MASTER THESIS ECONOMICS (EBM877A20)
THE EFFECT OF PSYCHOLOGICAL BIASES ON DUTCH HOUSEHOLD SAVINGS
Author: A.A.B. (Bert) Aalberts1 Student number: 2572672 Supervisor: Dr. P. (Pim) Heijnen Date: 08.06.2018
Organization: University of Groningen
Abstract:
Saving too little can have negative consequences, such as being inadequately protected against adverse shocks and having suboptimal retirement consumption. The life-cycle framework is the hallmark of studying intertemporal choice, but it has some undesirable elements when it is used to explain saving behavior (Thaler, 1994). I investigate whether psychological variables affect savings of Dutch households. For that purpose, I make use of the Dutch Household Survey (DHS). Survey-based psychological variables along with household characteristics are employed to explain savings. I find that self-perceived financial literacy and peer effects positively influence savings. Making financial education compulsory in high school curriculums may lead to the desired increase in financial literacy and, ultimately, savings. Furthermore, the formation of saving clubs might increase savings via the peer effects that are associated with it.
JEL-codes: D14, D15, D87, D91, H55
1 Contact: a.a.b.aalberts@student.rug.nl
I would like to thank Dr. Heijnen for his valuable comments on intermediate versions of the thesis. Furthermore, I appreciate that Prof. Dr. Alessie provided me with code that facilitated the
development of the dataset used in this thesis.
2 1. INTRODUCTION
According to a NIBUD report of van der Schors and van der Werf (2017), almost one out of three Dutch households is saving less than the minimum recommendation of NIBUD2. A potential explanation for this is the existence of a widespread ignorance of the merits of saving. Van der Schors and van der Werf indicate that this line of reasoning is incorrect; a large majority of adults understands the importance of saving. The question, then, is why so many Dutch households fail to have adequate savings.
Before considering why savings may be too low, it is vital to understand why one would save. Keynes (1936) provides an exhaustive list of saving motives. Over the past decades, a vast literature quantifying the importance of those motives has emerged. Still, studies find widely varying results on the magnitude of those savings motives. For instance, Kennickell and Lusardi (2004) distinguish between three categories of studies regarding the precautionary savings motive: studies that find either no, a modest or a large effect of the precautionary savings motive on wealth accumulation. I believe that assessing psychological biases might help in mitigating the difference in outcomes among studies on savings. For instance, it might be different studies consider different types of persons. If psychological biases do affect saving, including these biases in the regressions might result in coefficients of precautionary savings becoming more similar.
Employing psychologic concepts in economic research is relatively new. For a long time, economists have assumed that people are fully rational, taking correctly into account general economic variables such as the interest rate. Using those variables, people would then be able to calculate optimal levels of consumption in each period. However, there are at least two reasons why full rationality seems to be an invalid assumption. First, it seems unlikely that persons consider all available information. Second, individuals might use the available information in an incorrect way, as the calculations that are associated with finding an optimal consumption plan are complicated. Therefore, it might be necessary to depart from assuming rationality. In contrast with traditional economic thinking, the psychologic view on human behavior is radically different, as this profession finds that human behavior usually is not so rational (Loewenstein et al., 2008). The fields of behavioral economics, and subsequently neuroeconomics, aim to incorporate psychological concepts in economic research (Loewenstein et al., 2008). By doing so, results stemming from economic research might become more accurate. Regarding saving research, Thaler (1994) is one of the researchers that points out the importance of considering psychologic elements. According to Thaler, the life cycle model fails to explain savings satisfactorily. Part of this is due to its assumption of rationality, as just discussed.
2NIBUD is a Dutch organization that studies household budgeting decisions. Furthermore, it provides
3 Notice that the savings rate of an individual might be either too high or too low compared to the optimum. Although saving too much also has its downsides, I will focus on what can be done to increase savings3. However, before policy measures can be designed that might lead to a higher savings rate, it should be investigated which factors affect peoples’ savings decisions. Besides some observable variables, such as the income of a household and the interest rate, less clear-cut factors may also affect the savings rate. Even though some consumers prefer a certain savings rate, a variety of psychological biases may prevent them from reaching this. Concluding, the objective is to investigate which biases have what effect on the savings rate.
Therefore, the research question is as follows:
What effect do psychological biases have on Dutch household savings?
In order to obtain an answer to this question, the DHS (Dutch Household Surveys) surveys of CentERdata (Tilburg University) will be employed. Psychological variables are derived from answers to survey questions. Using those variables jointly with non-psychological variables enables to observe whether psychological variables affect savings or not.
This thesis will proceed as follows. Section 2 contains an overview of the relevant literature. Section 3 will take a closer look at Dutch savings and the Dutch pension system. Section 4 explains the dataset that will be employed. The econometric model follows in Section 5. Consequently, in Section 6, the results of the regressions are given. Section 7 will point at some of the limitations that this thesis has. In Section 8, potential solutions arising from the results will be discussed. Section 9 concludes.
2. LITERATURE REVIEW
2.1 Life cycle model
This literature review commences with a loose description of the life cycle model that was developed by Modigliani and Brumberg (1954). The life-cycle model is important, as it is a main tool for explaining consumption and, therefore, savings decisions. Although this section does not cover all elements of the model, the main insight is similar. In the finite life-cycle model, a rational, utility-maximizing consumer faces the decision on how much to consume in each period. This consumption is usually financed by income. Furthermore, it is assumed that consumers have a concave utility function. Concave utility implies that utility in each period increases in consumption, albeit at a decreasing rate. Consumers derive
4 utility from consumption in each period, but utility in later periods is discounted, due to impatience of consumers. In optimum, consumers smooth the marginal utility of consumption over the life-cycle. Individuals will thus choose a consumption level that smooths marginal utility of consumption in all periods, while the other part of their income constitutes savings in all periods. These savings generate interest. Under the assumption that the real interest rate equals the discount rate, consumption would be constant over the life-cycle. This prediction does not fully correspond with observed behavior, however. One of the major distinctions between the model and reality is that consumers often experience a drop in their consumption level at retirement. A potential explanation for this is that people have not saved enough during their life, as Hamermesh (1984) mentions.
The latter observation leads to the question why people are saving too little. Perhaps, their working-life income is just inadequate so that they cannot afford to save enough. Alternatively, persons might struggle to save due to psychological biases. The former explanation may be especially relevant for developing countries. This thesis centers around one of the most prospering countries of the world, where individuals have access to all kinds of financial institutions. Therefore, I will concentrate on psychological biases.
2.2 Traditional explanations for saving
Despite intuitiveness being a strong feature of the life-cycle model, the predictions that the life-cycle model makes do not always correspond with actual saving behavior. This is a first indication that assessing psychological biases might assist in studying saving decisions. To disentangle the effect of psychological biases on saving, it is necessary to understand which other factors influence saving first. The factors that affect saving are closely related to the reasons why people save. Keynes (1936) poses eight reasons why one would want to save, of which some will be discussed below. The motive names, which are underlined, are borrowed from Browning and Lusardi (1996), although most motive names are also mentioned by Keynes.
5 their expected lifetime income when they start working, and accordingly make decisions on consumption and savings. This expected lifetime income consists for a large share of wages, but especially in developed countries, various institutions can increase lifetime income considerably. Social Security benefits and occupational pensions are two examples of such institutions. When one expects an increasing wage profile, one may consume more now than if one expects a decreasing wage profile. If the permanent income hypothesis would not hold, this would likely be different. In that case, consumption might be the same for a given level of current income, regardless of wage profiles. Notice, though, that wages may be rather invariant over the working life, such that the wage profile is relatively flat. Still, wages fall to zero after retirement. Pensions partly fill this void by providing at least some income after retirement. However, individuals might want to maintain or even increase their consumption level after retirement. To do so, they may have to save some of their income during working life. Therefore, if people perceive that their pension income is low, I would thus expect higher savings during the working life. The third reason relates to the interest that accrues on deposits, the intertemporal substitution motive. One might expect that a higher interest rate will lead to higher savings, but due to opposing substitution- and income effects, this need not be the case. The fourth reason relates to the assumption that individuals want to increase their consumption over the life-cycle. An implication can be that people save less when they grow older. This is the improvement motive. A fifth and rather important savings motive is the bequest motive. One might expect that only those with children will have such a motive, but that might not be true. For instance, it could well be possible that childless people want to leave money to their friends.
Apart from the reasons provided by Keynes, I would like to mention two other reasons why individuals might want to save. For instance, Browning and Lusardi (1996) mention the downpayment motive. Saving via this motive is due to a desire to acquire durable goods such as cars or houses. If people plan to purchase such a good, I would expect savings to be higher
.
Finally, Hazlitt (1959) mentions another rationale for saving: people might expect future prices to fall. If this is the case, people may postpone consumption and save more. Consider for instance the purchase decision on durable goods like electronics. Some individuals may want to wait for a sale, rather than buying the good at the regular price. These persons will only buy the good when it is on sale. If people know when such a sale takes place, such an effect will be more pronounced the closer the sale becomes.2.3 The way in which people save
6 be plausible that everyone would hold their personal savings in one or more interest-bearing bank account(s), some consumers exclusively save at home. This may be due to several reasons. Firstly, opening a savings account may be costly and time involving. The second reason why people save at home relates to the instability of the financial system. The collapse of banks that occasionally happens feeds distrust among the population and most likely increases saving at home. On the other hand, burglary and fire are reasons for saving at the bank rather than at home. Therefore, both saving at home and at the bank have their drawbacks. From a risk diversification point of view, it is perhaps desirable to engage in both forms of savings. Besides personal savings, an important part of savings happens through contractual savings, especially via pensions. Typically, employees forgo a part of their wage in exchange for a stream of pension income after retirement. The main difference between contractual and personal savings is that the former are not accessible before retirement, while the latter usually are. Having no access to savings is beneficial in the sense that people, who suffer from temptation issues, cannot spend this money. However, a disadvantage is that one cannot address financial distress before retirement by using pension fund money.
Which form of savings is prevalent might depend on the pension opportunities that are offered. Employees that work at a firm offering only a modest pension plan or no pension plan at all are likely to engage more in personal savings. On the other hand, a generous pension plan may largely suffice for employees, who would then hardly make use of personal savings. The latter might not be advisable, due to reasons given in the previous paragraph.
2.4 Why do people save too little? The role of financial literacy
Now that the reasons and possibilities to save have been provided, I resort to the question why some persons fail to save enough, or do not even save at all. Some individuals just cannot afford to save. Some people might not care about saving. Some persons might think that they will not need savings. Others might however be unaware of how saving works.
7 Some researchers aim to evaluate financial literacy by employing survey questions. For instance, Lusardi and Mitchell (2011a) evaluate knowledge of financial and numerical concepts of a group of Americans aged over 50. The authors ask, among others, some rather basic questions on topics as compound interest, inflation and risk diversification. Here, the prominence of financial illiteracy is illustrated by evaluating the results of the compound interest question. Lusardi and Mitchell (2011a) formulate the question as follows:
‘’Suppose you had $100 in a savings account and the interest rate was 2% per year. After 5 years, how much do you think you would have in the account if you left the money to grow: more than $102, exactly $102, less than $102?’’ (Lusardi and Mitchell, 2011a, p. 4)
A third of the respondents does not know the correct answer to this question (Lusardi and Mitchell, 2011a). As compound interest is vital in the realm of saving, this result might hint at why many people are observed under-saving. Lusardi and Mitchell (2011a) also distinguish between people who have at least tried to calculate how much to save for retirement, and people who did not do. The authors find that the former are more likely to answer the question correctly than the latter are. This result might indicate that the persons responsible for saving decisions are financially relative literate.
In a closely related research, Lusardi and Mitchell (2011b) ask the same questions, but then to people from multiple countries. They find that financial literacy is also low in other well-developed countries than the US, like the Netherlands and Japan. Apparently, financial illiteracy is not unique to developing countries. Furthermore, the authors state that both the youngest and oldest age groups have the lowest financial literacy. Finally, it is observed that, on average, women are less financially literate than men are.
In another research on financial literacy, Lusardi et al. (2011) investigate what determines financial literacy of young Americans. Using the aforementioned question on interest rates, they find several differences between people who give the correct answer and people who give the incorrect answer. Again, men are financially more literate than women are, but Lusardi et al. find that peer effects and family characteristics also play a role. Peer effects indicate that people who have peers in college are more financially literate than people with less peers in college. Furthermore, family characteristics reveal that both the mother’s education and parents engaging in financial transactions lead to higher financial literacy of the child. Hence, financial literacy can also be increased by other means than education.
8 literate than men are. The question why this is the case is interesting, but I feel that the implications of this result are more relevant for this thesis. Bucher-Koenen et al. (2016) pose that, within couples, men are often in charge of financial decisions. One might therefore conclude that the lack of financial knowledge of women is not too much of a problem. There are two complications, however. First, the authors put forward that women tend to live longer than their husbands do. Therefore, there is often a period in life in which women must take their own financial decisions. Second, not all women have a partner. Especially for women that remain single throughout their entire life, financial education may be particularly important. These two factors indicate that improving financial literacy among women may be important after all.
Luckily, it is possible to mitigate financial illiteracy. Bernheim and Garrett (1996) find that employer-provided training on retirement saving leads to higher savings rates in US firms. If this holds more general, the savings rate could be boosted by devoting more resources to financial education.
2.5 Psychological biases that may affect savings
The previous section indicates that financial literacy may lead to people under-saving. It also mentioned that people might either have insufficient income to save or ignore the importance of saving. Still, some people might appreciate the merits of saving, have access to and knowledge of financial instruments, but save too little. It might be that part of this result is driven by psychological biases. Therefore, it may be beneficial to embark on some psychological analysis. This section will consider which psychological biases might affect people in shaping their decision on how to divide consumption over the life-cycle. Four biases will be discussed, status quo bias, anchoring, peer effects and time-inconsistent preferences.
2.5.1 Status quo bias
9 latter happens due to the default savings rate being low. Due to anchoring (see next paragraph) and status quo bias, people fail to adjust sufficiently away from the default rate (Choi et al. 2004). Hence, the effect of automatic enrolment on overall savings strongly depends on the default savings rate that is set.
2.5.2 Anchoring
Anchoring is another psychological bias that deserves attention. When presented a given level of a choice variable, anchoring implies that people choose a value closer to that default value than if they were unaware of this level (Rabin, 1998). If people did not face a default value, their choices would likely differ. The default value of the savings rate plays thus a role in shaping the saving decision, via anchoring. On one hand, posing a default, modest savings rate may induce participation in a savings plan. On the other hand, it may also cause people to stick with that, potentially too low savings rate. In designing those saving plans, anchoring thus imposes a tradeoff between the participation rate and the rate at which people save.
2.5.3 Peer effects
Another bias that may affect savings decisions relates to peers. Duflo and Saez (2002) claim that they are the first to investigate whether peer effects on saving decisions exist. Investigating contribution to a savings plan of university employees makes clear that increases in participation or contribution among all colleagues of an employee increase the contribution of that employee as well. This may indicate that peer effects do play a role in saving decisions. It appears to be natural that peer effects hold for non-participation as well. The latter would imply that one would save little if colleagues or relatives save little.
2.5.4 Time-inconsistent preferences
10 save if left on their own. Other people will correctly predict that they will frequently postpone saving if they do not save now. Therefore, these sophisticated hyperbolic discounters will save now. These people may find it easier to save on their own.
A way in which time-inconsistent preferences may lead to under-saving is via the choice of the saving instrument. O’Donogue and Rabin (1998) find that many consumers hold significant sums of money in low-interest bank accounts. Transferring those resources to a more generous interest paying account may increase retirement wealth by a large amount. The question is why some people do not transfer the funds. Since retirement benefits become only operable over a long period, people face ample opportunity to delay the transferring of funds. Delaying this decision by one day may not be that costly, but if the individual postpones the transfer many times, it will be costly. Only sophisticated people would decide to transfer the funds today, as they anticipate that they will have self-control problems in the future. Notice that status quo bias also plays a role here.
Hyperbolic discounting is closely associated with self-control problems. Self-control problems can hinder people in maximizing their well-being. Ameriks et al. (2004) construct a measure for such self-control problems. Participants of their research hear that they will receive some certificates for free visits to a restaurant, which remain valid for two years. The expected-ideal gap (EI-gap) is the difference between the number of certificates that the participants expect to use in the first year and the number of certificates that they would prefer to use in the first year. A positive EI-gap may be attributable to self-control problems, like temptation. Ameriks et al. run a regression that explains wealth through – among others – this EI-gap. They find a negative relationship between this gap and wealth accumulation. As higher savings are somewhat of a synonym of wealth accumulation, this finding implicitly means that self-control problems decrease savings.
2.6 Neuroeconomics
Hyperbolic discounting is an example of present-biased preferences. Hyperbolic discounting runs counter to the dominant view in economics, which assumes that people discount utility in an exponential fashion (cf. Koopmans, 1960). To address this inconsistency, discussing neuroeconomics may be helpful. The field of neuroeconomics aims to bring closer together the predictions that economic models make and the actual observed behavior of individuals (Loewenstein et al., 2008). Neuroeconomics utilizes assumptions that are accepted in the psychology world, although these assumptions challenge traditional economic thinking (Loewenstein et al., 2008).
11 systems in the brain consider all decisions (Loewenstein et al., 2008). As an illustration for this, McClure et al. (2007) find that people’s limbic regions become less active if the opportunity to consume a liquid is delayed considerably. In contrast, activity in the fronto-parietal region stays more constant with varying delays. According to me, a possible explanation is that people with hyperbolic discount functions have more active or better-developed limbic systems, as those systems place emphasis on short-term gratification.
2.7 Potential solutions that may increase savings
As a final part of this literature review, it is worthwhile to discuss some instruments that might lead to an increase in savings. I distinguish between:
2.7.1 Commitment devices
The way in which people behave might not correspond with the way that they would like to behave. Fortunately, self-commitment devices may help to solve this problem (O’Donoghue and Rabin, 1999). Examples of such devices are savings plans like Christmas Clubs, which are popular despite their relatively low payout (Strotz, 1956). Thaler and Benartzi (2004) describe an attempt to increase the savings rate in the United States, which is done by usage of the SMART program. Thaler and Benartzi investigate how this program has affected savings rates among participating companies. Although people can quit the SMART program at any moment without a penalty, it might still be qualified as a commitment device since quitting may be time-consuming. Furthermore, the status quo bias works in a favorable direction here. An attractive feature of this program is that savings rate increases go along with pay raises. As such, loss-averse participants experience their extra contribution to the savings plan not necessarily as a loss in income. Comparing participants with non-participants indicates that participants of the SMART-program have much higher savings rates than non-participants.
2.7.2 Increasing the default savings rate
As was discussed above, most saving plans involve a rather modest savings rate. On one hand, having a default option is desirable, as it spurs participation. On the other hand, people tend to stick with the default option, which may be lower than the rate that they would choose for themselves. If this savings rate is relatively low, increasing the default may lead to some people opting out of the savings plan, but that is probably outweighed by the extra savings that are made.
2.7.3 Unconventional saving plans
12 rounded value and the price of the purchased good will go into a savings account. Although savings per transaction will be less than a dollar, over time, this plan could lead to substantial savings. This might appear to be a fine supplement of regular savings. I believe that two problems might come with such a plan, however. First, people might feel less guilty of spending much, as their debit card transactions also lead to savings. Consumers might even view this plan as their major savings vehicle, which could lead to a spending spree. Ultimately, one might save much less than without this savings plan. Secondly, if people use this plan, they might use it as a substitute for regular savings, potentially leading to lower aggregate savings than without the program.
2.7.4 Saving in a bank account rather than at home
Earlier in the literature review, a comparison between saving at home and saving at the bank was made. Although both have their merits, van der Schors and van der Werf (2017) provide another argument for saving at the bank rather than at home. The concept of mental accounting is relevant here. Shefrin and Thaler (1988) state that consumers have an order of preferences for spending different types of money. Consumers spend more easily out of cash on hand than cash from savings accounts, for instance. Not only because withdrawing cash from saving accounts involves an action, but also because people might feel guilty of it. As cash at home is readily available, people might perceive it less as savings money, and spend it more easily. Consequently, saving through a savings account may be more successful when one pursues higher savings.
3. THE DUTCH CASE
3.1 Dutch savings: descriptive statistics
As I focus on saving by Dutch households, it is natural to provide characteristics of saving behavior of Dutch households first. The main purpose is to provide some context. In this section, two sources are employed. First, data of DNB (De Nederlandsche Bank, The Dutch Bank) is used. Second, I employ the earlier mentioned report of van der Schors and van der Werf (2017).
13
Figure 1: Total savings of Dutch households in billion euros (DNB, 2018).
Figure 2: Outstanding savings per household in euros (DNB, 2018).
Now that it is established that savings per household increased over this period, one might wonder why this is the case. I will discuss five potential reasons. First, savings usually accumulate interest. If the in- and outflow of savings would be the same, the accrual of interest would still increase the total savings. The outflow is potentially even lower due to status quo bias: once people have deposited, they resist withdrawing money. Second, the possibility to save via the Internet might have led to higher savings. Trips to the bank might feel cumbersome, so that saving via internet may increase savings. This holds especially for those living in rural areas. A third reason relates to life expectancy. People perceive that they may live more years in retirement and thus engage in higher savings. A fourth reason is the expected decrease in the generosity of Dutch pension benefits. A combination of low fertility and increasing life
0 50 100 150 200 250 300 350 400
Outstanding savings in aggregate (bln)
0 5000 10000 15000 20000 25000 30000 35000 40000 45000 50000
14 expectancy renders the current pension system unsustainable. In order to retain the system, some sacrifice has to be made. People might expect this and increase their savings accordingly. The final reason for an increase in savings is the rise in real income. Over the past decades, the average real income has risen. Higher productivity of workers is one important reason for this.
After having considered the aggregate data, it is time to look at a more detailed level. After all, the dataset that I use consists of microdata. For retrieving the statistics on individual savings behavior, the report of van der Schors and van der Werf (2017) is being used. Some interesting facts stem from this paper. First, one out of five Dutch people does not save at all. This is rather surprising, given the previous discussion on the various reasons for saving. Apart from being surprising, this result is important. Non-savers are unlikely to smooth their consumption over time. They might even be at risk of financial distress when facing unexpected high costs. Therefore, it may be preferable for most of these people to engage in saving. A second result is less spectacular. The lower the household income, the higher the proportion of people that does not save, which makes sense.
After interpreting the results on saving behavior on both macro- and micro-level, one might wonder whether there is a problem. Does it matter that one out of five people is not saving? Is the importance of saving overstated, especially due to the rather generous pension system that is featured in the Netherlands?4 I argue that saving is still important. Saving is not only beneficial to prevent poverty, but also to smooth consumption in order to maximize utility. Furthermore, the proceeds of contractual savings become only available after retirement5. Hence, for adverse life events that precede retirement, one still needs to rely on personal savings. Examples of such events are disability and long-term unemployment.
3.2 Dutch pension system
One of the crucial goals of saving is smoothing consumption over the life-cycle. Apart from saving out of disposable income, the existence of pension systems may also help towards this goal. In fact, the contributions that working people make during their working life can be interpreted as savings, with the proceeds becoming available after retirement. There are different forms of pensions that in the economic literature are better known as pension pillars. Börsch-Supan (2004) distinguishes between the public, the occupational and the personal pillar. The author compares some European countries - including the Netherlands -, with the United States in terms of composition of retirement income. In the Netherlands, the public and occupational pillar are of similar importance, while the personal pillar is less important.
4 According to van der Schors and van der Werf (2017), savings is money that is not spent
immediately. Although it is not explicitly stated, I believe that the net rather than the gross wage is relevant here. In that case, some of the individuals that do not save out of their net wages may contribute to a pension plan out of their gross wages.
15 It is notable that, in the Netherlands, the occupational pensions form the largest share of retirement income in the Netherlands, compared to the other countries. Still, ten percent of post-retirement income is constituted of personal savings. Furthermore, the actual level of retirement income is unknown. If the first two pillars are dominant, it might still be that retirement income is too low to have the desired level of consumption during retirement. In that case, there is still all incentive to increase household savings, as the personal pillar is the only one that one can influence.
The rest of this section focuses on the Netherlands, in particular on its pension system. In the Netherlands, there is a quite broad and generous pension system. Regarding the public pillar, there is general coverage of those aged at least 66 through AOW (Algemene Ouderdomswet). AOW is a form of Social Security6. If receiving AOW does not suffice to receive an income level that is deemed socially acceptable, people can apply for AIO (Aanvullende Inkomensvoorziening Ouderen). The latter serves as supplementary income for the needy elderly. As such, even the public pillar will suffice in providing an income level that normally would meet everyone’s basic needs. The second pillar consists of occupational pensions, which provide another source of income for some of the elderly. One is eligible for occupational pensions if the employer offers such a scheme.
Due to the variety of pension schemes for the elderly, it is not surprising that Dutch poverty rates are among the lowest for the group of people aged over 65, at least between 2011 and 2014. Conversely, the group of people aged 55-64 exhibits a relatively high poverty rate during that period (NRC, 2016). The problem for those people is that they are relatively often unemployed. Due to their high age, they are less attractive for employers. This can lead to long unemployment spells, potentially lasting until retirement. Their age makes that they are not yet eligible for the pension benefits mentioned earlier, which readily explains why they often are relatively poor.
Although the presence of a pension system may assist in achieving optimal post retirement consumption, it also has a shortcoming. The presence of a pension system may have a negative effect on household savings. All in all, if some retirement income is provided by pensions, consumers may opt to consume more out of their income. This means that pension benefits may crowd out personal savings. The more people decrease their household savings in anticipation of benefits they receive after retirement, the less effective that benefits system is. To my knowledge, two studies have evaluated this. First, there is a study by Alessie et al. (1997). They find that social security crowds out private savings, whereas pension wealth has no such effect. Contrary results were found by Euwals (2000), although this study focuses at
6The age at which Dutch persons become eligible for AOW used to be 65 years, but it will increase as
16 savings motives rather than actual saving. Savings motives not necessarily explain actual savings behaviour, as Euwals points out.
The Dutch pension system seems rather generous. Not everyone can reap the full benefits of this pension system, however. A share of the Dutch population is self-employed. Although self-employed people are eligible for Social Security through the first pillar, they are not entitled to occupational pensions through the second pillar (Mastrogiocomo and Alessie, 2015). The self-employed must therefore largely rely on their own discipline to save.
4. DATA
4.1 Characteristics of the dataset
For the empirical analysis, I make use of household data. The DNB Household surveys that were executed by CentERdata will be employed. This section will highlight the main features of these surveys. In order to do so, I employ a report that was written by Teppa and Vis (2012). Teppa and Vis provide detailed information on the CentERpanel and the DNB Household Surveys. Teppa and Vis mention that the panel is in many respects representative of the Dutch population. A notable exception lies in education, as highly educated individuals are overrepresented in the panel. The surveys have been conducted on a yearly basis since 1993, where the last survey was established in 2017. That means that, in aggregate, there are 25 years – or waves - of data. In these surveys, household members of approximately 2,000 households are being asked all kinds of questions. When individuals leave the sample, CentERdata adds new households while aiming to retain the sample representative. For some households, more than one family member is filling out the surveys.
In the analysis, use is made of information from the waves 2004 up to 2016. On aggregate, that implies that thirteen waves of data will be considered.
4.2 Variables
17
Table 1: Non-psychological variables with descriptions.
Table 2: Psychological variables with descriptions.
There are nine non-psychological variables, with savings constituting the dependent variable. Six of those variables are binary dummies. Originally, selfemp, highed and livetogether were no binary dummies, but they have been adjusted to make them binary dummies.7 The psychological variables have different ranges, as can be seen in Table 3. The variables are based on survey questions. The answers to these questions are all measured on a Likert scale (Likert, 1932). For those questions, respondents ought to indicate how much the statement describes their behaviour. Higher values correspond with the statement being more characteristic of one’s behaviour. The minimum and maximum value correspond with total denial and total agreement, respectively. Notice that some of the questions are assessed on a 1-7 scale (the questions considering time preference and peer effects) whilst others use a 1-5 scale (the questions considering living according to schedules and the question considering tracking of expenditures), and the question on self-perceived financial literacy is assessed on a 1-4 scale. There are subtle differences if seven-point scales rather than five-point scales are used8. Therefore, I find it inappropriate to rescale the variables.
Regarding financial literacy, the survey does not ask explicit questions on financial concepts, contrary to some of the papers that were discussed earlier. There are some questions that consider basic numeric concepts, but I feel that those do not form a satisfactorily proxy for financial literacy. The respondents
7For missing values on completed education, I assume that the highest level of attended education is
also the highest level of completed education.
8 Matell and Jacoby (1972) indicate that a higher number of answer alternatives generally leads to less neutral answers. On the other hand, having more steps might lead to the respondent taking more time to answer the question. Therefore, there is a clear trade-off when deciding on the number of answers.
Variable Type Description
savings Numeric The total amount that a household has saved over the past 12 months, in euro's. nethhinc Numeric The net income of the household over the past 12 months, in euro's
baby Dummy Equals 1 if family has at least one child, equals 0 if not.
pensfund Dummy Equals 1 if one participates/participated in the pension fund, equals 0 if not. selfemp Dummy Equals 1 if someone is self-employed, equals 0 if not.
highed Dummy Equals 1 if someone completed HBO or University, equals 0 if not. male Dummy Equals 1 if someone is a male, equals 0 if not.
age Numeric The age of the individual, in years.
livetogether Dummy Equals 1 if one lives together with someone, equals 0 if one does not.
Variable Type Description
schedules Ordinal The extent to which one thinks he or she lives according to schedules.
timepref Ordinal The extent to which one is willing to sacrifice current utility to achieve future goals spfinlit Ordinal The self-perceived financial literacy of an individual
18 have to indicate how they feel about their knowledge of financial matters. Most individuals indicate that they are more or less knowledgeable, although a substantial share feels their knowledge is either good or bad. Interestingly, men have a higher self-perceived knowledge than women do. This result is in line with the results that Lusardi and Mitchell (2011a) found.
4.3 Hypotheses
In this section, I will provide my expectations regarding the effect of the psychological variables on savings.
4.3.1 Living to schedules
First, I expect living to schedules to have a positive effect on savings. The most obvious reason is that these persons may be more likely to stick to a plan than those who live less according to schedules. Another rationale for this hypothesis is that people who live more toward schedules may also be more likely to set up a savings plan in the first place.
4.3.2 Time preference
For the variable time preference, I expect that a higher value leads to higher savings. Perhaps, people that are more willing to sacrifice current well-being for future well-being are more likely to save than people who are less willing to do so. This is consistent with the life-cycle model of Modigliani and Brumberg (1954).
4.3.3 Self-perceived financial literacy
I expect that higher self-perceived financial literacy of the respondent leads to higher savings. The rationale behind this hypothesis is threefold. First, financially more literate individuals are more likely to understand the concept of compounded interest, as discussed by Lusardi and Mitchell (2011a). Therefore, they might find savings more important than financially less literate persons, who likely estimate lower rates of return to savings. Second, financially more literate people may know more types of savings instruments. To the extent that having more different savings instrument translates to higher overall savings, this would positively affect savings. Third, people that are less financially literate may begin saving for retirement too late. This also leads to financially more literate people saving more. 4.3.4 Peer effects
19 effects that are at least as large as under Duflo and Saez, so that expecting positive peer effects seems justified.
4.3.5 Tracking expenditures
Finally, I expect that tracking expenditures has a positive effect on savings. People that track expenditures become more insightful on their spending and might adjust accordingly. Assuming that people more likely subconsciously overspend than subconsciously underspend, tracking expenditures might thus lead to higher savings.
4.4 Descriptive statistics
After merging the datasets of 2004 until 2016, thirteen waves of data remain. In aggregate, there are 9,951 observations, with 2,847 unique households across these observations. That means that, on average, there are around 3.5 observations per household. These observations only consist of the household heads. The reason for this is that there is evidence that they have the biggest influence on financial decisions like the one under study.
Descriptive statistics of the variables of interest are depicted in Table 3. Notice that this table represents both the descriptive statistics for the non-psychological variables and the psychological variables. Those figures are calculated by using the whole sample. The sample in a given year is likely to be representative, but when considering the sample over a longer period, this may be different9. For instance, households that participate in more than one year may differ in some respects from households that only participate one year.
I will now turn to the correlation matrix, which is presented in Table 4. Some attention will be devoted to the correlations that exceed 0.2, which is a sign of, at least, moderate correlation. Notice that these correlations do not say anything conclusive about causal effects, but they might hint at a relationship between the variables. First, savings are positively correlated with both self-perceived financial literacy and net household income. This former corresponds with what is put forward in the literature review. It is also interesting to note that there is a negative correlation between age and time preference. Apparently, individuals become less patient when they grow older. This may also indicate why older persons save less on average, although the correlation coefficient between age and savings is fairly small. Consistent with the latter is that older people have less saving peers. Finally, there is a positive correlation between age and being highly educated. This makes sense, as the highest completed education level is monotonically increasing over each individual’s lifetime
20
Table 3: Descriptive statistics over the whole sample.
4.5. Limitations of the dataset
Although the DHS survey covers a vast number of topics and appears to be sophisticated, there are a few drawbacks of this survey. This section will consider three aspects of the survey that may impede research. First, as was mentioned before, participating households need not to participate in all surveys. This if unfortunate, as for several households, there is an incomplete set of answers. When I proceed to empirical research, it is necessary to restrict attention to those households that answered all relevant questions. Second, the number of participants per household may vary per wave, which can have various reasons. Household members pass away, some children may have become sufficiently old (sixteen years old) to participate, or household members either enter the panel or withdraw from the panel. I concentrate on the household’s heads, as I assume that they have the biggest influence on savings decisions. The drawback of this is that it may result in less observations than when considering all household members. Finally, comparing results of participants that take part in multiple waves indicates that some people give varying answers to the same question in subsequent waves. Although this is normal for questions regarding time-variant variables like household income, this pattern is also observed for psychological concepts. Albeit people’s behavior might gradually change over a long period, it is unlikely radical changes would happen over only a year. It is even stranger that, in subsequent years, adults give different answers to questions that relate to their youth. One way to circumvent this is to ask such time-invariant questions only once. Such practice also limits the length of the survey, which might lead to more reliable answers.
# obs Mean Std. Dev. Min Max
22 5. RESEARCH METHOD
5.1 Econometric framework
As was discussed in the data section, after merging the several datasets, an unbalanced panel data set remains. In deciding whether this data set should be made balanced, there is a tradeoff. On one hand, deleting all observations for which there is no full set of answers makes the panel balanced. However, doing so might introduce bias if the households that did not respond in all periods are not representative for the population. Verbeek and Nijman (1992) state that both keeping the unbalanced panel and deleting households for which there is no full data is vulnerable to bias. My hypothesis is that, among the household heads that did not participate in all years, there will be disproportionally many older household heads. Hence, it might be preferable to stick with the unbalanced panel data set.
The plan is to employ a random effects regression, describing which variables explain savings. Survey participants are being asked how much money their household has put aside in the past 12 months. The answer to this question constitutes the dependent variable ‘’household savings’’. Although the answer to this question might partially refer to savings made in the previous year (since the survey need not be filled out at the end of some year), for simplicity, it is assumed that the savings are done in the year in which the survey is undertaken. Notice that the answers to this question are in an interval format. In order to facilitate the analysis, the midpoint of those intervals will be taken. The same procedure applies to the entries on the income of the household10.
A household may consist of multiple members. For measuring variables like household income, this does not matter, but when considering psychological variables, there might be a problem. It is unlikely that decisions on saving are made by only one household member. Then, a decision should be made on how to weigh the household members’ answers. There may be no clear-cut answer to this question. Therefore, I proceed by only considering the answers of the person who is the household head. This is done because I assume that this person has the biggest influence in deciding on financial matters, such as savings.
5.2 Regression equation
I will employ a random effects regression, which is given in equation (1):
23
𝑆𝑖𝑡 = 𝛼𝑖𝑡 + β𝑋𝑖𝑡 + δ𝑌𝑖𝑡 + ε𝑖𝑡 (1)
where 𝑆𝑖𝑡 are savings of household i in year t, 𝛼𝑖𝑡 is a constant and ε𝑖𝑡 is the error term. 𝑋𝑖𝑡 is a vector of both household and personal characteristics and 𝑌𝑖𝑡 is a vector of psychological variables. The variables that constitute those vectors have been discussed above. I am particularly interested in the coefficients of δ: they measure the effect on savings of the psychological variables (that are in 𝑌𝑖𝑡) that I am after.
6. RESULTS
This chapter is split in three parts. I will start with an evaluation of the results that the regression of Equation (1) yields. Thereafter, it will be considered whether the inclusion of psychological variables has increased the explanatory power of the model. Finally, I will discuss the presence of endogeneity in this model.
6.1 Results Equation (1)
In the first specification, savings are regressed on the non-psychological variables only. The results are presented in Table 5. Income has a significant positive effect on savings. The same holds for high education. It deserves attention that living together, being older and being male all lead to higher saving. Starting with the former, couples might save more than singles as the former face an additional bequest motive (Hurd, 1999). Once part of the couple dies, the spouse is still alive. Hence, the couple does not only bequest to children or other relatives, but also to the surviving spouse. Therefore, it makes sense for couples to save more than singles. Secondly, being male would lead to higher savings. One explanation lies in the presence or absence of children in the household. If it is true that children of divorced parents tend to live with the mother, female-headed households spend more out of income, ceteris paribus. Finally, higher age leads to lower savings. This follows the prediction of the simple life-cycle model that savings fall after retirement.
24 (1) (2) (3) nethhinc 0.0201*** 0.0200*** 0.0731*** (8.13) (8.08) (8.83) baby -368.9 -401.0 -335.1 (-1.29) (-1.39) (-0.86) pensfund -39.29 -55.44 1149.3 (-0.09) (-0.12) (1.38) selfemp 1279.7 1274.5 -452.5 (1.74) (1.73) (-0.38) highed 2081.8*** 2110.4*** 835.4* (8.66) (8.71) (2.49) male 1143.5*** 1144.5*** 1209.4** (3.82) (3.82) (2.92) age -23.60** 26.20 -22.37 (-2.94) (0.49) (-1.94) livetogether 772.4** 776.7** 236.4 (2.80) (2.82) (0.58) age2 -0.449 (-0.95) schedules 166.5 (1.12) timepref 203.6 (1.87) spfinlit 1069.5*** (4.65) peer 356.5** (2.61) trackexp -30.52 (-0.19) _cons 3652.0*** 2407.7 -3720.5* (5.59) (1.65) (-2.50) N 4196 4196 1077
25 The third specification adds the psychological variables of interest. Income, being male and high education are all still predictors of savings. Both age and living together do not affect savings anymore, however. With regard to the psychological variables, self-perceived financial literacy and peer effects do influence savings. When comparing these results with those obtained in the first specification, some changes deserve attention. First, the magnitude of being highly educated has decreased a lot, although it is still significant. This is most likely due to including self-perceived financial literacy in this specification. Still, there might be endogeneity present, as ability is likely to influence both education and self-perceived financial literacy on one hand, and savings on the other hand. Second, having access to a pension fund is not significantly affecting household savings in both specifications. I would expect that being in a pension fund would decrease household savings, as there is less reason for the household to save by itself when it has a pension.
I will now compare the results with the hypotheses that were put forward earlier. The hypothesized positive effect of both self-perceived financial literacy and peer effects on savings corresponds with the results. However, the other psychological variables have no effect on savings, while they were hypothesized to affect savings positively.
6.2 Additional explanatory power of psychological variables
In this thesis, I have opted for using psychological variables to explain savings. I will now consider whether the inclusion of these variables has made a difference. In order to do so, I will compare the regression of column (3) of Table 5 with a regression of column (2) on the 1077 observations of column (3) of Table 5. This comparison reveals that the overall R2 increases after including psychological variables. In the specification without psychological variables, the overall R2 equals 0.1575, while in the specification that includes psychological variables, the overall R2 equals 0.1883. Therefore, the latter model explains approximately three percentage points more of the variance in savings. This might seem rather modest. Notice, though, that this increase is relatively large compared to the original overall R2. Therefore, the inclusion of psychological variables appears to be valuable after all.
6.3 Endogeneity
26 literacy on savings should be interpreted cautiously. The direction in which the bias lies is ambiguous. One might expect that higher ability leads to higher savings, such that the coefficients for education and self-perceived financial literacy are probably upward biased. However, Card (2001) mentions that this not need be the case. In his review, he finds that many studies that instrument for education actually obtain higher coefficients for the instruments than for education itself.
7. CRITIQUES
I will now discuss some drawbacks of my research. One limitation is that my research is largely based on answers to survey questions. Attrition and non-response are two problems that often come with survey data. These concepts are problematic as their existence may hamper the representativeness of the sample. According to Teppa and Vis (2012), the dataset under study is no exception. Attrition rates between 2004 and 2009 are in the range of 15 to 20 percent per year. Regarding non-response, Teppa and Vis indicate that around 19% of households did not respond in 2010.
I would also like to stress two aspects of employing survey data that are more peculiar. First, there might be a discrepancy between what individuals answer to survey questions and what they would do if they face a real-life decision (Rabin, 1998). For instance, a person stating in a survey that he is willing to take a risky gamble will probably be less adventurous when one’s money is at stake. Rabin argues, however, that situations with real repercussions closely resemble the results that were found in experiments. Therefore, this argument against survey data may have little bite.
A second shortcoming of using survey data is that repeated participation may affect outcomes. Crossley et al. (2017) discuss the effect of participation in surveys on household saving. Comparing two randomized groups, they find that persons that have taken part in surveys make different choices than those individuals who did not. An explanation for this is that survey participants are exposed to questions on topics that they normally would not think of. In the context of this thesis, people might only start thinking about retirement saving when they are being asked about it. Especially for young individuals, such an effect might be pronounced. If this effect would be present in my data, I expect households that participate in multiple waves of the survey would have higher savings, ceteris paribus, to households that participated in only one wave of the survey.
27 8. POLICY IMPLICATIONS
It follows from the results that both self-perceived financial literacy and peer effects tend to have a positive effect on savings. Based on this, what can be done to increase the savings rate?
8.1 Self-perceived financial literacy
I believe that, when considering financial literacy, it is optimal to target the individuals who are least financially literate. These persons would probably benefit the most from an increase in financial literacy. Not only because their low level of financial literacy, but also because they often save less than those who are more financially literate (see Table 4). In the literature review, it is mentioned that Lusardi and Mitchell (2011b) find that the youngest and oldest people often are little financially literate. Introducing financial literacy as a mandatory school subject is a possibility when targeting the youngest individuals. Tomaskova et al. (2011) opt for this in the case of a Czech university, as they find that many students answer financial literacy questions incorrectly. Urban et al. (in press) find that, in American states where high school curriculums contain financial subjects, pupils make better financial decisions later in life than pupils that did not follow those subjects. One major difference between those studies is that Tomaskova et al. study universities, while Urban et al. consider high schools. According to me, financial education in high school might be more worthwhile than in university. The reason for that is that I think that the former on average features a lower level of financial literacy. Therefore, I argue that more attention to financial subjects in high school might lead to the desired increase in financial literacy.
With regard to this policy recommendation, I would like to stress one disadvantage. Spending more time on financial education will inevitably lead to less attention for other subjects (Urban et al, in press). Therefore, in designing high school curriculums, one should balance the goal of increasing financial literacy of students with the goal of spending enough time on other subjects.
8.2 Peer effects
28 the peer effect of such a program might be even larger than my results indicate. Not only do saving club participants have fellow members that save, but these members also expect that other participants save.
What such a program would look like in the Netherlands is unclear. A savings account in a bank that a group could contribute to is a possibility. An advantage above the South-African stokvel is that these savings would also generate interest. Such a savings account would thus combine the merits of interest and peer effects, which both help increase savings.
8.3 Flexibility of the pension system
Finally, I would like to mention one potential policy solution that is specific to the Dutch pension system. This solution does not directly relate to the results in Section 6, but it does relate to the discussion on status quo bias and anchoring in Section 2.5. Delsen (2014)discusses whether changes in the setup of occupational pensions could have an effect. Delsen indicates that some individuals are not satisfied with the current way in which occupational pensions are determined. They would rather have more flexibility in choosing the height of their contributions. Delsen doubts whether the effect of more flexibility would be beneficial, however. Using several survey-based results, Delsen states that many people appreciate the compulsory nature of pensions. Those persons indicate that they might not participate in the pension fund if it were not compulsory. This can be taken as evidence that freedom of choice may decrease rather than increase aggregate pensions. Still, I think that it may be worthwhile to experiment with higher flexibility in pension schemes.
9. CONCLUSION
29 DHS survey. By doing so, they were able to find proxies for, among others, status quo bias. I would have liked to do so, but it is outside the scope of this thesis.
Moreover, further research might benefit from the establishment of more sophisticated psychological variables. In this thesis, each psychological variable was based on only one survey question. However, several survey questions cover the same concept. By averaging responses to multiple relevant survey questions, one could create psychological variables that are more reliable.
Finally, I am interested in the further development of neuroeconomics. Although I praise further economic research on saving, a parallel extension of the body of knowledge of neuroeconomics might assist in the development of useful policy solutions.
REFERENCES
Alessie, R.J.M., Kapteyn, A. & Klijn, F.E. (1997). Mandatory Pensions and Personal Savings in the Netherlands. De Economist, 145(3), 291-324.
Ameriks, J., Caplin, A., Leahy, J. & Tyler, T. (2004). Measuring Self-Control Problems. The American
Economic Review, 97(3), 966-972.
Angrist, J.D. & Krueger, A.B. (1991). Does Compulsory School Attendance Affect Schooling and Earnings? The Quarterly Journal of Economics, 106(4), 979-1014.
Bernheim, B.D. & Garrett, D.M. (1996). The determinants and consequences of financial education in the workplace: evidence from a survey of households. NBER Working Paper series, no 5667.
Börsch-Supan, A. (2004). Mind the Gap: The Effectiveness of Incentives to Boost Retirement Saving in Europe. OECD Economic Studies, 39(2), 111-144.
Browning, M. & Lusardi, A. (1996). Household Saving: Micro Theories and Micro Facts. Journal of
Economic Literature, 34(4), 1797-1855.
Bucher-Koenen, T., Lusardi, A., Alessie, R.J.M. & van Rooij, M. (2016). How Financially Literate Are Women? An Overview and New Insights. The Journal of Consumer Affairs, 51(2), 255-283.
30 Choi, J.J., Laibson, D., Madrian, B.C., & Metrick, A. (2004). For Better or for Worse: Default Effects and 401(k) Savings Behavior. Perspectives on the Economics of Aging. University of Chicago Press, 81-125.
Collins, D., Morduch, J., Rutherford, S. & Ruthven, O. (2009). Portfolios of the Poor: How the World’s Poor Live on $2 a Day. Princeton University Press.
Crossley, T.F., de Bresser, J., Delaney, L. & Winter, J. (2017). Can Survey Participation Alter Household Saving Behavior? The Economic Journal, 127, 2332-2357.
Delsen, L. (2014). Keuzemogelijkheden binnen en tussen pensioenregelingen: niet voor elk wat wils. (Options Within and Between Supplementary Pension Schemes: Not for Everyone). Netspar Discussion
Paper, no. 12/2014-071.
DNB (2018). Spaargeld huishoudens (Saving money of households). Retrieved from
https://statistiek.dnb.nl/dashboards/spaargeld/index.aspx#.
Duflo, E. & Saez, E. (2002). Participation and investment decisions in a retirement plan: the influence of colleagues’ choices. Journal of Public Economics 85, 121-148.
Euwals, R. (2000). Do mandatory pensions decrease household savings? Evidence for the Netherlands.
De Economist, 148(5), 643-670.
Friedman, M. (1957). A Theory of the Consumption Function. Princeton: Princeton University Press.
Hamermesh, D.S. (1984). Consumption During Retirement: The Missing Link in the Life Cycle. The
Review of Economics and Statistics, 66(1), 1-7.
Hazlitt, H. (1959). The Failure of the New Economics. Princeton: Van Nostrand.
Hurd, M.D. (1999). Mortality Risk and Consumption by Couples. NBER Working Paper, no. w7048.
Kane, T.J. & Rouse, C.E. (1993). Labor Market Returns to Two- and Four-Year Colleges: Is a Credit a Credit and Do Degrees Matter? NBER Working Paper, no. 4268.
31 Kennickell, A. & Lusardi, A. (2004). Disentangling the importance of the precautionary saving motive.
NBER Working Paper, no. 10888.
Keynes, J.M. (1936). The General Theory of Employment, Interest, and Money. London: Macmillan. Koopmans, T.C. (1960). Stationary Ordinal Utility and Impatience. Econometrica, 28(2), 287-309. Likert, R. (1932). A technique for the measurement of attitudes. Archives of Psychology, 22 (140), 5-55.
Loewenstein, G., Rick, S. & Cohen, J.D. (2008). Neuroeconomics. Annual Review of Psychology, 59, 647-672.
Lusardi, A. & Mitchell, O.S. (2011a). Financial Literacy and Retirement Planning: Implications for Wellbeing. NBER Working Paper series, no. 17078.
Lusardi, A. & Mitchell, O.S. (2011b). Financial literacy around the world: an overview. Journal of
Pension Economics and Finance 10(4), 497-508.
Lusardi, A., Mitchell, O.S. & Curto, V. (2011). Financial Literacy among the Young. The Journal of
Consumer Affairs 44(2), 358-380.
Mastrogiocomo, M. & Alessie, R.J.M. (2015). Where are the retirement savings of self-employed? An analysis of ‘unconventional’ retirement accounts. DNB Working paper, no. 454.
Matell, M.S. & Jacoby, J. (1972). Is There an Optimal Number of Alternatives for Likert Scale Items? Study I: Reliability and Validity. Journal of Applied Psychology, 56(6), 506-509.
McClure, S.M., Ericson, K.M., Laibson, D.I., Loewenstein, G. & Cohen, J.D. (2007). Time discounting for primary rewards. Journal of Neuroscience 27(21), 5796-5804.
Modigliani, F., & Brumberg, R. (1954). Utility Analysis and the Consumption
Function: An Interpretation of Cross-Section Data. in K. Kurihara, ed., Post Keynesian
Economics, Rutgers University Press, 388-436.
NRC (2016). CBS: Tussen 55 en 65 jaar grootste risico op armoede (Highest poverty rates among people aged 55-64 years). Retrieved from:
32 O’Donoghue, T. & Rabin, M. (1998). Procrastination in Preparing for Retirement. Behavioral
Dimensions of Retirement Economics. Henry Aaron, ed. Brookings Institution and Russell Sage,
125-156.
O’Donoghue, T. & Rabin, M. (1999). Doing It Now Or Later. The American Economic Review, 89(1), 103-122.
Rabin, M. (1998). Psychology and Economics. Journal of Economic Literature, 36(1), 11-46.
van Rooij, M. & Teppa, F. (2014). Personal traits and individual choices: Taking action in economic and non-economic decisions. Journal of Economic Behavior & Organization, 100, 33-43.
Samuelson, W. & Zeckhauser, R. (1988). Status quo bias in decision making. Journal of Risk and
Uncertainty, 1(1), 7-59.
Schors, van der A. & Werf, van der M. (2017). Geld achter de hand makkelijker maken (Make saving money easier). Utrecht, Nibud.
Shefrin, H.M. & Thaler, R.H. (1988). The Behavioral Life Cycle Hypothesis. Economic Inquiry, 26(4), 609-643.
Strotz, R.H. (1956). Myopia and Inconsistency in Dynamic Utility Maximization. Review of Economic
Studies, 23(3), 166-180.
Teppa, F. & Vis, C. (2012). The CentERpanel and the DNB Household Survey: Methodological Aspects. DNB Occasional Studies, 10(4).
Thaler, R.H. (1994). Psychology and Savings Policies. American Economic Review, 84(2), 186-192. Thaler, R.H. & Benartzi, S. (2004). Save More Tomorrow: Using Behavioral Economics to Increase Employee Saving. Journal of Political Economy, 112(1), 164-187.
Tomaskova, H., Mohelska, H., Nemcova, Z. (2011). Issues of financial literacy education. Procedia –
Social and Behavioral Sciences, 28, 365-369.
Urban, C., Schmeiser, M., Collins, M.J. & Brown, A. (in press). The effects of high school personal financial education policies on financial behavior. Economics of Education Review.
Verbeek, M. & Nijman, T. (1992). Testing for selectivity bias in panel data models. International
