Introduction

In recent years, Europe has witnessed a revival of liberal arts education (LAE), a model commonly seen in general undergraduate studies in the United States (Abrahám, 2023; van der Wende, 2011). The country that went the furthest in integrating LAE into its higher education system is the Netherlands, where ten university colleges have been established within reputable research universities since 1998. Prompted by the Bologna initiative and the desire to differentiate the massified and prematurely specialized higher education system, Dutch LAE colleges developed as internationally oriented, publicly funded programmes offering three-year liberal arts and science bachelor’s degrees (Cooper, 2018; Dekker, 2017; University College Deans Network [UCDN], 2017).

Dutch university colleges are characterized by several unique features that distinguish them from conventional bachelor’s programmes. Firstly, university colleges are general academic programmes that are not professionally oriented. These programmes offer a broad, interdisciplinary education that focuses on preparing students for advanced academic studies rather than immediate entry into the labour market. Their radically open curricula allow students to combine courses from a wide range of fields, resulting in a self-tailored academic profile. Furthermore, LAE programmes are distinguished by their student-centred learning environment involving small-scale teaching, extensive student-faculty interactions, diverse assessment methods, and a strong academic community. Lastly, university colleges are also different in terms of access. Admission to a LAE programme involves a selection procedure, limiting the enrolment to around 250 students a year per college. In contrast to this, conventional bachelor’s programmes in the Netherlands are typically more professionally focused and monodisciplinary, with a fixed curriculum structure, large-scale teaching, and non-selective admission (Cooper, 2018; Dekker, 2017; UCDN, 2017; van der Wende, 2011).

As a novel educational development and an exception to the norm of specialized undergraduate curricula, the emergence of European LAE bachelor’s programmes raises a number of important questions. One particular area of debate is the labour market value of LAE. For a long time, liberal arts degrees have been disputed and even ridiculed on the grounds of their alleged professional irrelevance. Critics often portray LAE as ‘learning for learning’s sake’ that does not prepare for a specific career, producing unemployable graduates who lack practical skills. Particularly in the US, colloquial remarks criticizing the liberal arts as ‘worthless courses that offer no chances of getting people jobs’ (Kingkade, 2013) can often be heard.

In responding to these critiques, LAE supporters point out that this educational model equips students with a range of 21st-century skills, allowing them to develop into ‘creative, critical thinkers and problem solvers who can cooperate in teams and communicate across the boundaries of languages, cultures and disciplines’ (van der Wende, 2013, pp. 296–297). In today’s volatile and rapidly evolving world of work, these skills and abilities are highly valuable assets that can help graduates to cope with change and complexity (van Damme, 2016). In the view of LAE proponents, therefore, liberal and professional education are not mutually exclusive (Gombrich, 2016).

The place of university colleges within the Dutch higher education system distinguishes the discourse surrounding them from the more prevalent discussions on LAE in the US. While the prevailing view in the US tends to narrowly associate LAE with studies in the arts and humanities, the Dutch LAE model also encompasses social science subjects and scientific disciplines, allowing students to choose a concentration in one of these three major areas. As a result, US-based studies examining the labour market implications of LAE mostly compare graduates from applied and technical fields to those majoring in liberal arts (Humphreys & Kelly, 2014; Schneider & Sigelman, 2018; Rajecki & Borden, 2011). Another prominent theme in the US is the ‘small LAE college versus big research university’ debate, where LAE advocates emphasize the benefits of a more intimate and personalized educational experience (Seifert et al., 2008; Pascarella et al., 2005). Yet, liberal arts majors are also available at large US universities, implying that the type of institution and chosen fields of study are not necessarily indicative of LAE (Rossman et al., 2020).

Consequently, identifying what constitutes a LAE programme proves more challenging in the US compared to the Dutch higher education landscape. In this sense, it is important to recognize that the primary focus of this paper is the comparison between general academic, interdisciplinary, flexible university college programmes and subject-specific, monodisciplinary, fixed conventional bachelor’s degrees. Additionally, one should keep in mind the contrasting norms regarding the timing of students’ transitions from higher education to the workforce in the two countries. In the US, it is more common for students to enter the labour market right after obtaining their bachelor’s degrees. In the Netherlands, on the other hand, the vast majority of students choose to pursue a master’s degree before starting their careers. According to earlier estimates, around 80–85% of students with a research university (WO) bachelor’s degreeFootnote 1 continue their education by enrolling in a master’s programme (Allen & Belfi, 2020).

Unlike the US, where academics, policy-makers, and the general public are continuously engaged in discussions about the value of the liberal arts, a wider debate on European LAE—including its labour market repercussions—is largely missing, and research on this topic is still scarce. Although there is a growing body of literature advocating the relevance of the LAE model in preparing students for work and life in the 21st century, evidence in support of this notion remains mostly rhetorical. The only empirical investigation of European LAE conducted so far is the 2017 survey of Dutch university college alumni. This research showed that only 7% of LAE graduates who entered the labour market were unemployed, while 93% held a job. Furthermore, around three-quarters of employed graduates expressed that they are ‘(very) satisfied with their job’ and have ‘(very) many career opportunities’ (Research Centre for Education and the Labour Market [ROA], 2018).

In spite of its favourable findings, the ROA study did not investigate how LAE graduates compare to their peers with conventional, discipline-focused bachelor’s degrees, providing insufficient evidence to strongly support the economic case for LAE. The lack of empirical evidence in the discourse on the economic value of European LAE constitutes a major research gap that needs to be filled in order to better understand the link between this educational model and labour market preparedness.

The present article aims to fill this gap by examining whether the employment outcomes of Dutch LAE graduates are comparable to those of their peers with a subject-specific undergraduate background. This study is an exploratory analysis, intended to provide preliminary insights rather than definitive conclusions. The research question it seeks to answer is: How do university college graduates fare in the labour market compared to their peers with a conventional bachelor’s degree from a Dutch research university? Through this comparison, the paper shows that holding a university college degree is not associated with any distinct advantages or disadvantages in the job market.

This study makes three main contributions. Firstly, a reliable account of the capacity of LAE programmes to prepare students for the labour market is an important aspect of their evaluation that is still missing. Although university colleges have been highly praised for their teaching quality and commitment to academic excellence, many students and their parents still worry that a LAE degree might expose them to unemployment risks (Godwin, 2015). In light of the atypical nature of LAE and its relative newness in the Dutch context, there is a strong need for emphasizing the connection between LAE degrees and professional life.

Furthermore, as a marked exception to mainstream higher education, Dutch LAE programmes serve as a great case to examine the relative weight of the bachelor’s degree in determining the employment outcomes of master’s degree holders. Although the proportion of graduates with a master’s degree is continuously increasing (OECD, 2021), existing studies mainly focus on examining the effect of having a bachelor’s versus master’s degree as the highest attained level of education. By comparing the labour market outcomes of master’s graduates from Dutch universities who hold different types of bachelor’s degrees, the present research also addresses this wider issue, which has not been sufficiently studied before.

Finally, as a case study of educational innovation, this research also adds to the growing debate on what kind of education is needed to optimally prepare graduates for the challenges of working in the 21st century.

The paper proceeds as follows. Section 2 discusses the theoretical mechanisms that may explain the relative labour market advantages and disadvantages associated with a LAE degree. Section 3 presents the empirical methods, data, and variables used in the study. The results of the statistical analyses are presented in Sect. 4. Section 5 discusses the findings and concludes the paper.

Theoretical Background and Assumptions

Examining the relationship between a university college degree and occupational attainment is complicated by the fact that the vast majority of LAE graduates do not enter the labour market directly after graduation. In fact, as many as 89% of them pursue further studies, usually by undertaking a master’s degree (ROA, 2018). This trend is not unique to LAE graduates but holds true for all research university bachelor’s programmes in the Netherlands, although with slightly lower percentages ranging from 80 to 85% (Allen & Belfi, 2020). Consequently, the effect of LAE on employment outcomes may be indirect rather than straightforward, and therefore difficult to assess.

What is more, because of their curricular breadth, university colleges produce graduates whose disciplinary profiles can vary significantly, ranging from natural sciences to the humanities. It is widely acknowledged in the literature that the field of study is a strong determinant of labour market performance. Especially when it comes to earnings and employment rates, graduates majoring in science and technical fields generally tend to perform better than their peers with degrees in social sciences and humanities (see, for instance, Reimer et al., 2008). Therefore, the career outcomes of LAE graduates are also likely to be heterogeneous, reflecting the different disciplinary choices that students make with regard to their major and master’s degree programme.

To address these issues, the paper focuses on graduates who completed a master’s programme at a Dutch research university, accounting for the effect of different academic fields. This should ensure the comparability of their labour market outcomes, allowing to assess the indirect effect of having a different type of bachelor’s degree. This approach is guided by the implicit assumption that the differences between LAE bachelor’s programmes and conventional undergraduate degrees would translate into different labour market outcomes of their students. This general assumption can be further divided into three sub-hypotheses.

Theoretically speaking, there are three main reasons why LAE graduates might have different employment outcomes than their non-LAE peers with equivalent master’s degrees. First of all, the relative labour market performance of LAE graduates can be understood as a result of the distinctive human capital they acquired through studying liberal arts. This explanation is in line with the human capital theory (Becker, 1964), according to which labour market outcomes are essentially determined by one’s stock of knowledge, skills, and abilities. Accordingly, the fact that the skills and knowledge offered by university college degrees differ from those of conventional undergraduate programmes should have consequences for the employment outcomes of their graduates.

Secondly, a potentially significant contributing factor concerns the selectiveness of LAE programmes, which may be associated with a difference in their students’ abilities, academic interests, and motivation, independently influencing job prospects and performance. Selectivity may also produce labour market benefits by sending a signal of prestige and high value to employers. This expectation follows from the signalling approach (Spence, 1973), which stresses that employers can only infer the human capital of potential employees, rather than being able to directly observe it.

Thirdly, labour market signalling might also be driven by employers’ unfamiliarity with LAE. Due to the novel, atypical, and small-scale nature of university college programmes, their degrees might be faced with a lack of visibility and recognition in the Dutch labour market, resulting in distrust and negative expectations among employers.

Hence, the theoretical mechanisms explaining the relative labour market success of LAE graduates relate to three main factors—university-acquired human capital, selectivity, and employers’ unfamiliarity with LAE. The following subsections provide a more detailed discussion of their possible implications.

The Skills Hypothesis

One may assume that the labour market outcomes of LAE graduates would differ because their bachelor’s degree equipped them with a distinctive set of skills and knowledge. Unlike conventional bachelor’s programmes, university college degrees are not designed to prepare students for a specific profession. Rather, their main purpose is to lay down a broad foundation of academic knowledge and skills, supplementing it with in-depth expertise that students acquire in their chosen area of concentration. This should enable LAE students to develop into well-rounded intellectuals who are also sufficiently prepared for further specialized studies. By postponing specialization, therefore, a LAE degree focuses on the development of general academic knowledge and skills. The liberal arts skillset could be both an advantage and a disadvantage in the labour market.

The economic argumentation in support of the liberal arts typically stresses the increasing demand for generic skills. Theoretical and empirical studies have linked LAE with the acquisition of skills and abilities such as lifelong learning (Jessup-Anger, 2012; Seifert et al., 2008), critical thinking (Dekker, 2020), interdisciplinary awareness, creativity and innovation (Dekker, 2017; Ming et al., 2023; van Damme, 2016), and self-directedness (Claus et al., 2018). In their comprehensive study, Pascarella et al. (2005) found that the pedagogical practices and experiences associated with LAE significantly affect students’ growth in a range of cognitive and personal outcomes, including reading comprehension, writing skills, and openness to diversity and challenge. It is often argued that these kinds of generic or 21st-century skills are crucial for success in the contemporary workplace, enabling graduates to readily respond to changing economic conditions and be prepared for work in a wide variety of jobs and careers (Gombrich, 2016; van Damme, 2016; van der Wende, 2013). Overall, the dominant premise in the LAE literature is that a high(er) level of generic skills should positively affect the labour market performance of LAE graduates, providing them with an advantage over pure specialists.

There are, however, several important caveats to this assumption. Firstly, it is important to keep in mind that generic skills are also developed in specialized higher education programmes. As explained by Perkins and Salomon (1989), generic skills are always learned within a specific context and function in contextualized ways. For this reason, it would be wrong to contend that these skills can only be learned in general programmes. Consequently, one would have to assume that due to their learning environment, university colleges do a better job in teaching generic skills compared to traditional programmes. Furthermore, it may be argued that there are certain types of (generic) skills vital for labour market success that are exclusively provided by LAE. This may be the case with interdisciplinary skills—the ability to combine knowledge from various disciplines and approach problems from multiple perspectives (Dekker, 2017; van Damme, 2016), creativity and personal initiative (Kovačević et al., 2023), as well as (lifelong) learning skills, or the capacity to acquire new knowledge and master new skills quickly (Kovačević, 2022).

Secondly, the assumed importance of generic skills in the labour market is a subject of ongoing debate, with multiple studies providing evidence that occupation-specific skills still have the primacy (Humburg & van der Velden, 2015; Rajecki & Borden, 2010, 2011). In this regard, a major shortcoming of LAE is that students acquire relatively less specialized skills. Since a broad undergraduate curriculum comes at the expense of subject-specific education, it is expected that in the master’s phase LAE graduates can make up for their relative lack of field-specific knowledge and skills (Kovačević, 2022). That way, university college graduates are supposed to strike a balance between the general liberal arts foundation, which provides the flexibility necessary to navigate labour market changes in the long run, and the specific knowledge necessary to develop their professional expertise and make a smoother transition to their first job (Humburg & van der Velden, 2017; Rözer & Bol, 2019). In the Netherlands, however, where the linkage between the higher education system and particular occupations is traditionally tight, a generalist bachelor’s degree may still be disfavoured by employers (Di Stasio & van de Werfhorst, 2016). Thus, the relative lack of occupation-specific skills could negatively impact the labour market outcomes of university college graduates.

The Selectivity Hypothesis

The differences in labour market outcomes of LAE and non-LAE graduates could also stem from the selectiveness of university colleges. Their selection procedure usually involves the assessment of prior academic performance, letters of recommendation from former teachers, a motivation letter and personal essay, as well as an admission interview (Reumer & van der Wende, 2010).Footnote 2 Selectivity can affect graduates’ labour market prospects through two main channels.

Firstly, the success of graduates from selective programmes can be influenced by pre-existing abilities and characteristics that students bring with them (see, for example, Dale & Krueger, 2014). Since university colleges are among the very few bachelor’s programmes in the Netherlands that have a selective admission policy, their student population may differ from that of conventional programmes. On the one hand, stricter selection procedures might result in an above-average student body in terms of academic capabilities and motivation, potentially leading to favourable employment outcomes regardless of the bachelor’s degree. On the other hand, university colleges typically attract individuals who are less decided about their future career paths compared to peers from conventional bachelor’s programmes (Kovačević et al., 2023), as well as students with distinct academic interests that cannot be accommodated in a monodisciplinary programme (Dekker, 2021). These differences might also be reflected in labour market outcomes, although the exact implications remain uncertain.

Secondly, attending a selective programme can give graduates an advantage in the job market by signalling a high level of competence to employers. Since the productivity of a potential employee cannot be directly observed, employers rely on educational credentials to guide their hiring decisions. Employers may value the rigorous admission process associated with selective programmes, viewing it as an indicator of high ability. Additionally, the belief that selective programmes provide education of superior quality can signal to employers that their graduates are well-prepared and possess valuable skills. This can positively influence employers’ perceptions of candidates from selective programmes and increase their likelihood of being considered for job opportunities (Drydakis, 2016).

For this effect to manifest in the context of Dutch university colleges, employers would need to be aware of the difference in selectivity between LAE and regular programmes, perceiving it as an indicator of students’ abilities and/or teaching quality. In that case, they might prefer to hire from more selective programmes, thereby improving the relative labour market prospects of LAE graduates. In the Netherlands, however, the higher education hierarchy is known to be rather flat, with minimal differentiation in resources and quality between universities (Allen & Belfi, 2020). Given the highly standardized education system in the country, it is questionable whether the selectivity of LAE programmes could act as a strong labour market signal.

The Unfamiliarity Hypothesis

Contrary to the positive signals associated with selectivity, the newness and unconventionality of LAE in the Dutch context might convey a negative signal to prospective employers. Although LAE is relatively well established in the Netherlands compared to the rest of Europe, it still represents a major exception to the norm of specialized, monodisciplinary curricula, enrolling a tiny fraction of the overall student population (Dekker, 2017). As employers may not be familiar with the LAE model, it is possible that they distrust university college degrees when making hiring decisions. The lack of information on the programme’s curriculum and resulting skillset could make a LAE graduate appear as a riskier choice for employers than a candidate from a recognized study programme.

Employers are risk-averse and typically want to minimize uncertainty in their hiring decisions. The biggest risk they face arises from the non-linear relationship between individual skill levels and productivity. Productivity experiences a sharp decline when skills drop below a certain minimum level. Accordingly, the negative impact of below-average skills on productivity is greater than the positive impact of above-average skills. As underperformance is more detrimental than overperformance is beneficial, over-skilled workers cannot compensate for the under-skilled. Due to this inability to pool risks, avoiding individuals below the minimum skill level becomes the foremost concern for employers during the hiring process (Glebbeek & van der Velden, in press).

This could explain why employers may be reluctant to hire from LAE programmes. The limited information on LAE fuels their risk aversion, adding to the uncertainty regarding the fulfilment of skill requirements. To minimize the risk, employers may favour candidates from familiar programmes, where they can be sure that all students have attained the minimum skill level, even if these programmes have a lower average quality compared to LAE.

Summary and Research Objectives

Overall, the examined theoretical arguments do not clearly suggest that LAE graduates would outperform their peers or be outperformed by them in the labour market. While both positive and negative effects are possible, the lack of empirical research on LAE in the Dutch context makes it difficult to hypothesize which of the effects would prevail. The actual rigour of selectivity at university colleges, Dutch employers’ awareness of LAE, as well as their demand for generic, liberal arts skills, are factors that have not been thoroughly studied before. Furthermore, it is questionable to what extent the bachelor’s degree influences the employers’ hiring decision when the job candidate is a master’s degree holder. For these reasons, there is not enough ground to assume that LAE graduates would experience a clear labour market advantage or disadvantage over their peers with subject-specific bachelor’s degrees.

With this in mind, the following sections set out to empirically examine how LAE graduates and conventional bachelor’s degree holders compare with regard to a diverse set of labour market outcomes. The main goal of the empirical analysis is to establish whether there are any differences between the two groups in terms of these outcomes. While subsequent discussions may delve into the possible role of the mechanisms presented in this section, the objective is not to empirically distinguish between these effects. Rather, the empirical investigation focuses on determining whether LAE graduates experience better or worse employment outcomes compared to their peers, while the underlying causes of these potential (dis)advantages will remain within the realm of theory. It is important to stress that the methods employed in this study are designed to explore associations and relations, not to infer causality. Hence, its results should be interpreted as suggestive evidence rather than definitive causal conclusions.

Methods

Data and Sample

This study uses data from the Dutch National Alumni Survey (De Nationale Alumni Enquête, NAE, formerly also known as WO Monitor). The NAE is a biennial survey that collects job market performance data from alumni of master’s programmes at Dutch universities approximately one and a half years after their graduation. Data from four rounds of the NAE, conducted in 2011, 2013, 2017, and 2019, were included in the analysis. The 2015 and 2021 rounds had to be left out because they lacked data on the respondents’ bachelor’s degrees, making it impossible to identify LAE graduates.

To ensure comparability in terms of undergraduate studies, the initial sample was restricted to respondents who held a WO bachelor’s degree from a Dutch research university and were admitted to the master’s programme on this basis. Furthermore, only respondents who completed their master’s programme as full-time students and were under the age of 30 at the time of the survey were included in the sample. Finally, respondents who did not meet the criteria to be considered part of the labour force were excluded from the analyses.

The final sample comprised 14,933 respondents, of whom 210 were LAE graduates. Individuals who reported working less than 12 hours per week or not having a job but being available for work (992 in total, 21 LAE graduates) were only analysed for employment status. The remaining analyses were restricted to respondents who reported working at least 12 hours per week (13,941 in total, 189 LAE graduates).

Variables

The following employment outcomes served as dependent variables in the study:

  • Employment status (0 = currently without a paid job but actively looking for work; 1 = has a paid job, regardless of weekly working hours);

  • Quick career start: time between graduation and first paid job (0 = more than 3 months; 1 = 3 months or less);

  • Vertical match, operationalized as working in a job that requires a WO master’s degree or higher (0 = vertical mismatch; 1 = vertical match);

  • Horizontal match, defined as the alignment between the respondent’s field of study and the field required for their work (0 = horizontal mismatch, i.e. entirely different study field or no particular field required; 1 = horizontal match, i.e. own field only or own and related fields required);

  • Combination of vertical and horizontal match, consisting of three categories (1 = full match; 2 = only vertical but no horizontal match; 3 = rest);

  • Hourly wage from regular work, adjusted for international price differences and inflation (2015 USD). OECD’s PPP and CPI indicators (OECD, 2023a, 2023b) were used to make wages comparable across countries and years. To mitigate the influence of implausible extreme values, the top and bottom 1% of earners were removed from the analysis. A natural logarithm transformation was then applied.Footnote 3

Since earnings can serve as a rough indicator of productivity in the labour market, following the human capital theory, any disparities in hourly wages between the two groups could be attributed to their distinctive skillsets. In other words, any significant differences in hourly wages would relate to the skills hypothesis. On the other hand, employment status, time to first paid job, and vertical and horizontal matches represent outcomes that are more closely linked with the concept of signalling. Hence, any differences in these outcomes could be seen as influenced by labour market signals associated with the selectivity and unfamiliarity hypotheses.

The main independent variable in this study was the type of bachelor’s degree (0 = conventional programme; 1 = university college). Additionally, to account for factors concerning the graduates’ personal background, work, and study characteristics that could potentially influence labour market outcomes and/or participation in LAE programmes, the following control variables were included:

  • Survey round cohort (2011, 2013, 2017, and 2019. These were treated as separate dummies, with 2011 as the reference category);

  • Gender (0 = female; 1 = male);

  • Age;

  • Field of master’s studies (1 = Agriculture and natural environment; 2 = Nature and technology; 3 = Healthcare; 4 = Economics; 5 = Law; 6 = Behaviour and society; 7 = Language, culture and education. These were treated as separate dummy variables, with ‘Agriculture and natural environment’ as the reference category)Footnote 4;

  • STEM vs. Social: a binary variable indicating the categorization of master’s study fields into two groups (0 = STEM: study fields 1–3; 1 = Social: study fields 4–7);

  • Self-reported master’s GPA (grading scale: 0–10, minimum passing grade: 5.5);

  • Graduates’ activities during their master’s studies: internship in the Netherlands, internship abroad, work experience in a relevant field, attending courses abroad, participation in an honours programme, experience in management or administration, starting a company (all dummies: 0 = no; 1 = yes);

  • Prior education: the educational background preceding the respondents’ WO bachelor’s studies (1 = HBO: university of applied science; 2 = VWO: pre-university education; 3 = HAVO: senior general secondary education; 4 = other. These were treated as separate dummies, with ‘other’ as the reference category);

  • Further education: following other academic programme(s) after master’s degree (0 = no; 1 = yes);

  • Migration status, indicated by the country of residence at age 16 versus the country of work/job search (1 = native in the Netherlands; 2 = foreigner in the Netherlands; 3 = native in another country; 4 = foreigner in another country. These were treated as separate dummies, with ‘native in the Netherlands’ as the reference category);

  • Country of work (0 = non-OECD; 1 = OECD).Footnote 5

Table S1 in the supplementary material provides a descriptive overview of the main variables under investigation, as well as two additional variables, employment sector and company size, which were not included in the statistical models. This table reveals some noteworthy differences between LAE and non-LAE graduates. Firstly, the number of LAE graduates has been increasing over time as a result of the establishment of more university colleges in recent years. Furthermore, university colleges have a higher proportion of foreign students, reflecting the international orientation of these programmes. This is why it was necessary to control for migration status, as previous research has shown that having an immigrant background can negatively affect labour market outcomes (see, for example, Gheasi et al., 2017). Likewise, graduate employment dynamics and job market characteristics can vary significantly between developed and developing countries (Battu & Bender, 2020), making it important to account for these distinctions. Finally, a striking difference is that twice as many LAE graduates are employed at universities, working towards a PhD. This discrepancy might lead to an underestimation of their relative labour market position, as salaries in academia tend to be lower than those in industry. However, a higher proportion of academic jobs can also be seen as a positive factor, indicating highly-skilled work that provides greater career possibilities in the future. In response, the study includes a variable for further education.

Analyses

Regression analyses were performed in Stata 17 to examine how holding a LAE bachelor’s degree relates to various labour market outcomes. Hourly wage was analysed using OLS regression (regress command), while multinomial regression (mlogit and listcoef commands) was used to assess the combination of vertical and horizontal match. Logistic regression (logit command) was used for all other outcome variables. The analyses employed robust standard errors. Listwise deletion was used to handle missing values.

Three regression models were estimated for each outcome. A base model (Model 0) including only the bachelor’s programme type independent variable was estimated first. This model served as a foundation for assessing the effect of holding a LAE bachelor’s degree, against which the results of more complex models could be compared. Models 1 and 2 included the full set of controls. The only difference between these models was that the second one used the binary study field categorization, incorporating it into an interaction term with the LAE degree variable. This made it possible to compare the performance of LAE graduates and their non-LAE peers within two broader areas: natural sciences and engineering (‘STEM’), and social sciences and humanities (‘Social’). To accomplish this, the margins command with the dydx option was utilized.

To ensure the robustness of the main findings, additional analyses were conducted using propensity score matching (PSM). PSM is a statistical technique that can help reduce bias by matching individuals based on their probability of receiving treatment, quantified as a propensity score. Using this score, LAE graduates (treatment group) were matched with comparable graduates from regular bachelor’s programmes (control group), excluding non-matched individuals from the analysis. The average treatment effect on the treated (ATT), indicating the average difference in outcomes between the two groups, was then estimated.

PSM aimed to balance the two groups on observed characteristics that may impact the graduates’ career outcomes. However, it is important to note that PSM does not account for unobserved factors. Theory and prior research suggest that propensity score models should ideally include covariates that are both strongly associated with treatment selection and the outcome, or those that affect only the outcome (Steiner et al., 2010; Austin, 2011). Given the lack of pre-programme variables related to LAE participation in the available data, as well as the fact that the selection mechanism into LAE is not entirely clear, PSM could not adequately address the issue of selectivity. Therefore, its results should not be interpreted as evidence of causality. Instead, the intention behind its use was to serve as an additional analysis, especially in view of the unbalanced sample, with a relatively small number of LAE graduates compared to a large number of graduates from other bachelor’s programmes. By matching on variables that additionally affect employment outcomes over and above the type of bachelor’s degree, PSM provided an alternative to controlling for these variables in regression analysis.

One-to-one, nearest-neighbour matching without replacement was performed via the psmatch2 command in Stata 17. To ensure sufficient overlap between the treatment and control groups, the common option was included, resulting in the exclusion of treatment observations with propensity scores above the maximum or below the minimum propensity scores of the controls. The ties option, which matches treatment observations with multiple nearest neighbours in case they have identical (tied) propensity scores, was also utilized. Bootstrapping with 1,000 replications was used to obtain the standard errors and confidence intervals for the estimated ATT. The balance of covariates between the treatment and control groups was evaluated using the pstest command. PSM was conducted for all outcomes except the vertical and horizontal match combination, as psmatch2 does not support multinomial dependent variables. Since PSM does not rely on functional form assumptions, the hourly wage outcome variable was not log-transformed. The covariates included in the propensity score estimation model align with those used in the main regression model.

Results

Regression Analysis

Logistic regression results are presented in Tables 1 and 2. Table 3 shows the coefficients for LAE across all pairs of vertical and horizontal match combination categories. Full results of the multinomial regression for match combination can be found in Table S2 in the supplementary material. OLS regression results for log hourly wage are displayed in Table 4.

Model 0 shows significant results for two outcomes: employment status and match combination. Firstly, conventional bachelor’s degree holders are more likely to be employed than LAE graduates. Secondly, the relative probability of having a vertical match and horizontal mismatch rather than a full match is greater for LAE graduates compared to graduates from conventional bachelor’s programmes. LAE graduates are also more likely than their non-LAE peers to experience only a horizontal mismatch versus only a vertical mismatch or a full mismatch. These results reflect the data presented in Table S1, suggesting that the observed differences, although statistically significant, are relatively minor.

The findings from Model 1 reveal that the type of bachelor’s degree has a significant effect only on one labour market outcome: employment status. Holding all other variables constant, the odds of being employed for a LAE graduate are (1 – e− 0.667) × 100 = 48.7% lower compared to a graduate from a regular bachelor’s programme. Although this may come across as quite substantial, the absolute difference is not very big (90.5% versus 94.0%, see Table S1). No statistically significant effects were observed in other outcomes. After adding the controls, the effect of LAE on match combination ceased to be statistically significant. Hence, there are no differences between LAE graduates and their non-LAE peers in terms of making a quick entry into the labour market, vertical match, horizontal match, match combination, and hourly wage.Footnote 6

In addition to the main findings, the control variables also yielded noteworthy results. Master’s GPA exhibited a strong positive impact on all employment outcomes. For example, a one-point increase in GPA raises the odds of being employed by 44.3% (e0.366 = 1.443) and the odds of a vertical match by 112.8% (e0.755 = 2.128). Likewise, doing an internship in the Netherlands and acquiring relevant work experience during studies seem to be highly important factors for a successful entry into the labour market. In line with a widely recognized pattern, being male is associated with a 3.4% higher hourly wage. Other outcomes are not significantly affected by gender. Both in the Netherlands and other analysed OECD countries, graduates of foreign descent earn less than their native-born counterparts.

Through Model 2, it was possible to conduct separate analyses of labour market outcomes for graduates in two master’s study areas: STEM and Social. Similarly to Model 1, the only statistically significant differences were found with regard to employment status. This is illustrated by Figure 1, which shows the average marginal effects of bachelor’s degree type on employment status within the two fields. Compared to their peers from a conventional bachelor’s programme, LAE graduates with a master’s degree in STEM have a probability of being employed that is 0.133 or 13.3 percentage points lower. In Social fields, the difference in the probability of being employed between LAE and non-LAE graduates is much smaller (-2.9 percentage points) and statistically insignificant. For all other outcomes, additional analyses on the interaction term did not show any statistically significant differences between the two groups in STEM and Social master’s field subsets.Footnote 7

Fig. 1
figure 1

Average marginal effects of a LAE vs. conventional bachelor’s degree on employment status by master’s study field (STEM/Social), with 95% CIs

Full size image

Robustness Check: Propensity Score Matching

To scrutinize the findings, PSM was also applied. As a first step, it was necessary to check whether an adequate balance of covariates has been achieved. Tables S5-S9 in the supplementary material show that the matching process resulted in a considerable reduction in bias, as evidenced by a mean decrease of approximately three to four times. In the vast majority of cases, the standardized bias is below or slightly above 10%. Most importantly, as indicated by the last two columns in these tables, there are no statistically significant differences in the mean values of covariates between the treatment and control groups after matching. This demonstrates that the matching process was effective in creating a well-balanced control group.

ATT estimation results are presented in Table S10 in the supplementary material. These findings are largely consistent with those from the regression analyses. The only difference pertains to employment status, where the ATT for attending a university college versus a regular undergraduate programme is statistically insignificant, with a p-value that is close to the 5% threshold. However, when the ATT is estimated separately for graduates in STEM and Social master’s study fields, there is a statistically significant negative difference of 10 percentage points in the former. In the matched sample, 87.7% of LAE graduates with STEM master’s degrees are employed, compared to 97.9% of their non-LAE counterparts. For graduates with Social master’s degrees, on the other hand, the difference in employment status is negligible (only 1 percentage point) and statistically insignificant. This confirms the results reached by the regression analyses. For all the remaining outcomes, the differences between LAE graduates and their peers from conventional bachelor’s programmes are not statistically significant. Overall, the results remain consistent across methods.

Table 1 Logistic regression results for Model 0 and Model 1
Full size table
Table 2 Logistic regression results for Model 2
Full size table
Table 3 Multinomial regression results for mismatch combination: The effect of LAE across outcome pairs
Full size table
Table 4 OLS regression results for log hourly wage
Full size table

Discussion and Conclusion

The results of the statistical analyses suggest that the labour market performance of LAE graduates is on par with their peers who pursued a regular undergraduate programme. The only negative result was found among graduates with a master’s degree in the STEM domain. In their case, holding a LAE bachelor’s degree proved to have a detrimental effect on obtaining employment. For all other labour market outcomes, no statistically significant differences were found, neither positive nor negative. These findings add valuable insights to the preceding theoretical discussion.

Firstly, the findings indicate the absence of a selectivity effect, as LAE graduates did not surpass their peers in any of the outcomes. On the one hand, this implies that students’ abilities may not be the primary determinant in the selection process for university colleges. Drawing on prior research, it can be argued that other factors, such as the students’ specific academic interests and undecidedness regarding career choices (Dekker, 2021; Kovačević et al., 2023), take precedence when it comes to (self-)selection into LAE. On the other hand, the evidence supports the notion that the selectiveness of university colleges does not generate strong positive signals in the labour market. Overall, selectivity-related labour market advantages appear highly unlikely in the context of Dutch LAE.

Secondly, the data indicates that holding a LAE degree adversely impacts one’s chances of attaining employment in the STEM sector. The preference of STEM sector employers for graduates with a subject-specific bachelor’s degree raises questions about the underlying contributing factors. Building on the research of Di Stasio and van de Werfhorst (2016), it can be inferred that employers in STEM occupations disfavour LAE graduates because they associate a generalist undergraduate degree with a lack of job-specific skills. Their unfamiliarity with LAE is likely to add distrust, resulting in a negative signal.

While STEM employers may view LAE graduates as lacking specialized skills, this does not necessarily reflect their true skill level. The fact that LAE graduates who secure employment in the field perform equally well as their non-LAE peers suggests that the problem lies at the entry point, where employers’ perceptions and biases might play a crucial role. The absence of statistically significant differences in hourly wages seems particularly relevant in this context, as a skill deficiency would likely affect this outcome. Hence, although the degree to which LAE graduates lack specialized skills remains uncertain, negative signalling appears to be the main factor hindering their entry into STEM jobs. A recently completed qualitative study on Dutch employers’ perceptions of LAE supports this conclusion, indicating that employers in specialized jobs often attach greater significance to having a discipline-focused bachelor’s degree in a related field (Kovačević, 2024). Either way, the fairly large negative difference in employment suggests that LAE graduates may be faced with obstacles when looking for jobs in STEM professions.

Thirdly, the study’s findings contribute to the discussion on the labour market significance of generic skills. In this regard, the results show that the LAE skillset produced a neutral effect, without notable positive or negative consequences. Even if LAE students have better-developed generic skills than their non-LAE peers, or possess certain unique skills such as interdisciplinarity, these did not result in additional advantages in the labour market. Alternatively, it is possible that the provision of generic skills is not exclusive to LAE, as other programmes may be equally good at teaching these skills. However, it is important to note that these conclusions only pertain to the short-term impact of generic skills, as the study looked at early career outcomes, one and a half years after graduation. The literature frequently stresses the value of generic skills in effectively responding to long-term shifts in the labour market (Rözer & Bol, 2019). As revealed in the study by Pascarella et al. (2005), a considerable portion of the intellectual and personal development fostered by liberal arts education tends to manifest itself more strongly in the long run. In that sense, LAE proponents like to point out that a liberal arts education gives people the skills ‘that will help them get ready for their sixth job, not their first job’ (Zakaria, 2015, p. 76). Nevertheless, the dataset used in this study only included short-term labour market outcomes, which is one of its limitations. To fully examine the impact of generic skills, future research should explore how LAE graduates fare over the longer term.

A major limitation of this study is the relatively small number of respondents from university colleges in the final sample. The limited representation of university college graduates in the sample is primarily due to the inherently small-scale nature of Dutch LAE programmes and the surveys’ low response rates. However, it is important to note that NAE surveys were distributed to all master’s graduates in the Netherlands, ensuring that every possible respondent from LAE programmes was approached. To increase the sample size, all available cohorts were aggregated, resulting in the inclusion of the maximum possible number of LAE graduates. Thus, despite its small size, the available sample represents the largest achievable for this study.

The study’s focus on alumni of Dutch master’s programmes is another limiting factor, as a sizeable portion of LAE graduates pursue their master’s studies abroad. One may argue that the brightest students are more likely to undertake graduate studies in a foreign country, and that overlooking them compromises the validity of the analysis. However, the high international ranking and quality of Dutch higher education challenge this claim. In addition, focusing on master’s graduates from Dutch universities facilitated the comparability of their outcomes, which would not have been possible otherwise.

The study is also limited by the absence of information on socio-economic background in the NAE survey. Socio-economic circumstances have been found to affect labour market outcomes (Burke, 2016; Flap & Völker, 2008), and could also influence entry into university colleges, which have higher tuition fees than regular bachelor’s programmes (NUFFIC, 2022).

Lastly, it is crucial to recognise the exploratory nature of this study, whose design may not have captured all relevant factors influencing graduates’ employment outcomes. In particular, the lack of pre-programme variables in the NAE dataset limited the ability of our analyses to produce strong causal evidence. The findings presented here should therefore be seen as preliminary evidence that can serve as a starting point for further investigation, rather than definitive conclusions. Additional research is essential to validate and build upon these initial observations.

Despite these limitations, the study’s findings support the overall conclusion that LAE does not lack the capacity to prepare students for the labour market. This counters common criticisms and undermines the belief that LAE graduates are faced with limited career prospects. Considering the lack of prior empirical inquiry, these findings are particularly valuable to LAE stakeholders, highlighting that the unconventional nature of LAE should not be a source of concern. As a distinctive and recent addition to the Dutch higher education system, LAE programmes have proven to be a successful innovation, catering to a specific subset of students who seek greater flexibility to explore diverse fields of study and pursue a self-tailored academic pathway.

While recognizing the absence of negative associations between a university college degree and early employment outcomes, it is also important not to overstate the merits of the LAE model. In that sense, LAE supporters should exercise caution, especially when promoting the notion of increasing need for generic skills in the workplace, as there is not enough evidence to support such claims. In the ongoing debate on optimal educational paradigms for the 21st century, LAE should not be perceived as the sole and definitive solution, but rather as an alternative pathway leading to the same destination.

Examining university colleges as an exception among Dutch undergraduate programmes has also provided preliminary insights into the impact of the bachelor’s degree on the employment outcomes of master’s degree holders. The discovery of a neutral effect of undergraduate studies calls for further research on the role of different combinations of bachelor’s and master’s degrees, along with their associated skillsets, in determining labour market success. To complement the results of this study, future research should explore this issue in other countries, particularly within the context of post-Bologna degree structure in European higher education. This could provide valuable perspectives on how the division of university studies into two cycles influences the transition of graduates from university to work and their further career paths.

Finally, the results of this study hold practical implications for LAE programmes, offering guidance for improvement. To enhance the employment prospects of their students within STEM, university colleges may consider ways to increase the visibility of the LAE educational model in this segment of the labour market. To achieve this, university colleges should actively connect and engage with STEM employers. This could include participation in industry-specific events, conferences, and career fairs, as well as organizing panels, forums, or information sessions for employers. Additionally, university colleges could facilitate collaborative projects between LAE students and industry partners, and establish relationships with companies to provide internships. Such initiatives should counteract any negative signals associated with LAE and enable university college graduates to make a smoother entry into STEM professions.