University-to-work transitions in Germany – do graduate job seekers benefit from migration and work experience?

ABSTRACT

This paper investigates the effects of migration and work experience on university-to-work transitions of German university graduates. We use a job search model, signaling and social network theory to discuss different links between the duration of labor market entry, graduate mobility and work experience. We apply event history analyses and make use of administrative social security records to examine whether work experience and pre-study as well as post-study migration accelerates the labor market entry of graduates. Our regression results stress the importance of both mobility and work experience for the length of the transition period. However, whether the effect is beneficial or adverse depends on the type of graduate migration and previous employment.

KEYWORDS:

• Labor market entry
• migration
• work experience
• graduates
• Germany

JEL CLASSIFICATION:

• C41
• I23
• J61
• R23

1. Introduction

The expansion of tertiary education in OECD countries has led to a growing supply of university graduates, but demand has increased at a slower pace (Reimer, Noelke, and Kucel 2008). These trends sparked an intense debate over graduates’ labor market outcomes and whether they are deteriorating (see Beaudry, Green, and Sand 2014; Lauder and Mayhew 2020). There is also an intense discussion on these issues in Germany (e.g. Alda, Friedrich, and Rohrbach-Schmidt 2020; Reinhold and Thomsen 2017; Reisz and Stock 2013), although German higher education graduates exhibit on average positive labor market outcomes such as low unemployment risks. Most graduates succeed in finding a regular job within a couple of months after final examinations (Böpple 2010; Van der Klaauw and Van Vuuren 2010). In particular, skill shortages for technical professions on the German labor market may contribute to a short transition phase of graduates in related fields of study (IW 2021; BMWK 2022).

However, there is also some indication that this transition phase has become more complex and less smooth (Schiener 2010). Rough and long transitions from education to work tend to have persistent adverse effects on subsequent working histories (Kunze 2002; Audas, Berde, and Dolton 2005). Mroz and Savage (2006), von Wachter and Bender (2006) and Waldorf and Yun (2016) provide evidence for scarring effects of a period of unemployment at the starting point of one’s employment biography. In turn, a short transition phase into the labor market is desirable for individuals to prevent skill deterioration.

This study aims to contribute to a better understanding of the factors that shape the speed of graduates’ transitions into the labor market. Most importantly, we focus on the role of graduates’ prior working experience and migration behavior for the duration of university-to work transitions. An extensive literature has evaluated graduates’ transition phase based on labor market outcomes such as pecuniary returns, (in)adequate employment or the degree to which skills acquired at university match skills required on the labor market. Baerts et al. (2021) find a higher probability for job interviews for graduates with an internship by conducting a field experiment. For work experience gained through internships they provide a review on a great number of studies confirming in most cases this positive relationship for various labor market outcomes by employing various methodological approaches such as experimental designs and regression analyses. This finding is also reported for graduates in Switzerland for other types of work experiences such as prior vocational education and training (VET) regarding initial wage, probability of getting an internship in the post-graduation year and entry duration after studies (Oswald-Egg and Renold 2021).

Apart from this study, there is only a small range of duration analyses having examined the length of graduate’s transition phase so far. These works provide robust evidence on the importance of different individual and study-related, labor market and institutional factors (e.g. Chuang 1999; Barros, Guironnet, and Peypoch 2011; Salas-Velasco 2007). Detailed knowledge on the relevance of work experience and mobility behavior of graduates, however, is still largely missing.

This is striking because the majority of graduates have several jobs before completing studies and thus gain considerable work experience before they enter the labor market. For instance, a considerable share of graduates in Germany does work to finance studies, thus having less time for studying and therefore facing probably greater difficulties when entering the labor market.¹ But social contacts and labor market information gained via student jobs may as well alleviate job search and thereby enable a swift career start (Haak and Rasner 2009). Access to such job-relevant networks may also depend on the pre- and post-study migration behavior of graduates, which are among the most mobile groups on the labor market (Faggian, McCann, and Sheppard 2007). In addition, there is ample evidence that graduate migration sustains their career success (e.g. Di Cintio and Grassi 2011; Ganesch, Dütsch, and Struck 2019).

In this paper, we analyze the relationship between graduates’ pre- and post-study migration behavior, different types of previous work experience and the length of labor market entry after studies. We use event history methods to investigate university-to-work transitions of German university graduates. The theoretical considerations behind the empirical analysis combine arguments from a job search model, signaling and social network theory to establish links between the duration of labor market entry, graduate mobility and work experience. Student record data from six German universities that is linked with administrative biographical data reported in social security records provides detailed information on work experience of students and on mobility of the graduates. Our sample consists of 19,860 students who graduated between 1996 and 2012, and whose employment histories we can trace up to 2016. The data set includes 22.959 graduate-year observations.

Our results indicate that immobile resident graduates enter the labor market more rapidly than all mobile graduates who take up a job outside the region of studies. However, return migrants who moved to the region of studies for attending university and return to their home region after graduation tend to have a shorter entry duration than other mobile graduates. These findings point to the importance of location-specific knowledge and contacts for swift university-to-work transitions. The findings also suggest a supportive role of work experience for the speed of graduates’ labor market entry. In particular, apprenticeship training before studying seems to matter, which may point to significant signaling and network effects. In contrast, marginal jobs, which are temporary student jobs with a maximum of 15 weekly working hours, appear to prolong labor market entry. The adverse effects of this type of employment on university-to-work transitions tend to be particularly strong shortly before graduation.

We complement the small set of duration analyses on university-to work transitions in several ways. While the existing literature differentiates merely between migration before/and or after studies and considers general work experience during higher education, we include different mobility and job types in our analysis. Most notably, the use of administrative biographical data on a daily basis enables us to measure the exact length of employment episodes either before or during studies. Specific information on places of work and residence allows us to combine pre- and post-study migration behavior in this study. As most of the previous duration analyses resort to survey data, related problems such as survey bias are of less concern in this study as we make use of register data.

The remainder of the paper is organized as follows: In section 2, we discuss theoretical arguments and review the empirical literature. We introduce the survival model and the data set in section 3. Section 4 discusses the empirical findings of the event history analysis, while the final section presents the conclusions.

2. Theoretical considerations and empirical literature

In this section, we combine a job search model with theoretical arguments from signaling and social network theory to establish a framework that highlights different direct and indirect channels through which work experience and mobility might influence the length of the transition from graduation to labor market entry. We then discuss related empirical evidence on labor market entry of higher education graduates provided by duration analyses.

2.1 Theoretical considerations

2.1.1 A job search framework

Van der Klaauw and Van Vuuren (2010) apply job search theory introduced by Mortensen (1970) to analyze the transition of graduates from college to work in the Netherlands. In particular, their model takes into account a trade-off between devoting effort to studying and to job searching in the final stage of academic studies, i.e. students not only spend time to prepare for final examinations, but also start looking for a job. Allowing for job searching while studying is in line with our data, since we observe that a significant percentage of graduates start working very shortly after final examinations.

Van der Klaauw and Van Vuuren (2010) stress that study-to-work transitions, labor market outcomes and academic achievement are jointly determined, and it is important to consider the inter-dependency between these variables. In their model, students might invest time in studying $\left(e\ge 0\right)$ and searching for a job $\left(s\ge 0\right)$. The two activities are supposed to be substitutes. Checking different media for vacancies and applying for jobs gives rise to job search effort and costs $c\left(s,e\right)$, which increase with both search effort and study effort. The authors assume that study effort improves grades $g$, while job search effort $s$ gives rise to job offers, described by the arrival rate $\lambda \left(s,g\right)$ with $0\le \lambda \left(s,g\right)\le 1$. The probability of receiving a job offer increases with search effort, but marginal returns for search effort decrease [$\mathrm{\partial \lambda }/\mathrm{\partial s}>0,\mathrm{\partial }{\lambda }^2/\mathrm{\partial }{s}^2<0$]. Students with zero effort do not receive any job offers, i.e. $\lambda \left(0,g\right)=0$.² However, there might also be an indirect adverse effect of search effort, since investing more time in job search reduces study effort, which likely lowers grades and might therefore give rise to an adverse signal for employers (Røberg and Helland 2017), thereby decreasing the arrival rate and delaying labor market entry. Study effort might also influence the timing of labor market entry via different channels: there is a prolonging effect, since search effort declines as study effort increases. However, study effort also improves grades and, in turn, increases the arrival rate.

The crucial feature of a job offer is the wage offered $w$, which is assumed to be a random draw from a wage offer distribution. A student who receives an offer decides immediately whether to accept or reject it. Acceptance means that the graduate enters the labor market in the next period. Graduates are supposed to start working in regular jobs only after graduation. They will enter the labor market and exit non-employment if the wage associated with the best job offer exceeds the student’s reservation wage ($w^r$). After graduation, therefore, students either become unemployed (or continue with non-regular employment) or start working in a regular job. Immediate labor market entry requires job searching to start before final examinations.³

In a search model framework, the duration of the labor market entry process is influenced by the job offer arrival rate and the reservation wage. Salas-Velasco (2007) argues that the signaling approach proposed by Spence (1973) might add important aspects to this setting and discusses a framework that integrates the search processes of workers and firms.

2.1.2 Signaling

As information is imperfect, employers use observable individual characteristics (e.g. age), educational credentials (e.g. field of study, grade) and practical credentials (e.g. work experience) as signals to assess the productivity of job applicants. Graduates with less attractive signals from an employer's perspective are likely to be disregarded or less likely to be considered in the selection process among potential applicants. For this reason, these applicants have to make more job enquiries to firms on average and are supposed to be longer unemployed after graduation than graduates with more favorable signals. For instance, lower exam grades or a longer duration of study can be assessed as signals of lower productivity. Reimer, Noelke, and Kucel (2008) argue that a degree in specific fields of studies may indicate ability, i.e. demanding fields of study such as mathematics and physics are supposed to be positive signals.⁴ Work experience might as well provide valuable information for the recruiting process and may influence the probability that an applicant receives a job offer. Besides, practical work experience might be considered as a favorable signal if it lowers the costs of initial skill adaptation training (Conelly et al. 2011).

However, the type of work experience likely matters if recruiting firms use this information to assess the productivity of candidates and costs of initial training. Geel and Backes-Gellner (2012) argue that student employment will only complement formal education and thereby increase future productivity of graduates if employment is related to the field of study (and in turn prospective field of work). Previous work experience indicated by the application documents might therefore only increase the probability of receiving a job offer if there is a sufficient overlap between the sector of student employment and prospective field of work.⁵

2.1.3 Social network theory

Caliendo et al. (2019) argue that search effort might also be reflected by the geographic distance between the region of residence and the location of potential employers, i.e. the spatial scope of the search.⁶ They assume that workers tend to have better knowledge about local labor markets and access to local networks. Applying this reasoning to our setting implies that it will be easier for graduates to search for a job in their study region. This may be especially true for resident graduates with location-specific advantages accruing through prior schooling and studying (DaVanzo 1981). However, labor market entry outside the study area might reflect an increased search effort, giving rise to a shorter transition period. If the job search of the graduates precedes migration, a change of residence likely reflects a more intense search in terms of spatial range, and we might expect that mobility should go hand in hand with more rapid labor market entry (see also Guglielminetti, Lalive, and Wasmer 2015). Yet, there are also theoretical arguments, which suggest that labor market entry outside the region of studies may be associated with longer transition phases, as mobile graduates, except for return migrants, are unlikely to be able to make use of local network-based information advantages for job placement after studies (Faggian, McCann, and Sheppard 2007).⁷

Moreover, the reservation wage might differ depending on whether acceptance of the job requires a move. The existence of significant migration costs suggests that the reservation wage increases with distance between region of study and prospective workplace. In the model by Caliendo et al. (2019), individuals choose their optimal search effort by equating the marginal costs of job searching with the marginal benefits associated with additional searching. The return of an additional search effort is an increased probability of receiving a job offer paying more than the reservation wage.

In the job search framework outlined above, working while studying will reduce search effort before graduation and, therefore, has a dampening effect on the job arrival rate, which in turn postpones labor market entry. It is important to note that in this case, there is no countervailing effect of an increased study effort, since working while studying may reduce both study effort and search effort. However, there are likely additional (beneficial) effects of work experience on labor market entry not addressed in a job search model, which might be based on network effects. Most importantly, social network approaches (Granovetter 1973; Sarcletti 2007) emphasize that job search does not take place in a system of anonymous actors, but makes use of existing social network relationships. Such relationships can be used, on the one hand, as a source of information and, on the other, to get access to more distant social contacts. Valuable information about job vacancies and relevant labor-market specific knowledge can be passed on through social contacts with co-workers and employers. Through social contacts, it is therefore possible to find a new job with less effort. The use of information from such social contacts thus results in lower search costs (Granovetter 1973; Preisendörfer and Voss 1988). Consequently, graduates who could establish such labor market-relevant contacts over time and gain access to job-specific information by carrying out jobs before and/or during studies may therefore be able to find a job more quickly after graduation. Moreover, having established such contacts and networks in the region of studies goes hand in hand with a certain degree of ‘territorial embeddedness’ (Krabel and Flöther 2014), which might alleviate local labor market entry, but implies relatively high social cost when leaving the study region.

2.1.4 Combining the different approaches

To sum up, combining the arguments from a job search framework, the signaling approach and social network theory, the probability of taking up employment and the timing of labor market entry becomes the product of three factors: graduates searching for a vacancy, firms offering jobs to an applicant and acceptance on the part of the graduate. In the following discussion, we assume that the job offer arrival rate and the reservation wage of the graduates capture these factors (Salas-Velasco 2007).

With respect to our empirical model, we presume that a high probability of a match, i.e. acceptance of a job offer, should correspond with a swift labor market entry. The probability of a match is influenced by the job offer arrival rate and the reservation wage, which, in turn, are likely determined by various factors, including our pivotal variables, work experience ($\omega$) and migration, i.e. the expanded spatial scope of search ($\mu$). Thus, we assume that their effect on the timing of labor market entry will be mediated via their impact on $\lambda$ and $w^r$. Below we discuss the different channels through which the two factors may affect the likelihood of a match in more detail.⁸

In order to describe the effects that operate via $\lambda$, we expand the arrival function of the job search model and include characteristics of the young workers $x$ and $\omega$. Moreover, we consider indirect effects, which might arise as $\omega$ and $\mu$ likely affect search effort and grades. The extended function is given by $\lambda \left(s,g,x,\omega \right)$ with $\mathrm{\partial \lambda }/\mathrm{\partial s}>0$ and $\mathrm{\partial \lambda }/\mathrm{\partial g}>0$.⁹ The arguments of social network theory outlined above suggest that work experience positively affects the arrival rate because knowledge on the (local) labor market and contact to employers should enhance job search [$\mathrm{\partial \lambda }/\mathrm{\partial \omega }>0$]. Moreover, $\omega$ may also increase the probability of receiving a job offer because it could act as a positive signal pointing to high productivity and ambition of the candidate. However, the latter effect might only occur if the work experience is related to the prospective field of work.

Apart from these direct effects of $\omega$ stressed by signaling theory and the social network approach there might be important indirect channels through which $\omega$ and $\mu$ influence the arrival rate. With respect to search effort we suppose that working while studying will reduce search effort before graduation and, therefore, has an indirect dampening effect on the job arrival rate [$s\left(e,\omega ,\mu \right)$ with $\mathrm{\partial s}/\mathrm{\partial \omega }<0$].¹⁰ In contrast, increased search effort is assumed to go hand in hand with an extended spatial scope of search $\mu$ as reflected by migration [$\mathrm{\partial s}/\mathrm{\partial \mu }>0$]. Finally, there is another indirect effect of work experience operating via grades as we expect that working during studies reduces study effort and, thereby, grades [$g\left(e\left(\omega \right)\right)$ with $\mathrm{\partial g}/\mathrm{\partial e}>0$ and $\mathrm{\partial e}/\mathrm{\partial \omega }<0$]. With lower grades, in turn, the job offer arrival rate declines since employers likely use exam grades to assess the productivity of the graduates.

The influence of work experience on job offer arrivals is therefore indeterminate and depends on the size of a positive direct and the negative indirect effects. Positive effects of work experience are, however, more likely if student employment is related to the prospective field of work. The impact of migration on the arrival rate is supposed to be positive in this setting.

However, in order to arrive at a net effect of both $\omega$ and $\mu$ on the duration of labor market entry, we need to account for their influence on the reservation wage. Study-related factors such as grades, individual characteristics of the graduates, migration and work experience might also impact the length of labor market entry via the reservation wage $w^r\left(g,x,\mu ,\omega \right)$. The arguments put forth above suggest that an extended spatial job search should increase the reservation wage [$\mathrm{\partial }{w}^r/\mathrm{\partial \mu }>0$] because a corresponding migration event involves migration costs. Work experience might increase the reservation wage if graduates assume that firm perceive experience as a positive signal. Moreover, we suppose an indirect effect of work experience on $w^r$ that operates via its impact on grades. The indirect effect of $\omega$ is likely negative if working during studies reduces study effort and grades. As the latter may increase reservation wages, we expect a negative indirect effect of work experience on the reservation wage.

To summarize, there are various channels through which work experience and migration may affect the transition from university to work, potentially giving rise to opposing direct and indirect effects. For instance, migration might increase the length of the transition period via its impact on the reservation wage while the effect operating through job arrivals is supposed to be negative. The setting for work experience is even more complex. In the regression analysis, we therefore expect to detect net effects. Furthermore, due to the coexistence of a range of positive and negative influences, the impact of $\omega$ and $\mu$ on the duration of graduates’ labor market entry is theoretically indeterminate.

2.2 Empirical literature

Thus far, several duration analyses have identified and evaluated - to different extents – factors determining the length of graduates’ university-to-work transitions. Apart from one European cross-country study (Salas-Velasco 2007), these studies refer to individual countries such as Italy (e.g. Biggeri, Bini, and Grilli 2001; Pozzoli 2009; Sciulli and Signorelli 2011), Canada (Betts, Ferrall, and Finnie 2000), Germany (Böpple 2010; Haak and Rasner 2009), France (Barros, Guironnet, and Peypoch 2011), the Switzerland (Oswald-Egg 2016; Geel and Backes-Gellner 2012) or Taiwan (Chuang 1999).

Empirical evidence of these studies is in line with the proposition of the job search model that many graduates already begin to look for a job before final examinations. This applies, for instance, to 53% of the students from Mannheim University in Germany (Böpple 2010). Only 20% of Dutch graduates had not looked for a position before finally leaving university (Van der Klaauw and Van Vuuren 2010). Moreover, most of the duration analyses corroborate that a significant proportion of graduates enter the labor market within a few months after final examinations.

The duration analyses provide robust evidence on the relevance of individual characteristics and study-related factors for labor market entry. For instance, female and older graduates tend to have longer non-employment spells after studies than their male and younger counterparts (Salas-Velasco 2007). There is some indication that high academic achievement is associated with an immediate career start. The shorter the duration of enrollment at university, the faster graduates take up their first job after studies (Sciulli and Signorelli 2011). The same applies to the completion of studies within the regular time scheduled for the course program and better final marks (Biggeri, Bini, and Grilli 2001). In contrast, Sciulli and Signorelli (2011) and Pozzoli (2009) ascertain a longer transition phase for better performing students. This may be due to an indirect effect, since better graduates likely have higher reservation wages and thus lower acceptance rates of job offers.

The theoretical considerations discussed above suggest a supportive role of local information and network advantages for labor market entry. Teichert et al. (2020) show that the likelihood of German graduates entering the local or extra-regional labor market depends on where they could gain access to relevant networks through previous work experience. However, according to our theoretical argumentation an extension of the spatial scope of job searching might also point to an increased search effort and foster a career start shortly after graduation (Caliendo et al. 2019; Guglielminetti, Lalive, and Wasmer 2015, see section 2.1). There is ample evidence that graduates’ labor market outcomes are positively affected by spatial mobility¹¹, but only a few studies consider the effect of migration on the length of the university-to-work transition. Betts, Ferrall, and Finnie (2000) do not detect a significant effect of whether graduates moved to the higher education region for attending university on the length of the transition. Sciulli and Signorelli (2011) explore the timing of labor market entry on the provincial labor market of the university in Perugia (Italy). The analysis shows a faster labor market entry of resident graduates. This finding confirms a positive effect of location-specific advantages. Yet, these results are biased due to unobserved transitions of graduates finding a job outside the province of Perugia. Controlling for this bias the advantage of resident graduates no longer exists.

Faggian, McCann, and Sheppard (2007) develop a migration typology, which is also employed in this paper (see for description section 3.2, Figure 1), to study the interplay of graduates’ access to local social networks, pre-study migration experience and their post-study migration decision with graduates’ labor market outcomes. Evidence on a significant relationship between the migration types and graduates’ labor market performance does not exist for the length of the transition phase, but for the wage level. The works of Jewell and Faggian (2014), and Kazakis and Faggian (2017), for instance, disclose the highest wage premium for repeat migrants who move for education and later for work. For Chinese graduates Zhao and Hu (2019) find a higher wage premium for return migrants than for repeat migrants. The former group might benefit from location-specific advantage compared to the latter.

Figure 1. Types of graduate migration. Source: Faggian, McCann, and Sheppard (2007).

The empirical evidence of most studies on the length of the university-to-work-transition relates to work experience acquired while attending higher education, but no differentiation is made with respect to the types of jobs that were carried out. On one side, various studies provide empirical evidence for the supportive role of generic work experience for swift labor market entry of the young highly educated (Sciulli and Signorelli 2011; Biggeri, Bini, and Grilli 2001; Salas-Velasco 2007; Mason, Williams, and Cranmer 2009; Gault, Redington, and Schlager 2000). On the other, Pozzoli (2009) and Barros, Guironnet, and Peypoch (2011) reveal that graduates with work experience gained during studies have a lower likelihood of exiting unemployment. However, the effect of work experience on job searching and the transition to employment likely differs depending on the type of employment. Regular employment might enable students to establish more useful job search networks compared with loose contacts acquired through casual jobs and marginal employment (Teichert et al. 2020). There is also evidence that, in particular, specific work experience linked to the field of study (Weiss, Klein, and Grauenhorst 2014) or the occupation of desired jobs (Hammen 2009) facilitates the labor market entry of graduates.

The only duration analysis to differentiate between different types of working experience stems from Haak and Rasner (2009). They compare the importance of work experience for the career entry of German tertiary graduates between different fields of study. They detect that different types of work experience, such as vocational training before studies, working while studying and an internship, shorten the transition phase. This applies particularly to humanities, but less to law, economics and engineering. Haak and Rasner (2009) conclude that graduates in humanities rely on additional ‘practical’ signals when applying for a job because they have a degree in a field of study with less occupation-specific curricula.

Böpple (2010), however, finds that an internship does not contribute to a shorter transition phase. There may be at least two reasons for these inconclusive results. The theoretical considerations discussed above point to different effects of work experience on labor market entry. Depending on the size of the partly opposing effects, the net effect of experience might be positive or negative. Moreover, the effect might differ across types of work experience. Unlike Haak and Rasner (2009), the other above-mentioned duration analyses do not distinguish in detail between different types of work experience. But the findings of different studies suggest that the access to labor-market-relevant information and network resources crucially depends on the type of experience (e.g. Robert and Saar 2012; Hammen 2009; section 3.1). Moreover, the effect of these resources on labor market entry may also depend on whether a student is employed prior to and during higher education. However, the above-mentioned duration analyses mostly use information on whether a student has gained work experience or not.

Even though some duration analyses provide detailed evidence on individual and study-related factors for the speed of graduate’s labor market entry, working experience as well as migration behavior have received much less attention so far. We contribute to this literature with addressing both various types of work experience acquired either before or during studies and different migration types by combining pre- and post-migration experience in our duration analysis.

3. Empirical model and data

3.1. Econometric approach

To model the labor market entry of university graduates, we examine the hazard rate of transition from higher education to employment. Formally, the hazard rate $h_i\left(t\right)$ is the probability¹² of being employed given that the graduate $i$ was not employed up to the period $t$ after graduation: (1) $h_i\left(t\right)=\mathrm{Pr}\left(T_i<t+1\vee T_i\ge t\right)$(1) where $T_i$ is the length of the non-employment spell. We define labor market entry as taking up a regular part-time or full-time job with a minimum duration of three months. Hence, non-employment captures all other types of labor market statuses such as unemployment, participating in measures of active labor market policy or marginal employment. In order to determine the transition event, we make use of continuous data that comprises information on the exact starting date of the first full-time or part-time employment after graduation (see section 3.2).

We estimate a proportional hazard specification in order to identify important determinants of labor market entry: (2) $h_i\left(t,x_i\right)=h_0\left(t\right)\exp \left(x_i\text{β}\right)$(2) where $h_0\left(t\right)$ is the baseline hazard and $x_i$ is a vector of explanatory variables that includes individual characteristics such as gender, nationality and age, as well as study information such as examination grade or field of study. In the regression analysis, we focus on the effects of graduates’ work experience before and during studies and on the impact of mobility before studying and after graduation.

We apply a parametric model and assume that the baseline hazard $h_o\left(t\right)$ can be described by a Weibull distribution: (3) $h_i\left(t,x_i\right)=p{t}^{p-1}\exp \left(x_i\text{β}\right)$(3) The regression analysis provides an estimate of the shape parameter $p$ that indicates whether hazard rates increase or decrease exponentially over time. The probability of labor market entry might increase or decline with the length of the non-employment period after graduation. The pressure to take up a job perceived by young workers is likely to rise, e.g. due to financial necessity or threat of stigma. In their model, Van der Klaauw and Van Vuuren (2010) assume that unemployed graduates do not change their search effort and reservation wage across periods. However, we might well suppose that the reservation wage declines and search effort rises as the period of non-employment after graduation increases. This behavior might give a rise to a positive duration dependence, pointing to an increasing probability of taking up a job as the period since final exams increases.

The detailed information on graduates and their studies available in our data set enables us to consider a wide range of factors that likely influence their labor market entry. However, the estimates might be affected by unobserved heterogeneity at the individual level. We are not able to preclude that migration and work experience capture the effects of other unobserved features of graduates such as motivation. These unobserved characteristics might be important factors for job search effort and may thus influence labor market entry. Neglecting these factors might give rise to biased estimates of the ‘effects’ of migration and work experience. Hence, our analysis allows statements about correlations between the labor market entry of graduates and the pivotal variables, rather than a causal interpretation of the results. Nevertheless, we include a frailty term that captures unobserved heterogeneity in our model. It is assumed that graduates differ randomly in a manner that is not fully accounted for by the observed characteristics, and that the frailty term is independent of these observed characteristics.

3.2. Data, sample and key variables

Our analysis is based upon a comprehensive micro-level database, which links information from student records of six medium-sized German universities¹³ with the Integrated Employment Biographies (IEB) of the Institute for Employment Research (IAB) (for detailed information see Teichert et al. 2020). The student records contain individual characteristics (e.g. sex, nationality, age, type of university entrance qualification) and study-related information (e.g. field of study, type of degree, examination grade, study length, graduation date). The IEB data provide information on starting and ending dates of different labor market episodes (i.e. periods of unemployment, benefit receipt, employment, participation in training measures) for each individual who is subject to social insurance contributions. Sociodemographic characteristics (e.g. sex, date of birth, nationality, qualification level) and job features (type of employment, occupation, industry affiliation, region of workplace) are also available. This continuous (daily) biographical data enables us to determine the exact length of the transition period between graduation date and the starting date of the first regular job. The availability of register data is a big plus because it tends to be less biased and is not yet used in previous studies. Moreover, our data set includes novel information on the exact length of work experience and labor market entry.

We define the labor market entry as the first full-time or part-time employment spell with a minimum length of three months¹⁴ after graduation.¹⁵ We only consider regular employment subject to social security contributions. We thereby seek to rule out transition through short-term employment such as internships and temporary jobs. Furthermore, we only consider graduates who do not take longer than 730 days for their labor market entry. This implies that our data is right-censored. This restriction enables us to rule out (at least to some extent) a prolongation of the labor market transition, which might be driven by unobserved heterogeneity among graduates with respect to their propensity to look for a job. For instance, graduates might not seek to look for a job first, but realize other plans such as traveling, sabbaticals, gap years, family phases or further education. Moreover, we cannot observe labor entry abroad or via employment as civil servants or self-employment. Non-random self-selection of graduates into these entry options might affect our results as well. Nevertheless, this restriction seems to be a reliable assumption, since existing evidence suggests that almost all graduates take up a job within two years after graduation (see section 4).

We restrict the sample to graduates who are 20–35 years old at the date of certification and completed studies within 20 semesters. While we exclude graduates with a bachelor or PhD degree, graduates with a Master, Diploma or other degrees (e.g. state examination) are considered.¹⁶ We use the latest tertiary degree to ensure that graduates subsequently enter the labor market. In addition, we had to exclude a few fields of study (teacher training, law, human medicine and dentistry), as the entry of graduates in these fields into the labor market is subject to specific restrictions.¹⁷ The graduates in our sample do not necessarily have to have lived in Germany before studying. Indeed, some graduates have obtained their higher education entrance qualification abroad. However, we only include graduates for whom we can observe their first employment within two years after graduation in Germany.

In the sample, graduates already employed in regular jobs before the date of certification are disregarded, because their post-graduation job search may be different from that of the rest of the sample, and they might be less interested in finding another job (Pozzoli 2009). We also exclude those graduates who have their first regular employment spell at universities or research and development institutions because they are likely to pursue a doctoral education after studying. The final sample consists of 19,860 young workers who graduated between 1996 and 2012, and whose employment histories we can trace up to 2016.

Two variables, mobility and work experience, are the focus of our analysis.¹⁸ Other studies investigating the role of prior work experience for career entry usually rely on survey data with limited information on the graduates’ employment histories. The advantage of our data set is its coverage of graduates’ labor market biographies before, during and after studies on a daily basis. In order to measure work experience, we cumulate all employment spells (number of days) in either marginal jobs or regular part-time and full-time jobs before and during studies. The so-called marginal jobs are temporary student jobs with a maximum of 15 weekly working hours. Moreover, we observe whether a graduate had completed a vocational training before studying.¹⁹ In a next step, we differentiate according to whether work experience is sector-specific or not. For this purpose, we check whether graduates had already gained work experience in the same industry in which their first job after studies is situated.

Furthermore, the data contain detailed information on the residence and workplace at the NUTS 3²⁰ level that enables us to determine the mobility before and after studies.²¹ We use functional labor market regions as the regional unit to determine mobility. These consist of several counties (NUTS 3 regions), which are connected via strong commuter flows. While the location where the graduates received their university entrance qualification (home region) is reported in the student records, the location of the workplace and the residence are documented in the IEB. Mobility before studies corresponds to a move from the home region to the university region. Mobility after studies is defined as a move from the region of studies to the region of the first regular job after graduation. The universities included in our analysis are located in medium-sized regions. The results on the migration behavior of graduates when entering the labor market might therefore not apply to young workers who graduate in large urban regions. However, most universities in Germany are located outside the major agglomerations, and the data should be representative for smaller and more lagging regions which tend to suffer from significant outward migration of graduates.

We employ a typology of graduate mobility (see Figure 1), which has been developed by Faggian, McCann, and Sheppard (2007) and applied in numerous studies (see e.g. Kazakis and Faggian 2017). This typology combines mobility before and after studies and allows thereby to examine the relationship between pre- and post-study migration on one side and between migration propensity and local social networks on the other. Graduates who studied in their home region take up the first job either in the same region (immobile) or enter the labor market elsewhere (migrants). Graduates who were already mobile before studies and find the first regular job inside the region of studies are staying migrants. Graduates who leave the home region for studies and move from the region of studies to the region of the new workplace are return migrants if the workplace is in the home region or repeat migrants if the workplace is located neither in the home region nor in the university region.

4. Empirical results

Graduates’ entry durations are rather short. The estimated survival function based upon the non-parametric Kaplan-Meier method (Figure 2) shows that the majority of the graduates (65.7%) in our sample enter the labor market within one year after graduation. This finding is in line with previous evidence (see, for example, Pozzoli 2009; Haak and Rasner 2009; Sciulli and Signorelli 2011; Salas-Velasco 2007).

Figure 2. Survival function (Kaplan-Meier estimates) for graduates. Source: University panel linked to the IEB of IAB, own calculations.

In this section, we examine the relationship between spatial mobility and work experience on one side and the length of the transition phase between graduation and taking up the first job on the other. The outcomes of the continuous-time duration models, which base upon a proportional hazard approach, are presented in the following sections. We consider the effects of spatial mobility (section 4.1), general work experience (section 4.2), sector-specific and non-specific work experience (section 4.3), work experience gained shortly before final exams (section 4.4) and further control variables (section 4.5). All models include individual characteristics and study-related factors as control variables.

4.1 Mobility

Regression results on pre-study and post-study mobility are shown in Table 1, model (1) to (3). The results in these models suggest that graduates who moved to another region for attending university (spatial mobility before studies) have a significantly higher likelihood of exiting non-employment after having taken the university degree. In other words, graduates migrating before studies seem to find a job faster after studies than young workers who completed their studies in the region where they obtained their university entrance qualification. The difference in the hazard between two graduates who vary only with respect to pre-study mobility amounts to 4.6% ([1 – exp(0.045 × 1)] × 100). We suppose that these mobile graduates are possibly particularly motivated and might show a relatively high job search effort. Moreover, employers might perceive pre-study mobility as a signal indicating motivation and productivity of graduates.

Table 1. Regression results – determinants of transition time into first regular full-time and part-time jobs.

	(1)		(2)		(3)
	coef	se	coef	se	coef	se
Individual characteristics
Female	−0.102***	0.026	−0.108***	0.026	−0.101***	0.028
Age	−0.049	0.091	−0.002	0.093	0.110	0.093
Age²	0.000	0.002	−0.001	0.002	−0.002	0.002
Foreigner	−0.157***	0.060	−0.154**	0.061	−0.194***	0.063
Spatial mobility
Spatial mobility before studies	0.045**	0.022	0.046**	0.023	0.043*	0.023
Spatial mobility after studies	−0.112***	0.027	−0.114***	0.028	−0.112***	0.031
Higher education
University (Ref.: university of applied sciences)	−0.242***	0.052	−0.244***	0.055	−0.293***	0.071
Examination grade (Ref.: sufficient)
Satisfactory	0.080	0.049	0.080	0.050	0.093*	0.052
Good	0.027	0.042	0.028	0.043	0.039	0.044
Excellent/very good	−0.060	0.047	−0.059	0.047	−0.054	0.048
Other degrees (Ref: Diploma/Master)	−0.360***	0.050	−0.357***	0.051	−0.366***	0.058
Study length	−0.016***	0.004	−0.016***	0.004	−0.022***	0.005
Work experience
Vocational training	0.233***	0.033	0.212***	0.033
Experience (in 100 days)
before higher education
total	0.006*	0.003
marginal empl.			−0.001	0.004	−0.004	0.005
regular empl.			0.015***	0.005	0.020***	0.005
during higher education
total	−0.026***	0.003
marginal empl.			−0.030***	0.004	−0.030***	0.004
regular empl.			−0.002	0.007	0.001	0.008
Previous employer	0.269***	0.041	0.256***	0.033	0.271***	0.048
Ln(p)	0.718***	0.054	0.723***	0.062	0.735***	0.091
Ln(θ)	−0.920**	0.468	−0.869	0.529	−0.753	0.753
Implied p	2.050	0.110	2.061	0.128	2.085	0.190
θ	0.399	0.187	0.419	0.222	0.471	0.355
Log Likelihood	−8,997.83		−8,989.18		−9,007.76
Number of students	19,860		19,860		19,860
Observations	22,959		22,959		22,959

Notes: * significance at the 0.1 level, ** significance at the 0.05 level, *** significance at the 0.01 level; robust standard errors in parentheses. All models include time effects and fixed effects for field of study. Our sample consists of 19,860 young workers. Our units of observation are individual student-year observations (22.959). Since a relatively large number of students enter the labor market quickly, the number of student-year observations is not significantly higher than the number of graduates.

p is the shape parameter from a Weibull distribution. A value larger than one implies a positive duration dependence. θ is the estimated frailty parameter.

marginal empl.: marginal employment.

regular empl.: regular (part-time or full-time) employment.

We find the opposite effect on labor market entry for mobility after studies. Differentiating between the four graduate mobility types provides more detailed information on the effect of post-graduation migration. Corresponding regression results are summarized in Table 2. The immobile graduates are the reference category in all models. The estimates indicate that migrants and repeat migrants are significantly less likely than immobile graduates to be quickly employed after graduation. Hence, immobile graduates seem to enter the labor market faster. Graduates starting their career in their home (university) region seem to benefit from cumulated location-specific knowledge and network advantages through previous schooling and studying as pointed out in our theoretical considerations in section 2. Staying migrants and return migrants do not significantly differ from immobile graduates. Staying migrants might have developed local networks during higher education. Returning migrants who accept a job offer in their home region could obviously benefit, unlike the other two mobile graduate groups, from the access to already established networks.

Table 2. Regression results – mobility types and transition time into first regular full-time and part-time jobs.

	(1)		(2)		(3)
	coef	se	coef	Se	coef	se
Mobility types (Ref: Immobile)
Staying migrants	0.050	0.040	0.053	0.041	0.046	0.046
Migrants	−0.109***	0.036	−0.110***	0.037	−0.130***	0.038
Returning migrants	−0.043	0.042	−0.042	0.042	−0.048	0.046
Repeat migrants	−0.074**	0.033	−0.074**	0.034	−0.092***	0.035

Notes: ** significance at the 0.05 level, *** significance at the 0.01 level; robust standard errors in parentheses. All models include time effects, individual characteristics and study-related variables.

Our finding of a longer entry duration of the two mobile graduate types is not in line with our initial assumption (section 2) that graduates who look beyond the region of studies increase their search effort and thus find a job faster. Many graduates might first search for a job in the university region because they seek to make use of their location-specific knowledge and perhaps try to avoid migration costs. However, those young workers who do not manage to enter the labor market quickly might extend the spatial scope of their job search as they increasingly face pressure to find an adequate job. In this setting, migration goes hand in hand with a relatively late career start due to reverse causality. However, this argument (alone) cannot explain the differences that we observe between the three migrant types and, in particular, the absence of a significant negative effect in case of the returning migrants.

Another reason for the delayed career entry of mobile graduates might arise from a more selective job search behavior that is driven by high (initial) reservation wages. Reservation wages are, as argued in section 2, determined by individual characteristics and study-related factors. For instance, graduates with high grades likely have high reservation wages and are thus choosier when searching for a position (Pozzoli 2009). Moreover, reservation wages might be relatively high if labor market entry involves migration costs.

It is noteworthy that all graduate mobility types that are likely to have some knowledge on the region in which they take up their first job after graduation (immobile, staying migrants, return migrants) seem to enter the labor market more quickly than graduates for whom we assume that they have no or only little information on the regional labor market (migrants, repeat migrants).

4.2. General work experience

In Table 1, model (1) comprises information on pre-study and post-study mobility and on general work experience, while model (2) and model (3) differentiate additionally between marginal and regular employment before and during studies. These regression results indicate that work experience is relevant for labor market entry of university graduates.²² The effect of vocational training in particular is fairly large in all models. Having completed vocational training before studying increases the likelihood of starting a regular part-time or full-time employment by 26.2% (model 1). This outcome confirms results by Haak and Rasner (2009) for German graduates. The finding might be due to a signaling effect if an apprenticeship is evaluated as a positive productivity signal by employers, thus increasing the probability of receiving a job offer. Moreover, an apprenticeship training likely gives rise to network contacts in the training firm and the local labor market that might facilitate job searching after graduation. Both findings are in line with our theoretical arguments in section 2.

We only detect a weakly significant effect of work experience before studies in addition to an apprenticeship in model (1). However, this outcome seems to be caused by opposing effects of regular and marginal employment before higher education. While there is a significant positive effect of regular jobs on labor market entry in models (2) and (3), the estimates also point to an adverse, but insignificant relationship of marginal employment of moderate size: increasing work experience via marginal employment before studying by 100 days decreases the hazard of labor market entry by 0.4%. The effect of regular employment increases remarkably once we exclude the vocational training dummy in model (3). Hence, vocational training captures an important part of the experience effect induced by regular employment prior to studies.

While work experience that is gained before studying tends to facilitate labor market entry, there seems to be a detrimental effect of employment during higher education. But again, it is important to distinguish between types of employment. The coefficient for regular employment does not significantly differ from zero in all models. This result might be due to opposing effects of working while studying discussed in section 2. When students are working during higher education, they might reduce either job search efforts or study efforts or may invest less time in both finding a job and studying. The job search model suggests that this gives rise to unfavorable direct and indirect effects on labor market entry. This adverse influence may counteract positive network and signaling effects of regular employment during studies. The opposing effects might offset each other and result in an insignificant net effect of regular employment.

Work experience that is gained during studies is primarily due to marginal employment. The negative effect of experience in model (1) is solely caused by this type of employment. The main motivation to take up these jobs while studying is probably financial necessity, and often they will not provide many helpful work-related contacts for job searching later on. This is in line with the significant negative coefficient of marginal employment during studies in the models (2) and (3). The adverse effect of marginal jobs is much stronger if students work during higher education compared to marginal employment before enrollment. This difference probably points to a delayed university-to-work transition as a result of less intense job searching and reduced study effort caused by marginal employment during studies. These adverse effects cannot be triggered by jobs before enrollment.²³

There is also direct evidence of important beneficial network effects on labor market entry of university graduates. Starting the first regular job after studies at a previous employer increases the likelihood of exiting non-employment after studies by 31% in model (1). This provides clear evidence on the importance of labor market contacts obtained via previous work experience as argued in section 2.²⁴

4.3. Sector-specific versus non-specific work experience

Prior work experience, which is related to the sector of the first job, is expected to be of particular importance when searching for a job after studies. If graduates have already worked in the same sector of their subsequent job, they could probably benefit from sector-specific knowledge and contacts making it easier to find employment (see section 2). We therefore include different types of sector-specific and non-specific work experience in the regressions (Table 3). Switching from model (1) to model (3), the definition of sector-specific experience becomes wider (from 3-digit to 1-digit sectors) and the measurement of sector-specific experience thus less accurate.

Table 3. Regression results – sector-specific and non-specific work experience and transition time into first regular full-time and part-time jobs.

	(1) 3-digit		(2) 2-digit		(3) 1-digit
	coef	se	coef	se	coef	se
Work experience
Sector-specific vocational training	0.334***	0.061	0.299***	0.056	0.304***	0.045
Non-specific vocational training	0.181***	0.034	0.187***	0.035	0.150***	0.038
Sector-specific experience (in 100 days)
before higher education
marginal empl.	0.009	0.016	−0.002	0.012	−0.001	0.010
regular empl.	0.023**	0.011	0.024***	0.009	0.012*	0.007
during higher education
marginal empl.	−0.049***	0.008	−0.046***	0.007	−0.037***	0.005
regular empl.	−0.002	0.016	0.000	0.015	0.004	0.011
Non-specific experience (in 100 days)
before higher education
marginal empl.	−0.002	0.005	−0.001	0.005	−0.001	0.005
regular empl.	0.013**	0.005	0.012**	0.006	0.016**	0.007
during higher education
marginal empl.	−0.026***	0.003	−0.026***	0.003	−0.026***	0.004
regular empl.	−0.002	0.008	−0.003	0.009	−0.007	0.010

Notes: * significance at the 0.1 level, ** significance at the 0.05 level, *** significance at the 0.01 level; robust standard errors in parentheses. All models include time effects, individual characteristics, study-related variables and mobility types.

The regression results in Table 3 suggest that work experience acquired in regular employment before enrollment enables graduates to enter the labor market faster after studies. By far the strongest positive effect is observed for sector-specific vocational training. A sector-specific apprenticeship training increases the hazard of starting the first job by 39.7% at the 3-digit level. The corresponding percentage for a non-specific apprenticeship is much lower and amounts to 19.8%. An apprenticeship accounts again (see previous section) for the largest proportion of the experience effect induced by regular employment before enrollment: the effect of specific experience is significant in all models and there is some indication of a beneficial influence of non-specific experience which is, however, lower than the effect of specific regular work experience before enrollment.

In contrast, the coefficients for all types of regular employment during higher education are insignificant in the three models in Table 3. Almost all types of marginal employment show a negative correlation with the probability of taking up employment after graduation. The corresponding regression results for the time before studies are not significant. During studies, this adverse effect is even larger for sector-specific experience than for the non-specific one.²⁵ Thus, for jobs carried out while studying, we do not find an indication that beneficial effects of sector-specific information and contacts could outperform the negative effects on efforts for job search and/or studies.

There are good reasons to assume that the magnitude of positive effects increases as sector-specific experience becomes more precise. However, the size of the coefficients does not significantly change across the three models (1-digit to 3-digit level) in Table 3. There is some weak indication that the quality of the match might matter for sector-specific vocational training.

4.4 Work experience during the last months before graduation

Besides, we examine whether the work experience graduates gained during the last months prior to final exams has a different effect on the speed of the university-to-work transition than experience gained earlier during higher education. For this purpose, we cumulate all employment episodes in marginal or regular employment during the last 3, 6 and 12 months before graduation, respectively. Accordingly, we take work experience graduates have made during the rest of their studies into account. The regression results in Table 4 show for regular employment insignificant coefficients for the 6- and 12 month-period before graduation, respectively. These two outcomes are in line with the overall insignificant net effect of regular employment during enrollment in models (2) and (3) in Table 1 (section 4.2). However, we find for regular employment in the 3 month-period before final exams a significant strong negative correlation with the probability of taking up employment after graduation. This result indicates that students who work shortly before their final exams seem to have no or little time for job search. Negative direct and indirect effects associated with labor market entry in the job search theory seem to clearly outweigh positive signal and network effects of these last jobs carried out before leaving university.

Table 4. Regression results – work experience during 3-, 6- and 12-month periods before graduation and transition time into first regular full-time and part-time jobs.

	(1) 3-month period before graduation				(2) 6-month period before graduation				(3) 12-month period before graduation
	coef	se	coef	se	coef	se	coef	se	coef	se	coef	se
Experience (in 100 days)
before higher education
marginal empl.	−0.000	0.004	−0.000	0.004	−0.001	0.004	−0.001	0.004	−0.001	0.005	−0.001	0.005
regular empl.	0.013***	0.005	0.013***	0.005	0.013***	0.005	0.014***	0.005	0.014***	0.005	0.014**	0.006
x months before graduation
total	−1.150***	0.072			−0.467***	0.040			−0.211***	0.023
marginal empl.			−1.269***	0.079			−0.520***	0.048			−0.237***	0.038
regular empl.			−0.290**	0.133			−0.091	0.063			−0.050	0.032
during higher education (without last x months)
marginal empl.	−0.000	0.003	0.002	0.003	−0.001	0.003	0.001	0.003	0.001	0.003	0.004	0.004
regular empl.	0.016*	0.008	0.004	0.008	0.017*	0.009	0.004	0.008	0.019**	0.009	0.006	0.009

Notes: * significance at the 0.1 level, ** significance at the 0.05 level, *** significance at the 0.01 level; robust standard errors in parentheses. All models include time effects, individual characteristics, study-related variables and mobility types.

For marginal employment, the coefficients are significantly negative in models (1) to (3) in Table 4. The magnitude of this effect increases the closer the considered period is to the examination date. Hence, marginal employment exhibits the strongest negative correlation with the probability of labor market entry for the 3 month-period. Moreover, this coefficient is a much larger in absolute size than the corresponding estimate for regular employment. Temporary student jobs – mainly to finance studies – seem to have the strongest adverse effects on study and/or search effort shortly before the date of certification. This finding is in line with a job search framework that assumes a trade-off between study and search effort on one side and working activity on the other. These effects are most pronounced in the final stage of academic studies.

4.5. Control variables

Finally, we will briefly discuss the regression results for the control variables and the evidence on duration dependence. The findings indicate that women and foreigners usually take longer to find their first regular job in Germany after graduation. While the correlation with age is insignificant, the coefficient of study length points to a prolonged period of non-employment after graduation for graduates who take a rather long time to finished studies. Individuals graduating from a university of applied sciences have a higher likelihood of entering the labor market fast than those leaving regular universities. This difference is plausible because universities of applied sciences provide more practically oriented curricula and often offer cooperation with companies during studies (e.g. internships, dual study courses). Beyond valuable contacts to firms, employers might therefore expect a shorter phase of skill adaption of these students (see Jacob and Weiss 2010).

Regarding the type of degree, the results indicate that the probability of entering the labor market is 30% lower for graduates with other degrees relative to master/diploma graduates. These are degrees such as magister artium, which are mainly awarded in subjects that provide less occupation-specific curricula.

In line with previous studies on entry duration (see, for example, Haak and Rasner 2009; Böpple 2010; Biggeri, Bini, and Grilli 2001), we find a great variance across fields of study (see Figure 3). Relative to the reference group (economics and business administration), graduates of agricultural sciences, psychology, social sciences and geography/meteorology show a lower likelihood of exiting unemployment. However, pharmaceutical graduates have an even lower probability (74%) of finding their first regular employment than these groups of graduates. One reason may be the specific examination rules in this subject. A ‘practical year’ is common in this course of studies, which concludes with an additional examination.²⁶ In contrast, mathematics and natural sciences graduates show a higher probability of finding a job after graduation than the reference group. Skill shortages for technical professions on the German labor market may enable faster labor market entry of graduates in these related fields of study (IW 2021; BMWK 2022).

Figure 3. Effect of field of study on time to first job. Source: university panel linked to the IEB of IAB, own calculations.

Figure 4 shows the effects of year of graduation on the likelihood of finding a job. The probability of finding a regular job after graduation between 1996 and 2002 did not significantly differ from the likelihood in 2007 (reference year). However, from 2003 to 2006, when unemployment rates in Germany were relatively high, it was more difficult for university graduates to enter the labor market. The likelihood of ending unemployment after studies was also much lower in 2009 during the economic and financial crisis.

Figure 4. Effects of graduation year on the likelihood of finding a job. Source: university panel linked to the IEB of IAB, own calculations.

The estimate of the share parameter of the Weibull distribution (implied $p$) is larger than 1 (see Table 1), pointing to a positive duration dependence. Therefore, labor market entry of the graduates becomes more likely as the time elapsed since final examinations increases. The positive duration dependence might be caused by graduates who adjust their job search intensity and reservation wage after a significant period of unsuccessful search due to financial necessity or the threat of stigma effects. All model specifications also consider unobserved heterogeneity. The estimated frailty variance θ is significant at the 5% level in most models, pointing to important within-student correlation.

4.6. Limitations

Yet, we acknowledge that our empirical analysis has several limitations. First, it cannot be entirely ruled out that both work and migration experience correlate with other unobserved characteristics of graduates such as motivation and ability. For this reason, the method allows assertions on a correlation between work and migration experience on one side and entry duration on the other. Second, the transferability of our results to other national contexts depends at least to some extent on whether the country relies on comparable education and employment systems especially regarding the close links of the education and employment system, and the opportunity to gain practical work experience during (dual) vocational training before studies. Comparable countries for education systems and the education-employment linkage include Austria, Denmark, Switzerland and the Netherlands.

Third, the data base for our study enabled us to study very detailed different types of working experience. Yet, one may argue that working while studying determines reservations wages of students when looking for a certain job after the completion of study. Unfortunately, we have no information on either the search behavior or the reservation wage of the graduates.

Fourth, although most universities in Germany are located outside large urban agglomerations, it is important to point out that the results shown should only be considered in the context of medium-sized university regions. All universities included in the sample are located in these smaller, mostly semi-urban areas, so it remains an open question whether the results can also be applied to graduates from large metropolitan areas.

5. Conclusions

We extend previous duration analyses of university-to-work transitions by focusing particularly on the effects of spatial mobility and work experience of graduates on the probability of entering the labor market. Our study shows that immobile graduates take up the first regular job more rapidly than mobile graduates. However, this result is mainly caused by those migrant groups who probably cannot make use of location-specific knowledge and networks. Hence, our findings support the hypothesis that location-specific knowledge and contacts might accelerate labor market transitions.

In particular, an apprenticeship before studies and taking up a job at a previous employer encourage a rapid career entry. These findings are in line with important signaling and network effects. Typical student jobs (marginal employment), especially during studies, tend to have a dampening effect on the speed of labor market entry. These adverse effects seem to be particularly strong shortly before graduation. In contrast, regular employment might be a door-opener to relevant labor market knowledge and networks for desired jobs. However, this beneficial effect of regular employment is likely counteracted by having less time for studying and job searching if work experience is gained during studies.

Public investments in higher education are under a high pressure of legitimacy. Therefore, a major concern is that labor market absorption of graduates should take place shortly after the completion of studies in order to decrease the risk of skill-deterioration and to ensure direct returns. Our main findings suggest that the matching process after graduation might benefit from an improved access to labor-market-relevant knowledge and contacts. Labor market entry after graduation might therefore benefit from establishing more opportunities to acquire this information – over and above existing measures in order to improve university-to-work transitions of graduates. As our study relates to smaller university regions being located outside of large urban agglomerations which often suffer from outward migration of young high skilled workers such opportunities offered e.g. by higher education institutions and labor market actors might be a promising approach to sustain labor market entry in the region of studies and to retain local graduates.

Universities have already invested much effort in this direction. For instance, local fairs between firms and students have become commonplace in most higher education institutions in Germany. However, many measures still lack the inclusion of more practical elements and learning opportunities. This may apply particularly to universities, where work experience is still often perceived as an extra-curricular element. One possibility is the inclusion of dual courses in the curricula. These courses may include the transfer of theoretical knowledge but also practical components and use of skills. In this respect, one option may be to combine knowledge transfer and practical work experience in firms. This may also enable students to better handle the trade-off between working on the one hand and efforts towards study and job searching on the other. This issue is of great importance for students. A recent national student survey revealed that 59.5% of higher education students in Germany had a paid job in the winter semester 2017/2018 (Studitemps GmbH and Maastricht University 2019). The trade-off might be a major concern especially for those students who rely heavily on student jobs to finance their studies. More opportunities to combine theoretical and practical elements may enable these students in particular to benefit from improved possibilities to establish more useful contacts than in pure student jobs.

Future research on individuals’ motivations and strategies underlying their career starts might provide more insights on the duration of job entry and the role of mobility, location-specific knowledge and labor market networks in this context. More detailed information on periods and intensity of job searching, on the channels used to obtain information on vacancies, on graduates’ expectations for their desired job and on their spatial preferences would be helpful to learn more about the underlying mechanisms.

Acknowledgment

The authors would like to thank two anonymous referees and the editor, Colin Green, for helpful remarks and suggestions. We have also benefited from useful discussions with seminar participants at the Winter Seminar 2019 of the Gesellschaft für Regionalforschung (GfR). Furthermore, we would like to thank Ingo Liefner for helpful remarks and suggestions. The usual disclaimer applies.

Disclosure statement

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

Notes

1 In the winter semester 2017/18, 59.5% of the students in Germany carried out side jobs during higher education. On average, these jobs make up 40% of their monthly budget and may therefore contribute to a great extent to the financing of the livelihood of many graduates (Studitemps GmbH and Maastricht University 2019).

2 In the empirical analysis, we consider that some graduates are directly hired by the firm they worked for during studies because employers might use student jobs as a screening device. These young workers receive a job offer and probably manage to enter the labor market very quickly although their search effort is likely zero.

3 In the model, students have an infinite horizon and they know the job offer arrival function, the cost function and the wage offer distribution. However, they have no information on the time of job offer arrivals and the corresponding wages in advance. Van der Klaauw and Van Vuuren (2010) assume that graduates maximize the expected present value of future utility, which is influenced by the costs of job search, arrival rate, discount rate, and the wage or welfare benefits, depending on whether the graduate accepts a job offer or becomes unemployed after graduation. We refrain from a more detailed description of the formal model and refer to the presentation by Van der Klaauw and Van Vuuren (2010).

4 Arcidiacono (2004) and Kinsler and Pavan (2015) provide evidence for the relevance of corresponding selection processes when choosing the field of study.

5 We are grateful to a referee for suggesting this additional argument.

6 Our theoretical arguments and our empirical model refer to job search within the country in which the young workers graduate. We do not consider international migration to take up a first job after graduation. Mobility therefore means leaving the region of study and moving to another labour market region in Germany.

7 Moreover, there is also an argument for reverse causality. Teichert et al. (2020), for instance, note that the pressure to migrate may rise as the length of a residence spell in the university region increases if the duration reflects the length of unsuccessful job search in the university region. A longer transition period might thus also cause migration.

8 We refrain from developing a formal model since this is beyond the scope of the present paper, which focuses on the empirical analysis of graduates’ labor market entry.

9 We assume that $\lambda$ increases as grades improve. The grades might capture the effects of various study-related factors that influence the likelihood of receiving a job offer. We do not discuss in detail the effects of graduate characteristics $x$ on $\lambda$ because $x$ is thought to represent distinct attributes whose effects on $\lambda$ are supposed to differ.

10 However, we might assume that this specific effect of work experience gained during studies on the transition time arises only during the first few months after graduation. We are grateful to a referee for this suggestion. Yet, as we identify net effects in the empirical analysis, it is not possible to provide evidence on this supposition.

11 E.g. Waldorf and Yun (2016) or Iammarino and Marinelli (2015) interpret migration in the context of a reduced probability to get an education-job mismatch, i.e. overeducation.

12 To be precise, the hazard rate is the probability divided by time and therefore may be larger than one. In the continuous case, it can vary between zero and infinity and is rather a rate than a probability. When we speak of probabilities later in the results section, it should be understood against this background.

13 The six universities included in the analysis are medium-sized higher education institutions which offer each a broad range of study fields, and attract on average 44% of the graduates in our sample from their own university region (Table A2).

14 We also estimated the models with a minimum duration of six and twelve months. The results confirm the findings discussed in this paper.

15 The data at hand do not allow us to control for the intensity and timing of job searching in our models. A significant number of graduates start searching for jobs before final examinations (Böpple 2010; Van der Klaauw and Van Vuuren 2010). In the empirical model, we cannot differentiate between early job search before graduation and job search that starts after final exams. However, our empirical model assumes (like most of the other duration analyses) that only after the date of certification the graduates are at risk of taking up a first regular employment.

16 We exclude graduates with a bachelor degree from the analysis because at least in Germany most of them do not immediately enter the labor market, but pursue a Master’s degree. Furthermore, the two groups have very different requirements for entering the labor market and are therefore not comparable. Another reason is that we cannot subsequently observe graduates with a bachelor’s degree in the data when they continue studying for a Master’s degree.

17 Most graduates in human medicine and dentistry complete studies with a professional doctoral degree, and internships of two years after studies are obligatory for teachers and graduates in law.

18 For a complete list of all control variables, definitions and summary statistics, see Tables A 1, A 2 and A3 in the Appendix.

19 For additional information on the German education system, please see Appendix A4.

20 There are 401 NUTS 3 regions in Germany, which consist of urban (‘kreisfreie Städte’) and rural counties (‘Landkreise’).

21 In this study, we refer only to internal migration of higher education graduates after labor market entry in Germany. We cannot observe external migration when graduates take up a job abroad after studies.

22 Our results are in line with findings in Baerts et al. (2021) and the studies discussed in their literature review. However, the authors focus on internships and do not further differentiate the type of work experience as we do. Their overview points to robust evidence on beneficial effects of different types of internships on various labor market outcomes. The considered studies use a wide variety of research designs and quantitative methods, suggesting that the influence of the chosen method on the results is negligible.

23 Unfortunately, we cannot control for work experience that graduates gained abroad because information on employment in other countries is not available in our administrative data. This might introduce a measurement error that will in particular affect work experience of graduates who obtained their university entry certificate abroad. Assuming that our information on work experience is subject to a systematic measurement error the estimated effect of work experience on the probability of entering the labor market will suffer from an attenuation bias. However, we are confident that this measurement problem should not severely affect our regression results because the corresponding group of graduates is rather small and it mainly concerns work experience acquired before enrollment.

24 However, the beneficial influence of previous employers could also point to a human capital effect, i.e. firm-specific knowledge.

25 A test shows that the difference between the estimates for marginal employment during higher education for sector and non-sector specific experience is significant. The results indicate that marginal employment generally exhibits a negative relationship irrespective whether it is sector-specific or not.

26 We have estimated the regression models without graduates in the pharmaceutical field. This does not change the regression results significantly.

Appendices

Table A1. Description of explanatory variables.

	Personal characteristics
	Female	1 if female, 0 if  at time of graduation
	Age²	Age (at time of graduation) squared divided by 100
	Foreigner	1 if foreign graduate, 0 if German graduate
	Spatial mobility before studies	1 if not studying in home region, 0 otherwise
	Spatial mobility after studies	1 if not working in study region, 0 otherwise
	Mobility types	Immobile Staying migrants Migrants Repeat migrants Return migrants
	Higher education
	Type of university	1 if studying at university, 0 if studying at university of applied sciences
	Examination grade	Sufficient Satisfactory Good Excellent/very good
	Study length	Number of semesters
	Type of degree	1 if other degrees (e.g. state examination), 0 if master/diploma
	Field of study	Agricultural sciences Humanities Geography/meteorology Mathematics/computer science Natural sciences/engineering Psychology Social sciences Economics & business administration Other subjects
	Work experience
	Vocational training	1 if graduate completed a vocational training before studying, 0 otherwise
	Experience (in 100 days)	Marginalª/regular employment work experience before studies Marginalª/regular employment work experience during studies
	Previous employer	1 employer-specific work experience before/during studies, 0 otherwise

Table A2. Summary statistics – mobility and control variables.

	Obs.	Mean	Std.	Min.	Max.
Personal characteristics
Female	19,860	0.44	0.50	0	1
Age at graduation	19,860	27.37	2.26	22	35
Age² at graduation	19,860	753.98	127.98	484	1,225
Foreigner	19,860	0.04	0.19	0	1
Mobility before studies	19,860	0.56	0.50	0	1
Mobility after studies	19,860	0.66	0.47	0	1
Mobility type
Immobile	19,860	0.22	0.42	0	1
Staying migrant	19,860	0.12	0.32	0	1
Migrants	19,860	0.22	0.41	0	1
Return Migrants	19,860	0.11	0.31	0	1
Repeat Migrants	19,860	0.33	0.47	0	1
Higher education
Type of university: University	19,860	0.84	0.36	0	1
Exam grade
Excellent/very good	19,860	0.25	0.43	0	1
Good	19,860	0.56	0.50	0	1
Satisfactory	19,860	0.12	0.33	0	1
Sufficient	19,860	0.06	0.25	0	1
Study length	19,860	11.43	3.29	0	20
Field of Study
Agricultural sciences	19,860	0.04	0.18	0	1
Humanities	19,860	0.16	0.37	0	1
Geography/Meteorology	19,860	0.04	0.19	0	1
Mathematics/Computer science	19,860	0.09	0.29	0	1
Pharmacy	19,860	0.03	0.17	0	1
Natural sciences	19,860	0.13	0.33	0	1
Psychology	19,860	0.06	0.24	0	1
Social sciences	19,860	0.05	0.22	0	1
Business administration & Economics	19,860	0.40	0.49	0	1
Others	19,860	0.00	0.07	0	1
Type of degree
Diploma/Master degree	19,860	0.85	0.35	0	1
Other degrees	19,860	0.15	0.35	0	1

Source: university panel linked to the IEB of IAB, own calculations.

Table A3. Summary statistics – work experience.

	Obs.	Mean	Std.	Min.	Max.
Vocational training
total	19,860	0.20	0.40	0	1
sector-specific (3-digit)	19,860	0.03	0.18	0	1
non-specific (3-digit)	19,860	0.16	0.37	0	1
sector-specific (2-digit)	19,860	0.04	0.19	0	1
non-specific (2-digit)	19,860	0.16	0.36	0	1
sector-specific (1-digit)	19,860	0.07	0.26	0	1
non-specific (1-digit)	19,860	0.13	0.33	0	1
Work experience (in 100 days)
Total experience: before studies	19,860	1.96	4.01	0	44.28
ME experience: before studies	19,860	1.17	3.10	0	34.61
RE experience: before studies	19,860	0.78	2.58	0	44.28
Total experience: during studies	19,860	4.83	5.84	0	34.26
ME experience: during studies	19,860	4.22	5.58	0	33.62
RE experience: during studies	19,860	0.60	1.71	0	25.19
ME experience: before studies, sector-specific (3-digit)	19,860	0.08	0.88	0	34.61
RE experience: before studies, sector-specific (3-digit)	19,860	0.11	0.96	0	44.28
ME experience: during studies, sector-specific (3-digit)	19,860	0.62	2.47	0	31.73
RE experience: during studies, sector-specific (3-digit)	19,860	0.13	0.92	0	25.19
ME experience: before studies, non-specific (3-digit)	19,860	1.09	2.96	0	33.33
RE experience: before studies, non-specific (3-digit)	19,860	0.68	2.39	0	40.96
ME experience: during studies, non-specific (3-digit)	19,860	3.60	5.16	0	33.62
RE experience: during studies, non-specific (3-digit)	19,860	0.47	1.44	0	22.73
ME experience: before studies, sector-specific (2-digit)	19,860	0.12	1.03	0	34.61
RE experience: before studies, sector-specific (2-digit)	19,860	0.13	1.10	0	44.28
ME experience: during studies, sector-specific (2-digit)	19,860	0.76	2.72	0	31.73
RE experience: during studies, sector-specific (2-digit)	19,860	0.15	0.96	0	25.19
ME experience: before studies, non-specific (2-digit)	19,860	1.06	2.90	0	33.33
RE experience: before studies, non-specific (2-digit)	19,860	0.65	2.32	0	40.96
ME experience: during studies, non-specific (2-digit)	19,860	3.46	5.07	0	33.62
RE experience: during studies, non-specific (2-digit)	19,860	0.46	1.42	0	22.73
ME experience: before studies, sector-specific (1-digit)	19,860	0.22	1.34	0	34.61
RE experience: before studies, sector-specific (1-digit)	19,860	0.26	1.53	0	44.28
ME experience: during studies, sector-specific (1-digit)	19,860	1.28	3.32	0	31.97
RE experience: during studies, sector-specific (1-digit)	19,860	0.24	1.16	0	25.19
ME experience: before studies, non-specific (1-digit)	19,860	0.95	2.75	0	33.33
RE experience: before studies, non-specific (1-digit)	19,860	0.52	2.06	0	40.96
ME experience: during studies, non-specific (1-digit)	19,860	2.94	4.74	0	33.29
RE experience: during studies, non-specific (1-digit)	19,860	0.37	1.26	0	22.73
ME experience: during studies, without last 3 months	19,860	4.01	5.38	0	32.72
RE experience: during studies, without last 3 months	19,860	0.59	1.68	0	24.60
Total experience: 3-month period before graduation	19,860	0.23	0.37	0	0.90
ME experience: 3-month period before graduation	19,860	0.21	0.36	0	0.90
RE experience: 3-month period before graduation	19,860	0.02	0.10	0	0.90
ME experience: during studies, without last 6 months	19,860	3.78	5.18	0	31.97
RE experience: during studies, without last 6 months	19,860	0.57	1.63	0	24.60
Total experience: 6-month period before graduation	19,860	0.49	0.72	0	1.80
ME experience: 6-month period before graduation	19,860	0.45	0.70	0	1.80
RE experience: 6-month period before graduation	19,860	0.04	0.21	0	1.80
ME experience: during studies, without last 12 months	19,860	3.29	4.76	0	30.92
RE experience: during studies, without last 12 months	19,860	0.51	1.52	0	24.60
Total experience: 12-month period before graduation	19,860	1.02	1.36	0	3.60
ME experience: 12-month period before graduation	19,860	0.93	1.34	0	3.60
RE experience: 12-month period before graduation	19,860	0.09	0.42	0	3.59
Previous employer	19,860	0.14	0.35	0	1

Notes: ME: marginal employment, RE: regular (part-time or full-time) employment.

Source: university panel linked to the IEB of IAB, own calculations.

A4: German education system

The German education system provides various options to pursue further vocational and/or academic education after leaving school. The direct transition from general schooling to university is the acquisition of a higher education entrance qualification and enrollment at university. Normally, pupils complete their university-entrance diploma at the upper secondary school and begin their studies at a regular university. Another possibility is to study at a university or a university of applied sciences after the completion of (dual) vocational training, given that the trainee has a university entrance qualification. Then, these graduates obtain a double qualification, a vocational degree and a tertiary degree. A vocational training normally takes two to three years, and during this time the apprentices are employed by a training company. Hence, they have already gained practical work experience during vocational training before starting their studies at university. In 2013, a quarter of the apprentices starting a vocational education in Germany had a university entrance diploma and would have been therefore also entitled to study. In 2011/2012, 11% of all German first-year students at university had already completed a vocational training prior to higher education (BMBF 2015). However, considering all first-year students at German universities and universities of applied science, 22% had already completed successfully a vocational training when entering higher education (Scheller, Isleib, and Sommer 2013).