Retrenchments in higher education and an equal burden of the costs? About the effect of the abolition of the basic student grant on students’ income composition in the Netherlands

F.L. Stuurwold, s1474790 Master Thesis Public Administration: Economics & Governance

06-06-2019 Words: 17.227 Thesis supervisor: E. Suari Andreu Second reader: Prof. dr. M.G. Knoef

RETRENCHMENTS IN

HIGHER EDUCATION

AND AN EQUAL

BURDEN OF THE

COSTS?

About the effect of the abolition of the basic

student grant on students’ income composition

in the Netherlands

Abstract

INDEX

INDEX ... 1

1. INTRODUCTION ... 3

2. INSTITUTIONAL CONTEXT ... 6

3. LITERATURE REVIEW ... 6

§3.1 Expected effects of the policy reform prior to enactment ... 6

§3.2 Research concerning the income of students after enactment of the policy reform ... 7

§3.2.1 European research on income composition of students in higher education ... 7

§3.2.2 Dutch research on income composition of students in higher education ... 8

§3.2.3 Dutch research on income composition of students in higher education one year before and after the abolition ... 10

4. THEORETICAL FRAMEWORK ... 13

§4.1 Financial funding of education ... 14

§4.2 Study behavior: Principal-Agent Theory ... 15

§4.3 Students’ income composition ... 16

5. METHODOLOGY ... 18

§5.1 Case-selection ... 18

§5.2 Operationalization of concepts and variables ... 20

§5.3 Method of analysis ... 21

§5.4 Validity and reliability ... 23

6. DATA ... 24

§ 6.1 Description of the data ... 24

§ 6.2 Method of data collection ... 24

§ 6.3 Descriptive statistics ... 25

§6.3.1 Financial Funding of Education ... 26

§6.3.2 Study behavior ... 26

§6.3.3 Income composition ... 28

§6.3.4. Housing ... 32

§7.2. Effect on Time-investment and Motivation ... 37

§7.3 Effect on income composition ... 40

§7.3.1. Borrowed loan ... 41

§7.3.2. Work: Self-earned income ... 44

§ 7.3.3 Financial support from parents ... 46

§7.4 Effect of social background – interaction effect ... 48

8. CONCLUSION ... 51

9. DISCUSSION ... 54

1. Introduction

Recently, the financial policy for students in higher education changed. This new policy, which started in the academic year ’15-’16, contains the abolition of the basic student grant which is a monthly gift of the government of around 290 euro for students in higher education. The government wanted to invest those savings in the quality of higher education (Parliamentary Papers II, 34035, no. 12, 2014). Besides, it was argued in politics, that increased future earnings of higher educated people should be funded by private investment rather than government investment through the basic student grant. During the policy reform and the years after, politicians are concerned about the accessibility of higher education and income effects for students, particularly for students with a lower financial background (Parliamentary Papers II, 24725, no. 113, 2013). This leads to the following research question: To what extent did the abolition of the basic student grant influence the study behavior and income composition of students in the Netherlands between 2011-2018? There is a focus on this time period because the abolition of the basic student grant and the new policy ‘sociaal leenstelsel’ entered into force at the start of the academic year 2015-2016. With this timeframe of 2011 to 2018, the situation in the years before and after the reform can be compared in order to analyze the effects of this new policy.

Recent research has shown that this new policy leads to a slight decrease in the accessibility of higher education (Researchned, 2018). Besides, Researchned (2018) and Nibud (2017, p.12) stated that the number of students living without their parents decreased after the abolition of the basic student grant. When focusing on different social classes, there were no differences found for study- and housing decisions for students with different financial backgrounds after the policy reform. This means that after the policy reform, the social economic background did not influence the decisions of students to go studying or to live without parents (Van den Berg & van Gaalen, 2018). While these relevant recent research about the policy reform, there is no multivariate analysis performed about the effects of this policy reform. Therefore, the topic of the thesis is firstly motivated by the lack of multivariate analysis of the relevant effects of this major policy reform. Secondly, the topic of the thesis is motivated by the actuality and political relevance of this topic and the major impact of the policy reform

In this thesis, the broader effects of the abolition of the basic student grant will be analyzed. These effects are subdivided into three parts; financial funding of education, study behavior, and income composition. This distinction is made because it contains the essential parts of the new policy. Firstly, the rationale behind the policy, which is the financial funding of education. Secondly, the difference in students’ behavior before and after the policy reform will be discussed. This will be analyzed by focusing on study time investment and motivation. Thirdly, the financial effects on the students’ income composition will be analyzed before and after the policy reform. It will be analyzed how students did financially compensate the monthly gift of the government. Thereby, in the last paragraph, there is a specific focus on the financial effects for students with different financial backgrounds.

These three main parts of the thesis; financial funding of education, study behavior and income composition, are based on previous research and theories that focus on these topics (Adamson, 2016) (Bishop and Wößmann, 2004) (Eurostudent VI, 2016) (Nibud, 2017) (Researchned, 2018). The previous research will be discussed further in detail in the literature review while the theories will be explained in the theoretical framework. The previous research focused merely on descriptive statistics for those topics, but lack of relevant statistical information through multivariate analyses. As this thesis performs a multivariate analysis of these broader effects, it is scientifically relevant. In that way, a gap in the literature can be filled. Furthermore, the results give more insights into the current unknown effects of the new policy and could lead to policy adjustments. Because the current policy affects the whole student population and their financial situation, the results also have practical relevance. With a complete overview of the effects of the current policy, the policy could be adjusted to the most necessary needs of the current student population, and it becomes possible to correct adverse side effects. A possible adverse effect could be that students of lower social backgrounds are financially more affected by the people in comparison with students with a high social background, which was not the rationale behind the policy reform. Therefore, this aspect will also be analyzed.

The central methodology that will be used is a regression discontinuity design. This design will be used as it measures the effect of a policy before and after a policy reform for students with the old and new policy, which is the main point of interest in this thesis. The independent variable is the policy reform. The dependent variables are based on the three main topics in this research, namely financial funding of education, study behavior and income composition. The operationalization of these topics is further specified in the methodology section. The data with which the effects are analyzed comes from the ‘Studentenmonitor 2011-2018’ which contains data of 151.032 observations of students in higher education between 2011-2018. This data is focused on the social-economic developments of students in higher education. The data is at the student level and collected with a survey which contains data of 18.000 students per year. With this methodology and data, this research is structured as follows; at first, the institutional context will provide an overview of the political playing field and rationales behind this policy reform. Secondly, a review of recent literature and relevant theories will be given in the subsequent two chapters. Thereby, based on the recent literature and relevant theories, the hypotheses are formed. Thirdly, the methodology and data of this research will be elaborated on. Fourth, the results for the three main topics in this research will be analyzed and the hypotheses will be either confirmed or rejected. At last, there will be a closing chapter with the conclusion and policy recommendations.

The results have shown that bachelor students with the new policy slightly perform better, are slightly more motivated, have a higher borrowed loan, do not work more but receive more financial support from their parents in comparison with students with the old policy. Focusing on those financial aspects, students with the new policy and with a low social background do not significantly borrow more, work more or receive less from their parents in comparison to students with the new policy and a high social background. There already existed a difference between loan, working hours and financial support for students with a high social background compared to students with low social background, but this difference did not increase after the policy reform. Therefore, there is no different effect found for students from high- and low social classes when focusing on their increased average loan, average working hours or financial support from parents after the policy reform.

2. Institutional Context

The coalition agreement between the political parties 'VVD' and 'PvdA', called ‘Bruggen slaan’, with support from the political parties ‘D66’ and ‘GroenLinks’, formed the foundation for the policy reform in higher education. The policy reform entered into force at the start of the academic year 2015-2016. This new policy contains the abolition of the basic student grant which was a monthly gift of around 290 euro for students in higher education. Consequently, students can now only borrow the money from DUO or need to generate another source of income, instead of receiving a monthly gift of the government. The idea behind this policy reform was that students should invest more in their own future through education; not mainly the government. The new policy was justified through the rationale that high-educated people have a higher future income and should therefore invest more by themselves and that this investment should not also be paid, for a big part, by all the civilians through the tax system. This new policy saves substantial costs for the government. Besides, the government could invest these savings in the quality of education (Van den Berg & van Gaalen, 2018). The minister of education at that time stated that the investment was necessary to improve the quality of higher education, and therefore students would also benefit from the policy reform (Parliamentary Papers II, 34035, no. 20-9-1/2, 2014)

3. Literature Review

This literature review will provide an overview of previous research concerning the effects of this policy reform in higher education. At first, the expected effects according to several scholars in the literature of the abolition of the basic student grant will be discussed. Thereafter, more recent research about this topic will be discussed, focused on (the composition of) the income of students in the Netherlands between 2011 and 2018.

§3.1 Expected effects of the policy reform prior to enactment

Within the current literature, there is an ongoing debate concerning the expected effects of the abolition of the basic student grant. Before this abolition was enacted, Jongbloed and Vossensteyn (2012) posed two claims on the effects. Firstly, they argued that the accessibility of higher education would be slightly reduced. This claim was based on CPB calculations. In 2013, the CPB (2013) performed research on the

expected effects on accessibility by the abolition of the basic student grant. The CPB (2013, p.5) expected a decrease of 1.5% - 2.1% in the total student population. Secondly, Jongbloed and Vossensteyn (2012) argued that the fear of borrowing among students would not become a problem. This expectation is based upon the claim that higher education will lead to a higher income in the future.

The lion’s share of the research performed prior to the abolition of the basic student grant focused on the expected effects on the accessibility of higher education and for fear of borrowing. Research performed after the abolition of the basic student grant also focused on the effects on the composition of the income of students and study-and housing decisions of students.

§3.2 Research concerning the income of students after enactment of the policy reform

§3.2.1 European research on income composition of students in higher education

The discussion on the composition of the income of students is also performed on a European scale. Within this European scope, Eurostudent focuses on social and economic conditions of student life in Europe (Eurostudent VI, 2016). The Eurostudent survey focused on students’ resources; for instance, the composition of the student income in Europe. According to Eurostudent VI (2016), the income compositions for students in the Netherlands is as follows; 32% from family/partner support, 24% is self-earned income and 33% is national public student support (Eurostudent VI, 2016 p.159). Furthermore, for the large majority of countries, more than 80% of all students receive the most substantial part of their total income from family or partner. For students not living with parents, the average share of the income of those who receive contributions from family/parent has increased across Eurostudent countries from 68% to 78% between the years 2016 and 2018 (Eurostudent VI, 2016 p.147). Compared to other countries, The Netherlands has one of the lowest family/partner contributions, namely 32%. Besides, the percentage of self-earned income is also relatively low (24%) in the Netherlands compared to the other countries, with an average of 34%. National public student support (loan or gift) is 33% in The Netherlands, on average 14% (Eurostudent VI, p.159).

This research provides insight into the situation in the Netherlands compared to other European countries and their study finance system and income composition of students in those countries. The Netherlands has a relatively high percentage of national public student support, and a relatively low percentage of self-earned income. Parents contributions are also relatively low in the Nehterlands compared to other European countries. Because of this comparison, this research gives a broader context to the income composition of students compared to other European countries. Similar to other research, this research contains descriptive statistics and comparisons between the European countries. However, it lacks the multivariate analysis of the data for analyzing any significant effects or causes. Besides, it compares public support between different countries, while this contains both loans and gifts. In the context of this thesis, it therefore does not give the essential information to see the effect of the new policy.

§3.2.2 Dutch research on income composition of students in higher education Researchned (2018) assessed the borrowing behavior of Dutch students before- and after the abolition of the basic student grant. ResearchNed (2018) used the information of the ‘Studentenmonitor Hoger Onderwijs’ between the years 2011 and 2018. This dataset entails data of around 150.000 students within the Netherlands. Researchned (2018) states that there was an increase in the percentage of people who borrowed money from DUO. For all bachelor students (HBO/WO together), the percentage which borrowed in 2012-2013 was 23%, in 17-18 this percentage increased to 49%. Even though, by measuring only WO (University) bachelor students, this percentage increased even more, from 27% to 57%.

2017 2016 2015 2014 2013 2012 2011 Hbo 43% 39% 23% 21% 21% 21% 19%

Wo 57% 50% 31% 28% 26% 27% 26%

Total 49% 43% 26% 24% 23% 23% 22%

Table 3.1 Percentage of students who borrow from DUO 2011-2018 Source: Studentenmonitor(2011-2018)

ResearchNed (2018) has also compared first-year students and older students between ‘06-‘07 and ‘17-‘18 who borrowed money from DUO (figure 3.1). The percentage of first-year students who borrowed increased after the abolition of the basic student grant from 27% to 48%. This percentage increased less for older year students after ‘15-’16, namely from 40% to 45% (ResearchNed, p.187). Even though,

these descriptive statistics show the data for different variables over time, like percentage of students who borrow from DUO. However, it does not control for any difference in the academic years, and therefore the difference could also be caused by other factors instead of the policy reform. Therefore, it is not possible to derive a direct causal effect from the descriptive statistics.

Figure 3.1 Percentage of students who borrow from DUO 2006-2017. Source: Researchned (2018, p. 187)

Figure 3.2, provided from the research of Researched (2018, p.190), shows the average loan of students between ‘06-‘07 and ‘16-’17. The reform year, ’15-‘16, shows an increase of the average loan from around €400 to €500 a month. It is therefore relevant to analyze how this trend developed in the following years after the reform (2017-2018) and whether this trend is caused by the policy reform. This thesis will therefore compare the years after the reform with the years prior to the reform, by means of a multivariate analysis to analyze this possible trend.

Figure 3.2 Average loan per month in euros between 2006-2016. Source: Researchned (2018, p.190)

First-year Third-year or older First-year Third-year or older First-year Third-year or older

Figure 3.3 (Researchned, 2018, p.198) shows the difference in sources from which students get their money and how this has changed over years for students with the old and new policy. It shows that in the years after the reform, bachelor students with the new policy get their income more often from a combination of borrowing and working (33%) in comparison to students with the old policy (28%). Figure 3.3 shows this possible trend, but it lacks a multivariate analysis of this trend to analyze direct causal effects. It could be the case that the policy reform has caused that students borrow and work more, but without using control variables, this trend could also cause by other factors, like more first-and second year students in the group who receive the new policy. This characteristic is likely due to the new policy that entered into force in the academic year ‘15-‘16 for first-year students (bachelor and master). Students who already started before ‘15-‘16, did not directly get into the new system. This makes that both groups are likely not equal to each other with regard to year in study. Thereby, without using control variables, it fails to compare the same groups of students with the old and new policy over the years. Therefore, these descriptive statistics cannot analyze a direct causal effect of the policy reform. While the income composition of students after the policy reform is one of the main interests of this thesis, the income composition will be further analyzed by a multivariate analysis.

§3.2.3 Dutch research on income composition of students in higher education one year before and after the abolition

Whereas the aforementioned recent research focused on the expectations of the effects of the abolition (Jongbloed & Vossensteyn, 2012) (CPB, 2013), research mentioned in §3.2.1 and §3.2.2, focused on the effects of the policy reform at the European and Dutch level. This paragraph will focus more on a specific time period,

Borrowing and working

Only

borrowing Only working Not borrowing /not working Old policy New policy

Figure 3.3 Income composition between 2015-2017 for students with old and new policy. Source: Researchned (2018, p.198)

namely one year before -and after the policy reform. Nibud (2017) performed research before and after this policy reform, so between 2015 and 2017. Nibud (2017) focused on Dutch students and compared the students’ financial situation between 2015 and 2017. This research represents the results of a research among 1383 fulltime HBO and wo-students between 16-30 year. However, it also did not contain a multivariate analysis of the data. The research performed by Nibud (2017) provides insights into the descriptive statistics of the borrowed amount of money, hours worked, the financial support of parents, the three main parts of income composition in this thesis. The following three sections address those three main parts of this income composition.

§3.2.3.1 Borrowed amount of money

Concerning the first part of the income composition (student loans) in 2017, 55% of the student population in the Netherlands received a loan from DUO. Students in the new system have more often a loan from DUO in comparison to students in the old system (63% to 43%) (Nibud, 2017). The average between those groups is also different. When only accounting for students who take upon a loan, the average loan is 512 euros a month for the first group, compared to 464 euros a month for the latter group. In 2015, 28% of all first-year students borrowed money from DUO, whereas 62% of the first-year students in the new system borrowed money from DUO (Nibud, 2017, p.11-12). So, this research (Nibud, 2017) shows an increase in the average monthly loan and an increase in the percentage of students who borrow.

§3.2.3.2 Work

Concerning the second part of the income composition, being self-earned income, in 2017, 70% of the student population had a job or internship. They worked on average 17 hours a week. In 2015, students worked 15 hours a week. Therefore, their average received income from work increased from 332 euro to 409 euro (Nibud, 2017, p.11-12). This shows that after the policy reform, students work two hours more per week. A disadvantage is that, in this case, groups of students with different education years were compared. This makes the comparison between those groups not entirely reliable for the effect of the abolition of the basic student grant on working hours. As mentioned before, this is a recurrent problem when showing only descriptive statistics without using control variables.

§3.2.3.3 Financial support Parents

For the third part of the income composition, financial support from parents, for 2017, 58% of the students received financial support from their parents with an average of 165 euros a month. Younger students received more often money from their parents compared to older students in the old system. This same is true for students who are in the new system ‘studievoorschot’ respectively 63% and 53%. This group did not receive a higher amount of money after the policy reform. Compared to 2015, students received on average 179 euro. In conclusion, a slight increase is perceived. However, this can be explained through the lower ratio of first and second-year students who lived on their own. In 2017 fewer students lived on their own, and therefore needed less money on average because those students did not need to rent student accommodation (Nibud, 2017).

§3.2.3.4 Housing situation

The three sections above touched upon the composition of the income of students. A possible relation to the income composition is the housing situation of students. Therefore, the housing situation of students is addressed. In 2015, 65% of the student population lived on their own. For the first- and second year students, this percentage was 61%. Compared to 2017, only 44% of the first- and second-year students lived on their own (Nibud, 2017, p.12). So, this research has shown that after the policy reform, the percentage of students who lived on their own decreased. This is relevant because the housing situation could influence the income composition of students, like the average loan or financial support from parents (Researchned, 2018, p.187).

To sum up this literature review, some scholars have performed research on the expected effects of the abolition of the basic student grant (Jongbloed and Vossensteyn, 2012). Scholars who performed a comparative analysis on the situation both prior and after the abolition, focused on both the European and on the Dutch level (Eurostudents VI, 2016) (Researchned, 2018) (Nibud, 2017). For all these scholarly articles, merely descriptive statistics are used, and no multivariate analysis is performed. As a consequence, different groups of students are compared despite their substantial differences. For example, as shown in this literature review, using year of higher education as a control variable is important because those students differ from

each other in the amount they borrow (Researchned, 2018, p. 187). Therefore, the results are not fully reliable when two different groups of students are being compared.

This literature review has shown the trends who are founded after the policy reform. Even though, these trends, for example the average increased borrowed loan, are not performed by a multivariate analysis which makes the results not fully reliable because different groups of students where compared. The methodology will therefore focus more in detail on those differences between students with the old and new policy and will explain the control variables in this thesis. While recent research lack of multivariate analysis, it is this gap in the literature that this thesis will fulfill. This thesis performs a multivariate analysis whilst controlling for the different characteristics of the groups of students.

4. Theoretical Framework

The theoretical framework focuses on the effects before and after the abolition of the basic student grant and is divided into three main topics, financial funding of higher education, study behavior and income composition.

The first topic, the kind of financial funding of higher education, will be analyzed by the theory of Adamson (2009) who stated that a higher private investment will lead to a lower student performance. This is relevant because the policy reform has also led to a higher private investment in education (Eurostudent VI, 2016). This will be tested in H1. The second topic that is studied is the behavior of students after the policy reform. Study behavior is subdivided into time-investment, according to the principal-agent theory, and motivation. In the principal-agent theory are two main actors, the government and the students. To analyze their behavior, the principal-agent theory will be tested in the context of public spending on education and time investment of students. The government gives students no gift anymore and therefore H2 predict that the there is a negative effect on time investment of students in this principal-agent relationship after the policy reform. The motivation of students is a related issue in this context because studying become also more expensive after the policy reform. Therefore, H3 focus on the effect of the policy reform on the motivation level of students.

The third topic in the theoretical framework focuses on the income composition of students. It focuses specifically on effects of the policy reform on the three major income sources; borrowed loan, working hours and financial support from parents. This effects of the policy reform on these three main income sources will be tested in H4 through H6. Besides these three main sources of income, the financial background could interact with those variables. In this context will be analyzed if the effects of the policy reform on the three sources of income are significantly different for students with a low and a high social background. The effects of these interaction variables are incorporated in the hypotheses H7 till H9.

§4.1 Financial funding of education

Recent scholars write about the debate between private or public funding of education and the effect on students’ performance. In this context of the policy reform and the rationale for quality improvement of higher education, it is relevant to look at the effect of kind of funding of education and the effect on student performance. Adamson (2016, p.1), stated that education should be funded by public investment instead of private investment to improve the student performance. Adamson (2016, p.9) stated in his article;

‘The data suggest that the education sector is better served by a public investment approach that supports each and every child than by a market-based, competition approach that creates winners…and losers.’

Year 2010 2016

Family or partner-support 23% 32% Self-earned income 24% 24% Private investment 47% 56% Public sources* 46% 33%

Table 4.1 - Income composition of students between 2010-2016 Source: Eurostudent IV, VI, Data for students not living with parents * Loan and gifts together

Adamson (2016, p. 9) conclude that high-quality public investment that uses equity-based processes and focuses on teacher professionalization is often accompanied by high education outcomes. In comparison, a market-based system of education privations if often accompanied by low and unequal student performance and public dissatisfaction (Adamson, 2016, p.9). It is therefore relevant to analyze if there is a

difference in student performance before-and after the policy reform because with this new policy, a bigger part of education is not funded by the government anymore, but with a private investment (table 4.1). With the rationale behind the policy reform, quality improvement, it is relevant to analyze if the conclusion of Adamson (2016) are also applicable for higher education in The Netherlands. According to the findings of Adamson (2016), the first hypotheses is;

H1: Higher private investment rate (self-earned or from family) in education leads to lower students’ performance.

Students performance is measured in average exam grades on a 1-10 scale.

§4.2 Study behavior: Principal-Agent Theory

Bishop and Wößmann (2004) made a model for study behavior in higher education, based on the principal-agent theory. In the principal-agent theory are two actors, the principal who make decisions about the agent, and the agent who make a decision upon the instruction of the agent. The behavior of the agent can be motivated by a general and a personal interest. When the personal interest in contrary to the interest of the principal and there is asymmetric information, the principal-agent problem occurs. Asymmetric information is the situation where both actors do not possess the same information. In this theory, the principal has no or few information about the behavior or decisions of the agent.

This principle of this theory can be applied to the education studies context. Bishop and Wößmann (2004) made a model based on this principle and applied to higher education context. The principal is the government who chooses the level of educational spending that maximizes its net benefits, given students’ effort. The government reflects the public interest. Students are the agents and choose the level of their effort that maximizes their net benefits, given the government’s spending choice. This is measured by the opportunity costs of the students’ time. With decreased government spending between 2010 and 2016, it is expected that student time investment is decreased. Therefore, the second hypothesis is:

H2: With decreased government spending on education, student invest less time in their study

Besides, when studying become more expensive after the policy reform, it is expected that the lowest motivated students have a lower change to start studying. According to the Second Camber debates in 2016 (Parliamentary Papers II, 90-9-8, 2016) about the new policy, politicians state that higher education must stay accessible for the students who are highly motivated to develop themselves. It is therefore relevant if there is an indeed an effect found on the average motivation level of the students after the policy reform. If less low motivated students started in higher education after the policy reform, it is expected that the average motivation level has increase after the policy reform. Accordingly, the third hypothesis is:

H3: A higher private investment rate (self-earned or from family) in education leads to higher motivated students

§4.3 Students’ income composition

Previous research focused on expected effects after the abolition of the basic student grant. Jongbloed and Vossensteyn (2012), expected that student needs to borrow more after the abolition of the basic student grant. Recent research after the abolition did also describe an increasing trend in the average borrowed loan (Researched, 2018). It is therefore relevant to analyze if this trend is actually found when analyzed by a multivariate analysis including control variables. The fourth hypothesis in this thesis is therefore:

H4: After the abolition of the basic student grant, students borrow more on average If students borrow more on average after the abolition of the basic student grant, it is interesting to research if there is a comparable effect after the abolition on working hours, another possible source of income. Nibud (2017) found that students do work more after the policy reform, but this is only analyzed by descriptive statistics on a quite small sample of students. It is therefore relevant to analyze if this trend is actually true, when analyzed by a multivariate analysis including control variables. Therefore, the fifth hypothesis is:

H5: After the abolition of the basic student grant, students work more on average A third source of income is the financial support from parents. While the abolition of the basic scholarship has caused a financial gap for students, it is relevant if parents did compensate this decrease in income. Nibud (2017) state that in 2017, one year

after the abolition, that the family support from parents increases on average after the abolition of the basic student grant. Based on this previous research, the sixth hypothesis is:

H6: After the abolition of the basic student grant, support from parents increased H4 till H6 gives no expectations about the financial effects for students from different social economics classes. It is therefore also relevant to analyze what the financial effects are for students with different financial backgrounds. In former research, there was no focus on difference between parents with high and low income. It could be that these groups experience different effects on their income composition. In the context of equal opportunities, it is relevant to analyze if there are different effects, according to their income composition, for students in the new system which have different financial backgrounds. Therefore, the following three hypotheses are formed:

H7: Students from a lower financial-social background, borrowed more on average than students from a higher financial-social background, this difference increased after the policy reform.

H8: Students from a higher financial-social background receive more financial support from their parents in comparison to students lower financial-social background, this difference increased after the policy reform.

H9: Students from a higher financial-social background work less besides their study in comparison to students with a lower social background, this difference increased after the policy reform.

As is mentioned in this theoretical framework, two relevant theories will be tested in this context of the policy reform; the theory about financial funding of higher education (Adamson, 2009) and the principal-agent theory (Bishop and Wößmann,2004). These will be tested in H1 and H2. Besides these theories, previous scholars researched the effect of the abolition of the basic student grant. Based on these previous researches, H4 till H9 summarizes these expected effects, even though there were mainly based on descriptive statistics instead of a multivariate analysis. The next chapter will describe the specific methodology that will be used to do the multivariate analysis to research the particular effects of the policy reform.

5. Methodology

This research has a deductive research method and is based on previous research and theories about higher education. The method to research the question of this thesis is based on a quantitative research design with statistical data as the main source for the research. A regression discontinuity design will be used because it is able measure the effect of a policy before and after a policy reform for students with the old and new policy, which is the main point of interest in this thesis. This method will be explained more in detail in paragraph 5.3. The data that is used for this regression discontinuity design in this thesis originates from the ‘studentenmonitor hoger onderwijs 2011-2018’, which is described in more detail in chapter six.

§5.1 Case-selection

In this research, full-time bachelor students who are studying at a university of applied science (HBO) or at a university and are questioned between 2011 and 2018 are in the researched population. There is a focus on fulltime bachelor students because bachelor students are the biggest group that experience the differences of the new policy. Besides, those full-time bachelor students study for a longer time compared to master students. Therefore, they experience the effects of the policy reform over a longer time period. Furthermore, when comparing master students before and after the reform, a large number of master students before the reform already did not receive the student grant of the old policy. Therefore, it would be inaccurate to compare master students as not all of them studied with the old policy conditions before the reform. It is therefore more accurate to compare cohorts of bachelor students before and after the reform who all studied with the old or new policy conditions.

By analyzing only bachelor students, the results will not be blurred by the averages for bachelor and master students together. If analyzing both groups together, it would not be possible to make conclusions for those particular groups later on. That would make that results less usable in comparison to results which focus on a particular group of students.

In this research, students who started in cohort of 2015-2016 or later, are in the treatment group. In this way, receiving treatment is dependent on the cohort where the

students are in. The first year that a student started with higher education, thus the cohort, is therefore the running variable. In the control group are students who have started studying before the policy reform, thus the cohorts before 2015-2016. Those students study with the old policy conditions. By making this comparison, there must be controlled for the fact that in the treatment group are more younger students than in the control groups and that the groups also differ in characteristics like gender, age, social class, kind of- and year of education. The performed t-test in table 5.1 shows that across these dimensions the treatment- and control group differ significantly in their means for these six control variables. Because those cohorts differ from each other significantly, it can affect the outcome variables. Therefore, it is chosen to control in the regressions for those differences across the cohorts.

Treatment group Control group Difference

Gender 0.661 (0.004) 0.652 (0.002) -0.009* (0.004) Age 19.99 (0.012) 21.35 (0.015) 1.36*** (0.025) Social class 5.82 (0.013) 6.43 (0.006) 0.605*** (0.014) HBO/UNIV 1.51 (0.004) 1.44 (0.002) -0.074*** (0.004) Year of education 2.099 (0.007) 2.47 (0.004) 0.373*** (0.009) Native/migrant 0.094 (0.002) 0.0822 (0.001) -0.012***(0.002) Sample size (n) 16335 59517

By using control variables for the aforementioned characteristics, the groups are made as equal as possible. It is important to realize this difference in characteristics in those groups, has two consequences. At first, this proves that previous research compares significantly different groups with each other and makes conclusions based on significantly different groups. Second, this research will therefore use a regression discontinuity design which controls for these variables where the groups otherwise will be significantly different. When controlling for these variables, the real effect will be researched, and not the effect of two significantly different groups. In that way, the only thing that changes is the receival of the treatment, namely the policy reform.

Table 5.1 Means and differences of the used control variables between treatment and control group. The difference column (third column) shows the difference between the means for that particular control variable, calculated by a two-sample t-test. Standard errors are shown in parentheses. The group means are significantly different as the p-value is less than 0.05. *** means that the group means are significantly different for the particular control variable at the 99,99% level (p=0.00), ** = 99% level (P<=0.01), * = 95% level (P<=0.05).

There is chosen to not control for the academic year because it otherwise has collinearity with starting year (that is used for determining the different cohorts) by using in the same regression year of higher education as a control variable. This is because the academic year has collinearity with start year because students with the new policy are automatically in the same academic years. So, when measuring starting year and academic year in one regression, there is collinearity between those variables which leads to false results. There is chosen to exclude academic year and including the year of higher education because it is expected that the composition over the academic years is not very different when still controlling for enough other control variables (gender, age, social class, year of higher education, UAS/UNI and immigrant/native). Therefore, is chosen to still control for the year of higher education because, over the years, this composition is significantly changed in the control and treatment group (shown in table 5.1). This is because the control group contains after the reform year a relatively a high number of students who are third- or fourth-year students and a relatively low amount of first-and second-year students (see Table 7.1). Therefore, it is essential to control for the year of higher education while the groups of students differ significantly in that characteristic.

Academic year First year Second year Third year Fourth year total 11-12 2762 2609 2851 1130 9352 12-13 4271 3567 3470 1982 13290 13-14 3520 2394 3048 1737 10699 14-15 3481 3241 3075 2048 11845 15-16 351 1811 2259 1335 5756 16-17 150 580 2069 2518 5317 17-18 140 375 886 1808 3209

Table 7.1: Number of students with old policy (control group) over the years. This table shows that year of higher education is an important control variable because those groups contain very different amounts of students with the old policy.

§5.2 Operationalization of concepts and variables

To operationalize the concepts from the theoretical framework and variables, figure 5.1 is built to create an overview of the used theories, concepts and variables. To analyze the effects of the policy reform on the dependent variables, a treatment group and a control group are formed.

Figure 5.1

*Souce: Data of Studentenmonitor 2011-2018

§5.3 Method of analysis

The regression discontinuity design (RD design1_{) will be used as it measures the effect}

of the policy reform. Translated to this thesis, it measures the effect before and after the policy reform for students with the old and new policy, which is the main point of interest in this thesis. In this context, two groups of students with the old and new policy are created to analyze the effect of the policy reform on students. A sharp RD design

1 _{“The regression-discontinuity (RD) research design is a quasi-experimental method that can be used to assess}

the effects of a treatment or intervention. Unique to the RD design is that participants are assigned to groups solely on the basis of a pretreatment cutoff score. The name ‘regression-discontinuity’ comes from the fact that a treatment effect appears as a ‘jump’ or discontinuity at the cutoff point in the regression function linking the assignment variable to the outcome. In its simplest form, the design has a pretest or pretreatment (the assignment variable) measure, two groups (those scoring above and below the cutoff), and a posttest or posttreatment (the outcome) Reform of the study finance policy in The Netherlands Finance of education Study behaviour Funding of higher education of Adamson (2016) Income composition Housing Social background Principal-Agent Theory Previous research about the effect of the abolition of the basic student grant Private investment rate Private investment in the income composition between 2010-2016. Source: Eurostudent VI

Independent variable

(Y) Dependent variabels (X) Theories Concepts Operationalization of concepts

Support from paretns Self-earned

income Borrowed loan

DUO Average monthly _{loan DUO*} Motivation

Time-investment Private investment rate

Hours worked per week* Monthly support from

parents* Level of social background (1-10

scale)* Living with or without

parents*

Private investment in the income composition between 2010-2016. Source: Eurostudent VI

Current Motivation (1-5 scale)* Study hours a week * Student

will be used were the treatment switches cleanly off or on as the running variable passes a cutoff, namely if a student is in a cohort before or after the policy reform. A sharp RD design means that the treatment (the policy reform) followed perfectly at the cutoff value (academic year ’15-‘16), without exception. In this context, the cutoff value is the academic year ’15-‘16. If the running variable, the variable who determines the treatment (in this case the cohort), passes this cutoff year (’15-’16), students receive the treatment (new policy) (Angrist & Pischke, 2015, p.151). For example: If a student start in the academic year ’16-’17 (so cohort of ’16-’17), he receives the treatment because his cohort passed the cutoff year ’15-’16. Receive the treatment if therefore randomly assigned because it only depends on the year a student did started in higher education. So, no special kinds of students are select to receive the new policy and therefore, the treatment is randomly assigned. Students who started in ’15-’16 or later, cannot manipulate their treatment because it is not possible to receive the old policy in those cohorts. On the other hand, students who started before ’15-‘16 receive the old policy and cannot manipulate their treatment, all those students receive the old policy. To sum up, with random assignment of the treatment were subjects cannot manipulate the treatment assignment (which is based on the cohort), the requirements for a proper regression discontinuity design are fulfilled (Angrist & Pischke, 2015, p.151).

The main equation that will be used in the RD design is as follows: 𝑌_"# = 𝑎 + 𝐵1 ∙ 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡_"# + 𝛽2 𝑋_"#+ 𝜀_"#

This equation analyzes the effect of the policy reform on the outcome variable (𝑌_"#). In this thesis, according to the theoretical framework, this regression will be used six times to analyze six different outcome variables, student performance, time-investment, motivation, borrowed loan, self-earned income and financial support from parents. 𝑎 shows the parameter for comparison of the group intercept at the cutoff (academic year ’15-’16). 𝐵1 shows the treatment effect. So, this coefficient shows the effect of the treatment (the policy reform) on the treatment group in comparison to the control group. The treatment group are bachelor students with the policy; the control group are bachelor students with the old policy. For example, the coefficient of B1 shows the average difference in borrowed loan for students with the new policy in comparison to students with the old policy. The 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡_"#variable determines whether the bachelor student is in the treatment or control group. This variable is a

dummy variable and equals 1 when a student has started its study in or after the academic year ’15-16. This dummy variable equals 0 when a student started its study before the academic year ’15-‘16.

𝛽2 shows that effect of the control variables. The main control variables (𝑋_"#) that are used in the RD design are gender, age, social class, year of study, HBO/UNI and migrant/native background. These control variables are chosen because the means are significantly different for the treatment and control groups, as shown by the t-test in table 5.1. Therefore, these differences can influence the predicted outcome in the regression. It is therefore important to control for this differences because in that way, the outcome that is predicted by the model, has a higher chance to be explained by the policy reform and is not caused by a significant difference in the means of the characteristics in the treatment and control group. The equation ends with 𝜀"#, which is

the error term in the equation. A further justification of the choices in this research method is explained in the used models (1, 2 and 3) in the first paragraph of the results chapter.

§5.4 Validity and reliability

Concerning this research method, by doing a regression discontinuity design with six control variables, there is a high chance that the results are induced by the policy reform instead of other causes, for example by different characteristics in the treatment and control group. Furthermore, there is a consistency in the case selection method, as explained earlier in this chapter. With these two aspects, a RD design with six control variables and a consistent case selection method, the interval validity of the results is high. This means, there is a high change that the measured results are caused by the policy reform.

Besides, external validity is high because the results are based on a large sample of around 18.000 students per year. This sample is representative for the student population (Researchned, 2018). Therefore, the results can be generalized for the whole student population. With a large representative sample, and also comparing the 95% confidence intervals of the coefficients of the cohorts, there is a very low chance that the results are based on coincidence what makes this research also reliable.

6. Data

§ 6.1 Description of the data

The main data source in this thesis is the ‘Studentenmonitor Hoger Onderwijs’. This database is introduced in 2000 by the Ministry of Education, Culture and Science. The purpose of this database is to analyze systematically the yearly social-economic developments in the field of higher education. The data at the student level is yearly amplified with new survey data of a representative group of students of around 18.000 students per year. Therefore, the dataset included in total 151.032

observations for the year 2010 till 2018. The dataset contains around 800 different variables. Since not all variables are relevant only 22 variables will be used to analyze the hypotheses of the thesis.

1. Academic year 9. Age 17. Motivation level (1-5 scale)(1: low, 5: high) 2. Starting year 10. Living with or without parents 18. Weekly hours spend

on study 3. Registered at

university or UAS

11. Average Financial support from parents

19. Social background (1-10) (1: high, 10: low) 4. HBO/UNIV 12. Average working hours a

week

20. Student performance (average grade 1-10) 5. Full time/

part-time

13. Average monthly loan form DUO in euros

21. Supplementary grant

6. Year in their studies

14 Education of parents 22. Policy (‘old or new study finance system’)

7. Bachelor or master

15. Financial situation parents

8. Gender 16. Native/Migrant

§ 6.2 Method of data collection

The data of ‘Studentenmonitor Hoger Onderwijs’ is collected by a stratified sample of students in higher education, partly through a selection from CRIHO/1Cijfer HO, partly by panels. These students fill in the online questionnaire, with 170 questions, every spring between May and June. The answers to the questions in the survey are based on the current academic year. This means a survey filled in in May 2015 covers the academic year 2014-2015.

§ 6.3 Descriptive statistics

Academic year Old policy New policy total

11-12 9352 0 9352 12-13 13290 0 13290 13-14 10699 0 10699 14-15 11845 0 11845 15-16 5802 2863 8665 16-17 5317 5725 11042 17-18 3212 7747 10959 Total observations 59517 16335 75852

Table 6.1 Number of bachelor students with old and new policy

Table 6.1 gives an overview of bachelor students in the dataset with the old and new policy divided by the academic year. It shows an increase in the number of students who study with the new policy conditions after the reform year ’15-’16. At the same time, it shows therefore also a decrease in the number of students who study with the old policy conditions. This table is shown to give an overview of the treatment and control group that will be compared within the regression discontinuity design. To notice, the 80.4432 _{students are less than the original amount of 151.032 observations.}

This is caused by the fact that all master students and the academic year ’10-’11 are excluded. There is chosen to exclude those observations because they were not in the scope of this research.

The next section will describe the different descriptive statistics on the three main parts, namely financial funding of education, study behavior and income composition. Income composition is subdivided into the borrowed loan, work hours and support from parents in the same way previous research, like Eurostudent (2016), did. In this way, an overview of the development of the six different dependent variables will be provided. The descriptive statistics shows the average and median for bachelor students and also separates also for WO and HBO. This is show by graph number 1 for each dependent variable. There is chosen to show also the difference between HBO and WO to show the importance of using this variable as a control variable, because the averages of WO and HBO are in most of the cases a bit different. By showing both types of education, the differences between those kinds of educations become clearer. The median shows only the score for all bachelor students that are given the most

often because the bachelor students in total are in the scope of this research. The second graph of each dependent variable shows the average for students with the old and new policy. This gives a first indication of the difference between students with the old-and -new policy for that particular variable. It is essential to check if the difference between those groups should be negative or positive, which helps to control if the results of chapter 7 are in line with descriptive statistics.

§6.3.1 Financial Funding of Education

§6.3.1.1. Private investment and student performance

Graph 6.1 - Students performance is measured in average exam grades from 1-10 Source: Data of Studentenmonitor Hoger Onderwijs 2011-2018.

Graph 6.1 shows a quite stable trend in student performance between 2011 and 2018. This means, that student performance remains quite stable after the policy reform. The policy reform did lead to an increased private investment in education (Eurostudent VI, p.159). This combination of more public and student performance is useful when analyzing the conclusion of Adamson (2016). The states that a higher public investment in education has a positive effect on student performance. Besides, private investment leads to a lower student performance, so a negative effect (Adamson, 2016). This will be tested in H1.

§6.3.2 Study behavior

Study behavior is subdivided into time-investment, according to the principal-agent theory, and motivation. Motivation is another possible outcome variable and incentive behind time investment and will therefore also be analyzed as a mediating variable

later on. This is because the policy reform could influence time-investment through the channel of motivation.

§6.3.2.1 Time investment

Graph 6.2 shows on average a quite stable trend in study time investment per week for bachelor students between 2011 and 2018. On average, HBO bachelor students invest around 3 hours more time in their studies in comparison to WO bachelor students. When focusing on the second graph, the trend line for bachelor students with the new policy shows a difference of two and a half hours less compared to bachelor students with the old policy. In this second graph, there is not controlled for any variable, so it difficult to determine if this difference is probably caused by the policy reform. This effect will be analysed more in details in the result chapter with H2.

Graph 6.2 - Source: Data of Studentenmonitor Hoger Onderwijs 2011-2018

§6.3.2.2.Motivation

The first graph (graph 6.3) shows the average current motivation (measured 1-5 scale; 1;low, 5;high) for all bachelor students and with a distinction between HBO and WO students between 2011 and 2018. It shows a stable trend for the average motivation. The second graph (graph 6.3) shows the average current motivation for bachelor students with the old and new policy. The average for bachelor students with the new policy is on average 0.1 points higher in comparison to students with bachelor students with the old policy but shows overall a quite stable trend. Therefore, these descriptive statistics shows no remarkable influence of the policy reform on the motivation of

bachelor students between 2011 and 2018.

Graph 6.3 - Source: Data of Studentenmonitor Hoger Onderwijs 2011-2018

§6.3.3 Income composition

The third topic, income composition, is subdivided into the three major sources of income, borrowed loan, work and financial support from parents. Therefore, the descriptive statistics will also address these three sources of income. The next two diagrams (figure 6.2) give an overview of the changed in the income composition for bachelor students with the old and new policy. It shows that the abolition of the basic student grant is mostly compensated by an increased loan (+17%) and financial support from parents (+5%). Self-earned income did increase (-3%), but that could also be explained by the fact that the group bachelor students with the new policy contain more younger students who work less on average (Researchend, 2018). This shows also the importance to control for the year of education in the regression later on to analyze the real effect of the policy reform on the income composition.

Figure 6.2 – income composition bachelor students with old and new policy. Source: Data of Studentenmonitor 2011-2018

§6.3.3.1 Borrowed loan from DUO

Graph 6.4 shows the average of borrowed loan for all bachelor students and with a distinction between HBO and WO students. The next three graphs divide this loan for three different categories, namely for students with the old and new policy, housing situation and social class. These categories are specified because these three contains the most relevant characteristics of students where they differ the most when focusing on their financial situation. Therefore, these characteristics show the best those differences between different types of students and the importance to control for kind of education and social class. The housing situation in number 3 (in graph 6.4) shows the difference between students who living with-and without their parents. The graph shows a difference in the averages between those students when focusing on the three different parts of their income. Because after the policy reform, the difference between those groups increased even more, it could be that the policy reform influenced the housing decision of students. Therefore, this variable is used as an outcome variable and will explained more in detail in paragraph 7.3.4. Number 2 (in graph 6.4) shows the research difference of students with the old and new policy. Besides that, these characteristics are the most politically debated and it also therefore politically relevant to subdivide between these three characteristics and are therefore

show in this section.

All the graphs (6.4) show an increase in the average borrowed amount of money after the policy reform. This is the case for all bachelor students, independent of their kind of study, housing situation or social class. In the reform year, students with the new policy borrow more in comparison to students in the old system. The third graph (6.4) shows that students who live without their parents borrow on average around 100 euro more per month in comparison to students who live with their parents. So, the housing situation could have an influence on the average amount of borrowed loan. The fourth graph (6.4) shows the average borrowed amount for bachelor students with different social backgrounds. Students with a low social background borrowed more on average in comparison to students with a high social background. After the reform, the difference between those two groups decreased according to these descriptive statistics. This will be analyzed further in paragraph 7.4 by using an RD design with control variables and comparing students of different social classes.

§6.3.3.2 Work: Self-earned income

Graph 6.5 (1) shows the average of working hours per week for bachelor students with a distinction between HBO and WO bachelor students. The next three graphs divide the average of working hours for three different categories, namely for students before-and-after the reform, housing situation and social class. On average there is an increase in working hours after the policy reform, in particular for HBO students where the average of working hours increases from 10 to 13 hours. The average for WO students shows only a slight increase from 6 to 7 hours. Students with the old policy work 3 hours more on average in comparison to students with the new policy. Besides that, the working hours for students who lived without their parents increased more in comparison with students who live with their parents. Furthermore, students with the lowest social background work one hour more on average in comparison to students

1 2

3 4

with the highest social backgrounds. This difference did not change after the reform.

Graph 6.5 - Source: Data of Studentenmonitor Hoger Onderwijs 2011-2018

§6.3.3.3 Financial support from parents

Graph 6.6 (1) shows the monthly support from parents on average in euros, for all bachelor students and subdivided between HBO and WO students. The following three graphs divide the average of monthly support from parents for three different categories, namely students with the old and new policy, housing situation and social background. On average, there is an increase in the average support that students receive from their parents, particularly for students with the new policy in combination with WO education. In the reform year, students with the new policy receive on average 30 euro more per month from their parents in comparison to students with the old policy. Besides that, the difference in monthly support from their parents, between students with different housing situations and different social backgrounds did not change after the reform according to these descriptive statistics (graph number 3 and 4, graph 6.6). This will be further analyzed in the results chapter in paragraph 7.4.

3

1 2

Graph 6.6 - Source: Data of Studentenmonitor Hoger Onderwijs 2011-2018 §6.3.4. Housing

Graph 6.7 gives an overview of the housing situation of all students between ’11-’12 and ’17-’18, based on the data of studentenmonitor 2011-2018. It shows a quite stable trend over the time period. When focusing on the reform year, there is an increase in bachelor students who lived with their parents from 43% to 47%. This difference of 5% decreased in the years after the reform. To analyze if, and to what extent, the policy reform has influenced housing decisions, these statistics will be taken into account in a later stage in the results chapter. The graph 6.7 gives an indication that the policy reform has influenced the housing decision of students and the housing variable will therefore analyzed as a possible outcome variable to improve the interpretation of the results.

2 1

4 3

Graph 6.7 - Source: Data of Studentenmonitor Hoger Onderwijs 2011-2018

The interpretation of the results can be improved by taking into account the effect of the policy reform on the housing decision of students. It could be that there is an indirect effect of the policy reform on the borrowed loan through the effect of the housing situation. For example; when it turned out that because of the policy reform, the income composition has changed (higher loan) (blue arrow in figure 6.1), it could be that the average loan after the reform is also influenced by the change in their housing situation (green and red arrow). When less students decide after the policy reform to live without their parents, they will borrow less on average. This is because students who live with their parents borrow less on average (Researchned, 2018). So, there are two possible ways in which the average borrowed loan after the policy reform could be influenced. This is also shown in figure 6.1.

Figure 6.1 – Direct and indirect effect of housing on income composition

While this thesis focus on the effects of the policy reform, the housing variable will be treated as an outcome variable to analyze the effect of the policy reform on housing. But, because this effect can influence the results, the mediation role of the housing variable will be taken into account in the results chapter by the interpretation of the results but will be no further statistically analyzed in depth because that is beyond the scope of this thesis.

33% 38% 40% 37% 42% 41% 35% 67% 62% 60% 63% 58% 59% 65% 0% 20% 40% 60% 80% 100% 120% 1 1 - ‘ 1 2 1 2 - ‘ 1 3 1 3 - ‘ 1 4 1 4 - ‘ 1 5 1 5 - ‘ 1 6 1 6 - ‘ 1 7 1 7 - ‘ 1 8

HOUSING SITUATION BACHELOR STUDENTS

2011-2018

Living with parents Living without parents

Housing (living without parents)

Income composition (Borrowed loan) Policy reform