Over-indebtedness within the Netherlands:
The role of self-control and optimism bias
Author: Denise Oenema1 Supervisor: Dr. C. Laureti
Master’s Thesis Economics (EBM877A20) University of Groningen
February 2020
Abstract:
The aim of this paper is to investigate how self-control and optimism bias may affect the likelihood of over-indebtedness. Data collected in 2010 provides rich information on Dutch households. The problem of over-indebtedness has many dimensions, which results in a wide
variety of operational definitions. This paper uses an objective and a subjective measure of over-indebtedness. Probit analysis has not led to robust significant findings regarding the relation between over-indebtedness and self-control and optimism bias, respectively. Rather
education, employment status and homeownership are shown to be prevailing factors.
JEL Classification: D91, D12, E21
Keywords: over-indebtedness, self-control, optimism bias
2
1. Introduction
Excessive borrowing may lead to over-indebtedness. This concerns situations in which households are no longer capable of meeting their financial obligations, including household bills and such. While borrowing is not undesirable, excessive borrowing is. At different levels, over-indebtedness affects society. At micro-level over-indebtedness has been negatively related to stress (Dunn and Mirzaie, 2016) and psychological well-being (Bridges and Disney, 2010). At meso-level employers are confronted with high costs having employees with financial problems (Schonewille and Van der Schors, 2017), as over-indebtedness decreases labour productivity (Haas, 2006). Moreover, the macroeconomy is negatively affected through, amongst others, decreased spending (Kukk, 2019) and lower economic growth (Law and Singh, 2014).
Increasing knowledge about the causes of over-indebtedness helps policymakers to take action in tackling debt problems of households. Intervention programmes could be more targeted and effective if more is known about the risk factors.
Traditionally, economists hypothesize that individuals aim to smooth consumption over their lifetime (Modigliani and Brumberg, 1954). This provides a rationale for the fact that individuals choose to (save and) borrow. Herein it is assumed that individuals face no credit constraints, act rational and have perfect foresight. Hence, the trade-off between current and future consumption leads to a welfare-increasing optimal outcome. The state of over-indebtedness obviously is not the optimal outcome. Betti et al. (2007) theorize that, assuming rational behaviour, over-indebtedness can only originate from unexpected life events. This explanation is not fully satisfactory, which is why behavioural economists introduce irrationality into the life-cycle model.
3
desire for immediate reward leads to self-control problems, as the individual has a hard time deferring the reward. Shefrin and Thaler (1988) explain these conflicting preferences by the inner conflict between the ‘farsighted planner’ and the ‘myopic doer’. The temptation to spend comes from the doer that is only concerned with the present. The farsighted planner wants the doer to stick to the optimal plan. This conflict presents the self-control problem. Adding these mechanisms explain the suboptimal outcomes of traditional life-cycle theories. Those who lack self-control show impulsive behaviour leading to choices ‘made on the basis of a temporary, and often sudden, change in preference’ (Kirby and Herrnstein, 1995, p. 83). Self-control problems thus lead to a tendency to buy an item now instead of later, even when this is only possible by taking on debt. Severe self-control problems could lead to too much debt, increasing the likelihood of over-indebtedness (Gathergood, 2012). Bear in mind that future costs are discounted at a lower rate as well, so future interest payments are mentally undervalued.
Besides self-control problems, individuals suffer from optimal bias. Optimal bias is the tendency to overestimate the chance of favourable events to happen and underestimate the chance of negative events to happen (Weinstein, 1980). This irrationality exists because of the lack of imperfect foresight.
Optimism bias causes individuals to expect future income to increase more than it will. The expectation of higher future incomes will increase the current consumption level (Friedman, 1957). Consumption levels may be raised above current income levels, as life-cycle theory predicts, and hence debt accumulation follows from future income increases. Debt may become unsustainable when it turns out that income does not increase by the amount expected beforehand. Therefore, optimism bias may explain part of the phenomenon of over-indebtedness.
4
Self-control (McCarthy, 2011; Gathergood, 2012) and optimism (Hyytinen and Putkuri, 2018) are suggested to have a strong influence on over-indebtedness. Both McCarthy (2011) and Gathergood (2012) measure self-control with a statement concerning self-reported impulsive buying. Impulsive individuals were more likely to report having difficulties meeting credit commitments.
Empirical evidence finds a nonlinear relation between optimism bias and financial decision-making as well (Puri and Robinson, 2007). A modest amount of optimism is beneficial to financial decision-making, while extreme optimism is not (Puri and Robinson, 2007). Hyytinen and Putkuri (2018) find that those who are too optimistic about their future financial situation can be subdivided into prudent optimists and non-prudent optimists. The latter are confronted with a deteriorating financial situation, while the first experience either no difference or improvement. Non-prudent optimists face an increased likelihood of perceived over-indebtedness.
I will study the relation between self-control optimism and over-indebtedness in the Netherlands. Over-indebtedness is a growing problem within the Netherlands and household debt comprises 102 percent of GDP (BIS, 2019). This is salient compared to the average within the EU, which is 58 percent. Most of the debt consists of mortgages, which are accompanied by fiscal benefits. Thereby, the Dutch accumulate a considerable amount of assets as well due to compulsory pension savings. Even with these mitigating remarks in mind, studying the problem of over-indebtedness has not become less relevant for the Netherlands. According to NIBUD, an independent Dutch organisation that provides budgeting education, estimates that 20% of Dutch households face over-indebtedness (Schonewille and Crijnen, 2019).
The research question is as follows: To what extent do self-control and optimism affect the likelihood of over-indebtedness within the Netherlands?
Data for this research comes from the Longitudinal Internet Studies for the Social sciences (LISS, hereafter). The panel provided information on demographic and socio-economic characteristics, planned and actual expenditures, and expectations of the future.
5
descriptive statistics will be provided. Subsequently, methodology will be described. Finally, this paper concludes with a discussion.
2. Literature review
In this section the definition and measurement of over-indebtedness will be discussed. Literature on self-control, optimism and other drivers of over-indebtedness will be connected to over-indebtedness and the section will round up with the hypotheses.
2.1 Definition and measure of over-indebtedness
Defining at what point indebtedness turns into over-indebtedness appears to be difficult. A growing number of studies engaged in over-indebtedness, yet a general operational definition is missing (European Commission, 2008; OECD, 2015; Hyytinen and Putkuri, 2018; Ferretti and Vandone, 2019). Therefore, most prevalent measures used in the literature will be discussed.
Most relevant in the search for a definition of over-indebtedness is the report Towards a common operational European definition of over-indebtedness (European Commission, 2008). The research was motivated by the need for a common operational definition in order to make cross-country comparisons in the mapping of over-indebtedness. All available definitions from EU countries are compared and lead to six in most definitions recurring elements (see Appendix A2). Research into the over-indebtedness of European households (Civic Consulting, 2013, p. 21) used these elements to arrive at the following definition of over-indebtedness:
“Households are considered over-indebted if they are having – on an on-going basis – difficulties meeting (or are falling behind with) their commitments, whether these relate to servicing secured or unsecured borrowing or to payment of rent, utility or other household bills.”
6
standard of living expenses are already minimised, the household is over-indebted. The definition excludes informal commitments.
The above-mentioned definition for over-indebtedness will be adopted in this paper.
The given definition is lacking measurable components. Empirical research asks for operational definitions; hence, a more pragmatic approach is needed. The literature distinguishes objective and subjective measures of over-indebtedness (Niemi-Kiesiläinen, 2009; Keese, 2012). The objective measures of over-indebtedness employ data on arrears, debt settlement and ratios that include debt, debt service payments, income and/or assets. The subjective measures of over-indebtedness are based on the self-reported degree of financial hardship and levels of arrears (Betti et al., 2007). Both kinds of measures will be discussed below with reference to the thorough investigation of the European Commission (2008).
In an attempt to find one universal measure of over-indebtedness, the European Commission (2008) systematically investigated all indicators used at that time within the European Union. Indicators were classified into four groups of measurement according to the statistics needed in measurement. Those groups made use of data containing information about, respectively, (i) arrears, (ii) debt settlement, (iii) assessment by households of their financial burden, and (iv) other indicators. I will discuss each of these groups below, and for the sake of argument, save the third group for last. The third group consists of subjective measures, while the other comprise objective measures.
7
The second group indicates over-indebtedness using data on court-arranged solutions to debt, people assisted with repayment plans by debt advice agencies or administrative bodies, and debt write-offs by creditors (number/values) (text literally from table 7; European Commission, 2008). This measurement type is often used in register-based studies. A strength of these kinds of measurements is that they are based on actual behaviour (Posthumus et al., 2019). However, not every household turns to debt assistance. In the Netherlands, approximately 17% of households have been registered with debt counselling (Schonewille and Crijnen, 2019). So, while this measure may indicate difficulties with meeting commitments, it fails to capture the group that avoid or cannot find the way to help.
The last group has no common denominator and involves statistics/data concerning users of credit advice agencies, borrowing to income ratios of households, borrowing to income ratios of households calculated from national accounts, Credit Service to disposable income (also called household debt-service burden). The European Commission mentions ‘borrowing to income’ to stress the fact that arrears are not considered in this measure. The OECD (2015) uses such ratios covering information on debt, assets and income to assess over-indebtedness within OECD countries. In its report In it together: Why less inequality benefits all a household is considered over-indebted when the to-asset ratio exceeds 75%. Alternatively, the debt-to-income ratio labels households over-indebted when the debt is three times as large as gross yearly income. Both measures amounted to similar conclusions regarding over-indebtedness within OECD countries. Drawing conclusions from these ratios is equivocal, as the thresholds set seem arbitrary. Betti et al. (2007) argue that the threshold may differ across households, since the optimal level of indebtedness is different for each household and each life stage.
8
income, debt or subsistence level (Keese, 2009). Wealthy households would, for example, be considered over-indebted by using the debt-service-to-income ratio, while they could sell assets in order to repay debt and bring down debt service payments. This issue has been addressed by D’Alessio and Iezzi (2013).
To account for the asset side of a household’s balance sheet, D’Alessio and Iezzi (2013) propose a measure in which measurement of debt burden contains (return on) assets, (payment on) debts and income altogether. D’Alessio and Iezzi (2013) assume that households will sell assets to pay off debt which will reduce debt burden. They construct a debt burden indicator which comprises debt burden after subtracting assets that could be sold. Ferretti and Vandone (2019) encounter that these indicators fail to cover financial commitments related to household bills.
So far, objective measures have been discussed. And, while convenient, the pragmatic approach has drawbacks. Over-indebtedness is such a complex problem, that every objective measure used until now always fails to consider at least one of many aspects involved. D’Alessio and Iezzi (2013) find that classifying a household over-indebted depends to a large extent on the choice of measurement. Measures that have the advantage to consider assets as well fail to implement arrears on household bills, for example. Hence, another kind of measure often used is the subjective one.
Subjective measures are the third group of measurements discussed by the European Commission (2008). These indicators are derived from questions asking people to give facts about their financial situation and self-reported levels of arrears (text literally from table 7; European Commission, 2008). Gathergood (2012) indicates over-indebtedness in the UK by self-reported payment arrears of one or three months on one or more credit items.
9
Not a single objective measurement used in the European Union can in itself indicate over-indebtedness. Empirical papers often apply several measures of over-indebtedness to strengthen the analysis and compensate for disadvantages of certain measures. Ratios mentioned in the study of the European Commission (2008) overlooked either debt service, assets or income. Having dealt with this issue, even proposing a new debt-burden indicator, D’Alessio and Iezzi (2013) conclude that the debt-poverty indicator serves best to identify over-indebtedness. Hence, this measure, as applied by Keese (2009), will be used in this paper. This measure considers households over-indebted when income after subtraction of debt (service) payments is reduced below the poverty line. As it is common to use more measures to better investigate the problem, in line with past studies I will employ the self-reported measure of over-indebtedness (Gathergood, 2012; McCarthy, 2011).
2.2 Non-behavioural drivers of over-indebtedness
Over-indebtedness is, in broad terms, a consequence of overactive participation in the debt market or may result from the experience of life events that affect income and/or expenditure levels considerably (Vandone, 2009). Demographic and socioeconomic characteristics have often been found to relate to the occurrence of over-indebtedness (Anderloni and Vandone, 2008). Empirical evidence on age, gender, marital status, number of dependent children, education, income, employment status, home ownership and financial literacy will be discussed below. Moreover, (unexpected) changes within the household or within the economy can cause over-indebtedness (Gutiérrerz-Nieto et al., 2017). Finally, the supply side of the debt market plays a role as well. Disney et al. (2008) mention credit restraints and Gutiérrerz-Nieto et al. (2017) highlights the role of financial institutions’ lending behaviour and their implementation of collection.
2.2.1 Demographic characteristics
10
Pauls, 2019; Oksanen et al., 2015). Interestingly, McCarthy (2011) reports that female more often feel over-indebted than male. Over-indebtedness seems to be more likely for singles and divorced (Oksanen et al., 2015). Presence of children has been associated with a higher likelihood of facing arrears (Madern, 2015; Jian Xiao and Yao, 2013), and the number of children relates positively to over-indebtedness (Oksanen et al., 2015). Considering the household size, higher debt-service-ratios are found for households with more members (Haq et al., 2018; Keese, 2012). Characteristics combined, Betti et al. (2007) find single-adult households with children particularly vulnerable. Jian Xiao and Young (2013) have investigated the likelihood of arrears for different lifecycle categories. With cross-sectional data on American families, multiple logistic regressions are run to predict that young parents of children above the age of six years are most likely to face arrears.
Regarding education, Oksanen (2015) find a higher likelihood of over-indebtedness for primary educated compared to upper-secondary educated and individuals with a degree. Madern (2015) finds a higher likelihood of arrears for low compared to high educated.
Low incomes have been associated with debt problems (Madern, 2015). The OECD (2015) finds over-indebtedness more prevalent for middle-income households. Tenants more often face arrears than homeowners (Madern, 2015).
Moreover, Keese (2009) investigates the role of life events in the likelihood of over-indebtedness. Childbirth, number of children and unemployment appeared significantly positively related. For the Dutch case, Schonewille and Crijnen (2019) report more severe payment issues among those who recently divorced, lost their job or got disabled. In addition, Madern (2015) finds that an income shock enlarges the likelihood of over-indebtedness. External adverse shocks are caused by unexpected macroeconomic changes. The economic crisis of 2007 painfully revealed the vulnerability of indebted households. Rising interest rates resulted in households’ inability to pay mortgage rent. Consequently, many houses had to be sold, and house prices dropped. The economic downturn led to decreasing asset worth and job loss. These events are hazardous for households with relatively high liabilities (Disney et al., 2008).
11
sufficient income or assets to repay and service these debts. In addition, the collection procedure applied could be as such that households getting into debts deeper. Aristei and Gallo (2016) even point out that weak payment measures of lending institutions could lead to more arrears.
2.2.2 Financial literacy
In the decision-making process of debt, financial literacy is of great importance. Financial literacy has been defined as ‘people’s ability to process economic information and make informed decisions about financial planning, wealth accumulation, debt and pensions’ (Lusardi and Mitchell, 2014).
Poor financial literacy has been found to be a driver of over-indebtedness in both qualitative (Disney et al., 2008) and quantitative research (Norvilitis et al., 2006; Lusardi and Tufano, 2015; Mitchell and Lusardi, 2011). Due to poor financial literacy, households fail to correctly estimate the costs of borrowing and are incapable of selecting the optimal credit products (Lusardi and Tufano, 2015).
With respect to debt decisions, Gathergood (2012) points out that some concepts tested in financial literacy surveys are not relevant. Instead of using the common concepts (interest compounding, inflation and risk diversification), Gathergood (2012) implements three other questions. These questions require application of interest compounding on debt size, calculation of time it takes for debt size to double and calculation of the length of the period of repayment. Obviously, an individual who is not able to answer (most of) these questions correctly, will not be able to make informed decisions about debt.
12
By introducing a two-stage-decision-making concept, Goedde-Menke et al. (2017) describe how basic economic skills could affect decisions in the first stage. At this stage, an individual decides whether to take on debt. Here, financial literacy is of minor relevance, rather knowledge of basic economic concepts appears useful. Obviously, the notion that higher consumption levels limit choices in future consumption levels (scarcity) etc. helps an individual to consider whether future income will be sufficient if currently taking debt against future income. After having decided to take a debt, an individual must decide what kind of debt to take. This is the second stage in which financial literacy does play an important role.
2.3 Behavioural drivers of over-indebtedness
Besides non-behavioural drivers, behavioural drivers may explain part of the problem of over-indebtedness. Here, financial attitudes, self-control and optimism bias will be discussed.
2.3.1 Financial attitudes
Financial attitude matters in debt behaviour and hence, affects the likelihood of over-indebtedness. Keese (2012) argues that this likelihood is considerably higher for those who experience little internal barriers taking on debts. Almenberg et al. (2018) relate feeling uncomfortable with debt to lower debt levels. Their results even suggest that these attitudes are passed on from generation to generation (also suggested by Lea et al., 1995). Unfortunately, the dataset used in Almenberg et al. (2018) lacks information about the indebtedness of past generations, so it could not be confirmed.
2.3.2 Self-control and over-indebtedness
13
Self-controlled individuals manage to stir behaviour as such that rationally optimal outcomes are achieved. Lack of self-control is reflected in impulsive behaviour, which is ‘understood as behavior that is not regulated and that results from an unplanned, spontaneous impulse’ (Baumeister, 2002, p.670). When it comes to spending behaviour, the purchase may be inconsistent with an individuals’ term plans, i.e. short-term preferences conflict with long-term preferences (Baumeister, 2002).
Theory explains this conflict of preferences (Strotz, 1955) with the notion of hyperbolic discounting. Traditional theories have always assumed that individuals exponentially discount future rewards and costs, resulting in time-consistent preferences. Strotz (1955) questions this and proposes that individuals discount future rewards at higher rates as time distance enlarges, hyperbolic discounting. He presents a model in which an individual fails to stick to his optimal plan due to disobedience of his future self. Reconsideration of the optimal plan at any future point in time will lead to other choices, as the individual values present consumption over future consumption. Utility from present consumption constantly being higher than delayed consumption will lead an individual to trade future consumption for present consumption, i.e. dissave or borrow. Present bias is explained by hyperbolic discounting and leads through lack of self-control to inconsistency between planned and actual behaviour.
14
This leads to hypothesizing that impulsive individuals are more at risk of over-indebtedness. Using the extent to which an individual is prone to impulsive buying behaviour as a measure for self-control, resulted in finding a negative relationship between the level of self-control and over-indebtedness (McCarthy, 2011; Gathergood, 2012; Ottaviani and Vandone, 2018). The first hypothesis is therefore as follows:
Hypothesis I: Self-control problems increase the likelihood of over-indebtedness.
2.3.3 Optimism bias and over-indebtedness
Within the field of psychology, Weinstein (1980) found how individuals may be unrealistically optimistic. A group of students was asked to estimate the chances to experience several negative and positive life-events. They were asked to fill out to what extent the chances for them were higher or lower than average. Weinstein (1980) concluded that students overestimated the chance of experiencing positive events and underestimated the chance of experiencing negative events. The tendency of individuals to underestimate the risk of negative life events, while overestimating the risk of positive life events is referred to as optimism bias. As this bias appeared useful in explaining economic decision-making, optimism bias has been investigated within the field of economics as well.
Individuals’ consumption level depends on the expectations on lifetime income. The permanent income hypothesis states that expectations on future income immediately affect current consumption, but only if this change in income is permanent (Friedman, 1957). From this it is reasonable to assume that optimism bias leads to unsustainable levels of consumption, as actual income will be lower than expected income. Hence, expectation errors may have severe consequences for the household’s financial position.
15
expectations of future income affect borrowing behaviour. Next to the experimental study, a survey-based analysis led to finding a positive relationship between optimism bias and reported over-indebtedness. This is in line with Hyytinen and Putkuri (2018) who also find that extreme optimists more often report feeling over-indebted. From this the second hypothesis is the following:
Hypothesis II: Optimism bias increases the likelihood of over-indebtedness.
2.2.4 Research objective and hypotheses
Much evidence has been documented on the role of self-control in the problem of over-indebtedness. Past studies have used the Barrat Impulsiveness scale (Ottaviani and Vandone, 2018) or a statement concerning impulsive buying (Gathergood, 2012; McCarthy, 2011). Contrast to these methods, I will use the difference between expected change in expenditure pattern (relative to past) and actual change in expenditure pattern (relative to past). I have chosen for this measure, as self-control problems have often been described as a deviation of actual from planned behaviour.
In addition to using another measure for self-control, I add optimism bias. A relatively underexposed driver of over-indebtedness. The variable is constructed in the same way as Hyytinen and Putkuri (2018) have done. This paper differs from theirs in also taking into account the impact of self-control, which have been proven non-negligible (Gathergood, 2012; McCarthy 2011).
16
3. Data
This section provides an overview of the source of data and the sampling method which lead to a longitudinal dataset. Furthermore, construction and descriptive statistics of the variables will be shown.
3.1 Sample construction
In this thesis use is made of data of the Longitudinal Internet Studies for the Social sciences (LISS, hereafter) panel administered by CentERdata (Tilburg University, The Netherlands). The LISS panel is a representative sample of Dutch individuals who participate in monthly Internet surveys. The panel is based on a true probability sample of households drawn from the population register. Households that could not otherwise participate are provided with a computer and Internet connection. A longitudinal survey is fielded in the panel every year, covering a large variety of domains including work, education, income, housing, time use, political views, values and personality.2
A survey on consumer emotions and behaviour3 has been quarterly distributed within the LISS panel. This survey provides insights on household’s current financial situation and household’s expectations about their future financial situation. The data used in this study has been collected in September 2009 and September 2010 in order to compare expectations with actual outcomes. In 2009, the response rate was 59.4% (2,257 individuals). In 2010, the response rate was 72.5% (2,706 individuals). Data on financial capability4, collected in August 2010, serves to construct a variable for self-control. This survey has been filled out by 5,451 individuals, corresponding to a response rate of 71.6%. To calculate the residual income, gross income after subtracting debt payments, questions from another survey5 are merged with the former. This survey on debt payments has been presented to households in September 2010. The response rate was 70.8% (5,337 individuals). Finally, the database keeps track of socio-demographic and socio-economic information monthly. I use the information collected in August 2010 as this corresponds to the survey presented first (about financial capability).
2 https://www.lissdata.nl/sites/default/files/afbeeldingen/Reference_LISS_3.0.pdf 3 dataarchive.lissdata.nl; survey 30. Tilburg Outlook Consumer Monitor
4 dataarchive.lissdata.nl; survey 52. Are Effective Emotion Regulation Strategies Associated with Financial Capability
17
The above response rates would lead to expect that the sample size is large. However, on household expectation’s only 883 observations are present. This is due to the merge of two waves of the concerning survey to calculate the difference between expected and actual household’s financial situation. Another key variable in this research, self-control, leads after merging to a loss of 87 observations. As the information on monthly debt payments is highly important, I only keep observations if the household members have participated in all three surveys, which leads to a sample with 781 observations.
As Haq et al. (2018) has done in the wake of earlier research, I only kept heads of households within the sample. Thus, variables like age and education level reflect information of the head of the household6. Before I dropped the heads, I checked if there were cases where a non-head filled out the monthly debt payments of the household and the non-head of the household did not. To avoid unnecessary missing data, I replaced the missing monthly debt payments by the amounts filled out by another household member. In this case missing data could be explained by the observation that approximately one third of the household heads are not the one within the household taking care of household finances. The same has been done for monthly mortgage. Moreover, surveys on income and housing have been used to complement missing information on mortgage and debt payments. If a household head reported to have paid less than 100 euros to service debt in 2010 and monthly debt data was missing (4 observations), the missing value has been replaced by zero. For there is only one household that would be identified over-indebted by implementing the upper limit (which is 100 euros, so 8,35 euros a month), this step can be carried out safely. By taking this step I assume that the question on monthly debt, which is asked at the individual level, covers the total amount a household spends monthly on debt service payments. There were 44 household heads who responded to this question in the survey on income, while having missing data for monthly debt payments.
A loan can be taken out only from the age of 18, therefore those individuals should be excluded. However, after dropping 349 non-heads of households, the age of the youngest individual in the sample is 19. This is step one in construction of the sample, which now consists of 432
18
Dutch household heads. The regression sample differs due to missing variables, which will be addressed below.
3.2 Missing variables
Due to missing variables, not all observations will be included in the analysis. In this section, the way of handling missing variables will be discussed. An overview of all missing variables is given in table 1.
The construction of over-indebtedness will be described in the next section. Here, the variables needed for construction; monthly mortgage payments, monthly debt payments and gross household income, will be discussed regarding missing variables. For monthly mortgage payments and monthly debt payments an answer possibility was ‘I really don’t know’. Recall, that it concerns only households in which both the head and the partner report lack of knowledge. These answers are regarded as missing. For monthly mortgage payments 34 households answered ‘I really don’t know’ and 55 households did not answer, resulting in 89 missing data for this variable. For monthly debt payments 25 households answered ‘I really don’t know’ and 11 households did not answer, resulting in 36 missing data. Non-response on the debt question fully coincides with non-response on the mortgage question. Furthermore, 21 of those who did not know the answer to the mortgage question also lacked knowledge on monthly debt payments. For 35 observations gross household income was missing. The missing data of these variables led to a reduction of 110 observations for the construction of over-indebtedness regarding subsistence levels.
19
Pigott (2001) suggests that complete case analysis is an option in handling missing data if the data are missing completely at random (MCAR). To find out if the data is missing completely at random, I ran the MCAR test (Little, 1988) on the missing variables in table 1. The null-hypothesis for this test is that the missing variables are MCAR. Based on the MCAR test I assume that the missing variables are MCAR (Prob > chi-square = 0.07, hence I cannot reject the null-hypothesis at the 5%-level). A disadvantage of dropping observations instead of keeping them and estimating the missing values is that it provides estimated values less close to the observed values (Pigott, 2001). However, based on the MCAR test and the fact that other methods do not necessarily avoid bias, I choose to exclude observations with missing data from the regressions. After dropping 19% of the observations, the sample counts 307 observations.
Table 1. Pattern of missing variables
1 2 3 4 5 6 7 8 9 10 % of cases 1 1 1 1 1 1 1 1 1 1 71 1 1 0 1 1 1 1 1 1 1 11 1 1 0 0 1 1 1 1 1 1 5 0 0 1 1 1 1 1 1 0 1 3 0 0 0 0 1 1 1 1 0 1 2 1 1 1 1 0 1 1 1 1 1 1 # missing 35 (8.1%) # missing 35 (8.1%) # missing 89 (20.6%) # missing 36 (8.3%) # missing 13 (3.0%) # missing 7 (1.6%) # missing 2 (0.46%) # missing 3 (0.69% # missing 33 (7.6%) # missing 2 (0.46%)
Note 1: Patterns representing less than 1% of the cases are not reported in the table.
Note 2: Variables are (1) gross household income, (2) net household income, (3) monthly mortgage payments, (4) monthly debt payments, (5) household’s expected financial position, (6) household’s actual financial position, (7) education, (8) lag of self-reported health, (9) lag of net household income, (10) lag of employment status.
3.3 Construction of key variables
In this section, I will first describe the construction of all variables. After the construction of each variable has been understood, the descriptive statistics will be discussed.
3.3.1 Over-indebtedness
20
be compared with the minimum subsistence level within the Netherlands (see Appendix, table A3). Although these amounts are self-reported and do not originate from register-based data, I consider this to be close to objective, as it is derived from objective questions not concerning feelings (also noted by Disney et al., 2008). If residual income is lower than the minimum subsistence level, the household will be considered over-indebted. The relatively objective measure of over-indebtedness is binary. It is equal to zero if the household is not over-indebted and equal to one if the household is over-indebted.
For reasons stated earlier, a subjective measure will be investigated as well. Ideally, the same question as Gathergood (2012) and McCarthy (2011) would be used here. However, the number of observed cases is too low for this question (see rejected subjective measure, A.4). Therefore, individuals are asked how well they can make ends meet with their current income. Individuals could answer on a 5-point-scale, in which 1 is ‘very difficult’ and 5 is ‘very easily’. This question still is justified as a subjective measure, since individuals with the same income, same assets and same amount of debts can have the feeling that it is hard to make ends meet. This difference could stem from living standard, of which some individuals may feel cannot be lowered, while others lower their living standard to more easily make ends meet.
3.3.2 Self-control
21
Table 2. Distribution of debt service payments and self-control problems (SCP)
The dummy for self-control equals 1 if individuals responded (4) to (6), and zero otherwise. The threshold may seem somewhat arbitrary. Sensitivity analysis will pay attention to this matter. McCarthy (2011) deals with a four-point-scale and categorizes “tend to agree” as being impulsive, i.e. having self-control problems. In the LISS survey the answer (4) is “agree somewhat”, which is comparable to McCarthy (2011). Hence, these individuals are indicated as having self-control problems (SCP equals one).
To learn about the debt behaviour, table 2 shows summary statistics of gross household income and monthly debt service payments for households with and without self-control problems. From this table it is remarkable that the households without self-control problems pay on average a higher amount to service debt. However, these households have on average a higher
Without SCP
All DSP > 0
Mean
(Std. Dev.) Min Max
Mean
(Std. Dev.) Min Max
Total monthly mortgage and debt service payments (DSP) 410.38 (496.97) 0 4570 634.22 (489.60) 10 4570 Gross household income 3376.93 (2045.83) 0 15800 4250.20 (2102.06) 976.10 15800 Observations 272 176 With SCP All DSP > 0 Mean
(Std. Dev.) Min Max
Mean
(Std. Dev.) Min Max
22
gross household income as well, so no preliminary conclusions can be drawn yet. Regarding over-indebtedness, those descriptive statistics will be covered in the next section.
3.3.3 Optimism bias
For operational purposes, optimism bias is defined as ‘the difference between a person’s expectation and the outcome that follows’ (Sharot, 2011, p. 942). Puri and Robinson (2007) used this type of measurement in their research into the relationship between optimism and economic choice. The difference between an individual’s life expectancy and actual life expectancy according to statistics has been used to measure optimism. The measure has been validated by correlating with expectations about the economy, about income growth and psychological tests of optimism.
Following Hyytinen and Putkuri (2018) optimism bias is measured by taking the difference between an individual’s expected financial situation and actual financial situation.
𝐹𝑜𝑟𝑒𝑐𝑎𝑠𝑡 𝑒𝑟𝑟𝑜𝑟 = 𝐴𝑖,𝑡− 𝐸𝑖,𝑡|𝑡−1 (2)
where the first component measures actual financial situation, from which the expected financial situation (given 12 months in advance) will be subtracted. Hyytinen and Putkuri (2018) construct a categorical variable named FE1, in which those who have a forecast error equal to or lower than -2 are labelled ‘clearly pessimistic’. If the forecast error equals -1, the household is classified ‘moderately pessimistic’. The outcome is zero for those who make no forecast error, 1 for those who are ‘moderately optimistic’ and equal to or greater than 2 for those who are ‘clearly optimistic’. I adopt their method in which the forecast error variable, FE1, represents optimism bias in the regression.
23
pessimists report a better financial position than expected, these two findings are surprising. This points in the direction that optimists are not labelled optimists due to experiencing health shocks or income drops more often than other groups.
Table 3. Distribution of forecast errors, income changes and health shocks
Obs. % of full sample Mean income
change (Std. Dev.) Health shock (% of FE1- category) Clearly pessimistic 16 5.21 -325.75 (688.59) 12.50 Moderately pessimistic 44 14.33 23.29 (463.16) 4.55 No forecast error 157 51.14 59.66 (306.78) 7.01 Moderately optimistic 79 25.73 48.80 (550.90) 7.59 Clearly optimistic 11 3.58 187.82 (481.92) 0.00 3.3.4 Control variables
The following dummies are created: age, female, partner, children, high education, employed, unemployed, inactive, retired, homeowner, debt outstanding and health. Central Bureau of Statistics (CBS) categories are used to construct age. As only six observations starting from age 19 exist for the youngest category (15-24), the first and second age category are combined into one (19-34). The other categories can be found in table A1 (Appendix). Each of these categories will be included in regression as a dummy variable. High education implies college (not junior college) or university. There were 12 cases of individuals indicating (partial) work disability. These are included in the dummy unemployed. In the construction of debt outstanding I assume that a household with zero monthly mortgage and debt payments does not have debts.
Health needs some clarification as well. The survey on expectations and actual outcomes also includes a question on health satisfaction7 as well. This question has been asked in 2009 as well as in 2010. If health deteriorated and lead to dissatisfaction in 2010, I assume the individual has experienced a health shock during the preceding 12 months. This might not be a completely pure measure, as the health shock could have occurred shortly before responding to the survey. And hence the observation need not differ that much from someone who does not report a health shock. However, this is the best measure, to my knowledge, available from LISS provided that as much data as possible is retained.
24
3.4 Descriptive statistics
The sample consists of 307 heads of Dutch households. Table 4 (p.25) gives an overview of the descriptive statistics. Table 5 (p.25) shows how over-indebtedness differs across the two measures of over-indebtedness for those with and without self-control problems and for pessimists and optimists.
Demographic characteristics of the household show that most household’s head are aged 55 or higher. Remarkable is that for over-indebted households the fraction of youngest and oldest heads is considerably smaller and higher respectively. For non-over-indebted households the distribution across age is comparable. This also applies to the shares of female household heads. Over-indebted households have a lower share of household heads with partner than non-over-indebted households. The proportion of households with children is almost three times higher for non-over-indebted households relative to over-indebted households.
Socio-economic characteristics show that the mean net income for over-indebted households is much lower than for non-over-indebted households. A two-sample t-test indicates that this difference is statistical significant. Furthermore, heads of non-over-indebted household are more often higher educated and also more often employed than heads of over-indebted households.
25
difficulties making ends meet. This could be explained by the finding of Keese (2012) mentioned earlier, who found that subjective measures of over-indebtedness seem to reflect expectations of one’s future financial situations as well.
Table 4. Descriptive statistics
Full sample Over-indebted* Non-over-indebted*
Mean (Std. Dev.) Mean (Std. Dev.) Mean (Std. Dev.)
Age (years)
19-34 0.10 (0.30) 0.04 (0.21) 0.11 (0.31)
35-44 0.16 (0.37) 0.15 (0.36) 0.16 (0.37)
45-54 0.18 (0.39) 0.09 (0.28) 0.20 (0.40)
55-64 0.27 (0.44) 0.28 (0.46) 0.26 (0.44)
65 years and older 0.29 (0.45) 0.43 (0.50) 0.26 (0.44)
Female 0.27 (0.45) 0.43 (0.50) 0.25 (0.43) Partner 0.64 (0.48) 0.43 (0.50) 0.68 (0.47) Children 0.22 (0.42) 0.09 (0.28) 0.25 (0.43) High education 0.36 (0.48) 0.13 (0.34) 0.39 (0.49) Net household income 2595.27 (1258.60) 1202.35 (421.45) 2840.76 (1195.75)
Lagged net household
income** 2259.10 (1239.66) 1244.02 (409.29) 2790.89 (1191.53) Employment status Employed 0.56 (0.50) 0.24 (0.43) 0.61 (0.49) Unemployed 0.05 (0.22) 0.15 (0.36) 0.03 (0.17) Inactive 0.06 (0.23) 0.22 (0.42) 0.03 (0.16) Retired 0.34 (0.47) 0.39 (0.49) 0.33 (0.47) Homeowner 0.69 (0.46) 0.39 (0.49) 0.75 (0.44) SCP 0.11 (0.32) 0.13 (0.34) 0.11 (0.31) FE1 3.08 (0.87) 3.24 (0.92) 3.05 (0.85) Observations 307 46 261
*(Non-)over-indebted according to objective measure (residual income < subsistence level), **For two observations lagged net household income is missing.
Table 5. Summary statistics on dependent and main independent variables
Full sample Without self-control problems
With self-control
problems Pessimists* Optimists*
OIobj. 46 (15.0%) 40 (14.7%) 6 (17.1%) 7 (11.7%) 19 (21.1%)
OIsub. 52 (16.9%) 39 (14.3%) 13 (37.1%) 17 (28.3%) 13 (14.4%)
Obs. 307 272 35 60 90
26
4. Methodology
The aim of this paper is to identify if and how self-control and optimism bias relate to the likelihood of over-indebtedness. The derivation of the measures has been discussed in the data section, here the methodology is covered.
In general, I regress both measures of over-indebtedness on self-control and optimism bias separately. After examination of both separately, I investigate how the result changes if both are included in the same model. The three models are formally written as:
𝑂𝐼𝑖𝜃 = 𝛽0+ 𝛽1𝑆𝐶𝑃𝑖+ 𝛽2𝑋𝑖+ 𝜀𝑖, 𝜀𝑖 ~ 𝑁𝐼𝐷(0,1) (1)
𝑂𝐼𝑖𝜃 = 𝛽0+ 𝛽1𝐹𝐸1𝑖 + 𝛽2𝑋𝑖 + 𝜀𝑖, 𝜀𝑖 ~ 𝑁𝐼𝐷(0,1) (2)
𝑂𝐼𝑖𝜃 = 𝛽0+ 𝛽1𝑆𝐶𝑃𝑖+ 𝛽2𝐹𝐸1𝑖 + 𝛽2𝑋𝑖+ 𝜀𝑖, 𝜀𝑖 ~ 𝑁𝐼𝐷(0,1) (3)
in which the dependent variable, 𝑂𝐼𝑖𝜃 represents the binary variable for over-indebtedness equal to one if the household is over-indebted and equal to zero if not over-indebted. The superscript stands for either the objective (obj.) or the subjective (sub.) measure. In all three equations 𝑋𝑖 represents a basic set of control variables8 for every household (i).
Since the measures of over-indebtedness are both binary, ordinary least squares (OLS) regression would provide estimates of the probability that the dependent dummy variable will be equal to 1, i.e. that the household’s living standard is below subsistence level. The disadvantage of choosing for a linear probability model (LPM) is that the estimated coefficients are not restricted to the interval [0,1] (Wooldridge, 2012). Therefore, regarding binary dependent variables, it is more common to choose for either Probit or Logit models that are estimated using maximum likelihood estimation (MLE) procedures. Regarding results there is not much difference between these models (Verbeek, 2000), so I choose to follow the literature within this field (see McCarthy, 2011) in this respect. A disadvantage of MLE is that the
27
estimated coefficients cannot be directly interpreted. Therefore, marginal effects should be analysed as well in order to draw proper conclusions.
The first equation (1) will be estimated first by leaving out the vector of controls. I will regress the dummy for having self-control problems on the objective and subjective measure of over-indebtedness respectively. Then a second estimation will include the vector of basic control variables.
The second equation (2) will be estimated first by leaving out the vector of controls. I will regress the forecasting error variable on the objective and subjective measure of over-indebtedness respectively. Forecasting errors may be due to job loss, health or income shocks. Hence, a second estimation will control for these factors. The third estimation will include the basic set of control variables that are mentioned for all three models; age, female, partner children, education, employment status and homeowner. In all specifications, the household who does not make a forecast error (FE1 equals 3) serve as the reference group.
To estimate how self-control and optimism bias jointly affect over-indebtedness, equation (3) will be estimated including the basic set of controls.
5. Results
In this section the results of the estimated models will be discussed. To interpret the coefficients, average marginal effects are calculated. The tables shown in this section contain average marginal effects, regression results can be found in the Appendix (table A5).
5.1 Self-control and over-indebtedness
28
percentage point higher probability of facing over-indebtedness. Moreover, being unemployed or inactive increases the likelihood of over-indebtedness as compared to employed by 28.7 and 40.6 percentage points respectively. Education and homeownership significantly decrease the likelihood of over-indebtedness.
The subjective measure of over-indebtedness does find significant evidence for the relation between self-control and the likelihood of over-indebtedness. Without controlling for other household characteristics, self-control problems relates to a higher probability of over-indebtedness. In the specification that includes the control variables self-control still appears significant. However, the magnitude and significance level are smaller. Compared to the objective measure, the significant effect of age disappeared.
29
Table 6. Average marginal effects from separate Probit regressions model for SCP
(1) (2) (3) (4)
(OIobj.) (OIobj.) (OIsub.) (OIsub.)
SCP 0.0234 -0.0202 0.180*** 0.0969* (0.0622) (0.0552) (0.0555) (0.0505) Age 35-44a 0.117* 0.0123 (0.0699) (0.0839) Age 45-54a 0.0114 0.0299 (0.0508) (0.0844) Age 55-64a 0.0630 -0.0797 (0.0512) (0.0786)
Age 65 and oldera 0.184** -0.0459
(0.0924) (0.109) Female 0.0414 0.0367 (0.0443) (0.0456) Partner -0.00862 0.00829 (0.0459) (0.0501) Children -0.0364 0.0563 (0.0623) (0.0563) Education -0.111** -0.0786* (0.0456) (0.0446) Unemployeda 0.287** 0.281** (0.125) (0.132) Inactivea 0.406*** 0.0336 (0.121) (0.0918) Retireda -0.0286 -0.00778 (0.0689) (0.0867) Homeowner -0.102*** -0.201*** (0.0386) (0.0390) Observations 307 307 307 307
30
Table 7. Average marginal effects from separate Probit regressions model for FE1
(1) (2) (3) (4) (5) (6)
(OIobj.) (OIobj.) (OIobj.) (OIsub.) (OIsub.) (OIsub.) Clearly pessimistsa 0.0601 0.0524 0.0339 0.422*** 0.350*** 0.284** (0.101) (0.108) (0.0898) (0.127) (0.135) (0.122) Moderately pessimistsa -0.0365 -0.0315 -0.0268 0.0417 0.0348 -0.0120 (0.0509) (0.0525) (0.0468) (0.0644) (0.0639) (0.0532) Moderately optimistsa 0.0878* 0.0774 0.0709 0.0118 0.00342 -0.0215 (0.0533) (0.0524) (0.0469) (0.0490) (0.0495) (0.0444) Clearly optimistsa 0.0544 0.0823 0.0450 -0.0492 -0.0461 -0.154*** (0.119) (0.125) (0.115) (0.0910) (0.0950) (0.0346) Job loss 0.132 0.0421 0.0282 -0.0644 (0.123) (0.113) (0.141) (0.144)
Negative income shock -0.0552 -0.0371 0.0785 0.0465
(0.0614) (0.0561) (0.0524) (0.0470)
Negative health shock 0.144** 0.0567 0.103 -0.0162
(0.0670) (0.0600) (0.0735) (0.0695) Age 35-44a 0.119* 0.0249 (0.0686) (0.0821) Age 45-54a 0.00188 0.0535 (0.0478) (0.0815) Age 55-64a 0.0626 -0.0685 (0.0495) (0.0770)
Age 65 and oldera 0.193* -0.0401
(0.100) (0.110) Female 0.0482 0.0400 (0.0440) (0.0449) Partner 0.00163 0.0178 (0.0460) (0.0497) Children -0.0405 0.0699 (0.0624) (0.0553) Education -0.106** -0.0800* (0.0445) (0.0444) Unemployeda 0.240** 0.377*** (0.121) (0.137) Inactivea 0.378*** 0.0175 (0.125) (0.0858) Retireda -0.0433 0.0159 (0.0727) (0.0919) Homeowner -0.0947** -0.218*** (0.0395) (0.0383) Observations 307 307 307 307 307 307
31
5.2 Optimism bias and over-indebtedness
Optimism bias is expected to be negatively related to the likelihood of over-indebtedness. This effect has not been confirmed by the average marginal effects (see table 7) of the Probit regression for the objective measure of over-indebtedness. Only the first specification returns a weakly significant positive relation between being moderately optimistic and facing over-indebtedness. These individuals, without controlling for other household characteristics, have a 8.8 percentage point higher probability than those households who do not make a forecast error. When controlling for job loss and negative income and health shocks, the significant effect disappears and health turns out significant. Those who have experienced a health shock are 14.4 percentage points more likely to have a residual income below subsistence level than those who did not experience a health shock. Including the whole set of control variables results in the same conclusions as for the model in which self-control has been tested. In this case, having a household head in the upper age category, education, employment status and homeownership are the only explanatory variables that remain significant.
Forecasting errors have much more explanatory power when regressed on the subjective measure. Across all three specifications, being clearly pessimistic has been found significant. However, the sign is unexpected. Compared to those who do not make forecasting errors, clearly pessimists have a significantly higher probability of facing over-indebtedness, i.e. reporting that they have difficulties making ends meet. The specification with the full set of control variables also shows that clearly optimists even have a significantly lower likelihood of facing over-indebtedness. This could again, as written in the section on descriptive statistics, be attributed to the possibility of optimists by their personality are less likely to feel over-indebted.
32
5.3 Self-control, optimism bias and over-indebtedness
The results for combining the self-control and forecasting error into the same model are given in table 8. The first specification, which includes the basic set of controls, does not return significant results for the objective measure of over-indebtedness. Compared to the reference group of making no forecast error, for the subjective measure the clearly pessimists have a 27.8 percentage point higher probability and the clearly optimistists have 15.3 percentage points lower probability of reporting over-indebtedness. As with the former models, high education and homeownership are significantly negatively related to being able to make ends meet.
The pseudo-R2 values can be interpreted by comparing models and see how this value changes. Most interesting is the change from a model which includes all control variables with a model that adds the value of interest. This would show the added value in explaining the dependent variable. For both variables, SCP and FE1, this change is small.
5.4 Robustness checks
To check for robustness, I have changed the threshold of having self-control problems. In this case individuals who answer to the statement (see Appendix, table A1) ‘agree somewhat’ are no longer considered over-indebted. Results (see Appendix, table A8 and A9) do not change for the objective measure. For the subjective measure, the effect of self-control is no longer significant. For the subjective measure, the results are thus sensitive to the set threshold level at which the head of the household is considered having self-control problems.
Regarding optimism bias, I ran the same regressions using a measure on a household’s expectations about the economic situation in the Netherlands (see Appendix, table A10 and A11). This way, external shocks are the same for every household. Using this measure has not resulted in any significant relation between forecasting error and the likelihood of over-indebtedness. The significant effect of being clearly pessimist on subjective over-indebtedness disappears when using this measure for optimism bias. Hence, the findings are not robust.
33
Table 8. Average marginal effects from joint estimation for SCP and FE1
(1) (2)
(OIobj.) (OIsub.)
Clearly pessimistsa 0.0179 0.278** (0.0798) (0.116) Moderately pessimistsa -0.0297 -0.000980 (0.0462) (0.0541) Moderately optimistsa 0.0732 -0.0228 (0.0470) (0.0433) Clearly optimistsa 0.0217 -0.153*** (0.108) (0.0335) SCP -0.0214 0.0827* (0.0545) (0.0496) Age 35-44a 0.119* 0.0528 (0.0695) (0.0783) Age 45-54a 0.00556 0.0868 (0.0497) (0.0789) Age 55-64a 0.0600 -0.0463 (0.0510) (0.0701)
Age 65 and oldera 0.190** -0.0132
(0.0946) (0.0983) Female 0.0497 0.0381 (0.0439) (0.0444) Partner 0.00334 0.0166 (0.0461) (0.0495) Children -0.0380 0.0616 (0.0625) (0.0555) Education -0.108** -0.0783* (0.0451) (0.0438) Unemployeda 0.278** 0.332** (0.125) (0.135) Inactivea 0.403*** 0.0151 (0.122) (0.0823) Retireda -0.0358 0.00554 (0.0676) (0.0850) Homeowner -0.0992** -0.213*** (0.0393) (0.0379) Observations 307 307
34
individuals with self-control problems. Also significant at 1%-level is the finding that the probability of this expression of over-indebtedness increases for female and households with children. The significant effect of education no longer is relevant here. Even the effect of homeownership disappeared.
5.5 Discussion
I find no robust significant evidence for the effect of self-control on over-indebtedness. This result is not in line with McCarthy (2011) and Gathergood (2012). One measure coincides with McCarthy (2011), which is the alternative measure on over-indebtedness for which I did find a significant effect of self-control. The other measures for over-indebtedness are different from the ones used in this paper, which may explain for this different result. In addition, the difference could possibly be explained by either the size of the sample in this paper, which is rather small as already mentioned before. Furthermore, McCarthy (2011) uses a different dataset consisting of UK households. Perhaps the effect of self-control on the likelihood of over-indebtedness could be explained by the legal structure of a country, the culture of a country or other country-specific factors.
For optimism bias, a relationship between the subjective measure and being clearly pessimistic has been found. This is in line with Keese (2012). However, this contradicts Hyytinen and Putkuri (2018) who find that optimism bias increases the likelihood of perceived over-indebtedness and has been related to higher ratios of debt with respect to income. I do not find this result, however, I did not control for macroeconomic factors, which they did. So, the difference may lie in the choice of variables. Besides, Hyytinen and Putkuri (2018) study Finnish households, while I have studied Dutch households. The same reasons as mentioned before regarding different datasets may apply to this case.
The findings appear to differ across different measures which is in line with the conclusion of D’Alessio and Iezzi (2013).
35
6. Conclusion
This study has attempted to show how self-control and optimism bias relate to over-indebtedness. In this section I will first discuss the limitations of the study, then the main conclusions and policy recommendations follow.
Limitations of the dataset are first of all that it comprises only 307 households. Moreover, the households are not followed over time, as some surveys were single-wave studies. The advantage of panel data would be that the severity of over-indebtedness could be taken into account by observing to what extent situations of over-indebtedness persist.
The size of the dataset also lead to excluding financial literacy as an explanatory variable, since this would even further decrease the amount of observations. Also the amount of assets, which would include another survey, leads to a decrease of observations. Since every added survey leads to an even smaller sample, I have chosen to keep the most important surveys and disregard levels of financial literacy and amounts of assets. I acknowledge that this has implications which will be discussed in the results section.
Moreover, in this field, missing data can affect the result. A considerable share of individuals have difficulties with openness about money issues or is unaware of the financial position of the household. Both could indicate problems of over-indebtedness as lack of planning is associated with it (McCarthy, 2011). Feelings of shame could hinder a household to reveal its financial position (D’Alessio and Iezzi, 2013).
As Katona (1975) states that future expectations build on past expectations, a lag of past expectations or a lag of past forecasting errors would be of interest. However, the dataset lacks this information as the survey only started as from September 2009.
36
additional explanatory variables containing partner characteristics. However, this led to many missing variables, as not all partners join the panel.
Conclusions apply to this study only and cannot be generalized as the sample size was small. This study showed that the level of self-control does not significantly add to the likelihood of over-indebtedness, rather demographic and socio-economic variables are of importance. Clearly pessimists more often feel that it is difficult to make ends meet, however, this result did not hold for the frequency they run out of money.
According to this study, self-control and optimism bias do not need specific attention in household debt policies. Recall that other studies do and the results of this study may not apply to all Dutch households. Hence, I would suggest, based on other studies, that behavioural factors receive attention in household debt policies. Furthermore, more research is needed on why unemployed households have a higher likelihood of becoming over-indebted. The results of such studies could lead to suggestions for government in how to avoid over-indebted situations for these households. Many reasons for this relations are possible. Not having a financial buffer could be one of the reasons or not being able to quickly align expenditures with the drop in income. The design of social security may also play a role in this relation. However, from this study, no conclusions can be drawn in this respect.
Education policy could address the role of having received higher education. The curriculum of individuals that follow the path towards university may contain useful elements regarding financial capability that should be included in curricula for lower education as well.
37
References
Almenberg, J., Lusardi, A., Säve-Söderbergh, J., & Vestman, R. (2018). Attitudes toward debt and debt behavior. Working Paper Series, 24935.
Anderloni, L., & Vandone, D. (2008). Households over-indebtedness in the economic literature. Universit’a degli Studi di Milano, Working Paper, 46, 775.
Aristei, D., & Gallo, M. (2016). The determinants of households’ repayment difficulties on mortgage loans: Evidence from Italian microdata: Households’ repayment difficulties on mortgages. International Journal of Consumer Studies, 40(4), 453-465. doi:10.1111/ijcs.12 271
Bank for International Settlements (BIS) (2019, December 8). Credit to the non-financial sector. Retrieved from: https://stats.bis.org/statx/srs/table/f3.1
Baumeister, R. (2002). Yielding to temptation: Self‐control failure, impulsive purchasing, and consumer behavior. Journal of Consumer Research, 28(4), 670-676. doi:10.1086/338209 Betti, G., Dourmashkin, N., Rossi, M., & Yin, Y. (2007). Consumer over-indebtedness in the
EU: Measurement and characteristics. Journal of Economic Studies, 34(2), 136-156.
Bridges, S., & Disney, R. (2010). Debt and depression. Journal of Health Economics, 29(3), 388-403. doi:10.1016/j.jhealeco.2010.02.003
Civic Consulting (2013). The over-indebtedness of European households: Updated mapping of the situation, nature and causes, effects and initiatives for alleviating its impact (Brussels).
D’Alessio, G., & Iezzi, S. (2013). Household over-indebtedness: definition and measurement with Italian data. Banca d’Italia, Occasional Paper, no. 149.
Disney, R., Bridges, S., & Gathergood, J. (2008). Drivers of Over-indebtedness. Report to the UK Department for Business.
Dunn, L., & Mirzaie, I. (2016). Consumer debt stress, changes in household debt, and the great recession. Economic Inquiry, 54(1), 201-214. doi:10.1111/ecin.12218
European Commission (2008) Towards a common operational European definition of overindebtedness (Brussels).
Ferretti, F., & Vandone, D. (2019). Personal debt in Europe : The EU financial market and consumer insolvency. Cambridge, United Kingdom: Cambridge University Press. (2019). Friedman, M. (1957). The permanent income hypothesis. In A theory of the consumption
function (pp. 20-37). Princeton University Press.
38
Goedde-Menke, M., Erner, C., & Oberste, M. (2017). Towards more sustainable debt attitudes and behaviors: The importance of basic economic skills. Journal of Business Economics, 87(5), 645-668. doi:10.1007/s11573-017-0854-8
Grohmann, A. & Menkhoff, L. & Merkle, C. & Schmacker, R. (2019). Earn More Tomorrow: Overconfident Income Expectations and Consumer Indebtedness. Rationality and Competit ion Discussion Paper Series, 152, CRC TRR 190 Rationality and Competition.
Gutiérrez-Nieto, B., Serrano-Cinca, C., & Cuesta-González, M. (2017). A multivariate study of over-indebtedness' causes and consequences. International Journal of Consumer Studies , 41(2), 188-198. doi:10.1111/ijcs.12324
Haas, Oliver J., (2006), Overindebtedness in Germany, ILO Working Papers, International Labour Organization. Retrieved from: https://EconPapers.repec.org/RePEc:ilo:ilowps:9939 13223402676.
Haq, W., Ismail, N., Satar, N., & Watson, D. (2018). Household debt in different age cohorts: A multilevel study. Cogent Economics & Finance, 6(1). doi:10.1080/23322039.2018.14554 06
Hyytinen, A., & Putkuri, H. (2018). Household optimism and overborrowing. Journal of Money, Credit and Banking, 50(1), 55-76. doi:10.1111/jmcb.12453
Jian Xiao, J., & Yao, R. (2014). Consumer debt delinquency by family lifecycle categories. International Journal of Bank Marketing, 32(1), 43-59. doi:10.1108/IJBM-02-2013-0007 Katona, G. (1975). Psychological economics. Elsevier.
Keese, M., (2009). Triggers and Determinants of Severe Household Indebtedness in Germany. SOEPpaper, 239. Retrieved from: https://ssrn.com/abstract=1518373 or http://dx.doi.org/10 .2139/ssrn.1518373
Kirby, K. N., & Herrnstein, R. J. (1995). Preference reversals due to myopic discounting of de layed reward. Psychological science, 6(2), 83-89.
Kukk, M. (2019). Debt repayment problems: Short-term and long-term implications for spending. Review of Economics of the Household, 17(2), 715-740. doi:10.1007/s11150-018 -9424-2
Laibson, D. (1997). Golden eggs and hyperbolic discounting. The Quarterly Journal of Economics, 112(2), 443-478.
Law, S., & Singh, N. (2014). Does too much finance harm economic growth? Journal of Banking and Finance, 41(1), 36-44. doi:10.1016/j.jbankfin.2013.12.020
Lea, S. E., Webley, P., & Walker, C. M. (1995). Psychological factors in consumer debt: Money management, economic socialization, and credit use. Journal of Economic Psychology, 16(4), 681-701.
