Business & Economics Master Thesis: 15 ECT
Exploring the Role of Financial Literacy as Mediating Variable in the relationship between Education and Risky Debts in The Netherlands
Yvonne Maartje Melcherts 10685588
15th of July, 2018 Word Count: 11.767
Supervisor: Dr. Adam Booij
MSc. Business Economics, specialization Managerial Economics & Strategy Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by Yvonne Maartje Melcherts who declares to take full responsibility for the content of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
The data used in this paper is given access to by CentER data that conducted the dataset for the Dutch Central Bank.
Abstract
A vast amount of literature has investigated the relationship between educational attainment and levels of income. However, increasing level of unsecured debts are becoming a more present concern. This paper aims to investigate the relationship between unsecured debts with high interest rates and the degree of theoretical educational attainment versus the year of education. To deepen the analysis, the correlation is tested and mediated through levels of financial literacy. The data is obtained from CentER data in Tilburg and makes use of the data from 2005. This paper cannot make any significant concluding statements on the mediation effects of financial literacy from education to financial behavior. This is probably due to the relatively small sample size, large standard deviations and measures with little precision. Nevertheless, financial literacy is significantly related to education. This paper adds to an underdeveloped field of research and addresses a societal concern that should be investigated to a deeper extent in the future.
1. Introduction
Over the past decades, lines of credit have become increasingly available due to the digitalization of the financial sector. Although rapid access to credit can aid in avoiding financial concerns, paradoxical consequences can result. Household debts have become national concerns. Recently, more attention is paid to increasing financial awareness by initiatives such as “Wijzer in Geldzaken”. This is one example of initiatives that aim to create a better understanding of savings and liabilities and hope that less people will have financial problems in the future. Simultaneously, the abolishment of financial aid (studiefinanciering) for students in The Netherlands received much criticism in response to the argument that college educated adults will receive higher incomes and will therefore be able to absorb potential debts to finance their studies. This argument has been a widely used in other countries such as the United States to motivate and tolerate the borrowing of large sums of money to finance education and has led to enormous student debts.
This trend above calls for an understanding of drivers of delinquency. More specifically, to explore the relation to level of education and financial behavior. Limited literature explores the relationship between these two concepts. However, much research covers the linkage between education, financial behavior and financial literacy. In order to deepen the understanding between these concepts, the research question of this paper is: “To what extent does the role of the degree of theoretical education and education in years influence harmful financial behavior in measures of risky debt mediated through financial literacy? This paper adds content to a field of research that has not been developed to its fullest extent. By identifying driving factors behind payment delinquencies, policies can be targeted more specifically in order to reduce the levels of payment delinquencies.
Additionally, this paper observes both the degree of theoretical education rather and educational attainment in years which has not been done before.
This paper finds statistically significant results in the relation between the degree of theoretical education and financial literacy as well as the relation of education in years and financial literacy. The paper finds no significant mediating effects of financial literacy on risky debts in any of the four measures. This means that no concluding statements can be drawn on the role of financial literacy in the relation between education and risky debts. The rest of the paper is structured as follows: section 2 gives a literary overview of the research that has been conducted until now and outlines the proposed hypotheses. Section 3 explains the conceptual framework and the empirical statistics that will be used in the
analysis. The following section, section 4, outlines the data of the various variables and provides an elaborate descriptive summary table. The results in section 5 are given in
twofold: firstly, graphical results are outlined and then the empirical results are demonstrated and explained. Section 6 provides a conclusion and discusses possible limitations involved in the study and suggestions for future research.
2. Literature 2.1. Education and Income Inequality
2.1.1 Income Inequality
Education and income inequality are two concepts in macroeconomics that are tied together. Income inequality is a topic that has been widely assessed since the start of the economics discipline. As Pikketty and Saez (2003) demonstrate, inequality can follow waves as industrial revolutions occur and can increase and decrease over time. The famous Kuznets curve is often used to describe these trends: following an inverse U-type as the developments progress. Remarkably, as declining inequality marked the post-World War II Era, increasing inequality has been prevalent since the 1970s. Over the past years, the topic of income inequality has received significantly more attention resulting from increased disparity in affluent countries. Concerns are grounded in perceptions of social justice: do we allow some to struggle to put food on the table meanwhile others enjoy luxurious lives? According to Gilbert (2008), three main trends can be observed if we look at the period since the 1970s. Firstly, income inequality has increased in most high-income countries since the 1970s. Secondly, this growing inequality results into the disappearance of middle-income households. Thirdly, the trend of increasing inequality is in contrast with the trend of the decline of income inequality between the second world war and the 1970s.
Aside from the social aspect of income inequality, the economic consequences of income inequality are also widely discussed in literature. Many studies are based on the Deiniger and Squire’s (1996) dataset. The study contains observations both over time but also across countries. The dataset is widely used, for instance, Barro (2000), Forbes (2000) and Banerjee and Duflo (2003) all use different samples from Deiniger and Squire’s dataset. However, conflicting results were found because different parts of the dataset were used in different studies.
Leading to, Grigor Sukiyasan (2007) conducted his own study. He empirically evaluates the relationship between income inequality and economic growth using the
transition economies of Central and Eastern Europe and the Commonwealth of Independent States. He finds a negative strong relationship between income inequality and economic growth (Sukiyasan, 2007). As many share the findings of Sukiyasan, concerns have led to research on the sources of income inequality. The study by Kim and Tebaldi (2013) observes trends of income inequality in the United States and aim to find explanations and driving variables of fluctuations in income inequality. According to Kim and Tebaldi (2013), earned income is the main driver behind income inequality: “the presented evidence altogether highlights the importance of earned income in understanding overall income inequality” (Kim & Tebaldi, 2013, p.8). It is outlined that income from earned income is more relevant than income from wealth accumulation. As earned income is income generated form paid work (labor income), a direct relationship between earned income and education is drawn.
Recent concerns do not solely focus on the dangers of income inequality, but also cover the topic of debts which become increasingly more prevalent. More specifically, digitalization of the financial world has made it easier to apply for a credit card, make installment payments and get extended lines of credit. This leads to a new concern, namely debts with high interest rates. This paper aims to turn to a novel angle by moving from establishing a relationship between education and income to revealing the correlation between education and debts with high interest rates.
2.1.2 Education
As demonstrated by the many studies focusing on the relationship between educational attainment and income, it is known that educational attainment is positively related with income (Morgan & David, 1963 ; Gregorio & Lee, 2002). Nevertheless, education systems are different across countries. Both studies mentioned before are focused on educational attainment in years. Globally, different educational systems focus on different relative importance in the attainment of theoretical knowledge versus learning practical skills. Limited literature is focused on these specific differences which is why this paper addresses this variable. This study aims to focus on two measures of education – the degree of
theoretical knowledge versus and the years of education.
The educational system in The Netherlands allows for the measurement of two different variables of education. To clarify, students in the Netherlands take a test, and based on this test continue to different types of secondary education. The options are VMBO, HAVO or VWO. These types of secondary education give access to higher education that can also be categorized into three types: MBO, HBO and university. According to Nuffic (2018),
MBO is mostly focused on practical skills, HBO has a mixture of practical and theoretical knowledge and WO is mostly theoretical. The novelty of this thesis is that minimal research has been conducted in this aspect of Dutch education. Please refer to Figure 1 to gain an understanding of the education system in The Netherlands.
2.2 Debts
The relationship explored in this paper is the relationship between type of education and level of debt. This section gives a literary overview on the topic of debt. Debts define a range of credit agreements including student loans, mortgages, credit cards and private loans, it is essential to differentiate between two different types of debts: differentiated by defining healthy credit and risky credit based on the levels of interest rates.
2.2.1 Healthy debts
This section aims to motivate why student loans and mortgages are defined as healthy debts in this paper. As tuition fees increase and more students attend college, many college students are compelled to take out student loans to finance their education just like adults take out mortgages to finance housing. Aggregate student loan debt was the only increasing form of debt together with auto debt, since the recession in 2008 (Edmiston, 2017). The weak economy has motivated students to aim for well paid jobs which require a college degree rather than choosing a lower-paying job in public interest such as joining the police
department (Rothstein & Rouse, 2007). Other characteristics of high student loans are found in a decrease of 7% in long-term probability of getting married. An increase in college students who plan to move back home after graduation is also linked with high student debts (Dickler, 2010). On the other hand, recent research convincingly finds that college adds value in the long-term. Although it is difficult to isolate the effect of college accounting for biases in characteristics, sociodemographic characteristics and opportunities, known as the ability bias, a sophisticated economic literature has developed to address this issue: Goldin and Katz find that “higher education is a worthwhile investment” (Goldin & Katz, 2018, p.14).
Additionally, Moretti (2004) demonstrates that higher education increases skilled labor and has a positive effect of productivity of the workforce (Moretti, 2004). Other studies find that obtaining a college degree has a positive effect on income (Angrist & Krueger, 1991; Card, 2001; Heckman et al., 2006).
In The Netherlands, the so-called “Leenstelsel” was introduced, reducing government grants and increasing student loans. The effectiveness of this new system is beyond the scope of this paper but should be investigated in future research. Although a research gap exists between the recent increase in average student debts, studies have convincingly argued that a college education is a worthwhile investment. Therefore, we identify student loans as
As described earlier, education is directly related to income and therefore to the size and number of mortgages one can obtain. With a higher income, one can take higher
mortgages and thus a higher debt. However, the credit-checks beforehand are proportional and therefore a higher mortgage debt for someone with a higher level of income does not necessarily translate into a risky or harmful debt. As Peñaloza & Barnhart (2011) define mortgages as “good debt” in the means that a mortgage triggers self-discipline to be paid back (Peñaloza & Barnhart, 2011). Additionally, historically, housing provides returns that are marked as one of the safest. Therefore, it is can be argued to define mortgages as healthy debts in this study.
2.2.2 Risky Debts
This section aims to define credit card debts, installment debts, post-order debts, private loans, extended credit lines and loans from family as potentially risky debts and focus on drivers of delinquency.
A well-known type of credit widely used is the use of credit cards. Solely observing the vast amount of literature that covers credit card delinquency, signals the difficulties people experience in repaying the credit back in the proposed time schedule. The issue often is that credit card companies allow consumers to pay very minimal payments of the balance as monthly requirement – often only 1% to 5% of the total balance. Therefore, it can take years to pay back loans and makes the balance less transparent (Bricker & Bucks, 2013). This then results into interest rates being added to the original credit amount, which increasing the amount to be paid back and the problems associated with it. For instance, at the end of 2014, $861 billion was revolving credit outstanding (Butaru et al., 2015). Secondly, credit cards are an easy manner to reach credit in a relatively short period of time. Many banks do credit checks beforehand and see if someone is eligible. The use of credit cards has significantly increased over the past decades (Bricker et al., 2013).
Not only “regular” people have made use of the opportunity to use credit card in order to close some debit gaps in their finances, credit cards have become readily available for students as well. Credit-companies receive a reported $13 billion in discretionary income generated by the college student market (Kara et al., 1994). As a matter of fact, Manning (1999) demonstrates that credit card debts under students have been increasing over the past years, marking an average credit card debt of $2000 per student (Manning, 1999).
Subsequently, Davies and Lea (1995) demonstrate that higher levels of debt result into higher level of tolerance of debt when these students become adults: getting used to debts from a
young age is worrisome for future financial behavior (Davies & Lea, 1995). Credit-card debts and installment debts are treated equally in terms of payment compared to other forms of debt (Wang et al., 2011). In this paper, any debts that are characterized by high interest rates are defined as potentially risky debts.
Delinquent payers do not only form an obstacle for themselves in terms or requesting loans in the future, investors of these credit are the ones most concerned with factors that drive delinquent behavior because of the outstanding credit. Defining drivers of delinquency and factors that indicate more leaning towards delinquent behavior can help to identify the problems and resolve these issues. Naturally, the problem is two-sided: on the one hand, credit-card companies make profits because people pay late and pay interest rates over these late payments. On the other hand, these companies want their consumers to at some point be able to repay otherwise they end up losing money. Therefore, identifying the factors that drive delinquent behavior is important.
A lot literature focuses on macroeconomic factors that lead to delinquent behavior, specifically unemployment is a factor that is widely discussed and claimed to be a driver of delinquency (Deng et al., 2000; Livshits et al., 2007). However, as Agarwal and Liu (2003) explain, even when economic growth is high, unemployment rates are relatively low (in the period 1995-2001), credit card delinquencies kept increasing in the United States (Agarwal & Liu, 2003). The literature on macroeconomic drivers give mixed results, which is why this paper focuses on sociodemographic factors such as education.
Literature finds that: “Older and more educated consumers also were found to have a lower probability of being behind on their payments, holding all other factors constant” (Stavins, 2000, p. 24). Therefore, this paper explores the role of education, measured in twofold, in relation to risky debts.
2.3 Education and Debts
After the establishment of the association between income inequality and education, this section continues to examine the relationship between education and debts. College graduates earn higher wages, have lower unemployment rates and are more likely to have employer-provided health insurance which are all factors that may have an influence on unsecured debts (U. S. Bureau of Labor Statistics, 2009). Some literature covers the relationship between educational attainment and levels of debt. To demonstrate, Zhan and Sherraden (2011) find that unsecured debt is negatively related to college completion (Zhan &
prevalent among young college-educated adults but unequally distributed among social classes and more prevalent among lower social economic household (Houle, 2014).
Thus, after a literary overview on educational consequences and the importance to define drivers behind delinquency we formulate the following hypotheses based on the two measures of education and the fours measures of risky debt:
1.a: A positive relationship is expected between the category of education and the level of
debt measured as dummy (𝑌")
1.b: A positive relationship is expected between the years of education and the level of debt
measured as dummy (𝑌")
2.a: A positive relationship is expected between the category of education and the level of
debt measured in amounts (𝑌#)
2.b: A positive relationship is expected between the years education and the level of debt
measured in amounts (𝑌#)
3.a: A positive relationship is expected between the category of education and the level of
intensive margin debt (𝑌$)
3.b: A positive relationship is expected between the years of education and the level of
intensive margin debt (𝑌$)
4.a: A positive relationship is expected between the category of education and whether a
debt belongs to the top 10% (𝑌%)
4.b: A positive relationship is expected between the years of education and whether a debt
2.4 The mediating role of Financial Literacy
Financial literacy is a concept undeniably associated with both education and risky debts which why the potential role financial literacy between education and debt is assessed in this paper. The literature provides evidence that many that factors lead to delinquent behavior. One factor often mentioned is financial literacy. Financial literacy is defined as “‘A
combination of awareness, knowledge, skill, attitude and behavior necessary to make sound financial decisions and ultimately achieve individual financial wellbeing” (Atkinson & Messy, 2012, p. 14).
One dimension of financial literacy is how financial education enhances financial literacy. Fox, Bartholomae and Lee (2005) succeed in providing an elaborate summary of financial education initiatives introduced in the US in the past years and believe financial education will increase financial literacy (Fox et al., 2005) Nevertheless, skeptic critics voice concerns on whether financial literacy has a direct impact on financial behavior as Ambuehl, Bernheim and Lusardi (2014) demonstrate in their research: no significant improvement in financial decisions can be explained by financial literacy. They explain that: “while financial literacy undoubtedly plays an important role in decision making, the associated mechanisms are complex and mediated by a variety of other factors” (Ambuehl et al., 2014, p. 4). To examine the relation between the two measures of education and financial literacy, the following hypotheses follow:
5.a: A positive relation between the category of education and financial literacy score is expected
5.b: A positive relation between the years of education and financial literacy score is expected
A different dimension is the way financial literacy affects risky debts. Almenberg and Widmark (2011) find that levels of higher financial literacy are positively correlated with market participation and mortgages (Almenberg & Widmark, 2011). On the other hand, Huston (2010) provides a review of the literature on the effect of financial literacy and concludes that findings are mixed (Huston, 2010). Johnson and Sherraden (2007) aim to find the way financial literacy affects financial behavior and conclude that increasing financial literacy will lead to an increase in financial capability but requires access to financial services, policies and instruments. They continue to propose policy changes to make access feasible (Johnson & Sherraden, 2007). Additionally, Lusardi and Mitchell (2006)
illiterate and make poorer decisions in their retirement planning (Lusardi & Mitchell, 2006). Van Rooij et al. (2007) find that lower level of financial literacy result into lower level of market participation and thus less opportunity to accumulate value from welfare (Van Rooij et al., 2007). Bay et al. (2014) aims to provide context to the term “financial literacy”. They explain that financial literacy is not merely the ability to understand financial processes but also the ability to apply them which is requires more than simply knowledge, also but not limited to character, time and place (Bay et al., 2014). From the above-mentioned findings, the following hypotheses based on the relation between financial literacy and risky debts follow:
6.a: A positive relation between financial literacy score and the level of debt measured as
dummy (𝑌") is expected
6.b: A positive relation between financial literacy score and the level of debt measured in
amounts (𝑌#)
6.c: A positive relation between financial literacy score and the level of intensive margin debt
(𝑌$) is expected
6.d: A positive relation between financial literacy score and whether a debt belongs to the
top 10% (𝑌%)
Thus, literature demonstrates mixed results on the topic of efficiency of financial education and the relation between literacy and financial behavior. Also, the relation between financial literacy and debts is not widely discussed. Little is known about the interactions involved in the relationship between type of education and debt levels and the interaction of financial literacy. Therefore, our aim is to investigate the possible mediating role of financial literacy in the relationship of education and debt and hypotheses are empirically developed in section 3.
3. Conceptual Framework & Empirical Specification 3.1 Conceptual Framework
The aim of the paper is essentially to understand the relationship between education and risky debts. Additionally, the possible mediation effect of financial literacy is assessed in this paper. Mediation analysis is defined as “the statistical approach used to understand how the effect of an independent variable on an outcome is transmitted through an intervening variable (the mediator)” (Tofighi &Thoemmes, 2013, p.93). The following reduced-form model (Figure 2) is generated to assess the relationship between the two measures of education and the four measures of risky debt including the mediating variable: financial literacy.
Figure 2: Conceptual Framework
The variables in the sharp squares denote the independent and outcome variables, education and risky debt respectively. Education is measured in two measures represented as 𝜏. The first one is the degree of theoretical knowledge (𝜏 = 1) used in part I of the empirical specification, the second one is education in years (𝜏 = 2) and is used in part II of the empirical specification. The circle indicates the suggested mediator, financial literacy. The round squared represent the error terms: 𝑢+ represents the unobserved characteristics
that are associated with the moderator: financial literacy. The relationships between the variables are denoted with 𝛾", 𝛿" and 𝛼". The relation between education and financial
literacy is represented by 𝛾" . The relation between financial literacy and the measure of risky debt is represented by 𝛼". The third variable, 𝛿", denoted the relationship between education
and risky debt. The term in the center of the figure, term 𝛽1," is defined as the total effect including the mediated effect of financial literacy and is represented by 𝛽1," = 𝛿1,"+
𝛾1," 𝛼1,". The blue dotted line symbolizes a possible relation between the two error terms 𝑢+ and 𝑢5.
3.2 Empirical Specification
In this section, empirical specifications to test the hypotheses are outlined. As mentioned, two measures of education are measured. To clarify, the first part marked of the empirical
specification, equations (1), (2), (3) and (4), is when the measure of education as degree of theoretical knowledge and when 𝜏 = 1. For this measure, dummy variables for education are generated. As explained above, education levels are grouped into four categories of
education. The reference dummy, stands for education is group 1. 𝐷# defines education is group 2, and 𝐷$ defines when education is group 3 and 𝐷% denotes education is group 4. The
second part, equations (5), (6), (7) and (8) are meant to use the measure of education in years of education when 𝜏 = 2.
3.2.1 Part I
Firstly, the relationship between X and Y, education and measure of risky debt, is assessed. This empirical equation is used to reject or confirm hypothesis 1.a, 1.b, 1.c and 1.d. The measure of education (Y) is measured in four ways which will be explained later in this paper. The four measures are now expressed as k which can take values 1 to 4. Therefore, the measure of education is denoted in the equations as 𝑌1. The association between education and risky debt are represented by 𝛽1,", 𝛽1,#, 𝛽1,$ and 𝛽1,%. Control variables age and gender are added as a vector variable denoted by W. The coefficient of the control variables is 𝜇. The error term is denoted by 𝑢+.
𝑌1 = 𝛽1,"+ 𝛽1,#𝐷#+ 𝛽1,$𝐷$+ 𝛽1,%𝐷% + 𝜇′𝑾 + 𝑢+ (1)
The following empirical equation is aimed to find the relation between the suggested mediator and the independent variable. Namely the relation between financial literacy and education. As demonstrated in the conceptual framework, the relation between education and financial literacy is denoted by 𝛾". The following empirical equation is generated to
demonstrate this relation:
𝑍 = 𝛾=+ 𝛾1,#𝐷#+ 𝛾1,$𝐷$+ 𝛾1,%𝐷%+ 𝜇′𝑾 + 𝑢5 (2)
Financial literacy is defined by Z. In this equation, 𝛾 denotes the association of the specific education category on financial literacy. To note, an additional control variable is added, risk aversion, in order to ensure exogeneity of the error terms. The control variables are denoted with the vector variable 𝜇′𝑾, and the error term is represented by 𝑢,.
This section continues with the development of empirical specification to compute the relationship between education and debts incorporating the effect of a possible mediator. Therefore, the following equation is computed:
𝑌1 = 𝛿1,=+ 𝛿1,#𝐷#+ 𝛿1,$𝐷$+ 𝛿1,%𝐷%+ 𝛼1,"𝑍 + 𝜇>𝑾 + 𝑢
? (3)
Where 𝑢? = 𝑢++ 𝑢,
Equation 3 dissects the total effect as described in equation (1). The first part of equation (3) (𝛿1,=+ 𝛿1,#𝐷#+ 𝛿1,$𝐷$+ 𝛿1,%𝐷%) represents the direct effect of education on risky debts.
Financial literacy is defined by Z. The second part (𝛼1,"𝑍) captures the mediated association
by accounting for the association that financial literacy is responsible for on risky debts. By formulating equation (3) the different paths of association become visible. The first error term (𝑢+) is responsible for any unobserved characteristics between education and risky debt,
and the second error term (𝑢,) for unobserved characteristics influencing financial literacy, which in turn influence risky debts measured in k. In order to account for these observed characteristics, risk aversion is included as extra robustness control variable.
Following naturally, in order to understand the role of the mediating variable, the following equation is computed. Using the definitions 𝛽1," = 𝛿1,"+ 𝛾1," 𝛼1," , it is computed that if 𝛿1," = 𝛽1," then 𝛾1," 𝛼1," is zero and proof of any mediation effect is
absent. However, if 𝛿1," ≠ 𝛽1,", then 𝛾1," 𝛼1," ≠ 0 and proof of a mediation effect is found. Thus, the fourth equation:
𝛿1,"≠ 𝛽1," 𝑎𝑛𝑑 𝛾1," 𝛼1," ≠ 0 (4)
Based on the empirical literature, it is expected that equation (4) follows and that the
mediating variable will capture some of the association. This train of thought is followed by the following hypothesis:
7.a: When financial literacy is added to the equation 𝛿"," ≠ 𝛽",", when education is
measured in degree of theoretical knowledge
8.a: When financial literacy is added to the equation, 𝛿#," ≠ 𝛽#,", when education is
measured in degree of theoretical knowledge
9.a: When financial literacy is added to the equation 𝛿$," ≠ 𝛽$,", when education is
measured in degree of theoretical knowledge
10.a: When financial literacy is added to the equation 𝛿%," ≠ 𝛽%,", when education is
measured in degree of theoretical knowledge
3.2.2 Part II
This section aims to find the empirical specifications for the same associations as in part I, however education is now measured in years rather than in degree of theoretical knowledge. Thus 𝜏 = 2. Building on this the following equation is computed:
𝑌1 = 𝛽1,=+ 𝛽1,"𝑋 + 𝜇′𝑾 + 𝑢+ (5)
Where 𝑌1 denotes the outcome variable, risky debt measured in k. The total effect of
education in years on risky debt (𝑌1) is denoted by 𝛽1,". The years of education measures in years is denoted by X. Control variables age and gender are added as a vector variable
denoted by W. The coefficient of the control variables is 𝜇. The error term, unobserved characteristics in the effect of education in years on risky debts is measured by 𝑢+.
Moving to the second association, the relation between financial literacy (Z) and education measured in years (X) is demonstrated by equation (6). Where control variables age, gender and risk aversion are added as a vector variable denoted by W. The coefficient of the control variables is 𝜇. The error term, the unobserved characteristics in the effect of education in years and financial literacy (Z), is measured by 𝑢,
𝑍 = 𝛾1,=+ 𝛾1,"𝑋 + 𝜇′𝑾 + 𝑢, (6)
The third equation of part II represents the total effect but dissected into different paths of association. The direct path between education measured in years (X) and risky debts measured in k is represented by the coefficient 𝛿1,". The path between financial literacy (Z) and risky debts measured in k is represented by the coefficient 𝛼1,". There are two error terms association with this equation, namely 𝑢+ and 𝑢,. The first error term responsible for any unobserved characteristics between education and risky debt, and the second one for unobserved characteristics influencing financial literacy which in turn influence risky debts measured in k. Risk aversion is included as control variable to ensure exogeneity between the error terms.
𝑌1 = 𝛿1,=+ 𝛿1,"𝑋 + 𝛼1,"𝑍 + 𝜇>𝑾 + 𝑢
? (7)
Where 𝑢? = 𝑢++ 𝑢,
Moving to the final equation of this section, we conclude with equation (8). Similarly, like equation (4) in part I, equation (8) aims to symbolize the association that this paper is interested in. Using the definition of 𝛽1,"= 𝛿1,"+ 𝛾1," 𝛼1,": if the coefficient 𝛽1," from equation (5) is equal to the coefficient 𝛿1," from equation (7) it would follow that 𝛾1," 𝛼1," = 0. As the literature review suggests that there is a mediating effect of financial literacy we expect there to be a relation:
This chapter concludes with the following hypotheses:
7.b: When financial literacy is added to the equation 𝛿"," ≠ 𝛽"," , when education is
measured years
8.b: When financial literacy is added to the equation 𝛿#," ≠ 𝛽#," , when education is
measured years
9.b: When financial literacy is added to the equation 𝛿$," ≠ 𝛽$," , when education is
measured years
10.b: When financial literacy is added to the equation 𝛿%," ≠ 𝛽%," , when education is
4. Data and Summary Statistics
4.1 Sample
The data used comes from the DNB Household Survey (DHS). This survey is focused on collecting data on both psychological and economic aspects of financial behavior. The panel data exists since 1993 and provides data on over 2000 households every year1. The dataset is designed to be representative of the Dutch population. The data is collected by CentERdata, a survey research institute at Tilburg University. The data set provides information divided into 6 sections: 1) general information about the household 2) Households and employment 3) living conditions and mortgages 4) Health and income 5) Possessions and loans 6) Economic and psychological concepts. This paper makes use of the 2005 data wave2.
4.2 Independent Variable: Education
The DHS survey provides very elaborate data on levels of education which are used to compute two measures of education used in this paper. The data on education is collected in twofold: the highest level of education ever attended, and the highest level of education completed with a diploma. As it is essential that one has gone through all years of a program, the latter collection of educational attainment is chosen. The DHS has categorized education into 7 groups: No schooling or special education, primary school, VMBO, HAVO, VWO, MBO, HBO and University. The first measure of education is the degree of theoretical education. Referring to Figure 1, the different levels of education are sorted into 4 categories:
1. No schooling or special education/primary school/VMBO/MBO 2. HAVO/VWO
3. HBO 4. WO
Ranked from 1 to 4, 4 indicating the most theoretical studies and 1 indicating the least theoretical studies. The level of education, thus the independent variables, are treated as dummy variables in the analysis.
The second measure of education is education in years. The seven educational attainment categories from the DHS are given years of education according to Figure 1. An
1http://www.uvt.nl/centerdata/en/. See Nyhus (1996) for a detailed description of this survey and an assessment of the quality of the data 2 Access is given by Gamze Demirel who is assistant research at the Survey Research Department at CentERdata
overview of the provision of education in years is presented in Table 1. As observed, education in years ranges from 8 to 19.
Table 1: Education in years
Education Primary Secondary Higher Total
Special / no education 8 8 Primary school 8 8 VMBO 8 4 12 HAVO/VWO 8 5 13 MBO 8 4 4 16 HBO 8 5 4 17 University 8 6 5 19
Thus, the two measures of education that are tested: education as the degree of theoretical education and education in years.
4.3 Dependent Variable: Level of Debts
As this paper assesses financial behavior specifically targeting debts that could potentially be harmful, the aim is to look at debt with high interest rates. This is naturally concluded from the fact that debts with the higher interest rates should be paid off first. Therefore, mortgages and study loans are excluded, and the analysis focuses on credit card debts, private loans, extended lines of credit, installment debts, loans from family and other debts with high interest rates. The DHS data includes these forms of debt. These forms of debt, from now on risky debts, are described in four manners. The first measure is a dummy variable whether an individual has any of these debts at all denoted by 𝑌". The second measure of risky debt is the sum of the amount of debt in any of the risky debt categories described before denoted by 𝑌#. Moving to the third measure of risky debt, 𝑌$, is named the intensive margin of debt meaning the size of the debt conditional on having a debt at all. The fourth measure is again a dummy variable, indicating whether a person falls in the top 10% of debt or not denoted by 𝑌%.
4.4 Mediator 1: Financial Literacy
As explained, financial literacy is suggested to be correlated with financial behavior. Van Rooij et al. (2007) dedicated their research on financial literacy and highlight the importance of financial literacy in their paper on literacy and stock participation. They make use of the DHS data set as organized by CentERdata in Tilburg. In 2005 and 2010 they added an extra module to the DHS survey to the “heads of the households” and measured the rates of financial literacy among these household. This is a valuable measurement to the DHS set because it provides information of financial literacy in combination with the extensive set of questions asked in the DHS panel data survey3. Financial literacy is measured with two subsets of questions: one to test basic financial knowledge, second to assess more advanced financial knowledge. In this paper, only the basic set of questions will be used as only those questions are applicable to risky debts as we are interested in. The second set more advanced questions focus on stocks, shares and bonds knowledge and is not relevant when assessing one’s personal debts. The questions can be found in Appendix A.
4.5 Control Variables: Age & Gender
When assessing the relationship between degree of theoretical education and financial behavior in terms of debt and mediated through financial, it is important to control for additional variables when regressing the estimates. In order to counter problems in internal validity, control variables are included.
Firstly, a wide range of literature covers differences between males and females concerning education, financial behavior and risk attitude. Since all three of these elements are variables in our analysis it is important to include gender as a control variable. Education cannot cause gender, but gender can influence educational attainment which is gender is included as a control variable.
Secondly, age can be a determinant to capture differences between generation in education and financial behavior and is therefore also important to capture the specific effect of age. In the dataset, birthyear is an available measure and a new variable, age, is generating by setting the variable age as 2005 and subtracting the birthyear.
4.6 Control Variable: Risk Attitude
Risk attitude is, similarly to financial literacy, a variable that often correlated with financial behavior. A concern when doing a mediation analysis is the exogeneity of the error term 𝑢,. It is assumed that 𝑢+ and 𝑢, are uncorrelated. In order to ensure this exogeneity assumption, risk aversion is included as a control.
Risk attitude is defined by Allison Rosen as “a construct that describes decision making over quantifiable outcomes under conditions of uncertainty” (Rosen et al., 2003, p. 5). Jung (2015) finds a negative significant effect of education on risk aversion. This is supported by a wide range of literature (Jung, 2015). In line with Jung, Riley and Chow find similar results in the sense that risk aversion and education are negatively related (Riley and Chow, 1992). Recently, Francois Outreville (2013) demonstrates a negative relationship between risk aversion and education (Outreville, 2013). However, Jianokoplos and Bernasek (1998) demonstrate a positive relationship between education and risk aversion (Jianokoplos & Bernasek,1998). Again, this paper differentiates because two measures of education are considered: the years of education one has acquired and the proportional theoretical versus practical education rather than solely the years of education. Thus, as many empirical studies show relationships between education and risk aversion, it is important to include the variable risk aversion as a control variable in this paper to exclude any confounding factors.
Risk attitude can be measured in degree of risk aversion. In the same study conducted by Van Rooij et al. (2007), risk aversion was measured asking 2 questions. These answers provide 4 categories of degree of risk aversion4. Nevertheless, only 833 subjects answered these questions which is why a degree of risk aversion is assigned to only 833 subjects. The original risk aversion questions can be found in Appendix B.
4.7 Descriptive Statistic
Table 2: Descriptive summary
Variable Observations mean Std. Dev. Minimum Maximum
General Characteristics Female 1146 0.57 0.49 0.0 1.0 Age 1146 50.59 15.11 22.0 90.0 Retired 1146 .25 .43 0.0 1.0 Number of children 1146 .65 1.02 0 5 Education 1. no schooling or special/ only primary/VMBO 1146 .25 .42 0 1 2. MBO 1146 .30 .46 0 1 3. HAVO/VWO/HBO 1146 .33 .47 0 1 4.University 1146 .12 .33 0 1 Debts number in debt
Checking account deficit 1146 132 n/a n/a n/a n/a
Private loan 1146 34 7 386 10 571 0 40 480
Extended line of credit 1146 115 6 351 10 260 0 62 000
Credit card debt 1146 42 1 910 3 209 0 20 000
Loan from family 1146 37 6 856 10 367 100 40 000
Installment debt 1146 8 35 686 59 568 638 177 000
Post order debt 1146 21 946 1 092 53 4 630
Other debt 1146 5 58 600 56 207 1500 141 500
Any of the above 1146 281 .25 .43 0 1
Total amount of debt 1146 281 1 670 9 400 0 177 000
Top 10 percent in debt 1146 115 .10 .30 0 1
Financial Literacy Number correct
Question 1 1143 1070 .93 .24 0 1 Question 2 1143 916 .80 .40 0 1 Question 3 1143 984 .86 .35 0 1 Question 4 1143 871 .76 .43 0 1 Question 5 1143 814 .71 .45 0 1 Total correct 1143 4.06 1.16 0 5 Risk Aversion Number per category
Least risk averse (1) 833 25 0.03 .09 0 1
Slightly risk averse (2) 833 85 .10 .30 0 1
Risk averse (3) 833 129 .16 .36 0 1
Very Risk averse (4) 833 594 .71 .45 0 1
From Table 2, several characteristics of the dataset are observed. Firstly, it is noted that a slight majority of the subjects is female (57%) and that the average age is 50. Also, 25% of the subjects are retired and the average number of children is 0.65. As gender is included as control variable, this should not influence the regression outcomes.
Continuing to the education descriptive statistics, 25% of the subjects have not acquired a degree of education after primary school or VMBO, 30% of the subjects have completed an MBO degree and 33% have either completed HAVO, VWO or HBO with a diploma. Only 12% have finished university and received a diploma.
Moving to the risky debt section, 25% of the subject has a debt of some kind in the categories defined above. The largest debt category being a deficit on the checking account balance. However, the amount of deficit is not available in the dataset. This limitation is addresses in the discussion in section 6. The second largest category is an extended line of credit: more than 10% of the subjects have an extended line of credit ranging from a debt of 0 to 62 000 euros. A dummy variable is created to indicate the subjects that belong to the top 10% of debtors. These debts range from 1 900 to 177 000 euros indicating that the range of these debts is relatively large. This is also noted when looking at the standard deviations. They are all fairly large indicating low level of precision.
The third set of characteristics, indicates levels of financial literacy among the
subjects. As can be noted, for every question, the majority of the subjects answered correctly, which is observed when looking at the mean of the total sample, which is 4.06.
The fourth category describing the data on risk aversion also shows some remarkable numbers. As a matter of fact, 71% of the subjects fall into the category described as most risk averse. As the measure of risk aversion is not normally distributed, it may cause difficulties to observe a trend when regressed on education.
Also note, the numbers of observations shrink moving from the category debts, to financial literacy with 3 observations (1146 to 1143). Furthermore, the number of
observations shrink further moving form financial literacy to risk aversion (1143 to 833). This means that not all participants have answered all questions regarding financial literacy and risk aversion.
5. Results
This section aims to demonstrate the results of this paper both graphically as well as empirically. This section starts out with a visualization of the relationship between the variables. As four measures of risky debt are assessed, the graphical section focuses on only one, merely to demonstrate different trends. The measure of risky debt used in the graphical analysis is risky debt in numbers (𝑌#). Six graphs show association between the independent variable and the outcome variable, the independent variable and the mediator, and lastly the mediator and the outcome variable. The first three graphs take education as a measure of degree of theoretical knowledge whereas the second three graphs take education as a measure of education in years. Section 5.2 continues with analysis empirically. The empirical section contains 8 tables outlining the relationship between the four measures of debt, education, mediators and control variables. The first four tables use the degree of theoretical knowledge as measure of education, the second four tables use education in years as measure of
education.
5.1 Graphical Observations
This section visualizes the three different relationships on which the empirical equations are based. The relations are the following: the independent variable and the outcome variable, the independent variable and the mediator, and lastly the mediator and the outcome variable. The first three graphs take education as a measure of degree of theoretical knowledge whereas the second three graphs take education as a measure of education in years.
Graph 1 visualizes the relationship between the independent variable and the outcome variable, education as a measure of degree of theoretical knowledge and risky debts
respectively. An upwards sloping linear trend line is visualized. Counterintuitively, this would translate into a positive relationship between risky debts and education. The mean average of risky debts is higher for more theoretically educated subjects than less
theoretically educated subjects.
To continue, in Graph 2, education and financial literacy are graphed. The linear trend line is upward sloping suggesting a positive association between education and financial literacy. In other words, as an individual is more theoretically educated, his or her financial literacy score increases. This is in line with the hypothesis, however no concluding
The third graph shows the relation between the financial literacy score and risky debt. This shows that the amount of risky debt increases as one’s financial literacy score increases which is an intuitive trend: the more knowledgeable a person is about finances, the better he manages his finances.
As Graph 2 and Graph 3 show no flat lines, we assume that financial literacy
contributes a mediation effect. Although these observations seem to provide evidence for the proposed hypothesis, these observations should be interpreted with much caution. In order to separate the effect of education and financial literacy on the amount of risky debt, it is
necessary to run regressions to separate these associations which will be calculated in section 5.2.
The following set of graphs visualize the same associations as before, but with a different measure of education, education in years. Since the third graph (Graph 3 and Graph 6) visualize the association between financial literacy and risky debts, no changes will be observed, because a different measure of education has no effect in this observation.
In Graph 4, an upward sloping linear trend line is observed interpreting a positive relationship between education measured in years and risky debts. This is the same
observation as when education is measured in degree of theoretical knowledge and is for the same reasons counterintuitive: more years of education would naturally lead to less risky debts as indicated in the empirical literature overview.
Additionally, the relation between education measured in years and financial literacy also appears to be positive just like Graph 2. This indicates that financial literacy scores increase as the years of followed education increase. This is in line with the hypothesis. However, no concluding statements can be drawn from this observation as we need empirical evidence whether this is the case or not.
To conclude, the Graphs 1 to 6 allow an observation of associations between education and risky debt, education and financial literacy and finally, financial literacy and risky debt all observed in both measures of education. This suggests that financial literacy may be a mediating variable in the relation of education and risky debt. Although these observations seem probable, they should be treated with caution. The following section explores these associations empirically and only then concluding statements can be drawn.
3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 1 2 3 4 Fi na nc ia l L it er ay S co re Education Education and Finanical Literacy
0 500 1000 1500 2000 2500 3000 3500 4000 1 2 3 4 Ri sk y De bt Education Education and Risky Debt
0 2000 4000 6000 8000 10000 0 1 2 3 4 5 Ri sk y De bt
Financial Literacy Score Financial Literacy and Risky Debt
Graph 1: Education (degree theoretical) and Risky Debt Graph 2: Education (degree theoretical) and Financial Literacy Score
Graph 3: Financial Literacy Score and Risky Debt
0 500 1000 1500 2000 2500 3000 3500 4000 7 9 11 13 15 17 19 Ri sk y De bt Education in Years Education in Years and Risky Debt
Graph 4: Education in years and Risky Debt
3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 7 9 11 13 15 17 19 Fi na nc ia l L it er ac y Sc or e Education in years Education and Finanical Literacy
Graph 5: Education in years and Financial Literacy Score
0 2000 4000 6000 8000 10000 0 1 2 3 4 5 Ri sk y De bt
Financial Literacy Score Financial Literacy and Risky Debt
5.2 Empirical Results
This section demonstrates the empirical results from the equations computed in section 3.2. The results are used to either confirm or reject the null hypotheses proposed. Table 3 outlines the relationship between the two measures of education and the suggested mediating variable: financial literacy including the control variables.
The consecutive eight tables equate the relationship between the two measures of education and the four measures of debt. The tables are ordered as follows: Table 4a looks at the first measure of education and the first measure of risky debt. Table 4b looks at the second measure of education and the first measure of debt. Then Table 5a looks at the first measure of education and then the second measure and so on. The first column represents the regression of education on risky debts without any controls or mediating variables. The second column represents the regression of education in degree of theoretical knowledge on risky debts including control variables age and gender. The third column illustrates the relation between the mediating variable, financial literacy and the outcome variable, measure of risky debt. The fourth column includes both the measure of education and the suggested mediating variable namely financial literacy. The measure of risk aversion in included in the fifth column as another control variable to financial literacy.
Table 3: Education and Financial Literacy Financial Literacy (1) (2) (3) (4) (5) (6) Education (cat) Category 2 .565*** .540*** .469*** (0.091) (.093) (.101) Category 3 .362*** .354*** .326*** (.096) (.098) (.107) Category 4 .947*** .903*** .790*** (.100) (.103) (.113) Education (years) .073*** .067*** .060*** (.012) (.012) (.013) Age - .001 -.004 .000 -.004 (.002) (.003) (.002) (.002) Female - .278*** -.321*** -.286*** - .336*** (.067) (.078) (.068) (.077) Risk Aversion -.016 -.024 (.077) (.078) Constant 3.658*** 4.103*** 4.466*** 2.974*** 3.462*** 3.942*** (0.076) (.196) (.252) (.184) (.275) (.311) N 1145 1145 833 1145 1145 833 R2 0.0679 0.0825 0.0902 0.0327 0.0487 0.0601 F-Test 32.15*** 24.09*** 15.76*** 38.26*** 19.95*** 13.35*** * Significance at a 10% confidence level, ** Significance at a 5% confidence level, *** Significance at a 1% confidence level
Table 3 observes the association between the two measures of the independent
variable (education) and financial literacy. It can be observed in column (1) that there exists a significant association significant at the 1% level in all four categories of education. The reference category of education is category 1 of education. Moving from category 1 to category 2, the financial literacy score increases with .565 unit point. The third category also finds an increase in financial literacy score of .362 points but slightly lower than category 2. The fourth category finds the highest financial literacy score: .947 units higher than the reference group, category 1. The coefficients remain significant and similar when control variables age and gender are added in column (2). Thus, it can be concluded that there is a significant association between the degree of theoretical education and financial literacy, therefore, hypothesis 5.a is confirmed. Adding the controls in column (2) show that gender also affects the financial literacy score. Females have a financial literacy score that is .278
unit point lower than males significant at 1% level. In column (3) the extra control of risk aversion is included but shows no significant results. The coefficient for the degree of theoretical education and gender remains significant at the 1% level in column (3).
Column (4) continues with the analysis of the relation between the second measure of education, education in years, and financial literacy. We find a small but significant in column (5). A year of additional education increases the financial literacy score with 0.067 unit points. Gender is also significant: females have a lower financial literacy score with .286 unit points. Both education in years and gender are significant at the 1% level. In column (6) risk aversion is added as a control variable but appears to be very small and insignificant. Education in years and gender remain significant at the 1% level. Therefore, we confirm hypothesis 5.b and conclude that education measured in years and financial literacy are significantly positively related.
Table 3 provides empirical evidence to conclude a significant relation between both measures of education and financial literacy. The coefficients for both measures of education remain significant when adding controls. Additionally, gender is also significantly related to financial literacy: females seem to have lower financial literacy scores than males.
Table 4a illustrates the first regressions results of the independent and the outcome variable. In this regression, education is measured as the degree of theoretical knowledge and risky debts are measured as a dummy variable. In the first column, a regression evaluating the effect of education on the measure of debt in dummy without controls (Y1) is observed. Only category 3 displays significant results, however these do not last when the control variables are added in column (2). Therefore, hypothesis 1.a is rejected because there is no significant relationship between and education and debt when measured as a dummy variable. Gender also seems to have a small but significant association (-.006, p<0.01) with debt when
measured as a dummy variable. Column (3) illustrates the relation between financial literacy and debt measured as a dummy variable. The effect is very small and not significant (-.011), therefore hypothesis 6.a is rejected. Column (4) illustrates the association between education and risky debt including financial literacy. In column (5), risk aversion as a control variable is added, the association of financial literacy and risky debt becomes even smaller and remains insignificant. Risk aversion also has no significant association with risky debt measured as a dummy variable. If hypothesis 7.a were to be confirmed, a significant effect in column 2 that becomes weaker in column (5) should be observed. This regression displays the opposite: the effects for group 1 and 4 only become stronger when moving from column 2 to 5. Therefore, hypothesis 7.a is rejected.
Table 4a: Education (cat) and Risky Debt (𝑌1)
Risky Debt (Y1)
(1) (2) (3) (4) (5) Education (cat) Category 2 .007 - .025 - .020 - .050 (0.033) (.034) (.034) (.038) Category 3 .082** .035 .038 - .028 (.033) (.034) (.034) (.038) Category 4 - .030 - .052 - .044 - .099*** (.042) (.041) (0.42) (.046) Age - .006*** - .006*** - .006*** (.001) (0.001) (.001) Female -.018 - .020 - .047 (.025) (.026) (.029) Financial Literacy -.011 - .010 -.004 (.012) (.012) (.014) Risk Aversion - .017 (.018) Constant .220*** .589*** .291*** .627*** .636 (.024) (.070) (.049) (.084) (.102) N 1145 1145 1145 1145 833 R2 0.0679 0.0541 0.0009 0.0547 0.0403 F-Test 32.15*** 14.27*** 0.95 12.04*** 6.08*** * Significance at a 10% confidence level, ** Significance at a 5% confidence level, *** Significance at a 1% confidence level
Table 4b illustrates the relation between education measured in years and risky debt measured as a dummy variable. In the first column there is no observation of a significant association between education measured in years and risky debt measured as a dummy variable. However, when adding the control variables age and gender in column (2), the association between education and risky debt becomes small but significant (-0.010, p<0.05). Therefore hypothesis 1.b is confirmed. The negative effect implies that an additional year of education reduces having a risky debt with 1%. Age is also significant (-0.007, p<0.01) at the 1% level. Being female translates into a reduction of having a risky debt with 0.7%. In column (3) financial literacy is added but is insignificant. In the fourth column, risk aversion is added which should account for some unobserved characteristics affecting both literacy and debt. Nevertheless, financial literacy remains insignificant. When looking at the
coefficient for education moving from column (2) to (4), a stronger association (.010 to -.011) is observed both significant at the 5% level. If financial literacy would act as a mediating variable, some of the effect would be captured by financial literacy and the association of education and risky debt would move closer to zero when moving from
column (2) to column (4). Nevertheless, this is not the case and therefore we reject hypothesis 7.b.
Table 4b: Education (in years) and Risky Debt (Y1)
Risky Debt (Y1)
(1) (2) (3) (4) Education in years - .003 -.010** - .009** -.011** (.005) (.005) (.005) (.005) Age -0.007*** -.007*** -.006*** (.001) (.001) (.001) Female -.022 -.024 - .050* (.025) (.026) (.029) Financial Literacy - .009 -.006 (.012) (.013) Risk Aversion -.025 (.031) Constant .290 .767 .798 .789 (.070) (.101) (.108) (.124) N 1145 1145 1145 833 R2 0.0004 0.0532 0.0537 0.0405 F-Test 0,43 24.17*** 18.31*** 8.49***
* Significance at a 10% confidence level, ** Significance at a 5% confidence level, *** Significance at a 1% confidence level
Table 5a illustrates the association between education measures as degree of theoretical education and debt measured in amount. Observing the first column, only significant values are found for the third category of education. Nevertheless, when controls are added in column (2), all of the values of education are insignificant. Therefore, hypothesis 2.a is rejected. Column (3) demonstrates the relation between financial literacy and risky debt measured in amounts. The value is negative (-.22) but not significant. Therefore, hypothesis 6.b is rejected and it is concluded that there is no significant association between financial literacy and risky debt measured in amounts. The values for education move closer to zero when moving from column (2) to column (5). This is in line with expectations as outlined in hypothesis 8.a. The intuition behind this is that financial literacy captures some of the effect
of education on risky debts and therefore the effects of education become weaker in column (5). However, no values are significant in column (2) and therefore hypothesis 8.a cannot be confirmed although Table 5a provides suggestive evidence.
Table 5a: Education (cat) and Risky Debt (Y2)
Risky Debt (Y2)
(1) (2) (3) (4) (5) Education (cat) Category 2 -.19 - .46 - .25 -.17 (.37) (.40) (.49) (.64) Category 3 1.18 ** .84 .97 .15 (.58) (.60) (.599) (.62) Category 4 2.43 2.18 2.57 1.78 (1.72) (1.74) (2.13) (2.46) Age - .044 - .044** -.048 (.022) (.023) (.029) Female - .69 -.79 -.99** (.58) (.50) (.487) Financial Literacy -.22 -.38 -.80 (.53) (.61) (.94) Risk Aversion -.77 (.76) Constant 1.04*** 4.49*** 2.55 6.054** 8.96** (.31) (1.23) (2.26) (2.99) (5.25) N 1145 1145 1145 1145 833 R2 0.0089 0.0140 0.0007 0.0158 0.0222 F-Test 2.92* 3.40*** 0.17 2.62** 2.07**
Note: coefficient in risky debt in 1000, * Significance at a 10% confidence level, ** Significance at a 5%
confidence level, *** Significance at a 1% confidence level
Moving to the second measure of education, education in years, and the relation with risky debt measured in amounts please refer to Table 5b. In the first column, a significant association between education in years and risky debt in amount (.21, p<0.01) Nevertheless, in column (2) control variables are added and significant effects are no longer observed. There is no significant relation between education in years and risky debt measured in amount therefore hypothesis 2.b is rejected. Age is significant in throughout columns (2), (3) and (4) meaning that every year in age translates into a reduction in debt of 43 euros. This association is significant at the 5% level. Gender also plays a significant role: females have a
debt that is 930 euros lower than males (-.93, p<0.05). Moving from column (2) to column (4), a weaker association between education in years and risky debt is observed (.15 to .13). This could suggest that financial literacy captures some of this effect. However, none of these associations are significant and therefore hypothesis 8.b is rejected.
Table 5b: Education (in years) and Risky Debt (Y2)
Risky Debt (Y2)
(1) (2) (3) (4) Education in years .21*** .15 .18 .13 (.12) (.12) (.15) (.18) Age -.041** -.04*** -.043** (.019) (.019) (.025) Female -.64 -.73 -.93** (.57) (.51) (.50) Financial Literacy -.34 -.77 (.56) (.88) Risk Aversion -.88 (.89) Constant -1.37 3582.44* 6.83** (1.66) (2102.60) (3.29) N 1145 1145 1145 833 R2 0.0036 0.0080 0.0095 0.0190 F-Test 2.78* 4.26*** 3.18** 2.66**
Note: coefficient in risky debt in 1000, * Significance at a 10% confidence level, ** Significance at a 5%
confidence level, *** Significance at a 1% confidence level
Table 6a regresses education as a measure of theoretical education on intensive margin debt (Y3). From the table, no significant variables are observed in column (1). In column (2) the control variables of age and gender are added. Again, in column (2) no significant value are found for the relation between education and risky debt which is why hypothesis 3.a is rejected. A negative coefficient is found for being female implying females have smaller debt given they have a debt (-.76). However, this value is not significant. Age is found to have a positive correlation meaning that older people have higher debts conditional on having a debt at all. However, both age and gender are not significant, so no conclusions can be drawn from these values. Additionally, there is no significant coefficient in column (3) meaning that, by itself, financial literacy does not have a significant effect on debt measured as intensive debt. Consequently, hypothesis 6.c is rejected. An increase in the coefficient of education is noted when moving from column (2) to column (5) in all four categories of education. If the
mediator were to capture some of the effect, these coefficients should have decreased, but we observe them to increase meaning that financial literacy does not capture any of the effects of education on intensive that and can it can therefore be concluded that financial literacy does not provide any mediating effect in this regression. The suggested mediating variable financial literacy does not have any significant effect in this regression, so hypothesis 9.a is rejected.
Table 6a: Education (cat) and Risky Debt (Y3)
Risky Debt (Y3)
(1) (2) (3) (4) (5) Education (cat) Category 2 -1.70 -1.64 -.067 5.11 (2.46) (2.47) (3.49) (5.32) Category 3 3.80 4.84 6.03 6.56 (3.13) (3.87) (3.94) (5.14) Category 4 15.45 15.73 17.93 19.05 (10.82) (10.77) (12.53) (14.01) Age .096 .084 -.197 (.17) (.18) (.27) Female -.76 -1.97 -6.95** (3.38) (3.06) (3.85) Financial Literacy -1523.0 -2.46 -9.98 (3349.9) (3.60) (6.96) Risk Aversion -1.03 (4.02) Constant 8.32*** 4.66 1.74 15.72 68.04 (2.10) (8.81) (14.40) (19.86) (41.59) N 170 170 170 170 98 R2 0.0535 0.0575 0.0058 0.0710 0.2178 F-Test 2.17* 1.99* 0.21 1.63 0.95
Note: coefficient in risky debt in 1000, * Significance at a 10% confidence level, ** Significance at a 5% confidence
level, *** Significance at a 1% confidence level
Table 6b, explores the second measure of education, education in years, and intensive debt. Column (1) shows a positive association meaning that education in years increases the debt given there is a debt at all. Moving to column (2), controls are added, and a large positive coefficient for education in years is noted (1.25). The value is not significant and therefore it cannot be concluded that there is
