Introduction
Whether institutional change accelerates, resists, or merely reflects broader ideological currents of globalization lies at the core of political science. While much research focuses on immigration and trade policy (see, for example, Margalit Reference Margalit2012; Hainmueller and Hopkins Reference Hainmueller and Hopkins2015; Ahlquist et al. Reference Ahlquist, Copelovitch and Walter2020; Kim and Margalit Reference Kim and Margalit2021), the ideological effects of globalization-friendly higher education policy remain understudied. This gap is significant. As universities reinvent themselves in response to global pressures, they train record numbers of sophisticated voters – if not cultural, political, or financial elites – at the peak of their political socialization. In this paper, I test how institutional responses of higher education (HE) to globalization redirect students’ politics.
Globalization’s economic and cultural forces have intensified both competition and co-operation in HE, driving convergence on the Anglo-American model and sparking debate over the kind of ‘global citizenship’ this transformation is producing (see, for example, Marginson and Van der Wende Reference Marginson and Van der Wende2007; Stromquist and Monkman Reference Stromquist and Monkman2014). The ‘neoliberal hypothesis’ holds that universities have primarily adjusted to economic globalization, thereby narrowing the student experience to career preparation, while the ‘humanist hypothesis’ holds that universities have fostered intercultural exchange, thereby raising students’ awareness of the increasingly global character of social injustices (Marginson and Van der Wende Reference Marginson and Van der Wende2007; Bourn Reference Bourn2011; Balarin Reference Balarin2016; Kraska et al. Reference Kraska, Bourn and Blum2018; De Wet et al. Reference De Wet, Bacher, Wetzelhuetter and Nnebedum2024). These hypotheses have their counterparts in the public arena. Conservatives and progressives blame universities for seemingly contradictory reasons, charging, on the one hand, that they are ‘prioritizing social justice over academic achievement’, and, on the other hand, that they are only acting to ‘prepare their students, largely uncritically, for the market’.Footnote 1 I put these arguments to causal test in the European context.
Identifying the effect of globalization-friendly institutional change in HE requires us to depart from the empirical literature on education and political attitudes in choosing both ‘y’ and ‘x’. Instead of focusing on the ideological positions on the left–right axis central to domestic politics, our ‘dependent variable’ captures the importance individuals attach to cultural values related to either personal achievement or global justice. Our ‘independent variable’ captures institutional variation within attainment levels, distinguishing university graduates in settings (a) and (b), instead of variation in attainment levels within institutional settings, distinguishing university graduates and non-graduates.Footnote 2
I leverage institutional variation in HE from a unique case study. With the explicit goal of turning the EU into ‘the most competitive and dynamic knowledge-based economy in the world’, the European Commission coordinated the harmonization of twenty-nine countries’ degree structures, credit systems, and quality criteria, following a standardized template inspired by the Anglo-Saxon model emphasizing students’ employability and mobility as key drivers of the EU’s global competitiveness (see, for example, Neave and Maassen Reference Neave and Maassen2007). While HE typically responds to globalization through decentralized, incremental change, the ‘Bologna Process’ prompted signatories to fully comply with the agreed policy template between 1999 and 2010 (see, for example, Witte et al. Reference Witte, Huisman and Purser2009).
Taking advantage of the staggered adoption of the Bologna Process, I study the reform’s effects on political attitudes and labour market outcomes using a generalized difference-in-differences (DID) model (see, for example, Arold Reference Arold2024). After accounting for country and cohort fixed effects as well as country-specific time trends, the reform provides exogenous variation in individuals’ exposure to revised HE narratives and policies. This variation allows us to identify the reform’s effects by comparing differences in outcomes between cohorts (a) and (b) within the same countries who had enrolled in university before or after the implementation of the reform, with the differences between cohorts (a) and (b) in other countries that had not yet implemented the reform, and the differences between cohorts (a) and (b) who did not enrol in university.
I find that the Bologna Process had important and lasting effects on the political attitudes of European millennials. The reform significantly increased the salience of personal achievement, measured by the importance attached to social status and wealth, by approximately three to five per cent on a 0–1 response scale among treated alumni interviewed on average approximately five years after graduation. The reform, on the other hand, did not affect the salience of global justice measured by the importance respondents attach to worldwide egalitarianism and care for nature. Using a two-way fixed effects model estimator, a more conservative event study, or an even more conservative regression discontinuity design, I consistently find results supporting the ‘neoliberal perspective’.
Further analysis suggests that treated graduates prioritize personal achievement because of early ideational change at university, not because of later adaptation to socio-economic conditions (see, for example, Hainmueller and Hiscox Reference Hainmueller and Hiscox2007; Gelepithis and Giani Reference Gelepithis and Giani2020). I find that students and alumni in pre- and post-Bologna settings display similar attitudinal patterns and that the reform did not affect income, status, skills, or employment variables. These results point towards a political socialization mechanism and reveal an ironic policy failure: the functional economic narrative of the Bologna Process fed into the students’ desired but not realized gains in socio-economic status.
While the increasing involvement of international organizations such as the UN, the OECD, and the EU in shaping the policymaking, governance, and evaluation of educational institutions is well understood (see, for example, Marginson and Van der Wende Reference Marginson and Van der Wende2007; Martens and Niemann Reference Martens and Niemann2010), its political economy consequences for the formation of political attitudes are not. The Bologna Process is emblematic of how globalization has reshaped HE. In 1988, the Magna Charta Universitatum’s first principle described European universities as ‘differently organized because of historical heritage’ and ‘morally independent of all political influence and economic power’. A few years later, the ‘political influence’ of the EU, moderating the ‘economic power’ of globalization, encouraged European universities to harmonize their practices, language, and structure in line with the Anglo-Saxon model to increase both graduates’ employability and countries’ competitiveness (see, for example, Marginson and Considine Reference Marginson and Considine2000; Phillipson Reference Phillipson2001; Slaughter and Rhoades Reference Slaughter and Rhoades2004; Marginson and Van der Wende Reference Marginson and Van der Wende2007). I show that this policy adaptation, while economically rational, permanently altered the value system of European graduates.
Higher Education and Political Values
Durkheim saw the division of labour in the nineteenth century as both an opportunity for economic growth and a threat to social cohesion. In Durkheim’s view, education could help countries seize the opportunity while mitigating the threat. Through its ‘selection into employment’ function, education had to supply skills needed for the industrial production process, while through its secondary ‘socialization’ function, it had to instil the universalistic values of social solidarity needed to hold together an increasingly atomistic society. Households could pass on the appropriate general skills and particularistic values in the simple pre-modern society characterized by agriculture and mechanical solidarity. In the complex society characterized by industrial production and organic solidarity, however, the appropriate skills became specific and the appropriate values became universalistic, calling for a formal education system (Durkheim Reference Durkheim1893, Book II, Chapter 7).
While the specificities and weight of education’s selection and socialization roles change over time and across educational stages, the notion that HE’s role was to provide individuals with good jobs and polities with good citizens remained popular as free markets and liberal democracies solidified. As modernization proceeded, the selection function broadened to encompass not only the formation of real skills but also the signalling of presumed ones, while the socialization function was streamlined to specifically serve ‘democratic citizenship’ (see, for example, Arrow Reference Arrow1973; Spence Reference Spence1978; Lipset Reference Lipset1959).
Globalization further sharpened these selection and socialization functions, leading HE to reinvent itself and thus to shape the students’ experience and resulting attitudes differently. By increasing skill premia and inequality, economic globalization prompts universities to compete in supplying increasingly specific skills, professional networks, and signalling devices by imitating the practices of the most successful institutions (Acemoglu Reference Acemoglu2003; Helpman et al. Reference Helpman, Itskhoki, Muendler and Redding2016; Olssen and Peters Reference Olssen and Peters2005; Marginson and Van der Wende Reference Marginson and Van der Wende2007). Meanwhile, by highlighting the interconnected nature of first-order collective challenges, cultural globalization prompts universities to increase international academic exchange and institutional focus on worldwide inequality, conflict, patterns of human migration, and the threat of climate change (Oxley and Morris Reference Oxley and Morris2013; Killick Reference Killick2014; Siddiqui Reference Siddiqui2014).
While scholars broadly agree that globalization has profoundly affected HE systems, they disagree about how the institutional adjustments of HE to economic and cultural globalization shape students’ politics. The disagreement begins with different interpretations of the world’s structural trends, proceeds with different explanations of how HE systems adapted to them, and ends with different hypotheses of its consequences for global citizenship (Kraska et al. Reference Kraska, Bourn and Blum2018).
The neoliberal hypothesis characterizes globalization as an economic phenomenon under which knowledge is framed as a key productive asset and HE increasingly as a commodity allowing students to compete in the competitive global labour market. The increased supply of opportunities for international academic exchange is seen as a tool to further foster one’s real and presumed professional competence and payoff-relevant networks. Meanwhile, universities’ increased focus on researching and debating global injustices, climate change, and international conflict is seen as a form of lip service to the expectation that universities cultivate students’ understanding of global citizenship values (see, for example, Barnett Reference Barnett2000; Humfrey Reference Humfrey2011; Balarin Reference Balarin2016). In the critical view posited by the neoliberal hypothesis, globalization-friendly institutional change in HE is predicted to raise the salience of personal achievement but fails to enhance the students’ global citizenship values.
The humanist hypothesis characterizes HE as evolving in response to both the economic and cultural dimensions of globalization. Universities are seen as having responded to globalization by promoting intercultural dialogue, global citizenship, and critical thinking rather than solely aligning with market-driven demand. Such an adaptation could reflect the evolving ideas of HE managers optimistic about the depth and potential of initiatives fostering ‘global citizenship education’ – the ‘global’ HE counterpart of civic education during compulsory schooling (see, for example, McCowan Reference McCowan2012; Guo and Larsen Reference Guo and Larsen2014). Other times, this adaptation may simply be forced on universities facing increasingly diverse bodies of students and staff (Stromquist and Monkman Reference Stromquist and Monkman2014, chs. 2 and 6). In turn, increasingly flexible productive skills and increased cross-cultural exchange can, if balanced through appropriate institutional arrangements, jointly foster informed global citizenship – ‘an awareness of self, the world and ones position within it’ – along with the competences that support positive change (Kraska et al., Reference Kraska, Bourn and Blum2018). In this mainstream, liberal view, the competences acquired through increased cross-cultural exchange in contemporary HE, while potentially fostering careerism, should increase students’ political sensitivity to global justice issues.
Some evidence from the United States directly addressing how institutional change affects students’ attitudes backs the neoliberal hypothesis.Footnote 3 Meanwhile, most empirical work in European political science, while abstracting from institutional change, indirectly backs the humanist hypothesis.Footnote 4 However, in both cases, the empirical evidence is based entirely on conditional correlations. I now evaluate the plausibility of the neoliberal and humanist arguments within our unique case study.
Institutional Background of the Bologna Process
The institutional adaptation of HE to globalization is typically carried out gradually by national executives, subnational authorities, or universities. The exception was the ‘Bologna Process’, described as ‘the most encompassing and profound set of reforms of European higher education ever’ (Witte et al. Reference Witte, Huisman and Purser2009, p. 206), ‘one of the greatest socioeconomic reform initiatives of the last decades’ (Vögtle and Martens Reference Vögtle and Martens2014, p. 44), and a ‘significant force in unifying, modernizing, and ultimately revolutionizing global higher education’ (Piro Reference Piro2016, ch. 7, p. 286).
The Bologna Process kicked off in 1998 as an intergovernmental initiative aimed at harmonizing degree structures to favour international mobility. It quickly expanded in remit to the supranational level, led by the European Commission, with the aim of fostering employability and competitiveness (see, for example, Neave and Maassen Reference Neave and Maassen2007; Gornitzka Reference Gornitzka2007; Witte et al. Reference Witte, Huisman and Purser2009; Martens and Niemann Reference Martens and Niemann2010; Cino Pagliarello Reference Cino Pagliarello2022). The EU had long envisioned HE as falling outside its jurisdiction until the nineties, when its objective of turning itself into ‘the most competitive and dynamic knowledge-based economy in the world’, developed as part of the European Employment Strategy, was enshrined in the Lisbon Agenda. With this change, HE was tasked with leading a crucial narrative shift: from ‘Europe as a knowledge society’ to ‘Europe as a knowledge economy’. Twenty-nine European countries signed the Bologna Declaration in 1999, intended to create the integrated European Higher Education Area (EHEA) by 2010 with the goal of ‘making higher education more responsive to the requirements and challenges related to the globalisation of societies, economy and labour markets’ (Marginson and Van der Wende Reference Marginson and Van der Wende2007, p. 19).
Making sense of the complex history of the Bologna Process – with its multiple institutional actors, evolving agenda, and expanding geographic coverage – is beyond the scope of this paper.Footnote 5 I limit the focus to the changes in narratives and policies that could have shaped the political attitudes of post-Bologna European university students.
Changes in narratives and policies
The Bologna Process came about in 1999 to uphold ‘a systemic, sustained effort at making HE more responsive to the requirements and challenges related to the globalisation of societies, economy and labour markets’ (Kalvemark and Van Der Wende Reference Kalvemark and Van Der Wende1997, p. 19). Its mission was articulated through a series of declarations intended as guidelines for policy implementation. Descriptive text analysis of all the declarations issued from the onset of the Bologna Process to its agreed deadline confirms a strong bias of the reform towards economic outcomes. Employment-related keywords are used overwhelmingly more than democracy-related keywords (panel (a) in Figure 1). Perhaps more meaningfully, panel (b) in Figure 1, focusing on root words, shows that words related to ‘knowledge’, ‘development’, and ‘society’, which can be linked with both employment and democracy, co-occur much more frequently with the former. Hence, ‘knowledge’ is conceived of primarily as an economic asset, ‘development’ as an economic objective, and ‘society’ is equated with ‘economy’.
This brisk quantitative description aligns with scholarly work by educational discourse analysts who have repeatedly detailed how the Bologna Process marked a decisive narrative shift in framing HE as an economic institution – references to culture are cosmetic, aimed at mitigating the strong emphasis on students’ success (Nokkala Reference Nokkala2007; for a meta-analysis, see Wihlborg Reference Wihlborg2019).Footnote 6
Figure 1. Analysis of declarations, 1998–2010.
Note: Based on all Ministerial declarations, namely the Sorbonne Declaration, 25 May 1998; the Bologna Declaration, 18–19 June 1999; the Prague Communiqué, 18–19 May 2001; the Berlin Communiqué, 18–19 September 2003; the Bergen Communiqué, 19–20 May 2005; the London Communiqué, 17–18 May 2007; the Leuven/Louvain-la-Neuve Communiqué, 28–29 April 2009; and the Budapest/Vienna Declaration, 10–12 March 2010. Each document is available on the EHEA website, ehea.info. The keyword search focuses on skill, employability, market, competence, mobility, economic, qualification, and knowledge for ‘employment’, and democratic, cohesion, freedom, welfare, equality, justice, citizenship, participation, and solidarity for ‘democracy’. The network analysis is based on community detection algorithms, specifically the cluster_walktrap function in R. I focus on roots, as printed. Nodes represent the frequency of keywords associated with ‘Democracy’ and ‘Employment’, and edges represent the frequency of co-occurrences.
Narrative change can play a causal role in changing students’ attitudes by actively constructing the academic experience and inducing consistent policy change. The change in narratives enshrined in the Bologna Process guided policy implementation towards promoting flexible skill formation, professionalization, and mobility through changes in degree structures, the credit system, and diploma recognition protocols.
The introduction of the two-cycle degree structure with a longer/generalist undergraduate programme followed by a shorter/specialized master’s programme allows students both to sharpen skill formation and to experience mobility between the cycles. The introduction or amalgamation of the two-cycle degree structure sharply increased the number and degree of specialization of available undergraduate and master’s programmes. It also increased internal mobility for entire one- or two-year degrees and international mobility for short exchange programmes (Teichler Reference Teichler2012, p. 41).Footnote 7 Footnote 8 Moreover, while making the HE sector English-speaking was not an official objective of the Bologna Process, the homogenization of degrees and the ‘diploma supplement’ ensuring their recognition forced universities to compete in attracting students, prompting an increasing number of European universities to adopt English as a teaching language, to the detriment of learning quality and development of the national language in countries with low English proficiency.Footnote 9 Overall, the increase in the number of programmes/courses taught in English was understood by the main EU-level student unions as a market-oriented choice (ESIB 2005, p. 57).
Relatedly, the introduction of transparent metrics translating hours of work into a formal credit system (ECTS) was not perceived as a neutral attempt to clarify workload expectations. It did not go unnoticed that teachers would have modules ‘credited’ rather than ‘approved’ and students would accumulate formative ‘credits’ or ‘debits’ – that the ECTS was the educational equivalent of the euro, embodying knowledge commodification (see, for example, Maltese Reference Maltese2021; Lorenz Reference Lorenz2012). Credits functioned as a currency that allowed universities to include compulsory internships, trading off general knowledge with employment-oriented experience.Footnote 10
Based on a survey of European youths (Flash Eurobarometer 502: Expectations of Europeans for 2023 2022), panel (a) of Figure 2 shows that approximately twenty-five per cent of pre-Bologna students deemed the university to provide the best support for them to find a job. Among post-Bologna students, the share rose to thirty-five per cent, overtaking ‘the state’ and topping the list, consistent with the narrative that ‘higher education institutions, together with governments, government agencies and employers, shall improve the provision, accessibility and quality of their careers and employment related guidance services to students and alumni’ (Leuven Communiqé, 2009).
Figure 2. Evaluation of the Bologna process, by stakeholders. Data for panel (a) are from the Flash Eurobarometer 502, Youth and Democracy in the European Year of Youth. The survey question is: ‘Which of the following institutions is the best for you to find a job?’ The possibilities are those used in the histogram. I merged ‘Youth associations’ with ‘social services’ for readability. The dataset includes N = 10,884 ‘youth’ Europeans born between 1978 and 1992 as I focus on the same cohorts later used for statistical inference. Data for panels (b), (c), and (d) are from the Flash Eurobarometer 198, Perceptions of Higher Education Reforms. Each graph plots the sample mean. The survey questions are: ‘The old higher education system was better than the new one’, rescaled so that higher values indicate a preference for the Bologna system, and ‘Competition improves higher education’. Agreement could be expressed on a 0–4 Likert scale, rescaled between 0 and 1. The identification of the relevant subgroups in (b), (c), and (d) is straightforward. The dataset includes N = 5,782 HE stakeholders from the countries identified by two-digit labels in (b).
Policy implementation
While the implementation of the Bologna Process is considered a surprising success story overall, the involvement of different stakeholders in diverse pre-existing systems introduced some implementation heterogeneity. Most European countries adopted the Bologna degree structure, credit system, and quality assurance measures between 1999 and 2010; both the date and the speed of implementation varied along lines not reducible to any standard European macro area.Footnote 11
Data from the Flash Eurobarometer 198: Perceptions of Higher Education Reforms (2007) allow us to describe European academics’ perceptions of the Bologna Process by country, field of study, and academic rank. Overall, approximately two-thirds of the interviewed academics supported the Bologna Process, agreeing that ‘the new higher education system is better than the previous one’ (first survey item) and ‘competition improves higher education’. Panels (b) and (c) of Figure 2 reveal that this aggregate result is remarkably homogenous across academic ranks within European universities (from deans to PhD students) and – perhaps surprisingly given that humanities and STEM were asymmetrically impacted by globalization – across scholarly fields (applied science to humanities). The aggregate country-level differences in support are bounded between medium and high levels of support with some heterogeneity (panel (d) of Figure 2). Academics from Eastern European countries were the most enthusiastic about the Bologna Process, with approval rates often close to eighty per cent. Meanwhile, academics in French-speaking Belgium and France were equally split in supporting the Bologna Process. The degree of heterogeneity among professionals appears limited enough to be rationalized by a unified hypothesis yet substantial enough to require inference strategies that account for cross-context variation.
As documented in this section, diverse scholars agree that the Bologna Process narrative and policies were overall skewed towards fulfilling economic goals. Hence, we might expect that globalization-friendly institutional change in HE would primarily increase the salience of self-interested personal goals without substantially affecting the perceived importance of global goals among cohorts that studied under the Bologna framework. I test this hypothesis in the next section.
Data and Identification Design
Coding of the Reforms
Kroher et al. (Reference Kroher, Leuze, Thomsen and Trunzer2021) code the date when countries joined the EHEA (1999–2005), the start of the reform’s implementation (1999–2008), whether countries had a single-cycle degree system – with programmes typically lasting four or five years – or a double-cycle system with separate undergraduate and postgraduate programmes prior to the reform (0–1), and whether the reform was sharply or gradually implemented (0–1, but also as a continuous measure 1–9). I extend Kroher et al. (Reference Kroher, Leuze, Thomsen and Trunzer2021) by adding the minimum and average entry age for graduates (17–20). Our analysis of official texts led us to make minor revisions to the data from Kroher et al. (Reference Kroher, Leuze, Thomsen and Trunzer2021), as discussed in our supplementary information (SI).Footnote 12
These pieces of information allow us to identify the cohorts, defined by birth year, that the Bologna Process targeted and to control our analysis for pre-reform heterogeneity (SI, Table T.1). I assume that the intention-to-treat applies to graduates in the first cohort that entered university when the implementation started. The timeline in Figure 3 shows the implementation timeline for each country in our analysis. I focus on respondents from the countries that implemented substantial change under the Bologna Process reform for which the European Social Survey (ESS) has data. Figure 3 shows the implementation timeline for all countries included in our analysis.
Figure 3. Staggered adoption of the Bologna process.
Microdata
Our main dependent variables measure the salience of a set of core political or cultural values (Schwartz Reference Schwartz1992). I measure the salience of personal achievement using the following proxies:
-
• Personal achievement (status): ‘Being very successful is important to me. I hope people will recognize my achievements’.
-
• Personal achievement (wealth): ‘Being rich is important. I want to have a lot of money and expensive things’.
Answers range from 0 (‘Not at all like me’) to 5 (‘Totally like me’). These proxies are typically associated with certain components of individualism: power, defined as the quest for social control over people and resources, and achievement, defined as personal success through competence (see, for example, Kilburn Reference Kilburn2009; Beilmann et al. Reference Beilmann, Kööts-Ausmees and Realo2018; De Wet et al. Reference De Wet, Bacher, Wetzelhuetter and Nnebedum2024). Their correlation is ρ = 0.43, p < 0.01 in our effective sample.
I measure the salience of global justice using the following proxies:
-
• Global justice (equality): ‘It is important that every person in the world is treated equally. Everyone should have equal opportunities in life’.
-
• Global justice (environment): ‘It is important to care for nature and the environment’.
The correlation between the two standard proxies, which are also measured on a 0–5 Likert scale, is ρ = 0.28, p < 0.01 in our effective sample. In the literature, the importance of global justice is sometimes referred to as moral universalism, that is, the extent to which one’s altruism is independent of their social distance from the victims of social injustice (De Wet et al. Reference De Wet, Bacher, Wetzelhuetter and Nnebedum2024; Schwartz Reference Schwartz1992; Enke et al. Reference Enke, Rodríguez-Padilla and Zimmermann2023).
Figure 4. Salience of personal achievement and global justice, by education level and cohort.
Note: Each graph plots the average score for each dependent variable by cohort, based on ESS data. I include the same countries that will be used in the main analysis below.
Based on eleven rounds of the European Social Survey Round 1–11 Data 2024, Figure 4 plots the evolution of the average value of each dependent variable by cohort (defined by birth year, from 1945 to 1995), education level (graduate or not), and type (Bologna or not). The cohorts I focus on in the main analysis are those born in the eighties and early nineties who, depending on which country they enrolled in, may have enrolled in pre- or post-Bologna programmes.
For both proxies of personal achievement, we observe an upward trend. The importance of achieving status is consistently higher among individuals with a university diploma, while the importance of achieving wealth is slightly higher among less-educated individuals. While these variables are generally slow-moving, we observe a sharp discontinuity among graduates in pre- and post-Bologna settings. Indeed, on average, individuals who enrolled in a post-Bologna degree display higher salience of personal achievement than both individuals from the same cohort who enrolled in a pre-Bologna degree and individuals who did not enrol in university.
Conventionally, HE has been linked by political scientists to more cosmopolitan and post-materialist attitudes (see, for example, Norris et al. Reference Norris2000). As Figure 4 shows, this link still holds: university graduates attach higher importance than those with less education to global justice struggles. Focusing on graduates, Figure 4 shows that graduates in post-Bologna settings attach lower importance to global equality than graduates in pre-Bologna settings. This is less the case if we focus on the importance of nature, for which the attitudinal pattern of pre- and post-Bologna graduate cohorts is smooth.
These cohort-varying aggregate statistics suggest that the Bologna Process magnified the importance of personal achievement while slightly decreasing the importance of global justice among European graduate cohorts. The next section assesses whether institutional change in HE is causally responsible for these descriptive patterns.
Identification Strategy
I take advantage of the staggered implementation of the Bologna reform using a two-way fixed-effects (TWFE) generalized difference-in-differences model as a baseline, coupled with an event study approach and a standard regression discontinuity design (RDD). In each specification, I estimate reform effects by comparing the outcomes of cohorts – identified by year of birth relative to the year of reform implementation – eligible to enter university either pre- or post-reform implementation. In line with the literature, I assume that the date of implementation of the reform is random relative to each respondent’s date of birth, hence prospective students just happened to enrol at university before or after the Bologna Process implementation (see, for example Arold et al. Reference Arold, Woessmann and Zierow2024; Di Leo and Giani Reference Di Leo and Giani2024).
TWFE’s additional identification feature is that cohorts in countries that had not yet implemented the reform in a given year are used as counterfactuals for cohorts in countries that had implemented the reform in that same year. The key identifying assumptions are: (i) parallel trends – that in the absence of the Bologna Process, the attitudes and outcomes of university students across countries would have followed similar trajectories – and (ii) the homogeneity of treatment effects over time. The standard approach to credibly addressing the identification challenges implied by these assumptions is to combine country and cohort (Cantoni et al. Reference Cantoni, Chen, Yang, Yuchtman and Zhang2017), to which I add interview year fixed effects (FE) and a country-specific cohort trend. The country-fixed effects absorb time-invariant differences between countries – including in academic practices or structures, such as hierarchization or emphasis on generalist culture as well as in social, political, or cultural norms that, by determining pre-reform systems and levels of output, may correlate with implementation timing. On the other hand, most of the time-varying country-specific shocks one can think of, for example the 2008 financial crisis and related country-specific austerity policies or the 2020 COVID crisis, are unlikely to affect adjacent cohorts asymmetrically. The cohort fixed effects absorb time-varying differences within countries’ cohorts, eliminating the concern that the estimated reform effects simply reflect attitudes and outcomes that were trending in the country. The country-specific time trends, akin to the running variable in RDDs, account for time-varying country-specific shocks that affect close cohorts asymmetrically but smoothly. This approach allows us to account for changes that are asymmetric among countries while trending across cohorts within countries, such as political attitudes related to populism, trust, or environmental issues which could at the same time be heterogeneous at the country level while also progressively growing or decreasing across close cohorts. Including these trends means that outcomes must ‘jump’ at the pivotal cohorts for the reform effects to be significant (Arold Reference Arold2024). Finally, the interview-year fixed effects absorb time-varying differences that are constant across all units within each interview year. Together, these identification features address the concern that our estimations spuriously reflect pre-existing attitudes or local shocks. This specification can be written as
$${Y_{i,c,t}} = 1\left( {{t_{i,c}} \ge t_c^*} \right){\beta _{{\rm{reform}}}} + \left( {{t_{i,c}} - t_c^*} \right){\beta _{{\rm{Trend}}}} + {\bf{X}}_i{{\beta _{{\rm{Controls}} (i)}}} {\rm{ + }}{\alpha _{{c}}}{\rm{ + }}{\gamma _{t}}{\rm{ + }}{\varepsilon _{{{i}},{{c}},{{t}}}}$$
where Y
i, c, t
is the output variable for respondent i who enrolled in university at time t in country c, 1(t
i, c
≥t
c
*) is an indicator variable taking the value of one if the individual enrolled in the year of the reform or after, and t
i, c
− t
c
* is the country-specific trend. β
reform is the ‘intention-to-treat’ effect (ITT).
${\bf X}_{i}$
includes the year of interview (2002–22) and dummies for whether the individual identifies as female and has country citizenship. Finally, α
c
and γ
t
are country and cohort fixed effects, and ϵ
i, c, t
is assumed to be white noise. Standard errors are clustered at the country level to account for within-country correlation across cohorts.
My baseline sample focuses on individuals born between 1978 and 1992 who completed a university degree in any European country that (i) experienced substantial policy change as a result of the Bologna process and (ii) participated in the ESS at least once before and after the implementation of the reform. In the DID design, I take the UK and Israel, which did not experience major curriculum changes driven by the Bologna Process while sharing similar characteristics to our sampled countries, as ‘never treated’ countries.Footnote 13
I now discuss my approach to dealing with measurement error, bandwidth selection, external validity, the parallel trends assumption, and heterogeneity.
Threats to Identification
In the main analysis, I focus on all individuals interviewed in a country who completed at least one cycle of academic studies. The primary source of measurement error is that individual i may be inaccurately assigned to treatment or control status if i lives in country x but studied in country y, where the reform may or may not have been implemented. To assuage this concern, I run a first robustness check after deleting respondents born in a different country (labelled ‘exclude immigrants’ in the results table).Footnote 14
Individual i may also be inaccurately assigned to control status if i lives and studied in country x but had one or more gap years before enrolling. Some students delay their entry to university because of failing entry exams, personal preference, or military conscription. The share of students taking a gap year is lower for the academic programmes I consider than for the vocational ones I exclude, is minimal in some countries (for example France or the Netherlands), and is considerable in others (for example Finland or Germany). I run an alternative specification where the ITT cut-off is based on the average rather than minimum age of university entry, based on OECD data coded in SI Table T.1 (column ‘average entry age’).Footnote 15
The average completion rate for individuals entering universities in the EU was seventy per cent, displaying some country-level heterogeneity (see OECD report, 2008, p. 79). While the main analysis focuses on individuals who successfully graduated, I run an additional analysis including all adult individuals who do not hold a university diploma but who studied at least one year more than the number of years required to complete upper secondary education (ISCED level 3) – ranging from twelve to fifteen, depending on the country.Footnote 16 The increase in the effective sample approximates well the thirty per cent discrepancy between enrolment and graduation in the aggregate data (column ‘include dropouts’).
Another concern is that the reform may have causally changed enrolment. The literature is scant and has yielded mixed results on this possibility (Horstschräer and Sprietsma Reference Horstschräer and Sprietsma2015; Cappellari and Lucifora Reference Cappellari and Lucifora2009). I run an RDD to test for reform effects on both graduation and years of education for each country. I find statistically significant jumps only in Estonia (p < 0.05 ) and the Czech Republic (p < 0.1) (SI Tables T.3 and T.4). I then replicate the main analysis after deleting these countries (column ‘exclude jump’).
To ease the interpretation of the effect sizes, I dichotomize all attitudinal dependent variables around the median value of the effective sample. To test the robustness of the main findings against coarsening bias (Marshall Reference Marshall2016), I also provide estimates for the original range of the output variables, which are on a five-point Likert scale (column ‘ordinal DV’).
The main analysis uses individuals born between 1978 and 1992 to stay very close to the pivotal cohort while retaining statistical power. I run an additional TWFE specification after increasing these bandwidths by five years (column ‘long bwidths’).
The most direct change brought about by the Bologna Process was the introduction of a two-cycle degree structure with undergraduate and postgraduate programmes along with transferable credits. I follow Kroher et al. (Reference Kroher, Leuze, Thomsen and Trunzer2021) in using this criterion to establish the key reform implementation date. However, while some countries implemented the reform sharply, others did so gradually. In an additional specification, I explicitly use a continuous version of the ITT variable, rising across the period from the beginning to the end of the implementation process (column ‘gradual reform’).
One issue that remains open is that the structural drivers of enrolment may be time-varying variables that are not state-invariant. To isolate the ITT from country-varying confounders, following Muralidharan and Prakash (Reference Muralidharan and Prakash2017), I estimate a triple-difference TWFE model that includes individuals who never enrolled in university as an additional control group. The parameter of interest becomes the interaction term between whether the respondent belonged to a cohort eligible for the Bologna reform and actual enrolment in university (U i ∈ {0,1}); this leads to the following specification (column ‘Triple DiD’):
\begin{align}{Y_{i,c,t}} = &1\left( {{t_{i,c}} \ge t_c^*} \right){\beta _{{\rm{reform}}}} \times {\rm{Graduate}} + {U_i}{\beta _{{\rm{Graduate}}}} + 1\left( {{t_{i,c}} \ge t_c^*} \right){\beta _{{\rm{reform}}}} + \left( {{t_{i,c}} - t_c^*} \right){\beta _{{\rm{Trend}}}} \\&+ {{\bf{X}}_i} {{\beta _{{\rm{Controls}} (i)}}}{\rm{ + }}{{\bf{Z}}_i} {{\beta _{{\rm{Controls}} (c)}}}{\rm{ + }}{\alpha _{\rm{c}}}{\rm{ + }}{\gamma _{\rm{t}}}{\rm{ + }}{\varepsilon _{{{i}},{{c}},{{t}}}}.\end{align}
The main analysis uses eleven rounds of the European Social Survey Round 1–11 Data 2024. I draw on several additional sources for further analysis, as specified where relevant and summarized in Table T.2 of the SI.
Political Economy Effects of the Bologna Process
I now present the main findings. Table 1 presents the reform effects and key statistics, the output mean, R-squared, and number of observations across the nine specifications previously discussed based on data from the European Social Survey Round 1–11 Data 2024. We report conventional significance levels at 0.10, 0.05, and 0.01 following disciplinary conventions.
Table 1. The Bologna reform and the neoliberal student
Note: Standard errors in parentheses. + p < 0.10, * p < 0.05, ** p < 0.01. Includes data from Albania, Austria, Belgium, Croatia, Czech Republic, Denmark, Estonia, Finland, France, Germany, Hungary, Ireland, Italy, Latvia, Lithuania, the Netherlands, Norway, Poland, Portugal, Slovakia, Slovenia, and Spain. United Kingdom and Israel are ’never treated’ (see Table A1 for reforms’ coding). Controls include sex (0: male, 1: female), migration status (0: native, 1: foreign born), and a country-specific trend capturing the distance of each cohort from the pivotal one. We also include dummies for whether the country had a single-cycle degree prior to the reform, whether it had a credit system for modules, and whether it is part of the EU at any point in time. Each specification is discussed in section ‘Indentification strategy’. The ordinary least squares (OLS) naive specification is based on the effective sample of the two-way fixed effects main specification for comparability. Column 3 runs the main analysis after deleting immigrants, column 4 after deleting three countries that experienced an increase in enrolment (Czech Republic with p < 0.05, Estonia and Spain with p < 0.1), column 5 includes respondents that did not graduate but completed at least fifteen years of education, and column 6 uses the average rather than minimum entry age at university to establish the intention to treat cut-off. Self-interest and universalism have been dichotomized around the median value of the effective sample for ease of interpretation in all specifications other than ’Ordinal DV’ in column 7, where they range on a five-point Likert scale expressing increasing levels of importance. Column 8 expands the bandwidths by five years, while column 9 considers a continuous intention to treat following the coding of the reform implementation. Column 10 provides the triple-staggered difference-in-difference specification discussed in the text. The full results including all covariates can be found in four different tables, one for each DV, in the supplementary material (Full outcomes for Table 1).
Panels (a) and (b) of Table 1 show that the Bologna Process significantly and substantially increased the salience of personal achievement, proxied by the importance respondents attach to achieving status and wealth. the first column provides naive ordinary least squares (OLS) estimates. The coefficients in the second column are the baseline reform effects. The Bologna Process increased European graduates’ concern with personal status achievement, proxied by agreement with the statement ‘it is important to have success and be recognized’, by approximately five per cent on the 0–1 response scale, significant at p < 0.01, adding a substantial approximately ten per cent to the unconditional sample mean of 0.488 in the main TWFE specification.
This finding remains very stable across very different specifications. It ranges from a minimum of 4.3 per cent when I consider the triple DiD specification (last column), further reassuring us that the key test is unlikely to capture education-specific attitudinal trends, to a maximum of 6.1 per cent when I exclude immigrants (third column). The coefficient is lower when I consider the ordinal variable in the sixth column, suggesting that the effect across extreme levels is moderate relative to that taking place between intermediate values. Finally, the standard errors and p-values remain similar when I regress the output variable as originally coded.
The reform effects are of smaller magnitude with similar standard errors when I consider the importance to respondents of personal wealth achievement, captured by their agreement with the statement ‘it is important to have money and buy expensive things’, in panel (b). In this case, the estimates range from 2.9 per cent (p < 0.1) in the ninth column, where I allow for a continuous treatment, to 4.5 per cent in the third column, where I exclude immigrants. Consistently with the findings in panel (a), these effects add approximately ten per cent to the unconditional sample mean of 0.397 in the main TWFE specification, and the size of the reform effect coefficient is similar under the triple DiD specification.
With a similar level of consistency, based on the reform effects presented in panels (c) and (d) of Table 1, the Bologna Process did not significantly affect the salience of global justice. Reform effects on egalitarianism, in panel (c), are negative but non-significant. Reform effects on care for nature, in panel (d), are all close to zero. Indeed, the reform did not change the salience of global justice in any specification. SI Tables T.5, T.6, T.7, and T.8 provide the full outcomes for Table 1.
Further Robustness Checks
A first potential concern is that TWFE models aggregate all 2 × 2 DiD comparisons, potentially biasing estimates when treatment effects are time-varying, as already-treated units may serve as inappropriate controls. The Callaway-Sant’Anna (CS) estimator addresses this by excluding such comparisons (Goodman-Bacon Reference Goodman-Bacon2021; Callaway and Sant’Anna Reference Callaway and Sant’Anna2021). Event-study graphs show no significant pre-trends and similar patterns across TWFE and CS estimates, supporting the parallel trends assumption and robustness to treatment effect heterogeneity (SI Tables T.9 and Figure F.1). An even more conservative approach is to estimate local reform effects using an RDD, which has the additional advantage of relying on data-driven bandwidth and avoiding using synthetic control countries. Even in this case, I find broadly similar results (SI Tables T.10 and Figure F.2). A randomization test further confirms that our reform effects are highly unlikely to be spurious (Figure F.3).
A second concern that the TWFE model does not fully address is that policy adoption timing may be endogenous to the very outcomes under study, rather than merely non-random. Omitted variable bias may arise if differences in countries’ exposure to globalization, economic strength, investment in higher education, or political orientation influence both the timing of policy adoption and personal achievement values. Additionally, reverse causality may arise if aggregate trends in personal achievement themselves help predict when countries adopt the policy. In SI, I provide suggestive evidence that neither omitted variable bias nor reverse causality is likely to be pervasive (SI Tables T.11 and T.12).
Ideational and Socio-economic Mechanisms
The increased importance that Bologna graduates attach to personal achievement may reflect two mechanisms. The reform may also have had indirect economic effects on political values, as graduates affected by the Bologna Process may have adjusted their attitudes in response to the improved social status and economic outcomes it might have produced later in life, in line with its key premise of increasing employability (see, for example, Campbell Reference Campbell2009). However, the reform might have directly influenced European graduates’ political values through political socialization effects. Indeed, both explicit and implicit curricular changes shape the cultural norms that frame the student experience, thereby affecting the formation of political attitudes (Mendelberg et al. Reference Mendelberg, McCabe and Thal2017; Gelepithis and Giani Reference Gelepithis and Giani2025).Footnote 17
I provide strong evidence against the indirect economic mechanism by running the TWFE model from Equation 1 on a comprehensive set of socio-economic outcomes. The TWFE specification is particularly helpful here because the combination of survey year, country, and cohort fixed effects, coupled with country-specific trends, makes our estimates robust to the major labour market transformation that asymmetrically changed the labour market outcomes of millions of European youths around the years of the financial crisis. Panel (a) in Figure 5 shows that neither objective nor subjective measures of income and job quality differ significantly between the treated and control cohorts. The reform’s effects on both gross household income decile and subjective purchasing power are small and statistically non-significant. The latter estimates are based on responses to the question ‘How well are you coping with your present income?’, with responses recoded on a 0–1 scale from a four-point scale ranging from ‘Living comfortably on present income’ to ‘Very difficult on present income’. Similarly, the reform had no significant effect on job quality. This holds for both occupational prestige – derived from International Standard Classification of Occupations (ISCO) codes and converted into scores with the Standard International Occupational Prestige Scale, which ranges from twelve (for example shoeshiners) to seventy-eight (for example doctors) – and self-reported job satisfaction, measured on a five-point Likert scale.
Since the Bologna Process was primarily designed to enhance employability, this is a particularly important finding that warrants further scrutiny. Panel (b) focuses on key labour market outcomes that the reform explicitly sought to improve. It shows that the reform had no effect on a respondent’s likelihood of having experienced unemployment in the twelve months preceding the interview. Similarly, graduates affected by the reform are no likelier to be self-employed and display comparable levels of relative skill specificity, suggesting that the reform had only a minimal impact on labour market trajectories. The only notable difference in Panel (b) is that the affected cohorts report significantly higher labour supply, working approximately one additional hour per week (significant at p < .05). The full set of estimates and their robustness against all identification threats discussed in this section are presented in SI (Table T.13).
Figure 5. Socialization v. economic explanations: economic outcomes.
Note: Each graph plots the coefficient and ninety-five per cent confidence intervals from the main specification in 1. Data for the graph in panels (a) and (b) are from the ESS.
Read in conjunction with Table 1, panels (a) and (b) of Figure 5 reveal an ironic policy failure: The Bologna Process increased the salience of personal achievement without delivering corresponding gains in personal achievement itself.
The analysis supports a direct political socialization mechanism through which the Bologna Process increased the salience of personal achievement. To strengthen this claim, I analyse whether respondents’ attitudinal changes had emerged already during university years – before labour market exposure could play a role. Analysing data from Cumulative Eurobarometer, 2004-2023 2023, which focuses exclusively on young people, confirms that the emphasis of the Bologna Process on personal achievement likely took root during university years. Among the twelve political values examined, only ‘self-fulfilment’ is significantly affected by the reform (p < 0.05), while values such as equality, solidarity, tolerance, and human rights remain unchanged (SI, T.14). Similarly, the European Social Survey Round 1–11 Data 2024 allows us to imperfectly distinguish a minority of individuals who are still studying from the majority who are alumniFootnote 18 and demonstrate that there is no significant difference between Bologna students and alumni (SI, T.15).
Discussion
We have seen that the political economy legacy of the Bologna Process diverged from its original aims. Designed to make higher education across Europe more compatible and competitive to foster exchange and employability, the Bologna Process causally increased the affected cohorts’ valuation of status and wealth – without improving their actual socio-economic position. This seemingly counterintuitive finding highlights how shifting organizational cultures, not mere socio-economic outcomes, may have inadvertently emphasized the importance of personal achievement among European graduates. In the hope that this novel result will inspire further research, I now discuss the limitations of my approach for testing the neoliberal and humanist hypotheses and examine potential heterogeneity in my findings, both within and beyond our sample.
First, on the neoliberal hypothesis, a key question regards the extent to which the disclosed neoliberal bias is context-dependent and/or socially stratified by type of academic institution, its culture, or the field of study students pursue. The ESS lacks any granular information on the type, prestige, quality, or organizational culture of the specific academic institution attended by the sample respondents. As a second best, I explore this question by collecting country-level data allowing us to identify countries with relatively high average tuition fees and a relatively high share of programmes with compulsory internships. While I find no evidence that tuition fees moderate the reform’s effects, I find that the countries where the compulsory internships are popular experienced significantly stronger reform effects. This confirms once more that the institutional culture of the university, specifically the extent to which it embraced the vocational aspect of the Bologna process, played a major role in patterning the students’ values. Like much of the literature, this paper omits variation by academic subject as a moderating factor in the education–politics link (but see Hooghe et al. Reference Hooghe, Marks and Kamphorst2024). This omission may be especially relevant given the uneven exposure of academic disciplines to globalization. Regrettably, data on academic subjects are available only in three out of eleven ESS waves. While I do not find evidence that academic subjects substantially moderate reform effects, our analysis is underpowered (SI and Tables T.15 and T.16).
Second, focusing on the humanist potential of HE, one may argue that the unique focus on abstract morally universalist values prevents us from conclusively ruling out that the Bologna Process, which made international exchange central, fostered the kind of cosmopolitan civic values envisaged by advocates of global citizenship education. Since I test only a reduced form of the ‘humanist hypothesis’, my analysis could still conceal a fruitful interplay between HE internationalization and students’ cosmopolitanism. At first glance, the results from supplementary analyses align with this conjecture: the Bologna Process slightly increases support for freedom of movement. However, in line with our discussion of the Bologna Process’s rationale, support for freedom of movement increases significantly only when immigration is presented as an asset for economic growth (‘Immigration is good/bad for the country’s economy’), but does not change when freedom of movement is presented as an opportunity for either cultural or overall social improvement (‘Immigration harms/improves the country’s culture’ or ‘Immigration makes the country overall worse/better’) (SI T.17). These additional results suggest that the Bologna graduates are ‘pragmatic globalists’ who appreciate global integration mostly insofar as it aligns with increased economic opportunity – in line with the critique that HE’s emphasis on mobility is too narrowly directed at career opportunities (see, for example, Parker and Jary Reference Parker and Jary1995; Nóvoa and Lawn Reference Nóvoa and Lawn2002; Štech Reference Štech2011). While economic individualism is associated with conservative ideology (see, for example, Gelepithis and Giani Reference Gelepithis and Giani2020), I find no reform effects on left–right voting (see SI T.18).
Finally, how HE institutions respond to globalization and how individuals respond to HE institutions may hinge on specific geographic areas, time frames, or HE institutional features within which policies are implemented. As a result, this study has important limitations in terms of generalizability.
First, while the number of Bologna Process signatories rose from twenty-nine initially to eventually forty-nine, this analysis is limited to twenty-three of these countries, with Central, Southern, Western, and Northern Europe fully represented but the Balkans and the Caucasus almost entirely absent.
Furthermore, while my focus is on the Bologna Process, similar supranational initiatives exist elsewhere – such as the AUCE Strategy in Africa, ENLACES in Latin America, and the Brisbane Communiqué in the Asia-Pacific region. Exploring their ideational premises and political economy effects presents a promising avenue for a macro-comparative approach to advance our understanding of the interplay between globalization, HE, and worldwide patterns of political attitudes.
Moreover, I restrict the focus to values that capture a shift from ‘solid’ communal commitments to a more ‘liquid’ mindset of flexible self-interest (Bauman Reference Bauman2013). These values are likely relevant for the European millennials sampled in this study, who were born between the mid-1970s and early 1990s. Meanwhile, left–right divisions or different interpretations of value pluralism may resonate more with boomers or Gen Z.
Conclusion
While the causes of the Bologna Process have been widely studied by education scholars (for a review, see Wihlborg Reference Wihlborg2019), evidence about its consequences remains ‘surprisingly small, selective, and ambiguous’ (Kroher et al. Reference Kroher, Leuze, Thomsen and Trunzer2021). In documenting the political-economy causal effects of the Bologna Process, I contribute to some important conversations about higher education and political values that animate international organizations, scholars, and the public square.
First, by linking the evolution of higher education to the development of a global citizenship, I speak to an important policy debate. In framing economic growth as a condition for political togetherness in trade or HE policy alike, the EU frames personal and global goals as mutually reinforcing. By contrast, UNESCO frames them as mutually reinforcing only conditional on appropriate policy: while education should help students to ‘pursue [their] dreams and find purpose in life’ while ‘building a sense of belonging to a common humanity’ (UNESCO, Implementation Report 2024, p. 53), moral universalism does not come built in; instead, ‘broad reforms [are necessary] to ensure that learners are equipped with the attitudes and competencies to be engaged and responsible global citizens’ (Global Citizenship Education, mission statement). My evidence supports the view that the development of cosmopolitan solidarity requires an active institutional design, and my identification strategy provides a tool to study comparable policy initiatives worldwide (see, for example, Marginson and Van der Wende Reference Marginson and Van der Wende2007; Altbach et al. Reference Altbach, Reisberg and Rumbley2019).
Second, by shifting the focus from the quantity to the quality of education, I advance extant scholarly debates about the role of explicit or hidden curricula rather than enrolment (see, for example, Cantoni et al. Reference Cantoni, Chen, Yang, Yuchtman and Zhang2017; Mendelberg et al. Reference Mendelberg, McCabe and Thal2017; Paglayan et al. Reference Paglayan2025), focusing on higher instead of compulsory education. In the EU, median HE enrolment only rose from fifty-four per cent in 2016 to fifty-five per cent in 2021 (EU, Directorate General for Education and Culture 2023). As enrolment levels slow following decades of massification, my complementary angle and findings suggest that policymakers interested in social justice should revise their priority from broadening students’ access to HE to broadening their political outlook while relaxing the emphasis on career goals.
Third, by exploring ideological patterns beyond conventional left–right political attitudes, I help qualify an essential public debate about the ideological input of HE in contemporary Western liberal democracies. ‘University’ and ‘universalism’ share the Latin root ‘universus’ – what does their divergence reveal about contemporary universities? Conservatives and progressives often blame higher education for seemingly contradictory reasons: ‘prioritizing social justice over academic achievement’ on the one hand, or only acting to ‘prepare their students, largely uncritically, for the market’, forging ‘student-customers’, on the other.Footnote 19 These critiques are not specular: conservatives accuse academics of biasing the students on the left–right ideological scale, whereas progressives accuse universities of biasing students on the individualist–collectivist value hierarchy. My findings suggest that progressives are increasingly accurate. Conservatives get the premise right – indeed, most academics are left-wing (Van de Werfhorst Reference Van de Werfhorst2020) – but the inference wrong: it is the institutional context of the university, not the personal views of its employees, that shape students’ politics.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0007123425100963.
Data availability statement
Replication data for this paper can be found in Harvard Dataverse at https://doi.org/10.7910/DVN/U0ZMOR
Acknowledgements
The author thanks three anonymous referees of British Journal of Political Science for their remarkable engagement with the paper and constructive critiques, Ben Arold for interesting methodological tips, and all participants to a Quantitative group seminar at the Faculty of Education, University of Cambridge.
Financial support
The author acknowledges support from a KCL’s ARF research grant.
Competing interests
None declared.
Open access
