Financial behaviour of students:
Debt aversion, work and spatial (im) mobility
Stijn Koops1Master thesis MSc. Economics
University of Groningen, 19 January 2018
Supervisor Prof.dr. R.J.M. Alessie
ABSTRACT
Students in higher education have several methods to finance their education. Working part-time is common under Dutch students, while debt aversion makes borrowing less popular. This paper focuses on whether the housing decision is another financial method for students, which might be more present than before. The research is executed with a survey-based dataset which include financial and demographic information about Dutch students. The main result is that socioeconomic background influences the financial behaviour. Low socioeconomic students prefer living with their parents and working over borrowing. Moreover, the removal of the basic grant strengthens this result and will likely deter the accessibility of higher education for these students.
Keywords: Student behaviour, Educational Investment, Inequality, Government
policy
JEL-Codes: I22, I24, I28, D91
1 To contact the author s.a.koops@student.rug.nl
I would like to express my sincere gratitude to my supervisor Prof. dr. R.J.M. Alessie for providing support, expertise and interest during my writing process.
2
Section I. Introduction
Milton Friedman published in 1962 Capitalism and Freedom where he justified the involvement of the government in the education market. Rationale for the government to intervene in the education market is that education creates public returns besides private returns. The most economic rational argument for government intervention is the existence of a credit market failure. Without intervention, education is only available for rich families as tuition fees are too expensive. Banks are not willing to provide loans to students due to no collateral and uncertainty whether the loan can be repaid. Student grant or loans offered by the governments solve this problem, such that all individuals can decide to enter higher education. As education leads to an increase in human capital, productivity and wages are higher for educated persons. Consequently, this lead to a higher welfare level of society. Moreover, policy makers justify intervention in the education market to provide equal opportunities for every person. As education is a normal good, higher-income families would buy more education for their children, resulting in a persistence of inequality. Public access to education stimulate income mobility and redistribution. A higher educated population also leads to positive externalities. According to Groot and Maassen - Van den Brink (2010), the probability of committing crimes such as shoplifting, vandalism and threat, assault and injury decreases with the years of education. Silles (2009) found that education and health status are related, where she found estimates that one more year of education increases the probability of good health between 4.5 and 5.5 percentage points. All these arguments in favour of a higher educated population justify government intervention as the social returns might be larger than the cost of providing student grants and loans.
3
are an inefficient and inequitable use of public funds as higher-income families would enter education anyhow. Started in the study year 2015-2016, the Dutch system changed to a loan system that replaces the basic grants which might improve the micro- and macro-efficiency. On the other side, Schwartz (1985) found for the United States that a basic grant system has a significant positive effect on enrolment of individuals of lower income households compared to loans, indicating that the behaviour of students are subjective to the method of aid. Therefore, recent protests state that the new system deter entry for students from lower social welfare.
4
In relation to that paper, this paper attempts to contribute to the literature by focussing on the interaction between work and borrowing by taking into account the endogenous choice of the living situation. Additionally, the effect of the reform on the behaviour of students will be investigated, where the option of living with your parents might be preferred after the reform for students from low-income families. Therefore, the paper aims to investigate the borrowing, working and living behaviour of students while taking into account their social background. The structure of the paper is as follows: The next section will discuss the relevant literature regarding work and borrowing behaviour of students. Section III discusses the Dutch institutions regarding financial aid. Section IV and V describe the methodology the data respectively. Section VI provides the results and discussion. Finally, section VII concludes the paper and suggest advice for public policy makers.
Section II. Literature Review
5
from low socioeconomic families might experience difficulties with perceiving the value of education as their parents did not enter higher education. Moreover, if those students have to borrow to cover the cost of education, debt aversion might lower the value of education even more.
A useful microeconomic theory to analyse the financial decision of students is the life-cycle model of Ando and Modigliani (1963). The life cycle model assumes that individuals rationally decide their consumption behaviour over their life-cycle limited by the income available, assuming perfect capital markets. An implication of the model is that individuals prefer to have a stable level of consumption, indicating that people borrow and save to smooth their consumption over their lifetime. In the case of students, this would imply that students are willing to borrow during their study time as they might expect an increase in future income as a result of higher human capital. Davies and Lea (1995) found for a sample of undergraduate students that they are indeed relatively tolerant towards debt. Although student have low incomes, they perceive this as a temporary state, such that they accept to borrow to improve their lifestyle. Nevertheless, in that respect there is no consensus between literature as some find that students from lower social classes are more debt averse and are more likely to deter higher education out of fear of debt (Callender and Jackson, 2005). Cunningham and Santiago (2008) define debt aversion as ‘’an unwillingness to take a loan to pay for college, even when that would offer positive long-term return’’. The rationale is that students underestimate the future payoffs from obtaining a college degree compared to the current cost of studying.
6
7
performance. Therefore, the social loan system might indirectly affect the health and performance of students from low socioeconomic families.
The reason of working part-time during their academic life might not only have financial reasons. A survey executed by Robinson (1999) indicated that students work part-time for several non-financial reasons; students enjoy the work, believe that job history could gain employment in later life and do not experience it as a negative impact on their academic results. Van Essen (2016) shows for the Netherlands that the proportions of students that work part-time is around 85%, a lot higher than the average OECD student. Figure A2 in the appendix displays this high participation rate in the labour market
Research on how part-time labour supply affect students during their academic life show mixed results. Häkkinen (2004) state that entering the labour market increases future job payoffs and Darolia (2014) found an improvement in skills such as time efficiency, personal responsibility and communications. Negative effects related to part-time employment are increased probability of dropping out, lower academic performance and delays (Hovdhaugen, 2015; Singh 1998; King, 2002). More probable is that working part time has a nonlinear effect on academic success. Joensen (2009) indicates that working moderate hours improve academic success and future labour opportunities, while working more than 19 hours deter academic performance. Moreover, Darolia (2014) found that part-time employment does not crowd out study time for a moderate amount of hours.
Section III. Dutch Institutions
Grants and travel support
8
is different. When students work part-time and their income is larger than a certain threshold the government gives a fine. Since 2010, students need to repay the amount of money earned above the threshold as a fine. The amount of SF over the years is listed below in table 1.
For students with low income parents there is an additional grant available. All of the components listed above are at first a loan, but will be converted into a gift when students obtain their certificate within 10 years. This time-span is created to have an incentive to graduate. Performance is also an important factor, as the basic and additional grant is available for the nominal time of their study. In most cases this is 4 years, the same amount of years the travel support is available. In case the student cannot complete their study at the nominal time an interest-bearing loan is available for another 3 years and travel support for 1 year.
Possibilities to borrow
9
Policy changes
With the beginning of the study year 2015-2016 students who start a new study, either HBO, bachelor or master, do not receive a SF gift anymore. A social loan system is introduced in 2015-2016 to improve the quality of education as the government saves money which is reinvested in the quality of higher education.
Table 1: SF amount per month from 2004 to 2017
Basic grant Supplementary grant Maximum loan
Tuition fee
Earning threshold Living situation Living situation
Year With parents Independent With parents Independent
04-05 74 228 199 217 771 1476 9847 05-06 76 233 203 221 787 1496 10200 06-07 89 248 207 226 796 1519 10460 07-08 91 253 206 225 810 1538 10530 08-09 92 256 209 228 819 1565 10630 09-10 93 260 212 231 832 1620 11640 10-11 96 266 219 239 853 1672 12000 11-12 96 266 221 241 853 1713 12000 12-13 96 269 225 246 853 1771 12060 13-14 98 272 230 250 853 1834 12210 14-15 100 279 237 258 895 1906 12388 15-16 103 286 245 267 854 1951 13989 SL system 378 378 854 1951 - 16-17 104 289 251 274 863 1984 14216 SL system 384 384 863 1984 -
Numbers are nominal and subtracted from Dienst Uitvoerend Onderwijs (2017)
10
deteriorated. As a result, one might anticipate an increase in the probability to live with your parents to avoid lending.
A negative externality of this might be that students who experience more financial restrictions do not attend specific universities that are located less central in the Netherlands as one should live individually. As a result, access to the technical/research universities of Enschede, Delft, Eindhoven and Wageningen might be deterred for some students.
Next to abolishing the grant, the earning threshold is also removed for students entitled to the new system and the repayment period is extended from 15 years to 35 years. Regarding the repayment of the debt, the monthly payment is halved due to the longer time span. In the same period the ministry of education started the campaign ‘Studying with a plan’ to inform future students and parents about the changes in the system. The campaign was introduced alongside a website where students and parents can use a tool to calculate their future costs of studying. Although the result of the campaign shows an increase in need for knowledge and attitude, the amount of persons that use the tool is only 21%.
Section IV. Methodology
Working and borrowing as decision
11
The two-part model is explained by two equations. The first equation is a probit regression of the probability of using a financing method on exogenous covariates and a binary endogenous covariate. The second equation is an ordinary least squares regression based on a subsample of individuals that participate. As the decision whether to live individually or with your parents is endogenous, the model includes instrumental variables. The housing decision is a binary variable, such that it is measured by a probit regression. Formally, the model is:
𝐹1𝑖(𝑡)∗𝜃 = 𝛼 1′𝑥𝑖(𝑡)+ 𝛽1𝐻𝑖(𝑡)+ 𝜀1𝑖(𝑡) (1) 𝐹2𝑖(𝑡)∗𝜃 = 𝛼2′𝑥 𝑖(𝑡)+ 𝛽2𝐻𝑖(𝑡)+ 𝜀2𝑖(𝑡) (2) 𝐻𝑖(𝑡)∗ = 𝛾1′𝑥 𝑖(𝑡)+ 𝛿′𝑧𝑖(𝑡)+ 𝜐𝑖(𝑡) (3) 𝜇𝑖(𝑡)= (𝜀1𝑖(𝑡), 𝜀2𝑖(𝑡), 𝜐𝑖(𝑡))′ , 𝜇𝑖(𝑡)|𝑥𝑖(𝑡), 𝑧𝑖(𝑡) ~ 𝑁 (( 0 0 0 ) , ( 1 0 𝜎𝜀1,𝜐1 0 𝜎𝜀2 𝜎𝜀1,𝜐2 𝜎𝜀1,𝜐1 𝜎𝜀1,𝜐2 1 )) (4) where 𝜃 indicates the method of financing, either 𝑊 or 𝐵, and ∗ indicates that dependent variables are latent variables. The variables are inferred from the data as follows: 𝑂𝑏𝑠𝑒𝑟𝑣𝑎𝑡𝑖𝑜𝑛 𝑟𝑢𝑙𝑒𝑠 𝑤ℎ𝑒𝑟𝑒 𝐼(… ) 𝑖𝑛𝑑𝑖𝑐𝑎𝑡𝑒 𝑡ℎ𝑒 𝑖𝑛𝑑𝑖𝑐𝑎𝑡𝑜𝑟 𝑓𝑢𝑛𝑡𝑖𝑜𝑛: 𝐹1𝑖(𝑡)𝑊 = 𝐼(𝐹2𝑖(𝑡)∗𝑊 > 0) = 1 𝑖𝑓 𝑎 𝑠𝑡𝑢𝑑𝑒𝑛𝑡 𝑤𝑜𝑟𝑘𝑠 𝑎𝑛𝑑 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 𝐹1𝑖(𝑡)𝐵 = 𝐼(𝐹2𝑖(𝑡)∗𝐵 > 0 = 1 𝑖𝑓 𝑎 𝑠𝑡𝑢𝑑𝑒𝑛𝑡 𝑏𝑜𝑟𝑟𝑜𝑤𝑠 𝑎𝑛𝑑 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 𝐻𝑖(𝑡) = 𝐼(𝐻𝑖(𝑡)∗ > 0) = 1 𝑖𝑓 𝑎 𝑠𝑡𝑢𝑑𝑒𝑛𝑡 𝑙𝑖𝑓𝑒 𝑖𝑛𝑑𝑖𝑣𝑖𝑑𝑢𝑎𝑙𝑙𝑦 𝑎𝑛𝑑 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 𝐹2𝑖(𝑡)𝑊 = 𝐻𝑜𝑢𝑟𝑠 𝑜𝑓 𝑤𝑜𝑟𝑘, 𝐹2𝑖(𝑡)𝐵 = 𝐿𝑜𝑎𝑛 𝑎𝑚𝑜𝑢𝑛𝑡 𝐹2𝑖(𝑡)𝜃 = 𝐹 2𝑖(𝑡)∗𝜃 𝑖𝑓 𝐹1𝑖(𝑡)∗𝜃 > 0
The 𝑖(𝑡) subscripts indicate that the observations are independent cross sections only available for one period. 𝑥𝑖(𝑡) is a vector of explanatory variables, control
12
housing situation and is equal to one if a student lives independently. The housing decision is an endogenous choice, such that the model uses a vector of 𝑧𝑖(𝑡) variables
where some variables in 𝑧𝑖(𝑡) are not in 𝑥𝑖(𝑡). In other words, a variable that
influence the housing decision, but has no direct influence on the decision whether to work or borrow. The instrumental variables will be discussed on the following page. Equation (4) is a vector of the error terms where 𝜎1𝜐 indicates the correlation
of the error term of respectively the living equation and probit equation and 𝜎2𝜐2
the correlation of the error term of the living equation and OLS equation. If these are significantly different from 0 it leads to the conclusion that the living decision is indeed endogenous.
Apart from the sign of the coefficient the probit model cannot be interpreted. Therefore, to interpret the coefficients the average marginal effects are estimated for each explanatory variable:
𝜕Pr (𝐹𝑖(𝑡)𝜃 =1) 𝜕𝑥𝑖𝑘 = 𝜕ϕ(𝑥𝑖′𝛽) 𝜕𝑥𝑖𝑘 = 𝜙(𝑥𝑖 ′𝛽)𝛽 𝑘 (5)
13
Instrumental variables
As the living decision is a way to alter the cost of living for student it is likely a substitute or complement to borrowing and working. Therefore, living is expected to be endogenous, such that instruments are used. Variables that influence the housing decision, but has no direct influence on the decision whether to work or borrow are relevant instruments. Demographic dummies would be preferred as instruments, as they influence the living decision and not the decision to borrow or work. Due to an absence of demographic dummy variables in the data, different instruments are used for the borrowing and working equation.
The variables contact hours at school and time spend on extracurricular activities are used as instruments for the borrowing equation. The variable extracurricular activities contains the hours per week spend on administrative body at the university or administrative work at a student organization. Time spend on extracurricular activities is expected to increase the probability that a student will live individually, due to the fact that these activities are mostly scheduled after school time and in the university city. Contact hours might affect living individually negatively, as the value of living individually decrease if a student has less spare time and college students, who have on average more contact hours, life more often with their parents.
14
score slightly lower on study progress. These variables are expected to be independent of hours of work, as Robinson (1999) found that students do not experience work as a negative impact on academic performance.
Joint decision
The two univariate probit models explained above, where the living decision is endogenous, are likely to be a joint decision as the choice regarding working and borrowing are substitutes. If binary choices are correlated a multivariate probit model is convenient to describe the decision (Bel, Fok and Paap, 2016). As the housing decision is an endogenous choice, more likely is that both three methods affect the choices, such that a multivariate probit model describe the decisions of the students most efficient. Therefore, by means of a multivariate probit model the individual decision whether to work, borrow and live individually or not is estimated. Formally, the following three equations are estimated jointly where the observation rules are the same as the probit models above:
𝐹1𝑖(𝑡)𝑊 = 𝛼 1′𝑥𝑖(𝑡)+ 𝛽1𝐻𝑖(𝑡)+ 𝜀1𝑊𝑖(𝑡) (6) 𝐹1𝑖(𝑡)𝐵 = 𝛼 2′𝑥𝑖(𝑡)+ 𝛽2𝐻𝑖(𝑡)+ 𝜀1𝐵𝑖(𝑡) (7) 𝐻𝑖(𝑡) = 𝛾1′𝑥𝑖(𝑡)+ 𝛿′𝑧𝑖(𝑡)+ 𝜀𝐻𝑖(𝑡) (8) 𝑂𝑏𝑠𝑒𝑟𝑣𝑎𝑡𝑖𝑜𝑛 𝑟𝑢𝑙𝑒𝑠 𝑤ℎ𝑒𝑟𝑒 𝐼(… ) 𝑖𝑛𝑑𝑖𝑐𝑎𝑡𝑒 𝑡ℎ𝑒 𝑖𝑛𝑑𝑖𝑐𝑎𝑡𝑜𝑟 𝑓𝑢𝑛𝑡𝑖𝑜𝑛: 𝐹1𝑖(𝑡)𝑊 = 𝐼(𝐹2𝑖(𝑡)∗𝑊 > 0) = 1 𝑖𝑓 𝑎 𝑠𝑡𝑢𝑑𝑒𝑛𝑡 𝑤𝑜𝑟𝑘𝑠 𝑎𝑛𝑑 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 𝐹1𝑖(𝑡)𝐵 = 𝐼(𝐹2𝑖(𝑡)∗𝐵 > 0) = 1 𝑖𝑓 𝑎 𝑠𝑡𝑢𝑑𝑒𝑛𝑡 𝑏𝑜𝑟𝑟𝑜𝑤𝑠 𝑎𝑛𝑑 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 𝐻𝑖(𝑡) = 𝐼(𝐻𝑖(𝑡)∗ > 0) = 1 𝑖𝑓 𝑎 𝑠𝑡𝑢𝑑𝑒𝑛𝑡 𝑙𝑖𝑓𝑒 𝑖𝑛𝑑𝑖𝑣𝑖𝑑𝑢𝑎𝑙𝑙𝑦 𝑎𝑛𝑑 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 where 𝐹1𝑖(𝑡)𝑊 , 𝐹
2𝑖(𝑡)𝐵 and 𝐻𝑖(𝑡) denote the participation decisions in respectively
working, borrowing and living individually.
The error term 𝜀1𝑊𝑖(𝑡), 𝜀1𝐵𝑖(𝑡) and 𝜀𝐻𝑖(𝑡) are allowed to be correlated, which indicate
15
socioeconomic background. Students of low socioeconomic status might prefer living at their parents and working both, where the endogenous effect of living is taken into account.
Section V. Data and descriptive statistics
This section overviews the data and explain the sample selection and limitations. The data used for the research are subtracted from the Studentenmonitor, a yearly survey executed by ResearchNed. The dataset consist of cross-sectional data from 2001 to 2015. The period used in this paper is from 2005 till and including 2015, as the first years suffer from consistency problems and incomplete surveys. The year 2010 is excluded as well, as the survey was not held in this year. Each year contains information about the preceding study year, such that data from 2015 represent the study year 2014-2015. The survey covers several subjects including information about background characteristics, education, type of income, study progress, time expenditure and characteristics of their parents. The dataset is large enough to represent the population, however, there is no proportional distribution. Female students and university students are largely overrepresented in the dataset. This is corrected by applying a weighting procedure, such that the proportion represents the population (Studentenmonitor 2001 – 2015). The weighting procedure is based on four strata, namely type of education, sector, study year and gender.
16
Table 2: Descriptive statistics by financing mode and living situation
O B W BW LP LI
Mean Mean Mean Mean Mean Mean
Age 21.42 (2.47) 22.56 (2.56) 21.19 (2.30) 22.63 (2.37) 20.75 (2.12) 22.35 (2.44) Female 0.52 (0.50) 0.48 (0.50) 0.53 (0.50) 0.51 (0.50) 0.48 (0.50) 0.55 (0.50) Foreign 0.12 (0.32) 0.14 (0.34) 0.05 (0.23) 0.08 (0.28) 0.08 (0.26) 0.07 (0.26) Handicap 0.11 (0.31) 0.12 (0.32) 0.07 (0.25) 0.09 (0.28) 0.07 (0.26) 0.09 (0.28) University 0.51 (0.50) 0.57 (0.50) 0.38 (0.48) 0.50 (0.50) 0.27 (0.44) 0.57 (0.50) Repeated class 0.17 (0.38) 0.23 (0.42) 0.17 (0.37) 0.25 (0.43) 0.18 (0.39) 0.20 (0.40) Income parents
Far below median 0.07 (0.25) 0.08 (0.27) 0.04 (0.19) 0.05 (0.22) 0.04 (0.21) 0.05 (0.22) Below median 0.26 (0.44) 0.29 (0.45) 0.27 (0.44) 0.29 (0.46) 0.29 (0.46) 0.26 (0.44) Median 0.28 (0.45) 0.30 (0.46) 0.33 (0.47) 0.31 (0.46) 0.34 (0.47) 0.30 (0.46) Above median 0.26 (0.44) 0.23 (0.42) 0.25 (0.43) 0.24 (0.42) 0.24 (0.43) 0.26 (0.44) Far above median 0.13 (0.34) 0.11 (0.31) 0.11 (0.31) 0.11 (0.31) 0.09 (0.28) 0.13 (0.34) Socioeconomic class Low 0.40 (0.49) 0.39 (0.49) 0.49 (0.50) 0.44 (0.50) 0.55 (0.50) 0.39 (0.49) High 0.35 (0.48) 0.34 (0.47) 0.24 (0.42) 0.29 (0.45) 0.18 (0.39) 0.34 (0.47) Living independently 0.57 (0.50) 0.80 (0.40) 0.44 (0.50) 0.80 (0.40) - - Basic grant 149.82 (106.13) 175.64 (116.86) 139.76 (96.18) 155.12 (119.32) 92.15 (50.23) 190.00 (116.38) Supplementary grant 41.73 (89.00) 98.97 (163.90) 39.96 (85.61) 85.63 (153.11) 46.36 (93.51) 61.33 (127.23) Parental contribution 186.49 (236.26) 134.91 (171.95) 103.15 (154.37) 105.86 (144.95) 56.24 (109.34) 164.85 (191.55) Labour income montly - - 317.90 (289.07) 327.23 (258.05) 248.80 (269.14) 257.58 (290.73) Loan amount montly - 392.13 (217.57) - 363.10 (224.62) 40.91 (128.15) 165.23 (241.03) Observations 8.809 4.594 31.159 12.841 20.739 36664 Proportion of population 15.35% 8.00% 54.28% 22.37% 36.13% 63.87%
17
The variables used in the research have the following explanation. The financial variables basic grant and supplementary grant are self-reported amounts. The parental contribution variable is a self-reported amount that includes both cash as well as in kind amounts received from parents. Most other variables are dummies constructed from the survey questions. The female variable equals one if the individual is female and the foreign variable is one if one’s parents are born in another country. The handicap variable equals one if a student if students experience a physical or mental handicap or a learning disability and zero otherwise. The university variable equals one if a student attends university and zero if college. Repeated class equals one if a student repeated a class one or multiple times. The living independently variable equals one if a student life independent and zero if a student live with their parents. Socioeconomic status is determined by the education level of the parents. Students are of high socioeconomic status if both parents have obtained higher education and low socioeconomic status if both parents have not attended higher education. Students were one of their parents finished higher education serves as the reference group. Table A2 in the appendix describes the variables more explicitly.
18
partly, or due to the fact that debt-averse students are not willing to live independently.
Data limitations
Survey data lead to multiple limitations and challenges that need to be taken into account when interpreting the results. Firstly, the survey does suffer from sample selection bias as they do not provide information about non-users. Only students who attended higher education are included in the dataset. Students who would have entered higher education under the grant system, might postpone or not enter higher education at all due to debt aversion or work preference. Therefore, the effect of the policy of abandoning the grant system might be underestimated, such that there is an upward preference to borrowing and living individually. Moreover, the decision whether to borrow, work and live individually might all be influenced by the ability of a student. Some students might simply have high ability such that work does not affect the academic performance. However, this high ability might also imply that students expect high future incomes, such that borrowing increase as well. Low ability students on the other hand might experience difficulties combining working part-time and studying. Wagstaff et al. (2007) mention that survey responses might be biased due to fear of judgement or repercussions. Although the survey is anonymous, students might still prefer not to answer sensitive subjects such as borrowing amount, education level of parents and social class. Another option is that students mask their situation and report lower borrowing amount or better academic performance.
19
Section VI. Results
This section reports and discusses the effects of the independent variables on both financing methods, whether it affects the participation rate and the amount of loan and working hours. First, the results of the borrowing decision separately is discussed, where the living decision is discussed as an endogenous choice. Following is the outcome of the working decision. Finally, the work and borrowing decision are estimated as a joint decision, where living is treated as endogenous factor.
Borrowing behaviour
As discussed, the housing decision is likely to affect the decision to borrow. To test for endogeneity the instrumental variables contact hours at school and time spend on extracurricular activities are used. The average marginal effects from the housing probit equation are reported in table 3, including the test results for the relevance of the instrumental variables for the housing decision.
20
Table 3: Probit model of housing decision (3) Probit Contact hours -0.0019 (0.000)*** Extracurricular activities 0.0106 (0.000)*** Female 0.1014 (0.000)*** University 0.1610 (0.000)*** Handicap 0.0000 (0.997) Foreign -0.0810 (0.000)*** Study year1 Second year -0.0230 (0.000)*** Third year -0.0794 (0.000)*** Fourth year -0.0599 (0.000)*** Social class2
High social class 0.0488 (0.000)***
Low social class -0.0483 (0.000)***
Basic grant 0.0034 (0.000)*** Supplementary grant 0.0005 (0.000)*** Parental contribution 0.0010 (0.000)*** Repeated class -0.0384 (0.000)*** Observations 49,247 Test statistic Relevance instruments F(2, 49231) = 81.67 p = 0.0000
Standard errors are in parentheses where ***, ** and * indicate significance at the 1, 5 and
10 percent levels. The reference group of 1 and 2 are respectively the first study year and
medium socioeconomic class. (3) show the average marginal effects for the probit estimates.
as basic grant, additional grant and parental support all have positive effects. As the basic grant is removed after the reform while the additional grant does not rise proportionally, the probability to live individually decrease for students who cannot rely on parental contributions. The control variables show that the probability of living individually increases if a student is female and a university students, while it decreases if students have foreign roots.
21
interpreted as the change in the probability of the dependent variable when an explanatory variable is changed, while holding all other variables constant (at their means). The first-stage estimates are reported in table A4 in the appendix. Additionally, the table reports the test statistics whether living is endogenous for borrowing. The tests indicate that both instrumental variables, contact hours and extracurricular activities, are strong and exogenous instruments. Moreover, testing for endogeneity indicates that living is endogenous for the borrow decision, such that an IV approach is preferred above OLS for the borrowing decision. The probit estimates show the effect of the explanatory variables on the decision whether to borrow and the amount model, estimated by GMM, on the loan amount. Living is instrumented by the variables explained above. The correlation of the error terms of the living equation and borrowing equations is negative and significant at the 1% and 5% respectively, suggesting that the housing decision is indeed endogenous and a substitute for borrowing. All time dummies are positive while the cohort dummies are negative, both significant at the 5% level. Hence, borrowing is more severe during later years, while older generation borrow more often.
Firstly, the probability to participate in borrowing will be discussed. On average, if students live independently this increases the probability of borrowing by 24.32%, significant at the 1% level. This might indicate that the higher grant that student receive if they live independently does not seem to cover expenses such as rent. Therefore, some students might experience capital constraints such that they need to participate in borrowing if they live individually.
22
Table 4: Two part model for borrowing
(1) IV probit (2) IV amount Living independently 0.2432 (0.013)*** 282.53 (83.966)*** Female -0.0275 (0.005)*** -22.1412 (6.401)*** University 0.0525 (0.006)*** 5.755 (10.485) Handicap 0.0505 (0.009)*** 20.1807 (6.114)*** Foreign 0.0207 (0.009)** 39.8363 (9.455)*** Study year1 Second year -0.0103 (0.006)* 1.0807 (4.358) Third year -0.0282 (0.007)*** -6.6691 (4.742) Fourth year -0.0024 (0.008) 1.5688 (6.080) Social class2
High social class 0.0142 (0.006)** 1.9910 (4.386) Low social class -0.0509 (0.006)*** -5.8471 (5.859) Basic grant 0.0001 (0.000)*** -0.7532 (0.117)*** Supplementary grant 0.0009 (0.000)*** 0.1272 (0.013)*** Parental contribution -0.0002 (0.000)*** -0.1752 (0.036)*** Repeated class 0.0602 (0.007)*** 21.1807 (4.652)*** Constant 247.879 (54.150)*** Observations 49,247 13,905 Test statistics Relevance instruments F(2,12865) = 34.11 p = 0.0000 Exogeneity instruments3 Hansen J’s χ² = 1.18 p = 0.2575
Exogeneity Living4 GMM χ² = 4.81 p = 0.0283
𝝈(𝒍𝒊𝒗𝒊𝒏𝒈,𝒃𝒐𝒓𝒓𝒐𝒘𝒊𝒏𝒈) -0.1818 (0.0284)*** -0.0728 (0.0317)**
Standard errors are in parentheses where ***, ** and * indicate significance at the 1, 5 and
10 percent levels. The reference group of 1 and 2 are respectively the first study year and
23
(2008) who found that females are more risk averse compared to males. University students are found to lend more often, which is in line with the life-cycle model as university students participate a higher future income.
Some factors also influence the loan amount if a student decides to borrow. The loan amount is twice as high if students live individually compared to living at their parents. The estimate is significant at the 1% level, which strengthens the reasoning that students indeed need to cover their expenses of living individually. The estimates for parental contribution confirm these results, as students who receive 1 euro more financial help from parents are borrowing 18 cents less. Loans and grants are not perfect substitutes as indicated by the IV estimates. If the basic grant increases by 1 euro, students will decrease their loan by 75 cents. This indicates that students search alternatives to cover the reduction of the grant, for example living with their parents or working more hours. As the reform changes the type of financial aid from a grant to a loan system, expected is that students seek substitutes for borrowing in the form of working more or living at their parents.
Work behaviour
24
Table 5: Two part model for working
(1) Probit (2) OLS amount Living independently 0.0024 (0.005) 1.0731 (0.089)*** Female 0.0311 (0.004)*** -0.0871 (0.063) University -0.0402 (0.004)*** -1.3372 (0.070)*** Handicap -0.0956 (0.006)*** 0.0647 (0.118) Foreign -0.1303 (0.007)** 0.8902 (0.130)*** Study year1 Second year 0.0230 (0.005)*** -0.1673 (0.078)** Third year 0.0651 (0.005)*** 0.2522 (0.082)*** Fourth year 0.0834 (0.007)*** 1.0031 (0.107)*** Social class2
High social class -0.0396 (0.005)*** -0.3598 (0.081)*** Low social class 0.0126 (0.005)*** 0.3260 (0.074)*** Basic grant -0.0001 (0.000)*** -0.0074 (0.000)*** Supplementary grant -0.0002 (0.000)*** -0.0012 (0.000)*** Parental contribution -0.0003 (0.000)*** -0.0039 (0.000)*** Repeated class 0.0010 (0.005) 0.4710 (0086)*** Constant 11.2397 (0.424)*** Observations 53,734 40,325
Standard errors are in parentheses where ***, ** and * indicate significance at the 1, 5 and
10 percent levels. The reference group of 1 and 2 are respectively the first study year and
25
The results show that the housing decision indeed does not affect the participation in the labour market significantly. Although working is not affected by the living situation, the estimates show again a difference between students depending on their socioeconomic background. Students from lower socioeconomic backgrounds are more likely to work, while student from high social class participate less. Both estimates are significant at the 1% level. This in line with the expected larger debt aversion under students from lower socioeconomic background, such that participating in the labour market is preferred. Grants and support from parents decrease the probability of working, which would suggest that even more students will participate in the labour market after the reform. Female students are more likely to work, which is in line with the reasoning above that females are more debt averse if work and borrowing are substitutes. Attending university decreases the probability to work on average with 4%, which is also in line with the higher anticipated future income. Although most signs are significant, the economic values are small. The reason might be that almost the complete sample participate in the labour market, diminishing the effects of the explanatory variables.
26
Joint decision
The univariate models of borrowing and work show several contradicting signs for the explanatory variables, which supports the expectation that students choose between the different methods of financing their studies. The joint estimation allows the error term of work and borrow to be correlated, such that the estimates provide a more realistic view of the behaviour of students. The borrowing and work decision separately delivered different results regarding the influence of living. The living decision is endogenous for borrowing and has a negative correlation, indicating that living at their parents is a substitute for borrowing. Due to the different influence of living, the joint estimation is estimated imposing a constraint. The joint decision is estimated, using cmp program (Roodman, 2011), with the imposed constraint that the correlation between work and living is equal to zero.
27
Table 6: Multivariate Probit results
Standard errors are in parentheses where ***, ** and * indicate significance at the 1, 5 and 10 percent levels. The reference group of 1 and 2
are respectively the first study year and medium socioeconomic class. The joint estimation is executed by using cmp with the constraint that the correlation between work and living is equal to zero.
Borrow Work Living
Living independently 1.2009 (0.063)*** -0.0132 (0.017) Female -0.1495 (0.013)*** 0.1128 (0.013)*** 0.1193 (0.014)*** University 0.1203 (0.020)*** -0.1322 (0.015)*** 0.7330 (0.015)*** Handicap 0.1399 (0.023)*** -0.3466 (0.022)*** 0.1212 (0.026)*** Foreign 0.1147 (0.024)*** -0.4435 (0.023)*** -0.0374 (0.026) Study year1 Second year 0.0002 (0.017) 0.0793 (0.016)*** 0.1106 (0.018)*** Third year 0.0144 (0.024) 0.2167 (0.017)*** 0.1576 (0.019)*** Fourth year 0.1448 (0.024)*** 0.2698 (0.023)*** 0.5322 (0.024)*** Social class2
High social class 0.0155 (0.017) -0.1328 (0.017)*** 0.1179 (0.019)*** Low social class -0.1075 (0.016)*** 0.0422 (0.016)*** -0.0734 (0.017)***
28
This reluctance to borrow lead to efficiency costs if students would prefer to borrow if there is no debt aversion. The estimates for high socioeconomic class, who are likely to be less debt averse, show that they are more likely to borrow and live more often individually.
The estimates for the basic grant and supplementary grant affect the participation rate of both three methods different. The grants decrease the probability to participate in borrowing and working, while it increase the probability to live individually. Therefore, the reform might lead to an increase in students who decide to live at home. As the estimates for parental contribution are in line with the effect of grants, the movement from living individually to living with your parents might be higher for student who receive no contribution from their parents. As people with low education earn on average less, this would affect students from lower socioeconomic status more. The control variables show mostly the same effects compared to the univariate models. However, study years affect the probability that a student will live individually now positively. Hence, starting students will stay at their parents more likely. Given the reform in the financial aid system, this will decrease the entry of students who must move away as they live in a geographically unfavourable location to attend specific universities.
Section VII. Discussion and conclusion
29
in the decision making, children from parents with did not participate in higher education prefer working part-time and living at their parents over borrowing. The life-cycle model implies that students are willing to borrow during their study time as they might expect an increase in future income as a result of higher human capital. Although university students do borrow more, the debt aversion of low socioeconomic students is not in line with the life-cycle model. The basic and additional grants have a negative influence on the probability to borrow and work, while it increases the probability of living alone. Therefore, expected is that the reform lead to a movement from living independently to living with their parents. This movement is likely to be larger under lower socioeconomic students, as they are more debt averse.
The limitation of this study is that the research is realized from survey data. The self-reported numbers have a lower validity compared to administrative data. Fear of judgement about the information regarding borrowing or social status and different results over time might lead to a biased result. However, two larger drawbacks influence the results more. Firstly, students who do not attend higher education at all are not included in the group, as the survey is held under college and university students. Secondly, the dataset is from the pre-reform period, such that the actual change in behaviour cannot be studied yet. An extension of the paper is to investigate if the behaviour changes directly in future data sets from Studentenmonitor which include the post reform period. Future research could focus on last year student in high school regarding their choice whether to enter the labour market or obtain higher education. This is important as the results show that first-year students borrow and live individually the least of all students. As a result, expected is that debt averse and student living will deter entry under the current financial aid system.
30
if students stay with their parents or work more part-time due to financial reasons, it decreases the academic performance of students and efficiency of the system. The government and policy makers are advised to not underestimate this group. Policy makers can decide to restart the campaign ‘’studying with a plan’’ to improve the financial literacy of students, by explaining that the borrowing conditions are favourable for students. Consequently, more students will consider an efficient financing method by borrowing, such their academic performance will not decrease due to working excessively. Although the removal of the basic grant increases the efficiency, the removal of the additional grant need to be revisited in the short future. The additional grant was introduced to support children from low socioeconomic status, such that entry to higher education is secured. As the additional grant makes no difference anymore between living situations, entry is deterred for students that live in distinct areas. This is especially the case if these students desire to attend specific universities such as technical or research universities. Therefore, to improve the efficiency of this specific group, policy makers are advised to reintroduce the additional grant for student who need to live individually.
31
Section VIII. References
Ando, A., & Modigliani, F. (1963). The" life cycle" hypothesis of saving: Aggregate implications and tests. The American economic review, 53(1), 55-84.
Ball, S. J., Reay, D., & David, M. (2002). 'Ethnic Choosing': minority ethnic students, social class and higher education choice. Race ethnicity and education, 5(4), 333-357.
Becker, G. S. (1962). Investment in human capital: A theoretical analysis. Journal of political economy, 70(5, Part 2), 9-49.
Beekhoven*, S., De Jong, U., & Van Hout, H. (2004). The impact of first‐year students' living situation on the integration process and study progress. Educational Studies, 30(3), 277-290.
Bel, K., Fok, D., & Paap, R. (2016). Parameter estimation in multivariate logit models with many binary choices. Econometric Reviews, 1-17.
Bozick, R. (2007). Making it through the first year of college: The role of students' economic resources, employment, and living arrangements. Sociology of education, 80(3), 261-285
Burdman, P. (2005). The student debt dilemma: Debt aversion as a barrier to college access. Center for Studies in Higher Education.
Callender, C., & Jackson, J. (2005). Does the fear of debt deter students from higher education?. Journal of social policy, 34(4), 509-540.
Carney, C., McNeish, S., & McColl, J. (2005). The impact of part time employment on students' health and academic performance: a Scottish perspective. Journal of further and higher education, 29(4), 307-319.
32
Darolia, R. (2014). Working (and studying) day and night: Heterogeneous effects of working on the academic performance of full-time and part-time students. Economics of Education Review, 38, 38-50.
Davies, E., & Lea, S. E. (1995). Student attitudes to student debt. Journal of economic psychology, 16(4), 663-679.
Dienst Uitvoerend Onderwijs. (2017). Bedragen studiefinanciering. Retrieved from
https://duo.nl/particulier/student-hbo-of-universiteit/studiefinanciering/bedragen.jsp
Eckel, C. C., Johnson, C., Montmarquette, C., & Rojas, C. (2007). Debt aversion and the demand for loans for postsecondary education. Public Finance Review, 35(2), 233-262.
Eckel, C. C., & Grossman, P. J. (2008). Men, women and risk aversion: Experimental evidence. Handbook of experimental economics results, 1, 1061-1073.
Essen, C. V. (2016). Borrow or work for it? A study of individual funding methods of Dutch higher education.
Friedman, M. (1962). Capitalism and Freedom (Chicago, 1962). FriedmanCapitalism and Freedom1962.
Groot, W., & van den Brink, H. M. (2010). The effects of education on crime. Applied Economics, 42(3), 279-289.
Häkkinen, I. (2006). Working while enrolled in a university: does it pay?. Labour Economics, 13(2), 167-189.
Hartog, J., Odink, J., & Smits, J. (1999). Private returns to education in the Netherlands: A review of the literature. Returns to human capital in Europe: A literature review, Helsinki: ELTA.
33
Jacobs, B., & Canton, E. (2003). Effecten van de invoering van een sociaal leenstelsel in het Nederlands Hoger Onderwijs.
Joensen, J. S. (2009). Academic and Labor Market Success: The Impact of Student Employment, Abilities, and Preferences.
King, J. E. (2002). Crucial Choices: How Students' Financial Decisions Affect Their Academic Success.
Manning, W. G., Duan, N., & Rogers, W. H. (1987). Monte Carlo evidence on the choice between sample selection and two-part models. Journal of econometrics, 35(1), 59-82.
Mincer, J. (1974). Schooling, Experience, and Earnings. Human Behavior & Social Institutions No. 2.
Oosterbeek, H., & van den Broek, A. (2009). An empirical analysis of borrowing behaviour of higher education students in the Netherlands. Economics of education review, 28(2), 170-177.
Psacharopoulos, G. (1985). Returns to education: a further international update and implications. Journal of Human resources, 583-604.
ResearchNed (2017): Studentenmonitor 2001-2015. DANS. https://doi.org/10.17026/dans-zhz-dexe
Robinson, L. (1999). The Effects of Part-Time Work on School Students. Longitudinal Surveys of Australian Youth. Research Report. ACER Customer Service, Private Bag 55, Camberwell, Victoria 3124 Australia (Code: A109LSA; $22 Australian)
Roodman D. (2011). Fitting fully observed recursive mixed-process models with cmp. The Stata Journal 2011, 11(2), pp 159-206.
34
Schwartz, J. B. (1985). Student financial aid and the college enrollment decision: The effects of public and private grants and interest subsidies. Economics of Education Review, 4(2), 129-144.
Silles, M. A. (2009). The causal effect of education on health: Evidence from the United Kingdom. Economics of Education review, 28(1), 122-128.
Singh, K. (1998). Part-time employment in high school and its effect on academic achievement. The Journal of Educational Research, 91(3), 131-139.
Van Walsum, S. (2014, 6 november). Minister van de basisbeurs had vaak schoon pak nodig. Retrieved from https://www.volkskrant.nl/binnenland/minister-van-de-basisbeurs-had-vaak-schoon-pak-nodig~a3783975/
35
Section IX. Appendix
Figure A1: Participation in borrowing of Dutch students
Figure A2: Participation in working of Dutch students
0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015 TA KE -UP RA TE YEAR
Take-up rate of loans of Dutch students
University College 60% 65% 70% 75% 80% 85% 90% 2005 2006 2007 2008 2009 2011 2012 2013 2014 2015 LAB O UR PAR TI CI PATI O N RA TE YEAR
Participation rate of work Dutch students
36
Table A1: Sample Selection
Number of observation before selection (2005-2015) 138,748
Data is omitted from the raw dataset when Number of
observations
They are younger than 17 or older than 30 years old 10,917
Their nationality is unknown 178
They are studying part-time 3,396
The information regarding work is not consistent (i.e. reporting positive hours of work and not having a job and vice versa)
10,201
The information regarding work is not logical (i.e. reporting positive hours of work while receiving no labour income and vice versa)
12,147
The information regarding borrowing is not consistent (i.e. indicating not borrowing while receiving loan amount and vice versa)
3,019
The study year is unknown
The information regarding living is missing Students are from a cohort earlier then 1981
8,457 773 5,015
The education level of parents is unknown 5,161
The income level of parents is unknown 19,400
The information regarding borrowing is unknown 5,796
The information regarding work is unknown 554
Number of observations left (Two-part model working) 53,734
Data from instruments are missing 4,487
37
Table A2: Explanation of variables
Variable Range Description
Borrow 0-1 1 if the student borrows, 0 otherwise Work 0-1 1 if the student works, 0 otherwise
Living individually 0-1 1 if the student live individually, 0 otherwise Female 0-1 1 if the student is female, 0 otherwise
University 0-1 1 if the student attend university, 0 if the student attend college Handicap 0-1 1 if the student experiences a physical or mental disability, 0
otherwise
Foreign 0-1 1 if the student if at least one of the parents is born in a foreign country, 0 otherwise
Study year 1-4 Describe the study year a student attend Social class
- High 0-1 1 if both parents have obtained higher education, 0 otherwise - Medium 0-1 1 if one parents have obtained higher education, 0 otherwise - Low 0-1 1 if no parents have obtained higher education, 0 otherwise Basic grant 0-500 Monthly amount of basic grant a student receives
Supplementary grant 0-800 Monthly amount of supplementary grant a student receives Parental contribution 0-1700 Monthly amount student receives from parents; cash and in kind Loan amount 0-900 Loan amount a student receive monthly
Hours of work 0-39 Amount of hours a students work per week
Repeated class 0-1 1 if the student have repeated a class, 0 otherwise Contact hours 0-40.5 Weekly contact hours at the university/college
Extracurricular activities 0-22 Weekly hours spend on extracurricular activities, such as organizational work at student organization
Change of graduation 0-100 Perceived change of graduation in percentages
Academic performance 0-100 Ratio of the number of credits scored divided by the maximum that could have been achieved
Year 2005-2015 Year dummies
38
Table A3: First-stage probit results of housing decision (3) Contact hours -0.0059 (0.001)*** Extracurricular activities 0.0320 (0.003)*** Female 0.3053 (0.005)*** University 0.4851 (0.018)*** Handicap 0.0001 (0.030) Foreign -0.2441 (0.029)*** Study year1 Second year -0.0694 (0.020)*** Third year -0.2393 (0.022)*** Fourth year -0.1803 (0.028)*** Social class2
High social class 0.1471 (0.021)***
Low social class -0.1456 (0.019)***
Basic grant 0.0104 (0.000)*** Supplementary grant 0.0016 (0.000)*** Parental contribution -0.0031 (0.000)*** Repeated class -0.1157 (0.022)*** Constant -0.9866 (0.114)*** Observations 49,247
Standard errors are in parentheses where ***, ** and * indicate significance at the 1, 5 and
10 percent levels. The reference group of 1 and 2 are respectively the first study year and
medium socioeconomic class. (3) shows the estimates from the probit estimates. The null
39
Table A4: First-stage probit two-part model for borrowing and working
(1) Borrow (2) Work Living independently 1.0964 (0.046)*** 0.0080 (0.018) Female -0.1165 (0.025)** 0.1046 (0.013)*** University 0.0783 (0.016)*** -0.1351 (0.015)*** Handicap 0.1382 (0.023)*** -0.3214 (0.022)*** Foreign 0.0899 (0.025)** -0.4382 (0.022)*** Study year1 Second year -0.0185 (0.017) 0.077 (0.015)*** Third year -0.0444 (0.017)** 0.2189 (0.017)*** Fourth year 0.0897 (0.023) 0.2802 (0.023)*** Social class2
High social class 0.0194 (0.016) -0.1332 (0.016)*** Low social class -0.1198 (0.016)*** 0.0425 (0.016)*** Basic grant -0.0009 (0.001)*** -0.0005 (0.000)*** Supplementary grant 0.0021 (0.001)*** -0.0006 (0.000)*** Parental contribution -0.0011 (0.000)*** -0.0010 (0.000)*** Repeated class 0.1806 (0.018)*** 0.0032 (0.086) Constant -1.0144 (0.088)*** 0.9090 (0.083)*** Observations 49,247 53,734
Standard errors are in parentheses where ***, ** and * indicate significance at the 1, 5 and
10 percent levels. The reference group of 1 and 2 are respectively the first study year and
40
Table A5: IV model of working
(2) IV amount Living -0.3917 (2.881) Female -0.0396 (0.200) University -1.2055 (0.405)*** Handicap 0.0825 (0.135) Foreign 0.7806 (0.214)*** Study year1 Second year -0.2311 (0.079)*** Third year 0.2033 (0.137) Fourth year 0.8391 (0.146)*** Social class2
High social class -0.3015 (0.095)*** Low social class 0.3015 (0.126)***
Basic grant -0.0033 (0.007) Supplementary grant -0.0007 (0.001) Parental contribution -0.0030 (0.002)* Repeated class 0.4706 (0.103)* Constant 11.5369 (0.723)*** Observations 37,044 Test statistics Relevance instruments F(2,37004) = 18.90 p = 0.0000 Exogeneity instruments3 Hansen’s J χ² = 0.45 p = 0.5037
Exogeneity Living4 GMM χ² = 0.25 p = 0.6155
Standard errors are in parentheses where ***, ** and * indicate significance at the 1, 5 and
10 percent levels. The reference group of 1 and 2 are respectively the first study year and
medium socioeconomic class. (2) shows the estimates from the IV GMM estimates. The null
hypothesis of 3 and 4 are respectively H0: instruments are valid and H0: living is exogenous.
