0.219	***	0.042
2nd generation non-western	−0.390	***	0.026	−0.317	***	0.037	−0.278	***	0.037
Standard controls^a	no			yes			yes
Mediators
Field of study (ref. economics)
Agriculture							0.225	***	0.038
Education							0.435	***	0.019
Technical studies							0.351	***	0.017
Health studies							0.534	***	0.019
Social studies							−0.055	**	0.018
GPA							0.146	***	0.011
Extracurricular experience
Relevant work experience							0.126	***	0.012
Administrative experience							0.086	***	0.016
Mills ratio: lambda	−0.051	***	0.005	0.040		0.064	−0.067		0.064

Notes:

^a The standard control variables are gender, age (linear and quadratic), graduation cohort, study province, living abroad at time of survey, prior education and further education.

*p < 0.05; **p < 0.01; ***p < 0.001.

^c N = 104,424.

5.2.2 Medium-term analyses

Table 7 summarises the results of the analyses conducted on labour market outcomes in the medium term. For brevity’s sake, we only show the estimated associations between migration background and outcomes (see Appendix IV for the full results).

Table 7. Effects of migration status on outcomes in the medium term.

	No controls			Standard controls			Control for short-term			Control for mediators
	Estimate		S.E.	Estimate		S.E.	Estimate		S.E.	Estimate		S.E.
Odds of being employed
1st generation non-western	−1.7%		1.2%	0.0%		1.3%	−0.1%		1.3%	−0.4%		1.6%
2nd generation non-western	−2.7%	**	0.8%	−2.1%	*	0.9%	−2.0%	*	0.9%	−2.3%	*	1.4%
Hourly earnings
1st generation non-western	−7.6%	**	2.8%	−6.4%	*	2.7%	−7.3%	**	2.6%	−5.2%	*	2.6%
2nd generation non-western	−1.5%		2.1%	0.5%		2.1%	−1.0%		2.0%	1.1%		2.2%
Job satisfaction
1st generation non-western	−0.484	*	0.197	−0.367		0.203	−0.360		0.203	−0.332		0.207
2nd generation non-western	−0.212		0.150	−0.149		0.157	−0.129		0.158	−0.092		0.159
Skill shortage
1st generation non-western	0.324		0.205	0.244		0.211	0.229		0.211	0.320		0.215
2nd generation non-western	0.178		0.155	0.140		0.162	0.160		0.162	0.235		0.164
Skill utilisation
1st generation non-western	−0.359		0.204	−0.181		0.210	−0.193		0.211	−0.176		0.215
2nd generation non-western	−0.189		0.156	−0.068		0.163	−0.068		0.163	−0.055		0.165

Notes:

^a Controls for field of study, GPA, experiences during study, SES, gender, age (linear and quadratic), graduation cohort, study province, living abroad at time of survey, prior education and further education.

*p < 0.05; **p < 0.01; ***p < 0.001.

^c N = 5984.

When we compare the results in Table 7 with those in Tables 2–6, we see that there are fewer differences between migrants natives in the medium-term outcomes than in the short-term outcomes. First-generation migrants are no longer significantly less likely than natives to be in paid employment, and there are almost no significant differences between migrants and natives in terms of skill match and job satisfaction in the medium term. In terms of wages however, we now observe a significant penalty of 7.6% for first generation migrants compared to their native Dutch peers. Part of this effect (1.2 p.p.) is accounted for by the standard controls, but when we control for earnings in the short term, most of this (0.9 p.p.) is added back on. When we recall that this group had a significant earnings advantage in the short term (before controls), this finding makes sense. Adding the mediators to the model subsequently accounts for 2.1p.p., but a significant earnings penalty of over 5% remains unaccounted for.

We need to be cautious when interpreting the results in terms of the other outcomes. The far lower number of observations gives these analyses much less statistical power than those on short-term outcomes, so an absence of significance should not be too hastily interpreted as the absence of an effect. When we look at the size of the coefficients, the only outcome on which we see a clear reduction in effect size is that of employment chances. This also applies to the second-generation: although their employment chances relative to natives is still statistically significant, this effect is weaker both in terms of significance and effect size. Whereas in the short term second-generation migrants are 3.2 p.p. less likely than their native peers to be in paid employment after account for standard controls and mediators, in the medium term the gap has narrowed to 2.3 p.p.

In terms of subjective outcomes, although we mostly no longer see significant effects, the effect sizes are actually quite similar to those seen in the short-term analyses, especially for first-generation migrants. It may therefore be premature to conclude that this group has narrowed the gap in terms of skills match and job satisfaction compared to native Dutch graduates. We do see indications that second-generation migrants have not only retained earnings parity with their native peers, but also have narrowed the gap in terms of skill utilisation and job satisfaction.

6. Discussion

This study assessed and compared the early career trajectories of professional university graduates with a first- or second-generation non-Western migrant background with those of natives in the Netherlands. To get a complete picture of graduates’ initial work experiences, two objective and two subjective labour market outcomes were investigated one year and four to eight years after graduation. Holding constant for a large number of relevant background characteristics, we examined whether the associations could be partially or completely accounted by several mediators, such as GPA, extracurricular experience, parental social background and initial work experience.

Regarding the short-term findings, our results closely mirror those of Zwysen and Longhi (2018) in the UK. Like them, we find that migrants are more often unemployed, but have no earnings disadvantage as compared to natives. In the light of the noted differences between the Netherlands and the UK in terms of the higher education system and the composition of the migrant population, this correspondence is striking.

Unlike Zwysen and Longhi (2018), we looked at both first- and second-generation migrants, and examined not only objective labour market outcomes but also subjective indicators of labour market success. In both areas, this yielded interesting insights. Second- generation migrants did not perform systematically better in the labour market than their first-generation peers. This runs counter to both conventional wisdom and other research (Muñoz-Comet and Arcarons 2022; Van Tubergen and Kalmijn 2005), which suggests that the second-generation's greater acquaintance with the host country’s labour market and culture should give them an advantage compared to the relative newcomers. Muñoz-Comet and Arcarons (2022) found in Spanish context that while the later careers of second-generation migrants are not very different from those of non-migrants, first-generation migrants performed much worse in the labour market than natives. However, research has shown that there are large differences in the transition from education to the labour market between Southern and Northern Europe (Borghans and Golsteyn 2012). It is therefore not obvious that the findings by Muñoz-Comet and Arcarons (2022) should be expected in a country like the Netherlands, which has a labour market and education system that is more similar to those in other Northern European countries such as Germany, Belgium and Austria (Borghans and Golsteyn 2012).

With respect to short-term subjective labour market outcomes, we found that migrants experience significantly higher levels of skill mismatch and lower job satisfaction than their native peers. Worryingly, migrants are both more likely to experience a shortage of skills, and less likely to report that their existing skills are being utilised. First-generation migrants are most likely to experience a skill shortage, while low skill utilisation is more prevalent among second-generation migrants. The available measures do not reveal whether these differences result from lower skill levels on the part of first-generation migrants or higher skill demands in their jobs. Further research with more specific measures could perhaps shed more light on this. Both first- and second-generation migrants also reported significantly lower levels of job satisfaction than their native Dutch peers. Further, whereas the ethnic penalty in terms of objective labour market outcomes seemed to be only marginally mediated by field of study, grades and extra-curricular experiences during HE, the disadvantage in terms of skill utilisation and job satisfaction was partly accounted for by field of study.

These findings suggest that while migrants’ study choices do not strongly affect their employment chances per se, they may make it more difficult for them to find jobs that fit well with their skills and career goals. This could mean that significant gains could be had from better study and career guidance for young people with origins outside the Western world. Gains could also be had from increasing awareness among employers that a large and currently underutilised reserve of highly qualified young people from non-Western backgrounds exists, from whom they are currently failing to derive the fullest benefits.

In order to assess whether the problems facing migrants abate over time, we looked at the same labour market outcomes in the longer-term, more specifically four to eight years after graduation. Although limitations in the data mean that these results are indicative rather than definitive, the results suggest that many problems remain. First-generation migrants still have a significant disadvantage compared to non-migrants in terms of job satisfaction, and even after controlling for the short-term situation and mediators show an earnings disadvantage. For the second-generation we now only see a significant disadvantage as compared to non-migrants in terms of employment chances. It is not immediately clear why employment chances remain lower for second-generation migrants than for first first-generation migrants. It may be that this is because first-generation migrants are more prepared to settle for a lower-quality job, but this is speculative at this point; follow-up research will have to reveal the reason for this.

In both the short and the longer term, we were unable to account for all of the differences between migrant graduates and their native peers, so there is still much that can be done in future research to build on the results we have obtained. First, given the difficulty of reaching graduates long after graduation, other avenues could be explored to obtain a more robust view of longer-term outcomes, such as using register data to trace graduates’ career development over time. A second focus for future work is to meticulously explore the exact mismatches graduates experience between their own skills and required skills. Only then, can it be determined whether these are driven by deficits in terms of education, training and life experiences, or more by a failure to assign these graduates to jobs that do more justice to their skill set. Finally, more information is needed on the social, cultural and economic resources that are available to graduates in terms of social networks, access to relevant information, and awareness of Dutch society and the national labour market.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Although many immigrants experience relative downward social mobility: they end up in lower socioeconomic positions in their destination countries than they would have in their origin countries.

2 Agriculture, engineering, economics, health care, arts & humanities, social studies, education, law and science. The latter two fields are mainly restricted to academic universities, while the majority of teacher training takes place in HBO institutions.

3 For a more detailed account of the representativeness of the data used, see Appendix I.

4 We have estimated all models with robust and clustered standard errors within cohorts to account for the dependency in observations of respondents from a same cohort. In all models, the proportion of variance on the cohort level was very low and did not significantly differ from zero according to a likelihood ratio test. Therefore, we did not account for this clustering in our final models.

5 As a robustness check, all analyses presented in this paper were repeated with Japan and Indonesia included as non-Western countries. These models yielded similar effects as the final models in which these countries were included as Western countries.

6 As a robustness check, the analyses were also carried out using a stricter criterion for unemployment, namely a minimum of 20 h of work per week. This yielded highly similar results, as can be seen in Appendix II.

7 To eliminate possible differences in grading norms, GPA was operationalized as the deviation from the average grade of graduates who followed the same study programme at the institution.

8 To show how much difference the Heckman correction makes, Online Appendix III shows the results of all multivariate analyses without inclusion of the inverse Mills ratio as control variable.

9 Due to the relatively small number of cases it was necessary to restrict the number of control variables, to avoid instability in the estimates. We only omitted those control variables that had little or no relation to outcomes and/or migration background.