The effect of government cuts of doctoral scholarships on science
Abstract
EconStor is a publication server for scholarly economic literature, provided as a non-commercial public service by the ZBW.
Full text
Rossello, Giulia Working Paper The effect of government cuts of doctoral scholarships on science UNU-MERIT Working Papers, No. 2023-031 Provided in Cooperation with: Maastricht Economic and Social Research Institute on Innovation and Technology (UNU-MERIT), United Nations University (UNU) Suggested Citation: Rossello, Giulia (2023) : The effect of government cuts of doctoral scholarships on science, UNU-MERIT Working Papers, No. 2023-031, United Nations University (UNU), Maastricht Economic and Social Research Institute on Innovation and Technology (UNU-MERIT), Maastricht This Version is available at: https://hdl.handle.net/10419/326881 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-sa/4.0/
#2023-031 Theeffectofgovernmentcutsofdoctoralscholarshipsonscience GiuliaRossello Published19September2023 MaastrichtEconomicandsocialResearchinstituteonInnovationandTechnology(UNU‐MERIT) email:[email protected]u|website:http://www.merit.unu.edu Boschstraat24,6211AXMaastricht,TheNetherlands Tel:(31)(43)3884400
UNU-MERIT Working Papers ISSN 1871-9872 Maastricht Economic and social Research Institute on Innovation and Technology UNU-MERIT | Maastricht University UNU-MERIT Working Papers intend to disseminate preliminary results of research carried out at UNU-MERIT to stimulate discussion on the issues raised.
The Effect of Government Cuts of Doctoral Scholarships on Science* Giulia Rossello1,2 1REMARC Department of Economics and Management, University of Pisa Via Cosimo Ridolfi 10 56124 Pisa, Italy 2UNU-MERIT Maastricht University, Boschstraat 24 6211 AX Maastricht, the Netherlands September 19, 2023 Abstract I provide estimates of the impact of government cuts on PhD scholarships in Science. I leverage a unique quasi-natural experiment, the staggered cuts made by the Hungarian Government between 2010 and 2021 to expand Orb´ an’s political influence over the university system. The political aim of the cut ensures that it is exogenous to the economic cycle and to the scientific activity of universities. My analysis couples the complete enrolment records of doctoral students in the country around the years of scholarship cuts with a generalized difference-in-differences approach. I find that while government cuts of PhD scholarships have an ambiguous effect on students’ attainments, the policy has a clear negative effect on Science. That is, the severe reduction of scholarships increases the chance of completing the PhD by 1 pp, and the effect is stronger for female students. However, this positive effect is counterbalanced by a reduction of a similar amount of entry rates for females and non-traditional students. This suggests that besides training might improve, or the system might become more efficient, this is at the expense of social inclusion. Additionally, the effects of cuts on scientific production are negative both in terms of quantity and quality. The productivity of doctoral students drops by 2 pp while their scientific quality decreases between 0.2 pp and 1 pp. My results suggest that the reduction of doctoral scholarships might produce efficiency in terms of student attainment but at the expense of social inclusion, scientific production, and quality. JEL codes: H75, I23, I24, I25 Keywords: Government Appropriation, Higher Education, Doctoral Scholarships, Event Study, Difference in Differences *I would like to thank Prof. P` eter Mezei and his staff at the Szeged Law School for their help in gathering scholarship data and information about the Hungarian doctorate system. 1
1 Introduction Expanding government funding for higher education has been the pillar of 20th-century economic growth (Gennaioli et al., 2013; Sterlacchini, 2008; Goldin and Katz, 1999). However, despite increased demand for skilled labour (Blair and Deming, 2020), the last decades saw in most countries an opposing trend. A reduction of government appropriation and a decline in public funding per student. Past research warns about the danger of reducing State support for higher education concerning inequality, and human capital formation (Bound et al., 2010, 2019). Yet, the lack of causal studies and a blurry definition of budget cuts exacerbate the debate about government funds for higher education and their impact. For example, still is unclear under which conditions budget cuts at universities might improve efficiency. Part of the issue is that systematic studies with causal estimates are recent and U.S.-based. This empirical effort has documented that the reduction of government appropriation changes enrolment composition (Bound et al., 2020), reduces in-State student enrolment, students attainment, and scientific productivity (Bound et al., 2019), while it increases time to degree, debt, and deteriorates students’ socio-economic status (Chakrabarti et al., 2020). However, the U.S. system has unique characteristics: the coexistence of private and public universities, high tuition fees, and the ability to attract foreign talents. In contrast, most European and emerging countries have smaller systems, less international integration and fewer possibilities to adjust to lower appropriations. In most countries research universities are public and few receive inflows of foreign (extra-EU) students who may be charged with high tuition fees.1Differences imply that reduced support of government appropriation in those countries in theory might have more accentuated effects or opposite outcomes than those found in U.S. studies. On the one hand, cuts of appropriation might have more negative effects since many European and emerging countries cannot leverage tuition fees and have more difficulties in attracting private funding. On the other hand, in contrast, the effects might be less negative and promote efficiency. Most of those countries have small knowledge systems and less dynamic labour markets. That is, their ability to absorb graduates is limited. Thus, a decrease in appropriation resulting in a size reduction might create efficiency in the system, allowing the entrance of those required in the job market and academia. These potential contrasting results require further investigation of higher education systems outside North America. In this paper, I provide causal estimates of the impact in STEM of a specific cut of government appropriation to higher education in an Eastern European country, Hungary.2The system in Hungary is predominantly public and, in general, public Hungarian universities do not charge tuition fees.3The interest in the case of Hungary for the general literature is twofold. First, in contrast to the U.S. case, universities are less known internationally, and less able to attract private funding. Thus, the reduction of public funding might affect them more since it might impact the university’s status and prestige. For example, many universities are non 1Most European universities are public especially research-oriented institutions. Tuition fees are generally low and Universities cannot charge higher fees to European citizens. Some countries charge extra-EU citizens more but this is not a generalised practice and numbers are low. 2Hungary has one of the most ancient and prestigious university systems in Eastern Europe and, as demonstrated by the high number of Nobel laureates, is forefront in many scientific disciplines. 3There are some exceptions but in general tuition when present is very low, and international students are charged more between 1,200 and 5,000 a year. See this website for an overview https://www.masterspor tal.com/articles/2871/tuition-fees-and-living-costs-in-hungary-in-2023.ht ml; accessed June 2023. 2
listed in international rankings and their perceived prestige can be volatile, as found in other emerging economies (Gonz´ alez-Sauri and Rossello, 2023). Second, the specific case of the Hungarian universities after the second election of Orb´ an represents a perfect setting. It allows me to provide causal estimates of how severe government cuts of doctoral scholarships impact science by leveraging a unique quasi-natural experiment: the staggered cuts made by Orb´ an to increase his political control over the university system in 2010-2021.4Since cuts were made to persecute university deans who were in opposition to his government the cuts are exogenous to the business cycle, to university evolution, to students’ behaviour, and to the scientific agenda of the ministry. To provide causal estimates, I rely on the official data on PhD students and scholars collected by the Hungarian Doctoral Council (HDC) in the years around state-funded scholarship reduction with a generalized difference-in-differences empirical strategy. Notwithstanding recent improvements, the literature faces three main challenges that I address in this work: (1) the definition of cuts, (2) a lack of individual-level systematic studies, and (3) the endogeneity of changes in government higher education appropriations to the economic cycle. First, considering the definition of cuts, most studies address general funding reduction. For example, Bound et al. (2020) consider a reduction of the difference between total state-level appropriations and the university’s one, while Chakrabarti et al. (2020) use a calculation of shift-share of state-university appropriations as an instrument (i.e. the interaction of the baseline share of total revenue that comes from state appropriations at each public institution with yearly variation in state-level appropriations). However, general definitions do not address the impact of specific cuts. Indeed, each university decides how resources are allocated and, thus, responds differently to a reduction of appropriation, making a general cut often a non-homogeneous policy. Deming and Walters (2018) highlighted that institutions react to budget cuts through a mix of measures on prices and spending. They addressed this challenge by using the introduction of a tuition cup to isolate the effect of budget cuts from increased tuition. In Europe usually, the ability of universities to raise prices through tuition is limited, but the issue remains since budget cuts can impact the number of courses offered, infrastructures, or hiring. This makes it crucial to isolate the effect of a specific policy of cuts of government appropriation. In this paper, I overcome this issue. I examine the reduction of the number of government-paid scholarships for PhD students. This focus isolates the effect of a specific cut of appropriation without other confounding factors linked to universities’ adjustment to shrinking funding. Therefore, my contribution examines a specific cut of government appropriation that operates on spending, reducing sizes: shrinking the number of students in doctoral programs. This policy is relevant to the literature because it represents a budget cut without an obvious negative outcome, and potentially affects efficiency and competition. Second, systematic studies examining a university system focused on the aggregate impact of cuts on institutions but did not capture how reduced government appropriation affects individuals. Indeed, budget cuts might produce an internal re-organization of courses which potentially impacts students’ careers differently. For example, a general decrease in graduation rates might be the net outcome of the decreased number of students enrolled and a composition 4These cuts were made as a “revenge” against universities opposing Orb´ an’s plan to transform universities into private foundations (Kov´ ats and R´ onay, 2021). Such transformation was implemented in 2021. See the article in the Economist on May 1st 2021 titled “Viktor Orb´ an seizes control of Hungary’s universities” https: //www.economist.com/leaders/2021/05/01/viktor-Orbn-seizes-control-of-hunga rys-universities; assessed June 2023. 3
effect of course composition (courses can disappear, be merged, or split) but at the same time, individuals enrolled in traditional tracks might benefit for the reduced number of students per class. Unfortunately, the only work that looked systematically at individual-level outcomes is Chakrabarti et al. (2020) which examines how changes in state appropriations affect the education outcomes of enrolled students. They concentrate on bachelor and master students examining student debt, and degree completion. They found evidence that an increase in state appropriation increases completion rates for Bachelor students but not for Graduate students. In contrast, I consider the effect of scholarship cuts on enrolled PhD students, examining how cuts affect the demographic composition by looking at the entry of female and non-traditional students in doctoral programs, students’ probability of graduating, and students’ scientific output in terms of quantity and quality of publications. The effect of scholarship has been studied before but results are not systematic and concentrate on a specific university or funding scheme. Bin et al. (2022) study the effect of receiving the Sao Paulo Research Foundation PhD scholarship in Brazil. They find that the research performance of PhD receiving the scholarship was higher than those with a rejected application. Most of the existing research concentrates, instead, on scholarships at the bachelor and master levels. For example, Cohodes and Goodman (2014) use a regression discontinuity approach comparing students just above and below the eligibility threshold of a merit grant for first-year college students in Massachusetts. They find that since the scholarship shifted enrolment towards second-tier institutions the programme reduced completion rates of students. In a recent contribution Minaya et al. (2022), examined the effect of more stringent performance requirements for receiving aid scholarships in an Italian University, Politecnico of Milano. They found that the change improved the performance of higher and medium-ability students but it discouraged the others from continuing in their studies. However, there are no studies which study the impact of a reduction of governmentfunded scholarships. I contribute to existing research providing systematic and individual-level estimates of government cuts of PhD scholarships in an emerging economy. Third, the main issue I address is that government appropriations for higher education are endogenous to the economic business cycle. For example, cuts in government appropriation might follow a financial crisis that affects the willingness of students to enter higher education. In such a case, enrolment declines might cause budget cuts, rather than the other way around. An additional issue is that the government might invest in some universities to develop specific capabilities. For example, institutions specialized in specific topics (e.g. climate change, artificial intelligence, robotics, green transition) can attract political attention and receive more funding. Past research addresses this issue using an instrumental variable approach or a shiftshare identification strategy. Bound et al. (2019) use as an instrument the logarithm of aggregate state appropriation arguing that a State aggregate appropriations does not depend on a specific institution enrolment decisions or research output. Instead, Deming and Walters (2018) and Chakrabarti et al. (2020) use a shift-share approach considering the combination of the share of state-level and institutional-level appropriations, based on the idea that state-wide changes in appropriations for higher education will have different effects on institutions based on their underlying reliance on state fund. In contrast, I use a difference-in-differences empirical strategy exploiting the staggered massive reduction of doctoral scholarships made by the Orb´ an government in his attempt to seize control of Hungary’s universities. This policy was unexpected at universities and hit universities unaligned with the conservative government with different timing. The cut of government-funded scholarships has been the first policy made by Orb´ an to exert control over academia in an effort to root out opposition and to tame university deans to the privatization transition of universities. Other examples of Orb´ an’s acts reducing higher 4
education autonomy included shutting down the Central European University in 2019, banning gender studies in 2018, reducing the Academy of Sciences’ independence in 2019, and in 2022 passing control of the 11 main state universities to private foundations controlled by his allies. Because a conservative political agenda might impact research in Social Sciences and Humanities I focused on STEM disciplines. The specific case of Orb´ an’s cuts of doctoral scholarships is particularly apposite to provide causal estimates for state appropriation cuts because the cuts were punitive for political oppositions of university deans and can be considered exogenous to the economic cycle and the government investment agenda in university institutions. By taking these three issues highlighted by the literature I provide causal estimates on the effect of severe cuts of government appropriations operating with a shrinkage of sizes in government-funded PhD scholarship. My estimates are the first considering an entire university system outside the U.S. and the first study that examined a reduction of government appropriation which operates only by cutting budget reducing sizes. I found that the reduction of state-funded PhD scholarships had a negative impact on the diversity of new students enrolled. In particular, severe cuts in government-funded scholarships reduce the probability that female and non-traditional students enter PhD training in STEM by 1 pp. However, I found that size reduction improved the students’ attainment suggesting some potential effect on efficiency. In particular, I found that a severe reduction of PhD scholarship increased the probability of students to graduate by 1 pp. Further, my results support past evidence that a reduction in state support for higher education have a negative effect on research productivity, reducing the number of paper produced by 2 pp. Additionally, I found a negative, although moderate, impact on research quality that is reduced between 0.2 and 1 pp. My results highlight that reducing the support of the government for higher education by reducing the size of scientific doctoral programs has a potential trade-off. On the one hand, it might improve the condition of those entering the system by reducing competition, improving training, or fostering efficiency as underlined by the increased graduation probability. But, on the other, these improvements are counterbalanced by a reduction of diversity, social mobility, research productivity and quality. The latter, suggests that Science is a collective effort and operates in teams where often big is “better”, especially for what concerns metrics of quantity or “quality” of the research output. The paper is organised as follows, in Section 2 I highlight the institutional features of the Hungarian university system. In section 3 I describe the data construction and sources. Section 4 describes the methodology used and the empirical identification strategy. Section 5 presents the results. In section 6 I discuss the results and section 7 concludes. 2 Background In this section I will explain the context of the Hungarian university system and its recent changes. The Hungarian University System Hungary is an emerging economy of almost 10 million people with an old and prestigious university system. Indeed, the first Hungarian university was funded in 1367. While, in the contemporary era, until 1989 the Hungarian university system was modelled on the Soviet one where universities had little autonomy from 5
Figure 1: EVOLUTION OF THE HUNGARIAN UNIVERSITY SYSTEM 2008-2020 Notes: The black squares represent the government expenditure on tertiary education as a percentage of GDP(%) (left axis). Source: UNESCO Institute of Statistics. The gray triangles are the number of full-time students in higher education (right axis). Source: Hungarian Central Statistical Office 2021 (KSH). The light gray bars represent the total number of government-funded PhD scholarships. the central government. After the regime changed in 1989 the university system became more similar to Western economies. Today universities have high degrees of university autonomy. The system is relatively small, there are 35 universities and among those 22 have STEM doctoral programs. In terms of subjects, since the system is relatively small it achieved several specializations to compete internationally. The university system is specialised in aspects of mathematics, chemistry, medicine, and physics where Hungarian scholars obtained important recognition as well as 13 Nobel prizes. Today the Hungarian University system is public and almost free for Hungarian students. Indeed, public universities account for the 89% of researchers and the 80% of students (Kov´ ats and R´ onay, 2021). Figure 1 shows the evolution of funding and student enrolments between 2008 and 2020. The figure shows a significant reduction in the public support for higher education, the government expenditure on higher education as a percentage of GDP fell from 1% to 0.8% between 2008 to 2018. Kov´ ats and R´ onay (2021) underline that the government withdrew significant resources from the sector by 2010 which resulted in a decline in public support for higher education of almost 50% in real terms between 2008 and 2013. Looking at student enrolment over time the figure shows a decline, full-time students decreased by 40’000 units. In contrast, the total number of government-funded PhD scholarships stays pretty constant over time with a slight increase in the last period. However, besides aggregate figures, the number of funded scholarships was inconstant and subject to severe cuts and reallocation between fields and universities. In the next section, I will explain how this took place and why it represents a unique and apposite quasi-experiment setting to examine the impact of government scholarship cuts on Science. 6
reverse causality and omitted variable bias. For example, in such a context, the number of scholarships funded can decrease because the performance, popularity, or prestige of the university has changed. Alternatively, scholarships’ fluctuation might reflect the evolution of the university system, the growth of enrolments, or their re-organizations. Some fields can become less popular experiencing a drop in enrolments and, thus, doctoral programs can disappear or split across universities. Therefore, I leverage the staggered scholarship cuts made by the Hungarian government between 2010 and 2021 to obtain estimates that can identify their causal effect on the Science System. Under the assumptions described below, the staggered cuts of scholarships generate a quasi-experimental variation that allows the estimation of the causal impact of PhD scholarship reductions using a generalized difference-in-differences strategy. The strategy compares the before-after difference in outcomes between PhD students in universities that saw a severe cut of PhD scholarships and students in colleges that did not see such a reduction between the two periods. The baseline specification is the following two-way fixed effect (TWFE) model estimated using OLS with clustered standard error at the student level: Yi jugt =αg+δt+β×PhD Scholarship Cutgt +Xi×θ+Xj×γ+Xu×φ+εi jugt (1) where Yi jugt represents an outcome for individual iwho is supervised by advisor jat time tand was enrolled in university uthat belongs to the scholarship cut group cohort g.αg(or αu) indicates the the scholarship-cut-group cohort (or university) fixed effects. PhD Scholarship Cutgt is a dummy variable equal to 1 if in year tthere was a severe cut of PhD scholarship (a reduction ≥10 seats, i.e. the third quartile of the variation of scholarship cuts) and zero otherwise. Xi,Xj, and Xuare vectors of controls at the student-, advisor-, and university-level. Controls at the individual levels are student’s quality expressed as average Scimago journal ranking of her/his previous publications, student’s productivity measure as number of papers weighted the number of co-authors per publication. Advisor-level controls are the advisor’s gender (the associated probability that her/his name is a male name), the advisor’s quality expressed by the average Scimago journal ranking of her/his previous publications, the advisor’s productivity expressed as the number of previous publications weighed by the number of co-authors per publication, and the number of students supervised by the advisor in the previous 3 years including the actual one. University-level controls are the total millions of Forint (HUF) of national research funding grants won, the total number of PhD enrolled, and the total number of papers indexed in Scopus produced in year t. I estimate the equation using OLS and clustered standard error at the student level. In this way, the coefficient βidentifies the average treatment effect on the treated (ATT) of the scholarships cut on enrolment, graduation, and research performance under the assumptions that the university-level average treatment effect is homogeneous across treated universities over time and the parallel trend assumption. Under the described assumptions, the TWFE model can rule out the following concerns that impede to interpret results as causal. For example, one worry could be that more (less) prestigious universities have better (worse) student outcomes. I solve this and similar concerns driven by time-invariant differences in PhD outputs across universities including university fixed effects (or scholarship-cut-group cohort). A second issue could be that results are driven by the growth and evolution of the university system over time and that the latter affects student enrolments and outcomes across universities. For example, as the system matures the number of scholarships increases as the scientific production. Another example is that an economic crisis might negatively impact the career perspective of students and in turn, create 13
an adverse selection of PhD students into the system. To solve this concern I added year fixed effects. However, one concern might relate to the parallel trend assumption. In other words, universities that experience a severe scholarship cut at different time-frame might have different outcome trends of enrolment, graduation, and research performance. To address the issue I run the associated event study analysis estimating the dynamic TWFE model to account for the potential existence of pre-trend as follows: Yi jugt =αg+δt+β× 13 ∑ −8 Dk(gt)+εi jugt (2) Where Dk(gt)is a dummy equal to one if the scholarship cut group cohort gis kyears away from the cut in year tand zero otherwise. The baseline is the year before the cut (k=−1). To ensure the reliability of the results I provide in the event study plots the alternative estimators of Sun and Abraham (2021) and Callaway and Sant’Anna (2021). Both are additional checks for the parallel trend assumption and they have the advantage of providing consistent estimates in the presence of heterogeneous treatment effects across time and or treated units. Those models have been developed to relax the assumption of homogeneity of the treatment effects allowing units treated before/after to experience different trends. This is particularly apposite here since a cut of scholarship affects the treated cohorts for all of their enrolment period and effects might grow with time. As additional checks for the parallel trend assumption I added to the main tables (1) a model adding scholarship-cut-group cohort linear time trends that account for linear time trends; and (2) I ran a model that relaxed the assumption of linear time trend and uses the interaction of university and years fixed effects and that compares students within the same university-year who were exposed to cuts for different lengths of time based on the year in which they entered PhD. 14
5 Results Table 2: BASELINE RESULTS – FEMALE ENROLMENT A Female Name Enters PhD Training (1) (2) (3) (4) PhD Scholarship Cuts −0.0001 −0.0005 −0.002 −0.002 (0.004) (0.004) (0.004) (0.004) Observations 68,944 68,944 68,944 68,944 N Students 10,083 10,083 10,083 10,083 N Advisors 3,639 3,639 3,639 3,639 Year fixed effects ✓ ✓ ✓ ✓ University fixed effects ✓ ✓ Group-scholarship-cuts fixed effects ✓ Group-scholarship-cuts linear time trends ✓ Field dummy ✓ ✓ Controls ✓ ✓ Note: This table explores the effect of a government cut of funded PhD scholarships on the probability that a female name enters PhD training in STEM. Specifically, it presents estimates of coefficient β from equation (1), with the outcome variable representing the probability that a female name enters PhD in year t. The variable is greater than 0 and smaller than 1 if the student enters the sample and zero otherwise, where values closer to 1 are associated with the probability that the student has a feminine name. Column 1 estimates equation (1) without controls with university fixed effects; column 2 estimates equation (1) including instead cut-expansion groups fixed effects; column 3 includes advisorspecific and university-specific controls and scientific fields dummies to the previous specification in column 1; and column 4, instead, add to controls the cut-expansion groups linear time-trends. Our controls consist of the advisor’s gender (the associated probability that her/his name is a male name), the advisor’s quality expressed by the average Scimago journal ranking of her/his previous publications, the advisor’s productivity expressed as the number of previous publications weighed by the number of co-authors per publication, and the number of students supervised by the advisor in the previous 3 years included the actual one. University-specific controls are the total millions (HUF) of national research funding grants won, the total number of PhD enrolled, and the total number of papers produced which are listed in the Scopus database. Standard errors in parentheses are clustered at the student level. Significance codes: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 15
−0.2 −0.1 0.0 0.1 0.2 Periods Since the Event Average Effect − 95% CI −8 −6 −4 −2 0 1 2 3 4 5 6 7 8 9 11 13 Callaway Sant'Anna Sun & Abraham TWFE Figure 3: EFFECTS OF GOVERNMENT CUTS OF PHD SCHOLARSHIPS ON FEMALE ENTRY BASED ON DISTANCE TO/FROM CUTS INTRODUCTION Notes: The figure shows overlays of the event-study plots constructed using three different estimators: (1) a dynamic version of the TWFE model, (2) Sun and Abraham (2021), and (3) Callaway and Sant’Anna (2021). The outcome variable is the probability that a student with a female name enters PhD training in year t and zero otherwise. In estimations (1) and (2) the baseline period is -1 and they include control variables and standard errors clustered at the student level. The controls used are: advisor’s gender (the associated probability that her/his name is a male name), advisor’s quality expressed by the average Scimago journal ranking of her/his previous publications, advisor’s productivity expressed as the number of previous publications weighed by the number of co-authors per publication, and the number of students supervised by the advisor in the previous 3 years including the actual one. Additional university-specific controls are: the total millions (HUF) of national research funding grants won, the total number of PhD enrolled, and the total number of papers produced which are listed in the Scopus database. Estimation (3) uses doubly robust estimation and the baseline control group is the not yet treated one. In what follows I present how scholarship cuts affect the three sets of dependent variables on enrolment, graduation, and research performance showing the main regression and the event study plot described above. Enrolment Results – Table 2 shows estimates of βin equation 1 on the probability that a feminine name is enrolled as a PhD. Columns one and two are baseline difference-in-difference estimations with years fixed effects and university fixed effects or scholarship-cut-group cohort fixed effects. Columns three and four add field dummies and additional controls of students, advisors, and universities described above. Additionally, the model in column three controls (as column one) for years fixed effects and university fixed effects; while model four controls for scholarship-cut-group cohort linear time trends. I found consistent estimates that scholarship 16
cuts decreased the likelihood that a female name enters PhD training by 0.1-0.2 pp, however, the effect is not statistically significant at 10% significance level. In contrast, the dynamic analysis in Figure 3 shows that the effect of government appropriation for PhD scholarship is not immediate and grows over time. After the third year, the estimates become different than zero at 5% significance level. Additionally, the dynamic TWFE systematically underestimates the effect suggesting a heterogeneous effect linked to treatment cohorts. In particular, both estimations of Callaway and Sant’Anna (2021) and Sun and Abraham (2021) suggest an average effect of around 5 pp, corresponding to a decrease in female entry rates of 75%. Moreover, the effect growth over time from 5 pp in the medium run (3-9 years after the cuts) to more than 10 pp after 10 years. The growth and persistence of the effect over time is impressive and I will turn to this in the discussion. Table 3: BASELINE RESULTS – NON-TRADITIONAL STUDENT ENROLMENT A New Surname Enters PhD Training (1) (2) (3) (4) PhD Scholarship Cuts −0.014∗∗∗ −0.015∗∗∗ −0.010∗∗ −0.014∗∗∗ (0.004) (0.004) (0.005) (0.004) Observations 68,944 68,944 68,944 68,944 N Students 10,083 10,083 10,083 10,083 N Advisors 3,639 3,639 3,639 3,639 Year fixed effects ✓ ✓ ✓ ✓ University fixed effects ✓ ✓ Group-scholarship-cuts fixed effects ✓ Group-scholarship-cuts linear time trends ✓ Field dummy ✓ ✓ Controls ✓ ✓ Note: This table explores the effect of a government cut of funded PhD scholarship on the probability that a surname new to the university system enters PhD training in STEM. Specifically, it presents estimates of coefficient βfrom equation (1), with the outcome variable equal to one if a student with a surname that was not present before enters PhD training and zero otherwise. Column 1 estimates equation (1) without controls with university fixed effects; column 2 estimates equation (1) including instead cut-expansion groups fixed effects; column 3 includes advisor and university-specific controls and scientific fields dummies to the previous specification in column 1; and column 4, instead, add to controls the cut-expansion groups linear time-trends. Our controls consist of the advisor’s gender (the associated probability that her/his name is a male name), the advisor’s quality expressed by the average Scimago journal ranking of her/his previous publications, the advisor’s productivity expressed as the number of previous publications weighed by the number of co-authors per publication, and the number of students supervised by the advisor in the previous 3 years including the actual one. University-specific controls are the total millions (HUF) of national research funding grants won, the total number of PhD enrolled, and the total number of papers produced indexed in Scopus. Standard errors in parentheses are clustered at the student level. Significance codes: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Table 3 examined how the decrease of government appropriation for PhD scholarship affects social mobility presenting the estimates of equation 1 where the dependent variable indicates whether the student entering doctoral education has a surname new to the university system. The estimates show a seizable and significant effect at 1% significance level. The probability that a student with a new surname enters PhD training decreased by 1.4 pp after a severe cut of government scholarships. This is striking, corresponding to a 16% decrease in the entry rate of non-traditional students. The event analysis in Figure 4 shows a less significant effect when the heterogeneity of the treatment timing is considered and the peak of the effect 17
5-6 years after the scholarship reduction. Overall the analysis of enrolment outcomes highlights a negative impact of the decrease in government appropriation of PhD scholarships on the entry rate of females and nontraditional students with effect that persists for a decade after the cuts. This is not surprising given the well-known inertia and stratification of university systems in general and of those of emerging economies in particular (Rossello, 2021; Cowan and Rossello, 2018; Gonz´ alez-Sauri and Rossello, 2023). −0.2 −0.1 0.0 0.1 0.2 Periods Since the Event Average Effect − 95% CI −8 −6 −4 −2 0 1 2 3 4 5 6 7 8 9 11 13 Callaway Sant'Anna Sun & Abraham TWFE Figure 4: EFFECTS OF GOVERNMENT CUTS OF PHD SCHOLARSHIPS ON NEW SURNAME ENTRY BASED ON DISTANCE TO/FROM CUTS INTRODUCTION Notes: The figure shows the overlays of the event-study plots constructed using three different estimators: (1) a dynamic version of the TWFE model, (2) Sun and Abraham (2021), and (3) Callaway and Sant’Anna (2021). The outcome variable is equal to one if a student with a surname new in the university system enters and zero otherwise. In estimations (1) and (2) the baseline period is -1 and they include control variables and standard errors clustered at the student level. The controls used are: advisor’s gender (the associated probability that her/his name is a male name), advisor’s quality expressed by the average Scimago journal ranking of her/his previous publications, advisor’s productivity expressed as the number of previous publications weighed by the number of co-authors per publication, and the number of students supervised by the advisor in the previous 3 years including the actual one. Estimation (3) uses doubly robust estimation and the baseline control group is the not yet treated one. 18
Table 4: BASELINE RESULTS – GRADUATION PROBABILITY A Student Graduates as PhD (1) (2) (3) (4) PhD Scholarship Cuts 0.004 0.008 0.005 0.011 (0.009) (0.009) (0.009) (0.009) Observations 68,944 68,944 68,944 68,944 N Students 10,083 10,083 10,083 10,083 N Advisors 3,639 3,639 3,639 3,639 Year fixed effects ✓ ✓ ✓ ✓ University fixed effects ✓ ✓ Group-scholarship-cuts fixed effects ✓ Group-scholarship-cuts linear time trends ✓ Field dummy ✓ ✓ Controls ✓ ✓ Note: This table explores the effect of a government cut of funded PhD scholarships on the probability that a student graduates. Specifically, it presents estimates of coefficient βfrom equation (1) with our variable representing the probability that the student graduates in year t as the outcome variable. The outcome variable is equal to the probability of graduating in year t and zero if the student never graduates. Column 1 estimates equation (1) without controls with university fixed effects; column 2 estimates equation (1) including instead cut-expansion groups fixed effects; column 3 includes student, advisor, and university-specific controls and scientific fields dummies to the previous specification in column 1; column 4, instead, add to controls the cut-expansion groups linear time-trends. Our controls consist of students’ quality expressed as the average Scimago journal ranking of her/his previous publications, student’s productivity measure as the number of papers weighted the number of co-authors per publication, advisor’s gender (the associated probability that her/his name is a male name), advisor’s quality expressed by the average Scimago journal ranking of her/his previous publications, advisor’s productivity expressed as the number of previous publications weighed by the number of co-authors per publication, and the number of students supervised by the advisor in the previous 3 years including the actual one. University-specific controls are the total millions (HUF) of national research funding grants won, the total number of PhD enrolled, and the total number of papers indexed in Scopus produced. Standard errors in parentheses are clustered at the student level. Significance codes: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 19
−0.2 −0.1 0.0 0.1 0.2 Periods Since the Event Average Effect − 95% CI −8 −6 −4 −2 0 1 2 3 4 5 6 7 8 9 11 13 Callaway Sant'Anna Sun & Abraham TWFE Figure 5: EFFECTS OF GOVERNMENT CUTS OF PHD SCHOLARSHIPS ON STUDENT GRADUATION BASED ON DISTANCE FROM/TO CUTS INTRODUCTION Notes: The figure shows the overlays of the event-study plots constructed using three different estimators: (1) a dynamic version of the TWFE model, (2) Sun and Abraham (2021), and (3) Callaway and Sant’Anna (2021). The outcome variable is equal to the probability that a student graduates and zero otherwise. In estimations (1) and (2) the baseline period is -1 and they include control variables and standard errors clustered at the student level. The controls used are: students’ quality expressed as average Scimago journal ranking of her/his previous publications, student’s productivity measure as the number of papers weighted the number of co-authors per publication, advisor’s gender (the associated probability that her/his name is a male name), advisor’s quality expressed by the average Scimago journal ranking of her/his previous publications, advisor’s productivity expressed as the number of previous publications weighed by the number of co-authors per publication, and the number of students supervised by the advisor in the previous 3 years including the actual one. Estimation (3) uses doubly robust estimation and the baseline control group is the not-yet-treated one. Graduation Results – Table 4 shows the estimates, β, of equation 1 of the effect of scholarship cuts on the probability that a student has to graduate in year t. The Table shows an increase of the graduation probability by 1 pp, however, the ATT is not different than zero at 10% significance level. This might relate to the dynamic of the effect, by definition such an effect has a lagged impact on the dependent variable. In principle, the most affected students are those entering PhD in the year of the shock but since the doctoral education last in Hungary between 3 and 6 years the main impact is expected after such a lag. Indeed, this is exactly what is found in Figure 5 where the dynamic after the event shows a significant effect after 3 years and a peak of the effect after 6 years. After 6 years the graduation probability increases by 10 pp, a 37% increase in graduation rates. 20
Table 5: BASELINE RESULTS – FEMALE GRADUATION PROBABILITY A Female Name Graduates as PhD (1) (2) (3) (4) PhD Scholarship Cuts 0.014∗∗ 0.016∗∗ 0.010 0.015∗∗ (0.006) (0.006) (0.007) (0.006) Observations 68,944 68,944 68,944 68,944 N Students 10,083 10,083 10,083 10,083 N Advisors 3,639 3,639 3,639 3,639 Year fixed effects ✓ ✓ ✓ ✓ University fixed effects ✓ ✓ Group-scholarship-cuts fixed effects ✓ Group-scholarship-cuts linear time trends ✓ Field dummy ✓ ✓ Controls ✓ ✓ Note: This table explores the effect of a government cut of funded PhD scholarships on the probability that a female student graduates. Specifically, it presents estimates of coefficient βfrom equation (1) with our variable representing the probability that a female name graduates as PhD as the outcome variable. The outcome variable is equal to the probability of graduating in year t if the student has a female name (probability that the name is feminine >0.5) and zero otherwise. Column 1 estimates equation (1) without controls with university fixed effects; column 2 estimates equation (1) including instead cut-expansion groups fixed effects; column 3 includes student, advisor, and university-specific controls and scientific fields dummies to the previous specification in column 1; column 4, instead, add to controls the cut-expansion groups linear time-trends. Our controls consist of students’ quality expressed as the average Scimago journal ranking of her/his previous publications, student’s productivity measure as the number of papers weighted the number of co-authors per publication, advisor’s gender (the associated probability that her/his name is a male name), advisor’s quality expressed by the average Scimago journal ranking of her/his previous publications, advisor’s productivity expressed as the number of previous publications weighed by the number of co-authors per publication, and the number of students supervised by the advisor in the previous 3 years including the actual one. Universityspecific controls are the total millions (HUF) of national research funding grants won, the total number of PhD enrolled, and the total number of papers indexed in Scopus produced. Standard errors in parentheses are clustered at the student level. Significance codes: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 21
−0.2 −0.1 0.0 0.1 0.2 Periods Since the Event Average Effect − 95% CI −8 −6 −4 −2 0 1 2 3 4 5 6 7 8 9 11 13 Callaway Sant'Anna Sun & Abraham TWFE Figure 6: EFFECTS OF GOVERNMENT CUTS OF PHD SCHOLARSHIPS ON FEMALE GRADUATION BASED ON DISTANCE FROM/TO CUTS INTRODUCTION Notes: The figure shows the overlays of the event-study plots constructed using three different estimators: (1) a dynamic version of the TWFE model, (2) Sun and Abraham (2021), and (3) Callaway and Sant’Anna (2021). The outcome variable is equal to the probability that a student with a female name graduates and zero otherwise. In estimations (1) and (2) the baseline period is -1 and they include control variables and standard errors clustered at the student level. The controls used are: students’ quality expressed as average Scimago journal ranking of her/his previous publications, student’s productivity measure as the number of papers weighted the number of coauthors per publication, advisor’s gender (the associated probability that her/his name is a male name), advisor’s quality expressed by the average Scimago journal ranking of her/his previous publications, advisor’s productivity expressed as the number of previous publications weighed by the number of co-authors per publication, and the number of students supervised by the advisor in the previous 3 years including the actual one. Estimation (3) uses doubly robust estimation and the baseline control group is the not yet treated one. Table 5 and Figure 6 show respectively estimates for equations 1 and 2 where the dependent variable is the graduation rates of students with a female name. In the baseline specifications, estimates are consistent and statistically significant at 5% significance level. In particular, I found that the severe decrease in government PhD scholarship appropriation increases the graduation rates of female students by 1.5 pp implying an increase of 13%. The event study highlights again that the peak of the effect is after 6 years, reaching an increase of 5 pp which corresponds to a 42% increase in female graduation rates lasting 5 years. 22
or drastically reduces their chances of being selected for such programs. Moreover, once I accounted for the dynamic over time I found that the effect grows over time and government scholarship cuts reduce female entry rates more heavily in the medium-long run. This highlights, that female enrolment in STEM might depend on network effects linked to the number of previously enrolled female students. A similar mechanism has been underlined in the literature which examines the role of gender homophily in higher education (Rossello, 2021; Main, 2014). Overall, the contraction of doctoral scholarships might hinder social mobility. Given the importance of education on job attainment (Spilerman and Lunde, 1991; Mertens and R¨ obken, 2013), the persistent reduction of the entry rate of female and non-traditional students might hinder their ability to reach the top of the socio-economic ladder in the future and surely their chances to enter academia as professors. Second, besides this distressing result, I found that doctoral programs become more efficient after the scholarship cut. In particular, graduation probability increases substantially between 0.4 and 2 pp, particularly for female students. However, the dynamic over time is an inverted U-shape and the effect appears only in the medium run. This observation suggests that students are more likely to complete their doctorates but they do not reduce their time to graduation. The increased efficiency of doctoral programs in relation to scholarship cuts might relate to a selection effect of doctoral candidates or to the quality/quantity of the supervision received. For example, a selection effect might operate on the quality of candidates selected, allowing only the higher achieving students to enter doctoral training while those with lower achievement are excluded. In contrast, the effect might be irrespective of the selected candidates but linked to the quality of supervision. In fact, scholarship cuts are likely to reduce the number of doctoral students per cohort and this might imply that those who enter find better supervision and less competition for advisor’s time. Past research in higher education has shown a substantial positive impact of reduced class sizes and students’ attainment (De Paola et al., 2013). Globally my results have shown that there exists a trade-off between inclusion and efficiency exacerbated by the after-shock dynamics of entry rates and graduation of female and non-traditional students. In fact, I found that government cuts to PhD scholarships have an ambiguous effect on students’ attainments and potential unintended and long-lasting consequences in terms of inclusion. On the one hand, the reduction of scholarships increases the chance of completing the PhD, but at the same time, this positive effect is counterbalanced by a reduction of a similar amount of entry rates for female and non-traditional students. This trade-off is particularly relevant for the literature that examines gender or racial imbalances in science and higher education because it highlights a potential regression of the progress made in terms of equal representation of female and non-traditional students made in the last decades. Indeed, most higher education systems saw a contraction in government appropriation after the great recession and this dynamically might cause university systems to go back in terms of their demographic composition. The third consideration refers to the effect of scholarship cuts on the research output of doctoral students. I found that both the quantity and quality of the research are negatively affected. This is not surprising given the importance that funding has for Science (Stephan, 2010; Franzoni et al., 2022). Indeed it appears that size matters. Most of the scientific research in STEM is lab-based and organised around large scientific teams. PhD candidates are at the forefront of those teams and often are those who run the lab experiments in the first place. A lower number of PhD scholarships reduces sizes and improves the efficiency of doctoral programs but it reduces the scientific capacity of a department reducing scientific productivity 29
as well as quality. Looking at the dynamics over time, the effect is mostly in the short run. Both students’ productivity and quality of research drop immediately after the cuts and then they recover. The latter might suggest that research teams might adapt to smaller groups, increasing, perhaps, their external collaboration. 7 Conclusion Over the last two decades, university systems experienced substantial changes. A reduction of government appropriation linked to the reduced public support for public funding to universities and university marketization which culminated in austerity measures during the Great Recession. However, at the same time, enrolment has increased, peripheral university systems formalized doctoral programs, and number of doctoral graduates increased (Mangematin, 2000). These changes and transformations make the evaluation of budget cuts to higher education difficult to be identified. To overcome this issue, in this paper, I have considered the staggered cuts of PhD scholarships made by Orb´ an to expand his political influence over the university system. While past research has highlighted the general tendency of authoritarian regimes to exert control over the size and composition of the student body in higher education (Bautista et al., 2022; Gr¨ uttner and Connelly, 2005), in the case of Hungary, the cuts were mostly unexpected and staggered. Moreover, the political objective of Orb´ an’s government ensures that cuts were exogenous to the economic cycle and to the scientific activity of universities. I provided a causal estimation of the impact of scholarship cuts on Science using a generalized difference-in-differences approach. My results highlight an important trade-off. While the government’s reduction of PhD scholarships might improve efficiency by increasing the graduation probability of students; it does so at the expense of the inclusion of females and non-traditional PhD students, as well as, the quantity and quality of the scientific production. Future research is needed to understand the mechanisms behind this trade-off and examine more closely the complex dynamics of enrolments, graduation, and scientific production after the shock. References Bautista, M. A., F. Gonz´ alez, L. R. Martinez, P. Munoz, and M. Prem (2022). Dictatorship, higher education and social mobility. Available at SSRN (June 10, 2022). Bin, A., S. Salles-Filho, A. C. Spatti, J. P. Mena-Chalco, and F. A. B. Colugnati (2022). How much does a ph. d. scholarship program impact an emerging economy research performance? Scientometrics 127(12), 6935–6960. Blair, P. Q. and D. J. Deming (2020). Structural increases in demand for skill after the great recession. In AEA Papers and Proceedings, Volume 110, pp. 362–365. American Economic Association 2014 Broadway, Suite 305, Nashville, TN 37203. Bound, J., B. Braga, G. Khanna, and S. Turner (2019). Public universities: The supply side of building a skilled workforce. RSF: The Russell Sage Foundation Journal of the Social Sciences 5(5), 43–66. 30
Bound, J., B. Braga, G. Khanna, and S. Turner (2020). A passage to america: University funding and international students. American Economic Journal: Economic Policy 12(1), 97–126. Bound, J., M. F. Lovenheim, and S. Turner (2010). Why have college completion rates declined? an analysis of changing student preparation and collegiate resources. American Economic Journal: Applied Economics 2(3), 129–157. Callaway, B. and P. H. Sant’Anna (2021). Difference-in-differences with multiple time periods. Journal of Econometrics 225(2), 200–230. Chakrabarti, R., N. Gorton, and M. F. Lovenheim (2020). State investment in higher education: Effects on human capital formation, student debt, and long-term financial outcomes of students. National Bureau of Economic Research: Working Paper N. 27885. Cohodes, S. R. and J. S. Goodman (2014). Merit aid, college quality, and college completion: Massachusetts’ adams scholarship as an in-kind subsidy. American Economic Journal: Applied Economics 6(4), 251–285. Cowan, R. and G. Rossello (2018). Emergent structures in faculty hiring networks, and the effects of mobility on academic performance. Scientometrics 117, 527–562. De Paola, M., M. Ponzo, and V. Scoppa (2013). Class size effects on student achievement: heterogeneity across abilities and fields. Education Economics 21(2), 135–153. Deming, D. J. and C. R. Walters (2018). The impact of state budget cuts on us postsecondary attainment. Draft, Harvard University, 1567–1633. Franzoni, C., P. Stephan, and R. Veugelers (2022). Funding risky research. Entrepreneurship and Innovation Policy and the Economy 1(1), 103–133. Gennaioli, N., R. La Porta, F. Lopez-de Silanes, and A. Shleifer (2013). Human capital and regional development. The Quarterly journal of economics 128(1), 105–164. Goldin, C. and L. F. Katz (1999). The shaping of higher education: The formative years in the united states, 1890 to 1940. Journal of Economic Perspectives 13(1), 37–62. Gonz´ alez-Sauri, M. and G. Rossello (2023). The role of early-career university prestige stratification on the future academic performance of scholars. Research in Higher Education 64(1), 58–94. Gr¨ uttner, M. and J. Connelly (2005). Universities Under Dictatorship. Pennsylvania State University Press University Park. Kovarek, D. and G. Dobos (2023). Masking the strangulation of opposition parties as pandemic response: Austerity measures targeting the local level in Hungary. Cambridge Journal of Regions, Economy and Society 16(1), 105–117. Kov´ ats, G. F. and Z. R´ onay (2021). Academic freedom in Hungary. Open Society University Network Global Observatory on Academic Freedom at the Central European University in Vienna, Austria. 31
Main, J. B. (2014). Gender homophily, ph. d. completion, and time to degree in the humanities and humanistic social sciences. The Review of Higher Education 37(3), 349–375. Mangematin, V. (2000). Phd job market: professional trajectories and incentives during the phd. Research policy 29(6), 741–756. Mertens, A. and H. R¨ obken (2013). Does a doctoral degree pay off? an empirical analysis of rates of return of german doctorate holders. Higher education 66, 217–231. Minaya, V., T. Agasisti, and M. Bratti (2022). When need meets merit: The effect of increasing merit requirements in need-based student aid. European Economic Review 146, 104164. Rossello, G. (2021). Social transformations and labour market entry: an investigation into university systems in emerging economies. [Doctoral Thesis, Maastricht University]. ProefschriftMaken. Spilerman, S. and T. Lunde (1991). Features of educational attainment and job promotion prospects. American Journal of Sociology 97(3), 689–720. Stephan, P. E. (2010). Chapter 5 - the economics of science. In B. H. Hall and N. Rosenberg (Eds.), Handbook of The Economics of Innovation, Vol. 1, Volume 1 of Handbook of the Economics of Innovation, pp. 217–273. North-Holland. Sterlacchini, A. (2008). R&d, higher education and regional growth: Uneven linkages among european regions. Research Policy 37(6-7), 1096–1107. Sun, L. and S. Abraham (2021). Estimating dynamic treatment effects in event studies with heterogeneous treatment effects. Journal of Econometrics 225(2), 175–199. 32
Table 9: VARIABLE DESCRIPTION Variable Description Categories Percent/Mean (SD) Dependent Variables: Enrolment: A female name enters PhD training Probability that a female name enters PhD in year t 0.067 (0.245) A new surname enters PhD training Dummy equal to 1 if the student that enters in year t No 91% has a new surname and zero otherwise Yes 9% Graduation: A student graduates as PhD Probability that the student graduates in year t. It is equal to 0.270 (0.390) the number of years since enrolment over the total years of enrolment for students who graduate and zero otherwise A female name graduates as PhD Probability that a female name graduates in year t 0.119 (0.291) and zero otherwise A new surname graduates as PhD Probability that the student that graduates in year t 0.131 (0.303) has a new surname and zero otherwise Research Quantity and Quality: Student productivity The number of papers published in year t by the student 0.044 (0.309) weighted by the number of co-authors Average quality of papers published The average Scimago journal ranking of papers published 0.017 (0.203) by the PhD student in year t Treatment-Related Variables: PhD scholarship cut Dummy variable equal to one if the university saw a cut in No 54% PhD scholarship greater than the 3rd quartile of the distribution Yes 46% of scholarship cuts Group-scholarship-cuts The group cohorts when the first severe scholarship cuts occur Never 35% 2010 16% 2013 23% 2016 0.03% 2018 15% 2021 13% Universities 22 Universities where 6 are the largest DE 16% SZTE 15% PTE 13% ELTE 11% BME 10% MATE/SZIE 8% Small Uni. <5% 27% Student’s Controls: Student productivity average Average number of papers weighted by the number 0.099 (0.562) of co-authors per publication Student quality average Quality expressed as average Scimago journal ranking 0.033 (0.231) of her/his previous publications, Advisor’s Controls: Probability that the Advisor is Male Probability that the advisor’s name is masculine 0.777 (0.398) Advisor productivity average Average number of papers weighted by the number 1.484 (1.239) of co-authors per publication Advisor quality average Quality expressed as average Scimago journal ranking 1.098 (1.350) of her/his previous publications, Number of students supervised number of students supervised by the advisor in the 0.646 (1.036) previous 3 years including the actual one. University Controls: Number of PhDs Number of PhDs students enrolled in year t by the university 921 (546) Funds for research Total millions (HUF) of national research funding 586 (456) grants won or existing at year t in the university Scientific Production Quantity Number of papers in SCOPUS published by the university 839 (495) in year t 33
TheUNU‐MERITWORKINGPaperSeries 2023-01 Caninternationalmobilityshapestudents'attitudestowardinequality?The BraziliancasebyCintiaDeniseGranja,FabianaVisentinandAnaMariaCarneiro 2023-02 Demand‐ledindustrialisationpolicyinadual‐sectorsmallopeneconomybyÖnder Nomaler,DaniloSpinolaandBartVerspagen 2023-03 Reshoring,nearshoringanddevelopingcountries:Readinessandimplicationsfor LatinAmericabyCarloPietrobelliandCeciliaSeri 2023-04 Theroleofproductdigitizationforproductivity:Evidencefromweb‐scraping Europeanhigh‐techcompanywebsitesbyTorbenSchubert,SajadAshouri, MatthiasDeschryvere,AngelaJäger,FabianaVisentin,ScottCunningham,Arash Hajikhani,LukasPukelisandArhoSuominen 2023-05 Morethanafeeling:Aglobaleconomicvaluationofsubjectivewellbeingdamages resultingfromrisingtemperaturesbyStephanDietrichandStaffordNichols 2023-06 WasRobertGibratright?Atestbasedonthegraphicalmodelmethodologyby MarcoGuerzoni,LuigiRisoandMarcoVivarelli 2023-07 Predictingsocialassistancebeneficiaries:Onthesocialwelfaredamageofdata biasesbyStephanDietrich,DanieleMalerbaandFranziskaGassmann 2023-08 Relatedorunrelateddiversification:Whatissmartspecialization?ByÖnder NomalerandBartVerspagen 2023-09 Breachofacademicvaluesanddigitaldeviantbehaviour:ThecaseofSci‐Hubby GiuliaRosselloandAriannaMartinelli 2023-10 Theeffectoflobbies’narrativesonacademics'perceptionsofscientificpublishing: AninformationprovisionexperimentbyGiuliaRosselloandAriannaMartinelli 2023-11 Makingimpactwithagriculturaldevelopmentprojects:Theuseofinnovative machinelearningmethodologytounderstandthedevelopmentaidfieldbyLindsey Moore,MindelvandeLaar,PuiHangWongandCathalO’Donoghue 2023-12 GreenwindowsofopportunityintheGlobalSouthbyRasmusLemaandRoberta Rabellotti 2023-13 Thegreenanddigitaltransitioninmanufacturingglobalvaluechainsinlatecomer countriesbyRasmusLemaandRobertaRabellotti 2023-14 Impactofprestigious‐STEMEducationofcorporateboardmembersoninnovation effort:EvidencefromIndiabyRituparnaKaushik,SourabhBikasPaulandDanilo Spinola 2023-15 Reducingenvironmentalimpactthroughsharedownership:Amodelofconsumer behaviourbyFrancescoPasimeniandTommasoCiarli 2023-16 Welfarelosses,preferencesforredistribution,andpoliticalparticipation:Evidence fromtheUnitedKingdom’sageofausteritybyPatriciaJustino,BrunoMartorano andLauraMetzger 2023-17 Inequality,socialmobilityandredistributivepreferencesbyIsabelGüntherand BrunoMartorano 2023-18 Automation‐inducedreshoringandpotentialimplicationsfordevelopingeconomies byHubertNii‐Aponsah,BartVerspagenandPierreMohnen 2023-19 EthnicspatialdispersionandimmigrantidentitybyAmelieF.Constant,Simone SchüllerandKlausF.Zimmermann 2023-20 PublicR&DandGrowth:AdynamicPanelVECMDataAnalysisfor14OECD CountriesbyThomasH.W.Ziesemebyr
2023-21 Innovationandthelabormarket:Theory,evidenceandchallengesbyNicoletta Corrocher,DanieleMoschella,JacopoStaccioliandMarcoVivarelli 2023-22 RevisitingSchumpeterinEurope:Place‐basedinnovationandtransformative industrialpolicybyLucSoeteandJohanStierna 2023-23 IdentificationofFourthIndustrialRevolutiontechnologiesusingPATSTATdataby MaríadelasMercedesMenéndez,ÖnderNomalerandBartVerspagen 2023-24 Expectationsandthestabilityofstock‐flowconsistentmodelsbyHuubMeijers, JoanMuyskenandGiuliaPiccillo 2023-25 Criticalmineralsandcountries'miningcompetitiveness:Anestimatethrough economiccomplexitytechniquesbyJorgeValverdeCarbonell,Maríadelas MercedesMenéndez,andCarloPietrobelli 2023-26 InternalratesofreturnforpublicR&DfromVECMestimatesfor17OECDcountries byThomasH.W.Ziesemer 2023-27 Doesmoraltransgressionpromoteanti‐socialbehavior?Evidencefromlab‐in‐the‐ fieldexperimentsbyHalefomYigzawNigus,EleonoraNillesen,PierreMohnenand SalvatoreDiFalco 2023-28 Doesexposuretomarketspromoteinvestmentbehavior?Evidencefromrural EthiopiabyHalefomYigzawNigus,PierreMohnenandEleonoraNillesen 2023-29 Thevirtuousloopofqualityofgovernment(QoG)andinstitutionaltrustinOECD countries,2006‐2021andculturebyAhmedHussainandJoRitzen 2023-30 Alternativerelatedvariety,macroeconomicfactorsanddiversification:Extractive vs.non‐extractiveproductsbyBeatrizCalzadaOlvera 2023-31 TheeffectofgovernmentcutsofdoctoralscholarshipsonsciencebyGiuliaRossello