• No results found

MSc Finance Thesis

N/A
N/A
Protected

Academic year: 2021

Share "MSc Finance Thesis"

Copied!
26
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

MSc Finance Thesis

HAPPINESS AND RISKY FINANCAL INVESTMENT:

Evidence from Netherlands

Student No.: S2717026

Student: Yuting Zhou

(2)

Abstract

(3)

1. Introduction

(4)

Lyons, 2010; Lusardi, 2012). It is important to note that individuals’ happiness is as well determined by most of the adopted control variables, which to say the independent variable is endogenous to other regressors. Theoretically speaking, application of ordinary least squares (OLS) to a linear regression model where contains endogenous variables in regressors will cause biased coefficient estimates. To avoid the problems caused by endogeneity, we use an instrumental variable (IV) model by employing two instruments of happiness, namely the quality of family relations and leisure satisfaction. Diener (2002) have proposed that the most salient characteristics shared by the people who reported high happiness are their strong ties to friends and family, and argue for the quality of family relations as a strong genetic component to happiness. For leisure satisfaction, two theories have explored the relationship between leisure and happiness. One is comparison theory, and another one is livability theory. Nawijn and Veenhoven (2013) summarize from these two theories that comparing one’s leisure time and actives to others would have the influence to the extent of individual happiness, and after the meet of basic human needs, more recreational activities would be beneficial to happiness. Also, both Lloyd and Auld (2002) and Spiers and Walker (2009) find empirical evidence that satisfaction of leisure appears to associate positively with happiness. In this paper, we also provide evidence that the quality of family relations and leisure satisfaction are two powerful instruments to happiness.

(5)
(6)

2. Literature Study and Hypotheses Development

Easterlin (2004) investigate the relationship between health and happiness, marriage and happiness, and money and happiness. He finds that happiness at a national level does not increase with wealth once basic needs are fulfilled. This finding is also known as Easterlin paradox, and it is a key concept in happiness economics. Decades later, Stevenson and Wolfers (2008) refute the Easterlin paradox by presenting evidence that people in wealthy nations are not happier than those in poor nations. Since then, researchers in the field of happiness economics focus not only on people’s wealth but also a range of demographic characteristics and the emotional states.

Self-report survey on individuals’ feeling of happiness is the most common approach to measure happiness. The corresponding databases are derived from representative surveys of the same person repeatedly, controlling of unobserved individual-specific characteristics that might well be correlated with individual reporting behavior (Stutzer and Frey,2012). A variety of scales is used to assess happiness, such as the Positive and Negative Affect Scale (PANAS; Watson, Clark, & Tellegen, 1988), the Subjective Happiness Scale (SHS; Lyubomirsky & Lepper, 1999), or the Satisfaction with Life Scale (SWLS; Pavot & Diener, 1993). In the DNB Household Survey (DHS), researchers construct the happiness indicators by application to SHS, taking values from answers to a survey question, “in general, how happy would you say you are—very happy, pretty happy, or not so happy? “, and answers scale from 1(very unhappy) to 5(very happy). The aim of Happiness Economics is to find out the connection between individuals’ happiness and their financial behavior, which is inspired by some findings from psychological experiments that the behavior of happier people differs from that of less happy people. For the financial behavior, the importance of happiness has been verified by empirical studies. Such as, Mogilner et al. (2012) suggest that happiness influence people’s saving behavior. Guiso (2012) indicate that happiness affects individuals’ demand for insurance. Also, evidence for the relationship between happiness and the holding of risky investments can be collected from various experimental research, but the results for it are far from unanimous.

(7)

unhappy households’ higher risk aversion is caused by their less control over expenditures and investments. Isen and Patrick (1983) and Ifcher and Zarghamee (2011) propose the assumption that happier people have decreasing discount rates, so increase their risk aversion. Also, this reverse relationship, to some extent, is consistent to the Mood Maintenance Hypothesis (MMH), which conjectures that people in good moods will be less likely to take more risk. Therefore, it seems reasonable to propose a heuristic prediction is as follows:

There is a reverse correlation between individuals’ happiness and risky asset holding, which is to say happier people are less likely to hold risky financial assets.

In the investigation of the correlation between happiness and holding risk investments, control variables can ensure the outcomes without interference with other influencing factors. Reviewing literature, gender, marital status, education, labor status and wealth are most used control variables. Outside the interplay with happiness, a flouring literature proves the role of these demographic characteristics in individuals’ risky financial decisions. The findings from Charness and Gneezy (2007) illuminate the different attitudes towards risky investments between female and male, and females invest less and more risk averse. Brown et al. (2006) show in their empirical results that the interaction term between years of education and having risky financial investments is positive and highly significant in the US. For the marital status, Leimberg et al. (1989) hold the opinion that as individuals go through family transitions, such as getting married, their financial risk tolerance changes, which turns out to change their financial decisions. However, the findings from both Chaulk (2000) and Chaulk et al. (2003) indicate that there is only a statistically negligible moderating effect of marital status on people’s propensity to risky financial investment. These control variables are empirically proved to be one of the determinants of risky financial asset holding, and meanwhile, both Guven (2009) and Rao et al. (2014) also demonstrate significant correlations between adopted control variables and happiness.

(8)
(9)

3. Empirical model

Happiness is an endogenous variable, and endogeneity is a vital problem when Investigating the causal effect of happiness on the financial decisions. Fehr (2009) indicate that the endogeneity can arise for some possible problems—reverse causality, omitted variables, and measurement error. Besides, application of ordinary least squares (OLS) to a linear regression model where contains endogenous variables in regressors will cause biased coefficient estimates. In the other word, if 𝐸 𝑋#𝜀# ≠ 0, we cannot use OLS directly. Instead, instrumental variables can be used to overcome the problems caused by endogeneity. In this section, we will elucidate our model specification. Firstly, we introduce the IV model, and the general form of IV model is as follows:

𝑦 = 𝛽#𝑋# + 𝜀, 𝐸 𝜀𝜀- = Ω (1)

where 𝑦 is the individuals’ holding of risky investments; 𝑋# is the matrix of regressors, which includes 𝑋/011#2344 and all other control variables, specifically, control variables include gender, education, marital status, anger, anticipation,

monthly income, trust, urban indicator, age, home ownership and labor status; 𝛽# represents corresponding coefficient vector; 𝜀 is the error term. Ω denotes the covariance matrix of errors.

An IV model can be estimated as a two-stage least squares regression. In the first stage, happiness is regressed on the instruments and control variables to find the relationship between the instrument and the independent variable. In the second stage, we isolate the happiness variable from the regression, instead, use the instrumental variables. The selection of instrumental variables should satisfy the exclusion restriction that instruments have an effect on dependent variable (𝑦) only through the independent variable (𝑋/011#2344) but no other control variables. Thus, if the covariance matrix Ω is homoscedastic in the second stage, the estimates of instrumental variables are efficient and consistent. On the contrary, if Ω is heteroscedastic, the standard errors of the estimates are inconsistent, causing by invalid instruments. In our empirical model, as mentioned in section 1, we use the quality of family relations and satisfaction of leisure as instrumental variables. In section 5, we have checked the validity of these two instrumental variables. Firstly, we regress the independent variable with other regressors, to credit the use of IV model. Secondly, we apply Wald test of exogeneity to check for the exclusion restriction. Both outcomes have shown that the weak instrument is not an issue in our case.

(10)
(11)

4. Data

In this paper, we use the LISS panel data. The LISS database is the only one with available information for both the financials of the households, as well as for emotions and personalities variables which are needed in our studies, such as anger, trust and anticipation. And in Table 2, we list all definition and measures of our required variables.

(12)

Holding of risky investment Extremely unhappy 6.12% Unhappy 7.38% Neutral 7.12% Happy 14.82% Extremely happy 17.62% In line with Delis and Mylonidis (2015), we employ anger, trust, and the anticipation for future as the basic emotional states control in our IV model. The aim of the basic emotion states control is to assure the instrumental variables to have an effect on the holding of risky assets only through happiness and not through other emotional states. As for the proxy of anger, we directly extract from the LISS database, based on the agreement (from 0 to 5) to the statement “Get irritated easily”. The indicator of anticipation derives from the answers of the question “Do you expect your financial situation to get better or worse over the coming 12 months?”, and 1 represents much worse, 5 much better. For the proxy of trust, it is measured by the question whether most people can be trusted, and respondents take value from 0(low trust) to 1(high trust). Other than the basic emotions, we also hold a broad range of demographic variables constant to assess the relative relationship of the dependent variable (holding of risky investment) and independent variable (happiness). Those control variables are namely gender, age, education, marital status and urban indicator. Same as many household surveys, the total monetary values of wealth are often missing, because most of the respondents cannot precisely indicate the amount of their total current balances, and numerical values are easy to make up and fake. Instead of employing the individual aggregate wealth, we shorten the time frame and adopt the monthly income as the alternative to represent respondents’ wealth. The data of monthly income can be directly measured by the LISS survey, and there is much less missing data in comparison with the answer of total wealth amount. Besides, we also employ the home ownership as one of the control variables. The labor status is also included into control variables. We use the quality of family relations and the satisfaction of leisure as instruments for happiness. The quality of family relations is measured by the answer from the question “how would you generally describe the relationship with your family?”. The answer is ranging from 0(very bad) to 5(very good). As for the leisure satisfaction, it is constructed from the responses to the question “how satisfied are you with the way in which you spend your leisure time?”. The responses take value from 1(very dissatisfied) to 5(very satisfied).

Table 2 Variable Definitions

Notation Measure

(13)

Risky investment Dummy variable equal to 1 if the respondent possesses one or more of the risky Investments (growth funds, share funds, bonds, debentures, stocks, options, warrants, and so on), and 0 otherwise.

B. Independent Variable

Happiness Indicate to what degree the respondent consider himself/herself happy? Takes values from 0(very unhappy) to 10(very happy) C. Control Variables Trust “Generally speaking, would you say that most people can be trusted?” Dummy variable equal to 1 if the respondents report the higher levels of trust, and 0 with the answers of lower trust. Gender Dummy variable equal to 0 if the respondent is a male, and 1 is a female. Age The age of respondent Age^2 The square of respondent’s age Education Level of education in CBS (statistics Netherlands) categories. Take values 1 (primary school), 2 (VMBO), 3 (HAVO/VWO), 4 (MBO), 5 (HBO), 6 (WO) Marital status The marital status of the respondent. Never married, widow, divorced and separated respondents take value of 0; married respondents take 1. Urban indicator Urban character of the place of residence. Take values from 1 (very urban) to 5 (not urban) Labor status Dummy variable equal to 1 if the respondent has a paid job, and 0 otherwise Anger “Get irritated easily.” Take values from 0 (very inaccurate) to 5(very accurate).

Anticipation Indicate whether the respondents expect their financial situation will get better or worse in the coming 12 months. Takes values from 1(much worse) to 5(much better).

Monthly income(log)

The natural logarithm of respondent’s monthly income.

(14)

Table 3 tabulates summary statistics of key variables from 2013 wave. It shows that our sample consists of 2,875 observations after excluding invalid responses in the entire LISS samples.

Table 3 shows the mean happiness is 7.3 out of 10, which illustrates that on average respondents are happy. And the average risky investment participation rate is 16%. The mean age among entire observations is around 59, and more than 50% respondents are employed. Almost half of respondents consider their financial situation will be better in the coming 12 months. And the mean of trust is 0.6, which implies that most respondents are intended to trust others.

Table 3 Descriptive Data from 2013 Wave

This table reports the number of observations and summary statistics for the main variables of the empirical analysis. All data is from the 2013 wave of the LISS panel database. All variables are defined in Table 2

Variable Obs Mean Std. Dev. Min Max

Risky investment 2,875 0.164 0.371 0 1 Happiness 2,875 7.371 1.352 0 10 Gender 2,875 0.491 0.500 0 1 Marital Status 2,875 0.598 0.490 0 1 Home Ownership 2,875 0.733 0.442 0 1 Urban Indicator 2,875 3.008 1.270 1 5 Education 2,875 3.602 1.499 1 6 Anger 2,875 2.351 0.946 1 5 Trust 2,875 0.691 0.462 0 1 Anticipation 2,875 2.644 0.764 1 5 Labor Status 2,875 0.527 0.499 0 1 Age 2,875 59.52 14.46 22 96 Age^2 2,875 3387.0 1821.6 361 9216 Monthly Income(log) 2,875 7.796 0.726 0 12.12 Family Relation 2,875 3.982 0.773 1 5 Leisure Satisfaction 2,875 3.750 0.661 1 5

Table 4 show the distribution of happiness varied by gender, age, labor status, education, trust and marital status. In the LISS database, happiness is categorical variables taking values from 0 to 10, but in Table 4, it is recoded into five categories, namely extremely unhappy, unhappy, neutral, happy, extremely happy.

(15)

themselves unhappy, while only 2.97% of WO respondents are reported lower levels of happiness (extremely unhappy and unhappy). Respondents in unemployed status are also very unhappy. 3% of employed respondents report the lower values of happiness categories (extremely unhappy and unhappy), while almost 6% of unemployed people, two times of the employed, say that they are unhappy. Marital status matters the happiness too. Unmarried respondents (including never married, widow, divorced and separated) report lower levels of happiness, showing that 6% of unmarried respondents are unhappy, whereas just around 4% of married people feel so. As the factor of age, in Table 4, the simple comparison of means shows a U-shape relationship with happiness. People within 30 to 60 years old are most likely to be unhappy, with 5.64% proportion. By contrast, there is a less proportion of individuals who are younger (below 30 years old) and elder (above 60 years old) feeling unhappy, namely 4% and 3%. Concerning the trust, apparently, the high-trust respondents have reported the higher levels of happiness. 9.34% of low-trust respondents report themselves unhappy, whereas only 2.14% of high-trust ones are distributed in the unhappy groups. Needless to say the distribution of happy respondents, almost 90% of high-trust respondents are happy. Around 70% of low-trust ones consider themselves happy.

Table 4 Distribution of Happiness in 2013

This table shows summary statistics of happiness categories by gender, education, labor status, marital status, age and trust. The numbers are row frequencies, shown as percentages. The original happiness variable is a categorical variable taking values from 0(very unhappy) to 10(very happy), but in this table, we recode it as follows: extremely unhappy (0,1,2); unhappy (3,4), neutral (5,6); happy (7,8); extremely happy (9,10).

Happiness Extremely

unhappy

Unhappy Neutral Happy Extremely

(16)

Age 0-30 0.71% 3.31% 15.84% 69.27% 10.87% 10.80% 30-60 1.78% 3.86% 16.52% 65.05% 12.78% 44.35% 60 above 0.85% 2.33% 12.64% 70.69% 13.49% 44.86% Trust low trust 2.98% 6.36% 21.90% 61.40% 7.36% 30.89% high trust 0.48% 1.66% 11.49% 71.00% 15.37% 69.11%

Table 5 shows the distributions of each happiness level in four specified years. According to this simple comparison, we can see that there is no big difference among the happiness distributions in the four years. Averagely, around 80% of people report the higher levels of happiness in each year, and only around 10% of people consider themselves unhappy. In Chart 1, it provides a clearer and visual summary of the happiness distributions in the four years. The line in this chart represents the happiness distributions. It shows that four lines almost overlap, and illustrate the similar pattern of happiness distributions over years.

Table 6 tabulates the frequencies of risky investment participation rates in four years. It shows that the total participation rates are decreasing from 2007 to 2013. Meanwhile, we can see the comparisons on Chart 1. The bars in Chart 1 represent the distributions of risky investment participations. Obviously, unhappy individuals’ risky investment participations are lower than those of happy people. For happy individuals, the highest participation rate is distributed in 2007. Since then, happy people’s participations present a gradual stepdown trend.

Throughout the years, we can see that happy people appear to more likely to hold risky investments than unhappy people do. On average, more than 40% of happy people hold risky financial assets, whereas less than 15% of unhappy people own risky investments. Besides, based on this simple comparison, the financial crisis seems to affect the individual’ propensity to hold the risky assets, resulting from the fact that the participation rates slump since 2009. Nevertheless, the participation rates listed in Table 6 do not control for the interaction with other influencing factors, so it is not precise to predict the real relationship between happiness and risky asset holding. In the rest of our paper, we adopt a multivariate approach with an IV model to regress data from each year, in order to check the robustness of our finding about the correlation between happiness and risky asset holding over time.

Table 5 Distribution of Happiness

(17)

2007 2009 2011 2013 Extremely unhappy 1.09% 0.99% 1.26% 1.25% Unhappy 3.02% 3.25% 2.49% 3.11% Neutral 13.33% 13.81% 14.77% 14.71% Happy 69.25% 68.28% 68.79% 68.04% Extremely happy 13.31% 13.66% 12.69% 12.89% Table 6 The Risky Investment Participation Rate This table tabulates the risky investment participation rates in different levels of happiness. All numbers are row frequencies, shown as percentages. The original happiness variable is a categorical variable taking values from 0(very unhappy) to 10(very happy), but in this table, we recode it as follows: extremely unhappy (0,1,2); unhappy (3,4), neutral (5,6); happy (7,8); extremely unhappy (9,10). Risky Investment 2007 2009 2011 2013 Extremely unhappy 5.88% 2.27% 5.54% 6.12% Unhappy 8.50% 6.30% 9.71% 7.38% Neutral 9.91% 9.74% 9.47% 7.12% Happy 19.22% 16.98% 14.98% 14.82% Extremely happy 21.15% 19.24% 17.37% 17.62% Total 17.79% 15.81% 14.21% 13.71% Chart 1 Frequency Chart This chart displays a visual summary. All data are from Table 9 and Table 10. X-Axis is composed of five categories of happiness. The bars represent the frequencies of risky investment participations, and its corresponding Y-Axis is on the left. The lines represent the frequencies of different categories of happiness, and its corresponding Y-Axis is on the right. Legends are shown on the right side of this chart.

0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 0.00% 5.00% 10.00% 15.00% 20.00% 25.00% Extremely

(18)

5. Empirical Results

We examine the determinants of happiness by regressing control variables with ordinary least squares (OLS). This estimate, on the one hand, can serve as the multivariate approach to study the distribution of happiness. On the another hand, the regression outcomes can credit the validity of our IV model by proving the endogeneity of happiness. The results are shown in Table 5.

As for the demographic characteristics, we find that variables of gender and marital status positively influence happiness. That is to say, females are happier than males, and married people are happier than unmarried ones. The effect of age is significantly negative. Nevertheless, the coefficient of age-square is positive and significantly different from zero. Based on these two outcomes, we conjecture that the effect of age is not linear and might be a U-shape relationship. Table 7 also shows that the level of urbanization does not matter individuals’ happiness. The coefficient of education is positive and significant, denoting that the well-educated respondents are more likely to be happy. In case of individuals’ home ownership, our results suggest that respondents who own their dwelling report the higher level of happiness. The amount of monthly income also positively influences the degree of self-report happiness. The higher income, the higher level of happiness they report.

We have introduced two instrumental variables in our empirical model. Table 7 demonstrates that both the quality of family relations and leisure satisfaction coefficients are positive, and they are statistically significant in explaining happiness. Furthermore, we include the basic emotion variables in the IV model. Table 7 shows that the coefficients of anger, anticipation and trust all are highly significant and positive. Their pronounced correlations with happiness validate the necessity of the basic emotions control because the basic emotions approach enables to assure the instruments of happiness not to be interfered by other emotion states.

(19)

Anticipation 0.328*** Labor status 0.217*** Age -0.046*** Age^2 0.004*** Monthly income(log) 0.214*** Leisure satisfaction 0.130*** Family relation 0.276*** R-squared 0.20 Number of observation 2,875 Table 8 shows the results of our main investigation—impact of happiness on risky asset holding. In Table 8, column I presents the regression without interaction terms. In this regression, the coefficient of happiness represents the cumulative effect of happiness on individuals’ ownership of risky investments, and it is negative and statistically significant at 5% level. In a probit model, the negative sign is suggestive that happier people are less likely to hold risky investments. Compared to the findings from existing literature, our results are in line with Guven (2009) but contradict with Rao et al. (2014). Moreover, in Column I, the results show that the parameter of gender is negative and significant at 1% level, indicating that male is more likely to hold risky assets than female. Also, our results suggest that the marital status does not impact the individuals’ propensity to hold risky investments, which is contrary to Leimberg et al. (1989) that marital status changes individuals’ risk-taking behavior. The variable of home ownership has displayed a significant correlation with participation in the risky financial market. People who own their home have the higher possibility to hold risky assets. As well, employment has a positive effect on individuals’ holding of risky investments. For the coefficient of trust, it is reported positive and highly significant. Individuals characterized by higher levels of trust have a higher probability of holding risky assets.

(20)

is still positive and economically significant among women. For high-trust individuals, who belong to the group with high participation rates in the risky financial market, their propensity to risky assets are also proved to be determined by happiness, and the impact is positive. The group of employed individuals, which is also with the high participation rates, its interaction coefficient with happiness is not statistically significant. Above all, the disaggregated analysis implies that in subgroups of female and high-trust respondents, the happiness positively affects their holding of risky investments, whereas for employed people, changing their happiness level does not significantly change their propensity to hold risky assets.

Additionally, in the lower part of Table 8, it reports the Wald test of exogeneity and the first-stage results. The Wald test of exogeneity rejects the null hypothesis of no endogeneity, and the correlations between happiness and the two instrumental variables are reported positive and significant. That is to say, our results prove the evidence that the presence of weak instruments is not an issue in our empirical model. Besides, as the results are shown on the first stage, the coefficients of basic emotions all are statistically significant, indicating that capturing basic emotions in the first stage enables to improve the efficiency of the instrumental variables (Delis and Mylonidis,2015).

In summary, our findings support our proposed hypothesis that there is a negative relationship between happiness and individuals’ propensity to hold risky investments, implying happiness itself buffers individuals’ propensity to hold risky financial assets. Furthermore, though the aggregate effect of happiness is negative, in three given subgroups, the disaggregated effects of happiness are reported positive, and two of them are statistically significant. In the group of high trust, the happier are more willing to hold risky investments than the less happy. This finding rejects the hypothesis that the impact of happiness fades out in high-trust clusters. In the subgroups of female, the happier they are, the more likely they own risky investments, which also declined our hypothesis that women would not be influenced by happiness on risky asset holding. Among employed people, the results from 2013 wave show that their holding of risky investments is not determined by happiness. The disaggregated analysis, to some extent, confirms the importance of happiness effect on risky asset holding. Both the aggregate effect and disaggregating effects of happiness are significant in explaining individuals’ risky asset holding.

Table 8: Effect of Happiness on Risky Asset Holding in 2013

In this table, both Column I and Column II reports the marginal effects of the IV probit regression on the dependent variable—holding of risky investment, and the independent variable—happiness, which is instrumented with the quality of family relations and leisure satisfaction. Regression in column II includes three more given interaction terms. All variable definitions are provided in Table 2. All regressions are performed with robust standard errors.

(21)

Risky Investment I II Happiness -0.136** -0.410** Gender -0.286*** -1.611** Marital status -0.028 0.041* Home ownership 0.386*** 0.357*** Urban indicator 0.027 0.021 Education 0.168*** 0.171*** Anger -0.031 -0.037 Trust 0.118*** -2.133*** Anticipation 0.191*** 0.142*** Labor status 0.197** -0.522 Age 0.032*** 0.028* Age^2 -0.001* -0.001 Monthly income 0.110 0.093 High Trust*happiness 0.333*** Female*happiness 0.184** Employed*happiness 0.100 First-stage results Anger -0.132*** -0.043*** Anticipation 0.276*** 0.104*** Family relation 0.140*** 0.078*** Leisure satisfaction 0.281*** 0.083*** Wald Test of exogeneity 6.28*** 6.28*** Number of observation 2,890 2,890 Robustness Test In this section, we employ the data from four different waves, in order to check the robustness of our findings over time. The wave selection is up to the collected periods from the LISS. Since there are only four waves of the asset survey in the LISS, we use the data from waves of 2007, 2009, 2011 and 2013.

(22)

happiness turns to insignificant among employed people. As to discover the rationale for this change, we compare the unemployment rates among four years. In Appendix 1, it shows that the unemployment rate in 2013 is the highest and much higher than the other three years. According to the findings from Di Tella et al. (2003), they point out that the high aggregate unemployment rates have a negative effect on people who are even currently with a paid job. Our findings support the hypothesis that during the period of higher unemployment rates, the interaction of happiness with employed people is different from that in lower unemployment rates, by showing that the impact fades out in the time of serve unemployment rate. In case of the effect of happiness for high-trust people, Table 9 shows that only the interaction term in 2009 is reported negative, while in other three years, they are positive. However, the negative effect in 2009 is insignificant. Taken together, our findings indicate that the happier high-trust people are more likely to hold risky financial assets than the less happy high-trust people do. Moreover, our results demonstrate that among women, the effect of happiness on their holding of risky assets are positively and pronounced over four years.

In summary, the regression outcomes in Table 9 highlight the important role of happiness in explaining individuals’ propensity to hold risky assets. As the aggregate effect of happiness, they are reported negative over four investigated years. For the disaggregated happiness effects in three subgroups, most of them are positive and economically significant. Above all, the robustness test has verified our previous finding in even stronger manner, proving that the impact of happiness is the essential determinant of individuals’ propensity to hold risky assets, and the relationship between aggregate happiness and the holding of risky assets is negative.

Table 9 Relation Between Happiness and Risky Asset Holding Over Years

(23)

6. Conclusion

By using the data from the LISS, this paper examines the relationship between happiness and individuals’ risky financial assets holding. We have three major findings. First, we find that happiness affects individuals’ investing behavior regarding holding risky investments. In accordance with Mood Maintenance Hypothesis, happy individuals are less likely to hold risky assets because they do not want to undermine their happy feeling. Second, the disaggregated analysis has verified the robustness of our core findings by showing that happiness also significantly affects individuals’ holding of risky assets in different groups. Besides, our results also outline the difference between aggregate happiness effect and disaggregated effects on subgroups.

In this paper, we employ the probit model with continuous endogenous covariates. On one hand, the application to instrument variables enables to lessen the problems caused by the endogeneity. On the other hand, the OLS can no longer produce the best linear unbiased estimator (BLUE) with a binary dependent variable. Arguments are given for the rejection of linear regression that the statistical tests for linear analysis are inappropriate with a binary dependent variable, and linear coefficients can be meaningless when a predicted probability fall outside the range 0 to 1 (Hellevik, 2007). Instead, a probit model is common to regress with a binary dependent variable by applying probability distribution functions (Park, 2009). In a probit model, we report the marginal effects of happiness. As the results, the happiness explanation of individuals’ propensity to hold risky assets provides useful insights on the observed discrepancies in individuals’ financial investment. The disaggregated analysis helps to dig deeper into the role of happiness in financial decisions across character-based clusters. We have studies three clusters, the female is from the lower participations group, and both employed and high-trust individuals belong to the higher participations group. Our findings suggest that the effect of happiness are pronounced in these three groups, drawing the conclusion that happiness plays a key role in predicting the observed discrepancies in individuals’ risky asset holding. There are some limitations of our paper. In the robustness test, we use data from four waves of the LISS database. Three of them are in the period of financial crisis. Our results have shown that in 2009, the aggregate effect of happiness is insignificant, and the disaggregated effect in high-trust clusters turns to negative. These changes might be caused by the negative effect of the financial crisis. However, because of the limited data availability, this paper cannot confirm the impact on the happiness effect regarding individuals’ financial decisions making.

(24)
(25)
(26)

Referenties

GERELATEERDE DOCUMENTEN

Besides this, when looking at the regression results, statement 1, “I think it is more important to have safe investments and guaranteed returns, than to take risk to have a

IPO underpricing is lower when options are involved in the gold mining sector in Australia (Dimovski and Brooks, 2008). They find significant lower underperformance when

Furthermore, when the legal system dummy is interacted with cash in the OLS regression, firm profitability of civil-law firms is more negatively related at the

Graph IIX presents the monthly compounded portfolio outperformance (underperformance) of the market based on a strategy, which combines stocks which are momentum winners and have a

Excessive optimism as an indicator for overconfidence in this thesis, is tested by making an estimation of the economic climate which is subtracted from the subcategory of

Lagged NPL is impaired loans over gross loans at time t-1, lagged reserve ratio is the loan loss reserves over impaired loans at time t-1, Slope EU/US is the yield curve

Despite design features not having a significant effect on bank and systemic risk for a total period, the effects during a crisis might be significant and

Table 8: The effect of the four components of Corporate Social Responsibility on Corporate Financial Performance as measured by return on assets for European companies from the