Research Master Thesis The impact of the financial crisis on Dutch households

(1)

Research Master Thesis

The impact of the financial crisis on Dutch

households

July, 2014

Name student: Yao Pan Student number: s1939688

Student email: p.yao@student.rug.nl

(2)

2

The impact of the financial crisis on Dutch

households

Abstract

European economies have been affected by the great recession since 2007. The current financial crisis is said to influence various aspects of households such as income, cohabitation behavior, and subjective life expectancy. We focus on the wealth (financial distress, making ends meet, financial situation and economic situation), saving behavior and beliefs (income uncertainty, patience and risk aversion) of Dutch households. This study intends to explore how these variables change around the financial crisis and which groups of people are affected most severely by it. Data is drawn from the DNB Household Survey (DHS), covering both the pre-crisis and the pre-crisis period (2006-2013). Empirical outcomes reveal that this financial pre-crisis makes more people experience financial distress, hard to make ends meet, report a poor financial situation, slightly reduce their saving, and become more risk averse. Immediately after the crisis, people tend to get more pessimistic but later on become more optimistic. In addition, young and middle aged people, family without children, the self-employed, the widowed and the never married are the groups which are attacked the most by the crisis. The groups of the old people, employees, and the married have a protective role from the crisis. Consistent with the life cycle model, we also find that more patient people will save more, and they become more prudent right after the crisis but this trend then disappears.

(3)

3

1. INTRODUCTION

The current financial crisis has been regarded as the worst recession for European countries since the 1930s. While the crisis may be booming out, its influence may linger for quite some time. Households' wealth and income have plummeted, the unemployment rate has soared, and the stock market has collapsed. Not only young people are seriously hit by the current financial crisis, the elderly also suffer from it via poor health, income loss and pressures from young generation. For instance, Cavasso and Weber (2013) find that many older households have little wealth and financial distress ascends during the financial crisis in Europe. Many Europeans have faced a harsh situation and desperation they have never expected. How will this financial crisis affect Europeans and will this crisis have a long lasting effect? Today, several years after the crisis peaked, we still cannot give an answer to these questions and they still remain one of the most widely debated issues among economists.

Previous papers working on the effects of the financial crisis in the U.S. and Europe can provide some insight into this topic. Several macro studies suggest that the financial crisis indeed has a vital influence on the fiscal policy and the financial system (i.e. Reinhart and Rogoff, 2008; Goddard et al.,2009; Afonso et al., 2010; Bachellerie, 2011). Although these studies provide highly relevant evidence for the macroeconomic effects of the crisis, they do not discuss clearly the evidence for the micro-level effect. Considering the micro studies, the lack of adequate and timely available data for studying the impact of the crisis inspires some scholars to run simulation studies using pre-crisis data (i.e. Bosworth and Smart,2009; Chai et al., 2011 ). However, results from these studies can only provide the predictions for the crisis but cannot give solid evidence for the actual situation during the crisis. In order to solve this problem, the DNB Household Survey, focusing on the Netherlands, motivates us to analyze this issue. This dataset has an advantage that respondents are surveyed over a long time, covering both the pre-crisis and the pre-crisis period. Hence, we are able to observe how the actual behaviors of the Dutch households change during the crisis and the detailed survey questions provided by this dataset make it possible to investigate various aspects of households (i.e. saving behavior).

(5)

5

unemployment rate and household consumption expenditures per capita from 2000 to 2012 in the Netherlands. According to Figure 1, real GDP per capita drops by approximately 4.3% from 2008 to 2012. The unemployment rate rises from 2.8% to 5.3%. Furthermore, final consumption expenditures per capita declines by 6.1% since 2008. Hence, the Dutch economy faces a lower real GDP per capita, a higher unemployment rate and a drop in consumption expenditures during the current financial crisis.

Figure 1 GDP per capita , unemployment rate and final consumption expenditures per capita in the Netherlands

Source: World Bank and OECD database

(6)

6

and simulation work; second, we also investigate which groups of people are influenced most seriously by the crisis and how does the relationship between the preference parameters and wealth/ saving behavior change throughout the crisis, which are not yet studied before. Our findings indicate that this financial crisis makes more people experience financial distress, harder to make ends meet, report a poor financial situation and slightly reduce their saving, but with a lagging effect on these variables. By contrast, households get more pessimistic right after the crisis and the optimism seems to rise again in 2013. For the beliefs variables, we only find that people become more risk averse during the crisis. In addition, young and middle-aged people, family without children, the self-employed, the unemployed, the widowed and the never married are the groups which are attacked more severely by the crisis. Consistent with the life cycle model, more patient people tend to save more. People are more sensitive to income uncertainty right after the crisis but this trend then disappears after some time.

The rest of the paper will be organized in four additional sections. Section 2 provides a brief review of the relevant literature as well as the hypotheses. Section 3 will discuss the data and the methodological issues. The empirical facts and main results will be presented in Section 4. Finally, Section 5 will provide the conclusions, limitations and potential direction for further research.

2. LITERATURE REVIEW

We build on two main strands of literature in our analysis. One relates to how the financial crisis and economic downturn affects people’s wealth and saving behavior while the second focuses on the beliefs and expectations. Other relevant literature links two strands together.

Wealth and saving behavior

(7)

7

countries, covering the pre-recession (2006-07) and the recession periods (2010-11). They use a financial distress indicator as a signal for low liquid wealth for a given income. This indicator is constructed by considering distressed individuals whose financial wealth (value of net financial assets) is smaller than three months' income as well as whose equivalized income1 is not in the top third of the income distribution. Their results indicate that a higher education level, higher per-capita income, and home-ownership tend to lower the probability of being in financial distress while females are more likely to suffer from the financial distress than males. Besides, they also compare households conditions before and after the crisis and find a worsening of financial distress in all European countries except France and Italy. In the case of the U.S., Hurd and Rohwedder (2010) use the American Life Panel to investigate the effect of financial crisis and great recession on American households. According to Hurd and Rohwedder (2010), a household is financially distressed if "the respondent and/ or spouse is unemployed, or if the household is more than two months behind the mortgage payments (or in foreclosure), or if the value of the house is less than the amount of mortgage". They suggest that more households experience financial distress, have negative equity in their house and reduce spending during the financial crisis. Besides, households with lower income and younger people are more likely to experience financial distress. Lusardi (2010) uses survey data to assess how American households react to the economic shocks in terms of making financial decisions and of financial capabilities under current economic conditions. One main focus of this study is on the ability to make ends meet, measured by examining how people deal with every financial matters and the extent to which they balance monthly income and expenses to avoid overspending. This paper finds that the recent economic crisis has hindered American households' ability to make ends meet. Difficulty with making ends meet is especially high among young people and those with low income. In addition, Bosworth and Smart (2009) also analyze U.S. data. As they do not have the data of the crisis period, this paper uses the wealth holdings and price indices in 2007 to simulate the magnitude and distribution of wealth losses during the crisis period (2008-2009). They reveal that American households aged over 50 have lost almost a fifth of their net wealth from the financial crisis. Hence, according to the previous literature, their results show that households tend to have poor wealth status during the financial crisis and great recession.

1

(8)

8

Considering saving behavior, only a limited number of papers can be found on the effect of financial crisis or great recession on households saving behavior using micro data. As a rule most papers study saving rate at the macro level. For instance, Bachellerie (2011) shows that household's saving rate decreases in France since the financial crisis of 2007 while the structure of investment flows has no significant change. For the studies at the micro level, Van der Cruijsen et al. (2011) investigate whether the household experience with the crisis and knowledge on banking supervision will affect their saving behavior in the Netherlands by using the DNB household survey. The dataset asks detailed questions on respondents' saving behavior such as "Currently, do you have a saving account at a bank?". The authors then construct dummy variables of whether respondents' bank went bankrupt or bailout to indicate crisis experience. They reveal that households' experience with the crisis has a vital impact on their saving behavior. Households who were customer of a bank that went bankrupt or received government support gather more information than others and people with crisis experience are more likely to save at several banks.

Beliefs

(9)

9

measure risk aversion, they rely on both qualitative and quantitative measures. The former is based on directly asked survey questions2 while the latter infers an individual's relative risk aversion from his share of investments in risky assets. Similarly, by using the sample data from 20 emerging countries for currency crisis and 27 countries for stock market crisis between 1995 to 2005, Coudert and Gex (2008) argue that financial crisis often coincide with periods in which risk aversion ascends. In their paper, risk aversion is found to increase just before the stock market crisis and remain high during the crisis, but less so for currency crisis. Notably, Coudert and Gex use several quantitative measure of risk aversion, such as an indicator based on the correlation across different assets between price variations and their volatility; indicator using a principal components analysis on risk premia; indicator using implicit volatility of option prices. Concerning the measure of risk aversion, Kapteyn and Teppa (2011) provide a detailed comparison among different risk aversion measures. They include the measure based on choices over lifetime incomes, six direct questions about investment choices, direct questions on precaution and risk aversion and the measure based on the risk of the assets drawn from the DNB Household Survey. Kapteyn and Teppa first analyze all risk aversion measures as a set of background characteristics and then incorporate them into a household portfolio allocation model. This paper finally shows that qualitative measure based on a number of simple risk preference questions have the most explanatory power. Quantitative measures based on economic theory (i.e. measure based on choices over lifetime incomes) are less powerful as they may exceed the financial capability of respondents. Thus, we will use the more direct measure from the survey questions to test risk aversion and we will discuss it in more details in the next section. They also reveal that females are slightly more risk averse than males and have a higher precautionary motive. Better educated people are less risk averse. Age positively relates to risk aversion. In addition, Choi and Kim (2014) consider the case in Korea and estimate time preference parameters based on Euler equations. Their results demonstrate that Korean consumers become more patient in making consumption decisions during financial crisis in 1997 and 2008, which

2

Survey questions:

1. Which of the following statements comes closest to the amount of financial risk that you are willing to take when you make your financial investment: (1) a very high return, with a very high risk of losing money; (2) high return and high risk; (3) moderate return and moderate risk; (4) low return and no risk.

(10)

10

can explain a great decline in consumption during financial crisis. Overall, according to the prior findings, people tend to become more pessimistic, risk averse and patient during financial crisis or great recession.

Relationship between wealth/saving behavior and beliefs

Combining two strands of literature together, we can refer to the literature on the life cycle model. The basic idea of the model suggests that households will smooth consumption over the life cycle and hence save before retirement in order to offset the drop in future income. One important extension of the model is about the precautionary saving which emphasizes that people can also save against uncertain events, such as income shocks. According to the life cycle model and precautionary saving theory, saving increases with more risk aversion, income uncertainty, and patience (i.e. Skinner, 1988; Caballero, 1991; Browning and Lusardi ,1996). For the empirical results, Lusardi (1998) finds that households with more risk aversion accumulate more wealth while households with low rate of time preference (more patience) also accumulate more by using Health and Retirement Study (HRS) dataset. However, the empirical results on precautionary saving against income uncertainty are rather mixed. Lusardi (1998) indicates that people facing a higher income risk save more and accumulate more wealth. Carroll and Samwick (1997) reveal that people with higher income uncertainty tend to have higher wealth and precautionary motive is quantitatively relevant. Similarly, Skinner (1988) reveals that people respond to income uncertainty by accumulation assets, which accounts for a large fraction of savings. However, Guiso et al. (1992) argue that the influence of uncertainty on wealth accumulation is consistent with the precautionary saving theory but it only explains a small portion of saving.

(11)

11

There is still some space to research more by using real household data. Hence, in the following sections, we will give special focus on the Dutch households and examine how their wealth, saving behavior and beliefs change during the recent financial crisis.

3. DATA AND METHODS

3.1 Description of Data

The main dataset used in this paper is from the DNB Household Survey (DHS), which has been conducted by CentERdata, a survey agency at Tilburg University specialized in internet survey and experiments since 1993. Notably, although the Netherlands has an approximately 80% of the internet penetration, the selection of participants is not determined by the use and availability of internet. Households without a computer or an internet connection are provided with the necessary equipment (i.e. a set-top box to participate via their television connection). The CentERpanel forms a representative sample of the Dutch-speaking population. It consists of over 2000 Dutch household participants who answer questions concerning their psychological and economic aspects of financial behavior. Data are collected annually and the attrition is dealt with biannually refreshing samples that are drawn in view of keeping the panel representative of the Dutch population of sixteen years and older (excluding peoples in hospital, specialized care institutions or prisons). Moreover, the DHS includes five questionnaires: 1) work and pensions; 2) housing and mortgages; 3) income and house; 4) assets and debts; 5) economic and psychological concepts. These questionnaires (except the second one) should be filled out by respondents who are at least sixteen years old. The housing questionnaire is in principle filled out by the household head. However, CentERdata also provides the dataset 'general information of the household' which contains (mainly) demographic information on all members (including individuals aged below 16) of those households who responded to at least one of the 5 questionnaires mentioned above. The five questionnaires have been launched at different weeks of the year so that the number of responding households differs across the questionnaires.

(12)

12

behavior, risk perception and risk aversion, and expectations for the future when compared to the current situation. In addition, as this dataset has surveyed the households over a long period, we can observe their behavior both before and during the financial crisis. The sample we choose covers both pre-crisis and the crisis period from year 2006 to 2013. We drop all observations with missing values and the answers “Not applicable”, “Refusal”, “Don’t know”, and “No answer” in the analysis. In this study, we analyze the behavior at the household level and we only select the head of each family, who are often in charge of household finances, as a representative of one household. Initially, the sample consists of 14,653 household-year observations for which information is reported in the datasets ‘general information of the household’. As these datasets do not contain sufficient information on the marital status of the household head, we have retrieved these data from the questionnaires ‘household and work’, but consequently our final sample reduces to 9,982 household-year observations for which the values of all explanatory variables of the regression analyses are known. The actual number of observations depends on the response variables.

The first variable which we analyze is making ends meet. It is a widely used indicator of financial hardship. The original questionnaire states that “How well can you manage the total income of your household? It is very hard/ it is hard/ it is neither hard nor easy/ it is easy/ it is very easy”. Hence, the higher the number the easier people find it to make ends meet.

(13)

13

same way to get a dummy variable where 1 means being in financial distress and 0 refers to not suffer from financial distress.

The third variable is the current financial situation. The DHS includes a question “How is the financial situation of your household at the moment”. There are five choices: 1 "there are debts", 2 "need to draw upon savings", 3 "it is just about manageable", 4 "some money saved" and 5 "a lot of money can be saved".

The fourth variable is about people’s saving behavior. We select two survey questions to measure people's saving behavior. The first one is percentage of households who save, given by “Did you put any money aside in the past 12 months”, where 0 "no" and 1 "yes". The second survey question asks the respondents who have saved money to identify which category their household belongs to. The original question is "About how much money has your household put aside in the past 12 months (in euro)". The category includes "less than 1,500", "between 1,500 and 5,000", "between 5,000 and 12,500", "between 12,500 and 20,000", between 20,000 and 37,500", "between "37,500 and 75,000" and "75,000 and more". For the second measure of household's saving behavior, we construct amount of saving by adding one more category "0 euro" in order to avoid losing many observations. If people give answer "no" in the first question, they will belong to category "0 euro". Thus, we end up with eight categories for the second measure.

The last one is the expected economic situation. Respondents are asked about “How do you think the economic situation of your household will be in five years’ time in comparison to the current situation? Much worse/ worse/ (about) the same/ better/ much better”. This variable intends to test people’s expectation about their future economic situation. Higher number indicates a better expectation on the household's economic situation.

Beliefs

(14)

14

we measure risk aversion by referring to subjective risk aversion questions. The DHS has six questions concerning risk aversion but we select two most relevant questions to measure this variable. Respondents are asked to evaluate to what extent they agree with the statements on a 7-point scale, where 1 means totally disagree and 7 indicates totally agree. The first statement is “I would never consider investments in shares because I find this too risky”. As an alternative, the second measure, namely risk loving, is given by the statement “I am prepared to take the risk to lose money, when there is a chance to gain money”. The former one will be used in main analysis and the other will be applied for the robustness checks.

Next, we measure people’s patience. Patient people are more future-oriented and tend to save more and consume less at present than impatient people. This variable is a 1 to 7 scale with a survey question “Some people spend all their income immediately. Others save some money in order to have something to fall back on. Please indicate what you do with money that remains after having paid for food, rent, and other necessities”. Answer with 1 suggests “I would like to spend all my money immediately” and 7 indicates “I want to save as much as possible”.

The variable income uncertainty is a continuous variable. Respondents are asked to give the answer to the question “What do you expect to be the highest and lowest total net yearly income your household may realize in the next 12 months”. Then income uncertainty is computed by the difference between the highest and lowest total net yearly income of the households.

Other explanatory variables

(15)

15

temporal dynamics, later we will interact those with some socio-demographic characteristics to look at heterogeneous effects among different groups of the population.

Table A and B in Appendix gives a short description and summary statistics of the variables used in this paper. For instance, the average age of our sample equals 53.6 (ranging from 19 to 94 years old); the average family size is about 2.4 people; 72.3% of the respondents have home ownership. The correlation between each pair of main variables is shown in Table C in Appendix. Notably, this correlation is analyzed at the individual level. According to the table, the pairs of main variables are not highly-correlated. Thus, there is no serious multicollinearity problem.

3.2 Methodology

We are going to discuss the methodological issues in this section. For the regressions, we mainly apply linear models (OLS). We will use OLS to examine how wealth, saving and beliefs are affected by the financial crisis and how the relationships between beliefs and saving/ wealth holdings change during the crisis. We will account for the presence of a panel component in our regression analysis by providing robust clustered (at the household level) standard errors. In what follows, we use the index i to mean the individual (i=1,2..., N) while t denotes the survey year (t=2007,2008...2013).

Linear models

In this paper the method we use is the ordinary least squares (OLS) estimator. The regression function is as follows:

(1)

(16)

16

coefficient. We then extend this model by allowing for the interaction effects between key explanatory variables and year dummies, as stated by function (2).

(2)

where and are the main effects, denotes the interaction terms, is a column vector of other explanatory variables. The coefficients of the interaction terms are of our interest. Besides, the partial derivative of with respect to is . The interpretation of is the marginal effect of with respect to if all year dummies are equal to zero (i.e. in

2006). By means of a F-test, we will check whether the marginal effect of on is time specific. In other words, we will test the following null hypothesis: H0: =...= =0.

Moreover, as our dependent variable includes interval variable, such as amount of saving, we extend the OLS estimator by considering interval censoring. According to Cameron and Trivedi (2005), interval regression is a generalization of the model fit by tobit and takes into account of left-censored, right-censored or point data. Hence, since we know the ordered category into which observations falls but do not know the exact value of the observation, interval regression is applied.

4. EMPIRICAL RESULTS

In this section, we will first present some stylized facts about how wealth, saving behavior and beliefs change during the financial crisis in section 4.1. Then, in section 4.2, we analyze wealth, saving behavior and beliefs as a function of all background characteristics but excluding interactions. Section 4.3 will describe the complete regression outcomes with interactions to test which groups of people are influenced the most by the financial crisis and how the relationships between beliefs variables and wealth/ saving behavior change during the crisis.

4.1 Stylized facts

(17)

17

First, before running regressions, we present some figures, which depict the time evolution over time of our main variables. In Figure 2.1, we can see that the percentage of people who suffer from financial distress rises slightly since 2010 except a reverse in 2012. Generally, more people are financially distressed since 2010. Year dummies are jointly significant at the 0.01 critical level (refer to Table D.1 in Appendix). Recalling making ends meet, we combine category 1 and 2 into one category - "difficult" - while category 4 and 5 are combined into one category - "easy" in order to make the graph look more straightforward. Figure 2.2 compares three groups of "easy", "neutral" and "difficult". The outcome presents that year dummies are jointly significant at the 0.01 critical level. It reveals that the proportion of respondents who answer "difficult" increases by almost 3% from 2010 onwards and at the same time fewer people give answer "easy". Though this variable is not affected immediately by the crisis in 2009, more people found it hard to manage their household income after 2010 in the Netherlands. Regression result also shows that the estimated coefficients of year dummies reduce since 2010, which is analogous to the stylized fact in Figure 2.2.

Figure 2.1 Financial distress Figure 2.2 Making ends meet

Figure 2.3 demonstrates that the percentage of households who save money reduces mildly since 2009 but we do not find any statistical significance in Table D.1 in Appendix. According to Figure 2.4, the amount of saving of Dutch households declines a bit from 2010 onwards, which matches with regression results. In general, households' saving behavior does not change much during the crisis.

(18)

18 Figure 2.3 Percentage of households who save Figure 2.4 Amount of saving (conditional

upon saving)

Figure 2.5 depicts the current financial situation. Table D.1 in Appendix shows that the coefficients decline significantly from 2010 onwards. We then combine category 4 and 5 into one single category, called "have savings". The graph indicates that the proportion of people who have debts and use their saving increases significantly after 2010. Simultaneously, households with saving decline mildly since 2010. Therefore, more people tend to have a poor financial situation from 2010 and this trend lasts until 2013. Based on Figure 2.6, it reveals how the expectation on economic situation changes over the survey year. We combine category "much worse" and "worse" into one category while "better" and "much "better" into one category. We find that the percentage of people who think that the economic situation will become worse increases right after the crisis and reverse in 2013. On the contrary, the proportion of Dutch households who answer "better" decreases gradually since 2009 and ascends again in 2013. This finding is identical to Table D.1 that already in the run-up to the crisis expectations have declined and, as the crisis intensified, Dutch households have became increasingly pessimistic. In 2013 this downward trend reverses and we observe a significant increase in optimism, although it remains below its pre-crisis level.

(19)

19 Figure 2.5 Current financial situation Figure 2.6 Expected economic situation

Beliefs

For the beliefs variables, we observe how income uncertainty, patience and risk aversion change between 2006 and 2013. Year dummies are statistically significant for both income uncertainty and patience at the critical level 0.01 (refer to Table D.2 in Appendix). Figure 2.7 depicts that subjective income uncertainty rises in 2010 but then reduces and finally increases again in 2013. Thus, shortly after 2009, income uncertainty raises but this trend does not persist. According to Figure 2.8, patience ascends slightly since 2009 and then drops in 2012 and 2013. However, the magnitude of the change over time is tiny.

Figure 2.7 Income uncertainty Figure 2.8 Patience

0 20 40 60 80 1 0 0 p e rce n t 2006 2007 2008 2009 2010 2011 2012 2013 have debts draw upon saving manageable have savings

0 20 40 60 80 1 0 0 p e rce n t 2006 2007 2008 2009 2010 2011 2012 2013 worse the same

(20)

20

For risk aversion, the higher the number the more risk averse people become. On the contrary, lower number in risk loving means more risk averse. From Figure 2.9, there is a clear upward trend in risk aversion since 2010. We can observe a similar trend in Figure 2.10 that people become more risk averse throughout the crisis but less obviously. Regression results again confirm this trend.

Figure 2.9 Risk aversion Figure 2.10 Risk loving

In conclusion, according to the above figures we find the following facts:

(1) Initially, the financial crisis does not affect the financial distress, ability to make ends meet and current financial situation. After 2010, however, we observe that more people experience financial distress, find it harder to make ends meet and have a poor financial situation.

(2) Households' saving tends to decline slightly during the financial crisis.

(3) Shortly after the financial crisis, people are more likely to be pessimistic about their future economic situation. But they begin to become more optimistic after some time

(5) Dutch households become more risk averse throughout the crisis.

(6) Income uncertainty and patience also change, but without obvious trends.

(21)

21

4.2 Simple regression results without interactions

In this section, we run simple regressions to see how these variables are affected by demographic and socio-economic variables during the financial crisis. We include all key explanatory variables discussed in section 3.1.

(22)

22

Besides, only Model 4 shows that the amount of saving declines dramatically from 2010 onwards except a reverse in 2012. According to both Model 3 and Model 4, saving is influenced by the age groups, education level, marital status, employment status, home ownership, and net household income. In particular, higher education level, home ownership and rich family tend to save more while the unemployed, divorced, widowed and single are less willing to save.

Based on the regression outcomes in Model 5, current financial situation shows that people tend to have worse financial situation after 2009 as estimated coefficients drops substantially from 0.101 to 0.004, which again confirms our finding in Figure 2.5. Except age classes, other background variables significantly affect the current financial situation. For instance, household composition is found to be significant. In particular, larger family and the presence of underage and adult children tend to have worse financial situation. Regarding marital status, the divorced, single and widowed have a poorer financial situation than others. The self-employed and unemployed perform the worst in current financial situation. Higher education, richer family and home owners are more likely to get rid of poor financial situation during the crisis.

(23)

23 Table 1 Regression results of wealth and saving behavior

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

variables Financial distress Making ends meet Percentage of households who save Amount of saving1 Current financial situation Expected economic situation ageclass(36-45) -0.115*** (0.024) 0.031 (0.047) -0.007 (0.026) 598.2** (299.4) 0.015 (0.058) -0.305*** (0.046) ageclass(46-55) -0.139*** (0.024) -0.038 (0.049) -0.064** (0.027) -59.0 (296.8) -0.039 (0.061) -0.575*** (0.046) ageclass(56-65) -0.155*** (0.027) -0.019 (0.055) -0.053* (0.030) -60.6 (317.3) -0.010 (0.069) -0.839*** (0.048) ageclass(66-75) -0.174*** (0.033) 0.039 (0.070) -0.060 (0.040) -606.8* (362.9) 0.093 (0.086) -0.919*** (0.060) ageclass(75+) -0.190*** (0.038) -0.010 (0.083) -0.100** (0.047) -1,215*** (417.1) 0.052 (0.097) -0.882*** (0.068) underage child in HH 0.160*** (0.039) -0.152** (0.072) 0.024 (0.040) -771.4* (420.8) -0.157* (0.086) 0.040 (0.062) adult child in HH 0.119*** (0.035) -0.177*** (0.065) -0.057 (0.036) -627.5 (442.7) -0.207*** (0.079) 0.060 (0.057) family size 0.0297** (0.015) -0.052* (0.027) -0.035** (0.015) -133.3 (161.4) -0.063* (0.034) -0.015 (0.024) female 0.021 (0.018) -0.158*** (0.043) 0.024 (0.023) -659.8*** (208.8) -0.206*** (0.050) -0.130*** (0.035) pre-university -0.087*** (0.024) 0.178*** (0.058) -0.033 (0.032) 281.7 (227.8) 0.053 (0.064) 0.096** (0.045) senior vocational training -0.038 (0.024) 0.016 (0.049) 0.044 (0.027) 137.6 (222.9) 0.013 (0.055) 0.018 (0.040) vocational training -0.080*** (0.020) 0.222*** (0.043) 0.064*** (0.023) 1,090*** (223.2) 0.177*** (0.049) 0.081** (0.036) university -0.132*** (0.021) 0.411*** (0.055) 0.023 (0.027) 2,556*** (360.6) 0.227*** (0.061) 0.156*** (0.042) self-employed 0.005 (0.029) -0.131** (0.062) -0.067* (0.034) 635.1 (458.3) -0.220*** (0.075) 0.289*** (0.063) unemployed 0.111*** (0.030) -0.410*** (0.062) -0.183*** (0.032) -677.2** (317.5) -0.512*** (0.068) -0.010* (0.050) student,homemaker 0.101** (0.045) -0.045 (0.107) -0.134** (0.053) 8.878 (317.6) -0.030 (0.101) 0.019 (0.0730) retired 0.020 (0.023) 0.011 (0.052) -0.071** (0.029) -428.8 (275.9) -0.164*** (0.061) -0.006 (0.045) divorced 0.015 (0.028) -0.264*** (0.065) -0.106*** (0.036) -1,184*** (344.3) -0.186*** (0.072) 0.081 (0.052) widowed -0.052* (0.031) -0.087 (0.076) -0.081** (0.041) -309.8 (414.7) -0.110 (0.078) 0.083 (0.052) single -0.040 (0.027) -0.160*** (0.060) -0.109*** (0.032) -1,719*** (327.7) -0.188** (0.074) 0.115** (0.048)

living with partner 0.010

(24)

24 Table 1, continued

home ownership -0.123*** (0.018) 0.236*** (0.038) 0.085*** (0.021) 621.3*** (170.9) 0.201*** (0.042) 0.059* (0.030) ln_income -0.126*** (0.012) 0.333*** (0.026) 0.086*** (0.013) 2,013*** (186.1) 0.290*** (0.027) -0.007 (0.019) 20072 -0.003 (0.013) 0.019 (0.022) 0.005 (0.015) 36.53 (170.2) 0.056* (0.029) 0.060** (0.023) 2008 0.0004 (0.014) 0.062*** (0.0240) 0.026 (0.016) 294.4 (225.9) 0.107*** (0.032) 0.0004 (0.027) 2009 -0.014 (0.014) 0.034 (0.025) 0.011 (0.016) 404.4 (248.9) 0.101*** (0.033) -0.041 (0.028) 2010 0.011 (0.015) 0.072*** (0.027) -4.33e-05 (0.018) -120.1 (209.5) 0.041 (0.034) -0.037 (0.029) 2011 0.044*** (0.015) 0.017 (0.028) -0.004 (0.018) -34.70 (245.3) 0.031 (0.036) -0.062** (0.029) 2012 0.017 (0.015) 0.0003 (0.028) 0.008 (0.018) 8.05 (236.8) 0.028 (0.035) -0.138*** (0.030) 2013 0.033** (0.016) -0.017 (0.029) -0.012 (0.018) -260.3 (210.2) 0.004 (0.035) -0.071** (0.030) Constant 1.626*** (0.127) 0.0591 (0.264) -0.052*** (0.135) -16,166*** (1,880) 0.650** (0.292) 3.560*** (0.204) Observations 8,991 8,971 8,967 8,668 8,968 8,539 Adjusted R-squared 0.180 0.229 0.070 - 0.142 0.204 p-value of year dummies 0.003 0.004 0.448 0.001 0.004 0.000

Significance levels as follows: *** p<0.01, ** p<0.05, * p<0.1. Robust Standard Errors clustered at the household level. Notes: 1 interval regressions;

2 Year 2006 is set to zero as the base category.

______________________________________________________________________________

Beliefs

(25)

25

Female tend to be more risk averse than males. The higher the education, the less risk averse people become. In addition, we also observe that people who are home owners tend to be less risk averse than those who are not home owners. And richer family has lower degree of risk aversion.

For the income uncertainty, only age classes and employment status are found to be statistically significant. In particular, young people face a higher income uncertainty than the elderly. The self-employed report the highest income uncertainty compared to other groups. One potential explanation is that the self-employed need to undertake the risk by themselves while for those working in a company this risk will be taken by the company. Hence, the self-employed tend to have a higher income uncertainty than other groups. Patience is affected by the age groups, education level, family size, marital status and home ownership. Older people become more patient than the youth while the households with homeownership also tend to be patient. Regarding the education level, university will be more patient than other education levels. The group of the single is the most impatient category in terms of the marital status.

Table 2 Regression results of beliefs

Model 7 Model 8 Model 9 Model 10

variable Income uncertainty Risk aversion Risk loving Patience

(26)

26 Table 2 continued

senior vocational training 0.029

(0.022) -0.190 (0.122) -0.023 (0.088) -0.012 (0.067) vocational training 0.020 (0.015) -0.311*** (0.114) 0.157* (0.080) 0.030 (0.060) university 0.037 (0.023) -0.815*** (0.136) 0.533*** (0.096) 0.189** (0.077) self-employed 0.307*** (0.051) -0.330** (0.148) 0.252** (0.116) 0.126 (0.089) unemployed 0.032 (0.034) 0.007 (0.144) -0.059 (0.100) -0.054 (0.087) student,homemaker 0.038 (0.040) -0.011 (0.259) 0.096 (0.161) 0.089 (0.119) retired -0.016 (0.024) 0.171 (0.136) 0.008 (0.094) 0.096 (0.068) divorced -0.022 (0.032) -0.125 (0.161) 0.202* (0.118) -0.084 (0.088) widowed -0.001 (0.031) -0.186 (0.179) 0.035 (0.121) -0.007 (0.098) single -0.041 (0.026) -0.170 (0.143) 0.228** (0.103) -0.261*** (0.089)

living with partner 0.053

(0.037) -0.026 (0.133) 0.249** (0.105) -0.016 (0.086) home ownership 0.001 (0.015) -0.290*** (0.094) 0.306*** (0.064) 0.200*** (0.054) ln_income -0.007 (0.016) -0.274*** (0.065) 0.177*** (0.050) 0.048 (0.034) 20071 0.011 (0.022) 0.153** (0.065) 0.041 (0.050) 0.012 (0.038) 2008 -0.012 (0.026) 0.236*** (0.071) -0.093* (0.054) -0.043 (0.040) 2009 -0.032 (0.0221) 0.435*** (0.077) -0.188*** (0.055) -0.031 (0.042) 2010 -0.007 (0.028) 0.272*** (0.080) -0.272*** (0.056) -0.008 (0.043) 2011 -0.015 (0.024) 0.388*** (0.082) -0.271*** (0.059) 0.019 (0.044) 2012. -0.026 (0.021) 0.557*** (0.081) -0.184*** (0.058) -0.004 (0.046) 2013 0.021 (0.026) 0.603*** (0.081) -0.224*** (0.058) -0.023 (0.047) Constant 0.355** (0.159) 7.303*** (0.677) 0.858* (0.515) 4.27*** (0.367) Observations 8,392 8,435 8,431 8,967 Adjusted R-squared 0.041 0.071 0.108 0.053

p-value of year dummies 0.116 0.000 2.88e-08 0.791

Significance levels as follows: *** p<0.01, ** p<0.05, * p<0.1. Robust Standard Errors clustered at the household level. Notes: 1 year 2006 is set to zero as the base category.

(27)

27

In conclusion, the regression outcomes in this section are in accordance with the stylized facts we find before. Wealth and saving behavior are affected more by the crisis than beliefs. Besides, some background characteristics have significant impacts on both wealth/ saving behavior and beliefs, such as marital status, home ownership and employment status.

4.3 Regression results with interactions

In this section, we study whether the crisis have different influence on different groups of the population by adding, one at the time, interaction terms between the year dummies and some key explanatory variables to our main specifications. In particular, we include interaction terms between the year dummies and marital status, child, home-ownership, employment, age classes and education level at the first stage. If these interaction terms are significant, it means that the time trend is different across different groups of people. At the second stage, we study the interactive effect between beliefs variables and wealth/ saving behavior. Regression results are presented in Appendix and we will discuss the graphs of marginal effects of the interaction terms which are both statistically significant and economically meaningful in this section.

(28)

28 Figure 3.1 Time-specific marginal effect of age on financial distress

For the current financial situation, we can find both economically and statistically significant result for the interactions between year dummies and employment status. Figure 3.2 shows the financial situation of self-employed deteriorates strongly between 2008 and 2012. This result might be due to a composition effect, which means that due to the financial crisis it has become very difficult for job-seekers to find a job as employees. People therefore were 'forced' to become self-employed. The financial situation of especially the unemployed worsens between 2011 and 2013, while it remains fairly stable for retirees and employees.

Figure 3.2 Time-specific marginal effect of employment status on current financial situation

(29)

29

For the percentage of households who save, we find interesting outcomes for the interactions between year dummies and employment status, which are both economically meaningful and statistically significant. According to Figure 3.3, decision to save reduces dramatically for the self-employed since 2009 and hence this group of people are hit the most by the crisis. The unemployed, employed and retired remain quite stable. Besides, among all categories, the employed and retired are more likely to save money than other groups of people. In Figure 3.4, the group of the widowed has the most obvious downward trend of saving between 2009 and 2011. After 2011, the never married is the only group which observes a considerable decline in terms of the amount of saving. This may due to the fact that for both the widowed and never married they do not have a partner to share the loss from the crisis and could only reduce their saving. Among all groups, the married keeps stable over time and hence is affected less seriously by the financial crisis.

(30)

30 Figure 3.4 Time-specific marginal effect of marital status on amount of saving

Beliefs

For the beliefs variables, we can notice that the interactive effect between year dummies and child dummies is statistically significant for risk loving, at the 5 percent critical level. In Figure 3.5, family without children becomes less risk averse from 2009 onwards. However, this trend is less obvious in other groups.

Figure 3.5 Time-specific marginal effect of child on risk loving

2 0 0 0 3 0 0 0 4 0 0 0 5 0 0 0 6 0 0 0 L in e a r Pre d ict io n 2006 2007 2008 2009 2010 2011 2012 2013 year3 married divorced widowed never married living together with partner

2 .2 2 .4 2 .6 2 .8 L in e a r Pre d ict io n 2006 2007 2008 2009 2010 2011 2012 2013 year3

(31)

31

Focusing on patience, time-specific marginal effect of marital status is significant at the 5 percent level. The figure below shows that the widowed has a slightly upward trend through the crisis which infers to a higher patience. Such a trend is not visible for the other groups. On the contrary, never married group drops dramatically after 2011 on patience. Moreover, the level of patience is very stable over time for people who are married.

Figure 3.6 Time-specific marginal effect of marital status on patience

Relationship between wealth/saving behavior and beliefs

In order to be consistent with the basic idea of the life cycle model, we limit our dependent variables to focus on the saving behavior. We use OLS regressions to investigate the general relationships among preference parameters and saving behavior and examine how these preference parameters influence Dutch households' saving behavior during the financial crisis. In Table 3, we present OLS results by including income uncertainty, patience, time dummies and other explanatory variables. As both risk aversion and risk loving obtain unexpected signs from the regressions, we exclude two variables from now on by considering both the economic meaning and statistic significance. The complete outcome can be found in Table F.1. We can notice that patience leads to a higher probability of making decisions to save and a higher

4 .6 4 .8 5 5 .2 L in e a r Pre d ict io n 2006 2007 2008 2009 2010 2011 2012 2013 year3 married divorced

widowed never married

(32)

32

amount of saving, statistically significant at the critical level 0.01. For the income uncertainty, the estimated coefficients have a correct sign but insignificantly. Our empirical findings prove that more patient people tend to save more, which is in agreement with the life cycle model. Next, we include the interaction terms between year dummies and income uncertainty or patience to observe the trend during the crisis.

Table 3 OLS regression results with main effects (partial)

Model 1 Model 2 variables percentage of households who save amount of saving patience _0.099*** _902*** (0.006) (63.7) income uncertainty 0.007 141 (0.011) (123)

Significance levels as follows: *** p<0.01, ** p<0.05, * p<0.1. Robust Standard Errors clustered at the household level.

______________________________________________________________________________

Income uncertainty

(33)

33 Figure 4.1 Time-specific marginal effect of income uncertainty on percentage of households who

save

Figure 4.2 Time-specific marginal effect of income uncertainty on amount of saving

(34)

34

Patience

Unfortunately, we do not find any significant results for patience with respect to the interactive effects. The complete regression results are presented in Table F.3 in Appendix.

Consequently, the empirical outcomes show that the higher the patience the more people tend to save, which confirm the findings of the life cycle model. We also present that households are more sensitive to the income uncertainty in 2009.

5. DISCUSSION AND CONCLUSION

Motivated by the previous literature and the current situation, this paper intends to investigate how current financial crisis influences households in terms of wealth status, saving behavior and beliefs. The paper mainly concentrates on the Netherlands from 2006 to 2013 (covering both pre-crisis and pre-crisis periods). Two main research questions are stressed. First, we study how wealth status, saving behavior and beliefs are affected by the financial crisis and which groups of people are of greater impact. Second, based on the life cycle model, we examine whether the relationships between Dutch households' saving behavior and preference parameters change during the crisis.

(35)

35

pattern once the European economy will recover. On the other hand, if beliefs and preferences change, the economic downturn is more likely to have long lasting effects on households. Thus, as this crisis does not change the beliefs much, the empirical results imply that current financial crisis tends to have a more significant impact in the short run. In addition, some background variables also significantly influence wealth/ saving behavior and beliefs. Our findings are very consistent with the previous literature. Moreover, we find that different groups of people indeed are affected differently by the crisis. In particular, the middle aged group has an increasing trend of being financially distressed. The current financial situation deteriorates the most in the group of the self-employed. The self-employed, the widowed and the never married are the groups which reduce their saving more dramatically than other groups. Among all groups, family without children become less risk averse while the never married are less patient during the crisis. Second, in accordance with the theoretical models, more patient people will save more. The saving behavior of Dutch households is more sensitive to income uncertainty in 2009 but does not persist for other years.

(36)

36

APPENDIX

Table A Descriptive statistics for main variables

variable observation mean Std. dev min max

Financial distress 11,872 .26 .44 0 1

Making ends meet 10,616 3.38 .87 1 5

Current financial situation 10,616 3.40 1.00 1 5 Percentage of households who save 10,611 0.68 .47 0 1

Amount of saving 10,228 2.51 1.32 1 8

Expected economic situation 10,100 2.92 .79 1 5

Beliefs

Patience 10,607 4.98 1.21 1 7

Risk aversion 9,947 4.53 2.11 1 7

Risk loving 9,942 2.53 1.55 1 7

Income uncertainty 9,440 .22 .54 0 13.82

Other key explanatory variables

(37)

37

Table B Explanation of dummy variables used in the regressions

Variable Description

Age class

35- If aged below 35

36-45 If aged between 36 and 45

75+ If aged above 75

Child

Child 1 If no child in the household

Underage child in HH If has underage child (age<18) in the household Adult child in HH If has adult child (age>=18) in the household Education level

Low education If no education, special education, primary education and pre-vocational education and other education

Pre-university If pre-vocational education

senior vocational training If senior vocational training or training through apprentice system

Vocational training If vocational colleges university If university education Employment status

Employed If employed on a contractual basis

Self-employed If works in own business, free profession, freelance, self-employed and other occupation

Unemployed If looking for work after having lost job, looking for first-time work, (partly)disabled, unpaid work, keeping benefit

payments, works as a volunteer

Student, homemaker If students and work in own household

retired If retired

Marital status

Married If married or registered partnerships

Divorced If divorced from spouse

Widowed If widowed

Single If never married

Living with partners Living together with partner (not married)

female If gender is female

(38)

38

Table C Correlation Matrix

(39)

39

Table D.1 Regression results with year dummies of wealth and saving behavior

variables Financial distress Making ends meet Percentage of households who save Amount of saving1 Current financial situation Expected economic situation 20072 -0.020* 0.047** 0.002 131 0.086*** 0.026 (0.012) (0.021) (0.014) (154) (0.028) (0.023) 2008 -0.034*** 0.083*** 0.016 367* 0.131*** -0.063** (0.013) (0.024) (0.015) (201) (0.030) (0.026) 2009 -0.044*** 0.102*** 0.008 637*** 0.154*** -0.127*** (0.013) (0.025) (0.015) (219) (0.032) (0.027) 2010 -0.046*** 0.140*** 0.002 276 0.109*** -0.130*** (0.014) (0.027) (0.016) (191) (0.033) (0.028) 2011 -0.020 0.092*** -0.012 501** 0.079** -0.192*** (0.014) (0.027) (0.016) (230) (0.033) (0.027) 2012 -0.029** 0.086*** -0.006 401* 0.085*** -0.269*** (0.014) (0.027) (0.016) (220) (0.033) (0.028) 2013 -0.002 0.053* -0.015 126 0.067** -0.117*** (0.015) (0.028) (0.016) (199) (0.034) (0.029) constant 0.283*** 3.31*** 0.682*** 3,284*** 3.31*** 3.03*** (0.012) (0.024) (0.013) (161) (0.029) (0.023) Observations 11,872 10,616 10,611 10,228 10,613 10,100 Adjusted R-squared 0.001 0.002 0.000 - 0.002 0.013 p-value of F-test year dummies 0.001 0.000 0.625 0.070 0.000 0.000

(40)

40

Table D.2 Regression results with year dummies of beliefs

variables Income uncertainty Risk aversion Risk loving Patience

20071 0.011 0.118** 0.017 0.012 (0.020) (0.060) (0.047) (0.035) 2008 -0.022 0.192*** -0.128*** -0.016 (0.024) (0.065) (0.050) (0.037) 2009 -0.041** 0.388*** -0.231*** 0.018 (0.021) (0.072) (0.052) (0.039) 2010 -0.003 0.257*** -0.308*** 0.048 (0.027) (0.073) (0.051) (0.040) 2011 -0.021 0.386*** -0.339*** 0.080** (0.022) (0.072) (0.054) (0.040) 2012 -0.039** 0.560*** -0.252*** 0.032 (0.019) (0.073) (0.053) (0.041) 2013 0.008 0.599*** -0.225*** -0.009 (0.024) (0.073) (0.054) (0.043) constant 0.233*** 4.21*** 2.72*** 4.96*** (0.017) (0.060) (0.046) (0.034) Observations 9,440 9,947 9,942 10,607 Adjusted R-squared 0.001 0.009 0.006 0.001 p-value of F-test year dummies 0.0126 0 0 0.175

(41)

41

Table E.1 P-value of F-test of the interactions variables Financial distress Making ends meet Current financial situation Percentage of households who save Amount of saving Expected economic situation Interaction of year and

child

0.153 0.457 0.526 0.707 0.401 0.906 Interaction of year and

home ownership

marital status

employment status

age class

education level

0.677 0.173 0.151 0.007 0.248 0.606

Table E.2 P-value of F-test of the interactions

variables Income

uncertainty Risk aversion

Risk loving Patience Interaction of year and child 0.258 0.322 0.011 0.373 Interaction of year and home

ownership

0.634 0.688 0.166 0.162 Interaction of year and marital

status

0.263 0.684 0.261 0.028 Interaction of year and

employment status

0.172 0.207 0.110 0.282 Interaction of year and age

class

0.304 0.402 0.642 0.718 Interaction of year and

education level

(42)

42

Table F.1 Regression results with main variables

Model 1 Model 2

variables

percentage of households who save Amount of saving patience 0.099*** 902*** (0.006) (63.7) income uncertainty 0.007 141 (0.011) (123) ageclass(36-45) -0.024 276 (0.027) (321) ageclass(46-55) -0.078*** -270 (0.027) (315) ageclass(56-65) -0.084*** -456 (0.030) (332) ageclass(66-75) -0.093** -964** (0.039) (380) ageclass(75+) -0.142*** -1,606*** (0.047) (430) Underage child in HH 0.010 -899** (0.040) (451) Adult child in HH -0.040 -544 (0.035) (489) family size -0.031** -31.1 (0.015) (177) female 0.018 -674*** (0.022) (212) pre-university -0.025 270 (0.032) (232) senior vocational training 0.049* 196 (0.026) (232) vocational training 0.056** 925*** (0.022) (225) university -0.010 2,289*** (0.026) (359) self-employed -0.090** 403 (0.035) (492) unemployed -0.196*** -715** (0.031) (318) student,homemaker -0.178*** -72.3 (0.053) (359) retired -0.084*** -579** (0.028) (288) divorced -0.088** -1,064*** (0.037) (375) widowed -0.086** -214 (0.039) (439) single -0.087*** -1,416*** (0.031) (343)

(43)

43 (0.027) (459) home ownership 0.066*** 483*** (0.021) (177) ln_income 0.085*** 2,151*** (0.013) (207) 20071 0.004 95.8 (0.016) (179) 2008 0.040** 494** (0.017) (242) 2009 0.026 522** (0.017) (266) 2010 0.018 -42.1 (0.018) (228) 2011 -0.001 15.4 (0.019) (263) 2012 0.015 36.4 (0.018) (244) 2013 -0.011 -160 (0.019) (227) Constant -0.499*** -21,875*** (0.140) (2,167) Observations 7,855 7,640 Adjusted R-squared 0.134 -

(44)

44

Table F.2 OLS regression results of income uncertainty

Model 3 Model 4

variables Percentage of households who save Amount of saving income uncertainty _-0.011 _-219 (0.021) (200) ageclass(36-45) -0.0004 525 (0.028) (335) ageclass(46-55) _-0.056* _-15.5 (0.029) (333) ageclass(56-65) -0.045 -56.6 (0.032) (357) ageclass(66-75) _-0.051 _-536 (0.042) (400) ageclass(75+) -0.105** -1,218*** (0.050) (4523) Underage child in HH _0.011 _-872* (0.044) (467) Adult child in HH -0.050 -601 (0.039) (500) family size _-0.038** _-105 (0.017) (184) female 0.025 -613*** (0.024) (224) pre-university _-0.038 ₁₇₀ (0.034) (241)

senior vocational training 0.049* 175

(45)

45

living with partner _-0.037 ₂₈₇

(0.029) (484) home ownership 0.087*** 683*** (0.022) (186) ln_income _0.089*** _2,186*** (0.013) (212) 20071 -0.004 -63.2 (0.018) (213) 2008 _-0.047** ₃₂₈ (0.019) (244) 2009 -0.009 213 (0.019) (281) 2010 _0.011 _-71.0 (0.020) (224) 2011 -0.003 -32.5 (0.020) (264) 2012 _0.008 _-69.0 (0.021) (300) 2013 -0.018 -323 (0.020) (226) 2007*income uncertainty _0.031 ₆₇₄ (0.040) (476) 2008*income uncertainty -0.072 494 (0.045) (782) 2009*income uncertainty _0.171*** _1,463*** (0.033) (568) 2010*income uncertainty 0.017 31.8 (0.026) (225) 2011*income uncertainty 0.017 312 (0.033) (365) 2012*income uncertainty 0.011 305 (0.047) (917) 2013*income uncertainty 0.004 531 (0.028) (382) Constant -0.070*** -17,896*** (0.144) (2,151) Observations 7,861 7,646 Adjusted R-squared 0.076 -

p-value of F-test year

dummies 0.036 0.270

p-value of F-test of year

and income uncertainty 0.000 0.097

(46)

46

Table F.3 OLS regression result of patience

Model 3 Model 4

variables Percentage of households who save Amount of saving patience _0.094*** _805*** (0.011) (125) ageclass(36-45) -0.032 3567 (0.024) (283) ageclass(46-55) _-0.088*** _-287 (0.025) (280) ageclass(56-65) -0.094*** -434 (0.028) (294) ageclass(66-75) _-0.102*** _-989*** (0.037) (343) ageclass(75+) -0.141*** -1,594*** (0.045) (397) Underage child in HH _0.017 _-850** (0.036) (406) Adult child in HH -0.052 -620 (0.034) (432) family size _-0.026* _-45.0 (0.014) (154) female 0.019 -710*** (0.021) (197) pre-university _-0.020 _382* (0.030) (217)

senior vocational training 0.045* 162

(47)

47

living with partner _-0.020 ₂₂₂

(0.026) (406) home ownership 0.065*** 440*** (0.019) (163) ln_income _0.081*** _1,960*** (0.012) (1801) 20071 -0.003 75.6 (0.074) (646) 2008 _0.010 ₂₉₀ (0.080) (860) 2009 -0.082 -450 (0.079) (911) 2010 _-0.109 _-612 (0.081) (672) 2011 -0.068 -1,968** (0.079) (1,001) 2012 _-0.009 _-370 (0.078) (717) 2013 0.076 71.0 (0.078) (673) 2007*patience _0.001 _-12.9 (0.014) (136) 2008*patience 0.004 7.52 (0.015) (184) 2009*patience _0.019 ₁₇₆ (0.015) (198) 2010*patience 0.022 98.8 (0.015) (150) 2011*patience 0.012 377* (0.015) (226) 2012*patience 0.003 74.5 (0.015) (162) 2013. *patience -0.017 -64.9 (0.015) (150) Constant -0.445*** -19,415*** (0.139) (2,005) Observations 8,959 8,660 Adjusted R-squared 0.132 p-value of F-test year

dummies 0.257 0.473

p-value of F-test of year

and patience 0.144 0.490

(48)

48

REFERENCES

Afonso, A., Grüner, H. P., & Kolerus, C. (2010). Fiscal policy and growth: Do financial crises make a difference?.Working Paper Series1217, European Central Bank.

Alessie, R., Brugiavini, A., & Weber, G. (2006). Saving and Cohabitation: The Economic Consequences of Living with One's Parents in Italy and the Netherlands. In NBER International

Seminar on Macroeconomics 2004 (pp. 413-457). The MIT Press.

Arrow, K. (1965). Aspect of the Theory of Risk Bearing. Yrjo Jahnsson Foundation, Helsinki. Bachellerie, A. (2011). Household savings behavior in 2010. Banque de France. Quarterly

Selection of Articles 22: 67-78.

Bosworth, B., & Smart, R. (2009). The wealth of older Americans and the sub-prime debacle.

Working Paper 2009-21. Chestnut Hill, MA: Center for Retirement Research at Boston College.

Browning, M., & Lusardi, A. (1996). Household saving: Micro theories and micro facts. Journal

of Economic literature, 1797-1855.

Caballero, R. J. (1991). Earnings uncertainty and aggregate wealth accumulation. The American

Economic Review, 859-871.

Cameron, A. C., & Trivedi, P. K. (2005). Microeconometrics: methods and applications. Cambridge university press.

Carroll, C. D., & Samwick, A. A. (1997). The nature of precautionary wealth. Journal of

monetary Economics, 40(1), 41-71.

Cavasso, B., & Weber G. (2013). The Effect Of The Great Recession On The Wealth And Financial Distress Of 65+ Europeans, in Boersch-Supan et al. Active Ageing and Solidarity between Generations, Berlin: DeGruyter.

(49)

49

Choi, Y., & Kim, S. (2014). Hyperbolic Discounting and Consumer Patience During Financial Crises in Korea.

Coudert, V., & Gex, M. (2008). Does risk aversion drive financial crises? Testing the predictive power of empirical indicators. Journal of Empirical Finance, 15(2), 167-184.

Cruijsen, C. V. D., De Haan, J., Jansen, D. J., & Mosch, R. H. (2011). Household savings behaviour in crisis times. DNB working paper.

Goddard, J., Molyneux, P., & Wilson, J. O. (2009). The financial crisis in Europe: evolution, policy responses and lessons for the future. Journal of Financial Regulation and Compliance, 17(4), 362-380.

Guiso, L., Jappelli, T., & Terlizzese, D. (1992). Earnings uncertainty and precautionary saving.

Journal of Monetary Economics, 30(2), 307-337.

Guiso, L., Sapienza, P., & Zingales, L. (2013). Time varying risk aversion (No. w19284).

National Bureau of Economic Research.

Hurd, M. D., & Rohwedder, S. (2010). Effects of the financial crisis and great recession on American households (No. w16407). National Bureau of Economic Research.

Kapteyn, A., & Teppa, F. (2011). Subjective measures of risk aversion, fixed costs, and portfolio choice. Journal of Economic Psychology, 32(4), 564-580.

Lusardi, A. (1997). Precautionary saving and subjective earnings variance. Economics Letters, 57(3), 319-326.

Lusardi, A. (1998). On the importance of the precautionary saving motive. American Economic

Review, 449-453.

Lusardi, A. (2010). Financial capability in the United States: Consumer decision-making and the role of Social Security.MRRC Working Paper 2010-226.

Pratt, J. W. (1964). Risk aversion in the small and in the large. Econometrica: Journal of the

(50)

50

Reinhart, C. M., & Rogoff, K. S. (2008). Is the 2007 US sub-prime financial crisis so different? An international historical comparison (No. w13761). National Bureau of Economic Research. Skinner, J. (1988). Risky income, life cycle consumption, and precautionary savings. Journal of

Research Master Thesis The impact of the financial crisis on Dutch households