- 1 -
The effect of happiness on stock market participation
Author: B.J.A. Oude Avenhuis1 Student number: S2545675 Supervisor: dr. C. Laureti Master Thesis University of Groningen 12th of June 2019 Abstract
Using data from the Dutch Household Survey from De Nederlandsche Bank in a period ranging from 1993 to 2018, this paper examines the relationship between happiness and stock market participation. This while controlling for a broad range of control variables known to influence stock market participation. The main results of this study show that happiness has a negative impact on stock market participation. These results remain valid after performing various robustness checks, such as clustering our data on household level. This paper does not find evidence that people in a happier country participate more in the stock market than people from a less happy country.
Keywords: Stock Market Participation, Household Finance, Happiness
- 2 -
Acknowledgements
I would like to express my gratitude to dr. Carolina Laureti for the support and engagement she gave throughout the process of writing this thesis. Her relevant feedback and guidance have been stimulating in order to accomplish this paper.
- 3 -
1. Introduction
Individuals have become more and more active within financial markets. In the last couple of years, researchers have examined reasons and explanations for individuals to participate in the stock market. Mostly, these explanations are based on microeconomic variables, such as prices, quantities, wages, etc. Most often however, researchers do not take into account emotional factors, such as confidence or happiness. This opens a new field of research and makes us question how such emotional factors can influence stock market participation for individuals. Nowadays, this field of research increasingly catches attention from researchers, who stress the importance of psychological factors (other factors than classical microeconomic variables) to understand participation in the stock market. This paper tries to contribute to this field of research by investigating the effect of happiness on stock market participation.
In the past, some determinants of stock market participation already have been identified by means of academic research. For example, Vissing-Jorgensen (2002) found that stock market participation greatly increases with family wealth. Furthermore, Van Rooij et al. (2011) found that financial illiterate are less likely to participate in the stock market than their financial literate counterparts. In addition, other studies have been performed on reasons for individuals to participate in stock holding as well. We can think of Rosen et al. (2004), who related stock market participation to health, or Hong et al. (2004), who came with a relationship between social interaction and stock market participation. Overall, participation in the stock market sometimes is difficult to understand. Researchers often refer to this phenomenon as the “participation puzzle” or as the “equity premium puzzle”. People who can understand this participation puzzle might attract a tremendous amount of new participators to the stock market.
The literature highlights multiple emotional reasons for individuals to participate in financial markets. Among those, Rao et al. (2016) performed a study on happiness and stock market participation within the Chinese market. They used data on households, from the China Finance Household Survey (CFHS). The main findings of their study are that household’s propensity of investing in stocks or mutual funds are strongly associated with happiness. Chinese households who are happier invest more in the stock market. Furthermore, emotions are an important driver in investment behavior (Van Winden et al. 2011), which is strongly related to stock market participation.
- 4 -
Netherlands will give us insight whether people in a happier country participate in the stock market even more. According to the World Happiness Report, The Netherlands is placed 5th
on a ranking out of 156 countries with a score of 7.488 out of 8. So, average happiness in The Netherlands is on a high level worldwide. Whereas China is placed 93th, with a score of 5.191 out of 8, so a much lower score than The Netherlands. Ever since 2012, this World Happiness Report (Sachs et al., 2019) comes out yearly and measures well-being of people all around the world. It shows that quality of people’s lives can be assessed coherent, reliable and valid with the use of different subjective well-being measures. The report captures these features as “happiness”.
Using data from the Dutch Household Survey (DHS) from De Nederlandsche Bank (DNB) this study investigates the relationship between happiness and stock market participation of individuals within The Netherlands. The DNB started this household survey in 1993 and repeated it yearly. Each year approximately two thousand households participate in the survey and to participate, household individuals at least have to be at the age of 16 or older. Our dataset contains all 26 years the survey is collected.
This study contributes to the literature on happiness and stock market participation in different ways. First of all this study contributes to a better understanding of the role emotions (happiness) play in stock market participation. We do so by researching a large panel dataset of 26 years, which is new when considering the effect of happiness on stock market participation. Furthermore, we get insight in whether people a happier country participate in the stock market even more. Lastly, our insight in the role happiness plays in stock market participation might help us better understand the stock market participation puzzle.
- 5 -
2. Literature Review
This section provides an overview of existing literature in the field of both happiness and stock market participation. First, we provide determinants of stock market participation as well as describing the well-known concept of the “participation puzzle” in section 2.1. Thereafter we illustrate the relationship between happiness and stock market participation that already is known in section 2.2. Then we focus on previous research, both on stock market participation and happiness using the DHS survey from De Nederlandsche Bank in section 2.3 and lastly we build our hypothesis in section 2.4.
2.1 Stock market participation
Stock market participation has a direct effect on the equity premium. Therefore, it is important to understand the determinants of stock market participation for solving the equity premium puzzle (Moskowitz and Vissing-Jørgensen, 2002). The equity premium puzzle describes the higher historical real returns of stocks over government bonds. Moskowitz and Vissing-Jørgensen (2002) also look at why individuals are willing to make investments and similar to our study, they use household data to assess whether or not we can explain individuals’ investment behaviour.
A first interpretation of the literature makes us wonder why there are so many households that do not hold stocks either directly or indirectly through mutual funds or pensions funds. Guiso and Sodini (2013) found that only half of the U.S. households participate in the stock market. Furthermore, in Austria, Italy, Spain and Greece participation rates for households in the stock market are even below 10%. Non-participation is actually a puzzle because you would expect that households should hold some stocks when they have expected utility preferences if their non-stock income is uncorrelated with stock returns (as long as the equity premium is positive). Of course, this same theory holds for individuals within the stock market, since we assume a basic economic principle that individuals want to maximize utility. The first ones to point out this non-participation puzzle were Haliassos and Bertaut (1995). In addition, when households do not hold stocks, stock returns have zero covariance with their marginal utility, so they should be risk-neutral with respect to a small additional stock position. Therefore, holding zero stocks cannot be optimal (Beshears et al., 2018) and we wonder why participation rates in the stock market are this low.
- 6 -
Furthermore, participation rises with wealth; this is consistent with the aspect of fixed costs. Hagerty and Veenhoven (2003) argue that greater wealth goes with greater happiness. So greater happiness might indicate more participation in the stock market. In addition, the findings of Vissing-Jørgensen are also in line with the findings of Briggs et al. (2015), they performed a study on Swedish lotteries and found that winning $150.000 increases stock market participation by 12 percentage points among those not previously participating. In addition to Briggs et al. (2015), Lindqvist et al. (2018) performed a similar research where they linked lottery winnings in the Swedish market to life satisfaction. Veenhoven (1991) argues that happiness is relative and that lottery winners are not happier than for example paralyzed accident victims. Relating happiness to wealth (Hagerty and Veenhoven, 2003), lottery winnings to happiness (Veenhoven, 1991) and wealth to stock market participation (Briggs et al., 2015) makes us wonder whether there is a relationship between happiness and stock market participation as well.
So, wealth is related to stock market participation (Vissing-Jørgensen, 2004; Brigges et al., 2015; Hsu, 2012). However, Guiso and Sodini (2013) argue that among the top 5% of the wealth distribution, more than 65% of Austrian, Spanish and Greek households hold no stocks. Recall that household participation rates in the stock market are below 10% for these countries. These findings of Guiso and Sodini (2013) might question the findings of Vissing-Jørgensen (2004) that fixed costs are likely to be an explanation for non-participation in the stock market.
In addition to wealth, socioeconomic status also explains non-participation in the stock market. People with low socioeconomic status are less likely to participate in the stock market (Kuhnen and Miu, 2017; Das et al., 2017). Both studies are motivated by neuroscience research on how adversity affects the brain’s response to subsequent outcomes. A study by Gerdtham and Johannesson (2001), on the effect of happiness on socioeconomic variables argues that there are several socioeconomic indicators affecting happiness. The link between these three studies makes us question if there exists a relationship between happiness and stock market participation itself, since social relations are a necessary cause for happiness (Diener and Oishi, 2005). Furthermore, Hong et al. (2004) argue that households who are more social are more likely to invest in the stock market. Again, the link between happiness and stock market participation seems reasonable, however is not directly related to each other by empirical evidence.
- 7 -
with findings about expectations people have in the stock market. Hurd et al. (2011) and Kézdi and Willis (2011) both performed similar studies and found that survey respondents who report higher expectations for stock market returns are more likely to participate.
Lastly, stock market participation is also related to knowledge. For example, Van Rooij et al. (2011) state that there is a positive relationship between financial literacy and stock market participation. They use data from the Dutch Household Survey from De Nederlandsche Bank, the same dataset that we use in this study. In addition to Van Rooij et al. (2011), Cole et al. (2014) states that an additional year of education increases the probability of stock market participation by 4 percentage points. Not only education but also IQ is positively correlated (after controlling for income, wealth, age, occupation and family effects) with stock market participation (Grinblatt et al., 2011). By contrast, a study performed by Veenhoven and Choi (2012) states that there is no correlation between IQ and happiness on an individual level. In conclusion, wealth, socio-economic status, optimism etc. are all determined to be important drivers of stock market participation. However, an emotional factor such as happiness might also be an important driver. Therefore, this study examines this relationship, helping us better understand the equity premium puzzle on which stock market participation has a direct effect (Moskowitz and Vissing-Jørgensen, 2002).
2.2 The relationship between happiness and stock market participation
- 8 -
to address the potential endogeneity of happiness. Furthermore they created three instruments as measures of happiness: a happiness index of a city, ecological environment and leisure entertainment. The results show that all these three instruments are significant which denotes that their estimation of the relationship between happiness and stock market participation does not suffer from a weak instrument problem.
Rao et al. (2016) try to explore mechanisms behind the correlation between happiness and stock market participation. Happy households might be less risk averse, they might be more optimistic and they might be more trusting. Rao et al. (2016) created proxies to test whether these mechanisms are an issue and they found that risk tolerance and optimism are of no influence between happiness and stock market participation. Furthermore, trust (or social capital) is also of no influence to the relation between happiness and stock market participation (Rao et al., 2016). These findings are conflicting with those of Hurd et al. (2011) and Kézdi and Willis (2011), who state that stock market participation is affiliated with optimism and trust.
2.3 The Dutch Household Survey in practice
To study whether or not stock market participation in the Netherlands is influenced by happiness, we use the Dutch Household Survey (DHS) from De Nederlandsche Bank. The DHS survey has been used for multiple studies already. The next three paragraphs provide an overview of some previous performed studies in a similar field of study, using the DHS survey as well. Furthermore, we assess the methods that have been used to perform these studies. In the end, this might give a well thought view on the possibilities there are when using this dataset.
Firstly, Hurd et al. (2011) performed a study on the effect of households’ stock market expectations and whether they participate in the stock market or not. Similar to this study, stock market participation is measured. Hurd et al. (2011) study this relationship with the use of panel data in a period of two years. In particular, they state that household stock market participation is a puzzle, just as we described in this section above as the equity premium puzzle or the non-participation puzzle. The main finding of their study is that there exists a correlation between households’ stock market expectation and households’ stock market participation. Hurd et al. (2011) state that on average the Dutch population holds a rather pessimistic view about the stock market, which could be a reason for the low participation rate in the stock market, explaining the non-participation puzzle.
- 9 -
time jobs and has no effect on women’s happiness. In their study, happiness is the dependant variable, which they study by using a linear model. Furthermore, the same as Hurd et al. (2011), they use panel data on the Dutch Household Survey consisting of multiple waves. An important drawback Collewet and Koning (2011) mention is that happiness is self-rated, and those whose happiness does not change over time have to be dropped out of the estimation sample. Collewet and Koning (2011) state that this of course leads to fewer observations but also might lead to selection effects. They argue that people who rate themselves happy are less likely to experience a change in their level of happiness over time than people who are closer to extremes. It is important to keep this insight in mind, since our study also uses panel data to evaluate the effect of happiness.
Thirdly, in addition to the previous two mentioned studies, various papers by Van Rooij appear. For example, the relationship between financial literacy and stock market participation (Van Rooij et al., 2011). In this research, they do not make use of panel data, however they take a closer look at one specific year (2005) in the survey and the model they use in their study is based on a linear probability model (OLS). Compared to Van Rooij et al. (2011), Guiso et al. (2008), also does not make use of panel data in their model. They take the 2003 wave of the Dutch Household Survey to estimate their model and find that trust positively correlates with stock market participation.
2.4 Hypothesis building
Although there has been research on the relationship between happiness and stock market participation before, this previous research only focused on Chinese households. Rao et al. (2016) studied this relationship and found that happier households in China participate more in the stock market. According to The World Happiness Report, people in The Netherlands are on average happier than people in China (Sachs et al., 2019). Therefore, we hypothesize that people in The Netherlands would participate in the stock market even more.
Besides the results of The World Happienss Report, it is interesting to study happiness and stock market participation in The Netherlands since The Netherlands and China are both very different countries. For example when taking into account risk-taking behavior. China is much less risk-averse than countries that have a more individualistic society (Hsee and Weber, 1999), such as the Netherlands2 (Hofstede, 2011). Risk-taking behavior and stock market participation positively correlate to each other (Vissing-Jørgensen and Attanasio, 2003). Furthermore, Helliwell et al. (2009), state that differences across countries, cultures and regions are all different factors linked to happiness.
- 10 -
To get a better understanding of the role emotions play in stock market participation we consider the role happiness plays. Understanding this role helps us better understand the stock market participation puzzle and might bring us one-step closer in solving it. Multiple microeconomic variables have been linked to stock market participation already, such as financial literacy, trust, knowledge, status, wealth, etc. (Gerdtham and Johannesson, 2001; Diener and Oishi, 2005; Hagerty and Veenhoven, 2003). However, happiness is an interesting emotional aspect that might influence stock market participation as well. Therefore, we come up with the following hypothesis:
H1: Happiness will have a positive effect on stock market participation.
3. Data
This section describes the dataset used in this paper. Furthermore, we describe our dataset construction and present key variables used in the empirical analysis. Lastly, we present descriptive statistics of our dataset.
3.1 Data source
- 11 -
3.2 Data set construction
In order to construct the final dataset, a unique identification number is created for each household individual. This number consists the number of the household as well as the number of the individual within the household. This unique identification number is used to merge all five questionnaires for each single year the survey is distributed. Thereafter, all single years are appended using the same unique identification number in order to arrive at the final dataset that consists of 26 years (survey’s). This dataset contains 131.438 single observations on 32.900 individuals. So, we have an unbalanced dataset since not every individual is observed for all 26 years the survey was distributed.
As a final step, we remove outliers from our dataset. Individuals who reported an age of 15 or lower (or 95 or older) are dropped from the dataset. In the first case, ages of 15 or lower are not allowed to participate in the survey, since the minimum respondent age is 16, so these participants where either not of interest or filled in an incorrect age. In the second case, some participants reported ages of for example “2018”, which does not fit to an actual appropriate value, therefore these observations are dropped as well. Furthermore, observations who reported a negative income are also dropped. Lastly, observations who reported “Don’t know” answers on happiness are dropped as well, since these answers correspond with values of -9 and are seen as outliers in our dataset where happiness is ranked from 1 to 5. This caused 316 observations to be dropped. After organizing our dataset and dropping these outliers, the dataset ultimately contains 19,443 observations.
3.3 Variable construction 3.3.1 Dependant variable
- 12 -
- 13 -
Table 1. Distribution of stock market participation
(1) (2) (3) (4) (5) (6) (7) (8)
Total Bonds MutualFunds Shares Put1 Put2 Call1 Call2
Freq Freq Freq Freq Freq Freq Freq Freq
Stock market participation (Percent) (Percent) (Percent) (Percent) (Percent) (Percent) (Percent) (Percent)
No 50,813 61,152 54,741 56,979 62,803 62,785 62,515 62,699 (80.61) (97.02) (86.85) (90.40) (99.64) (99.61) (99.18) (99.47) Yes 12,219 1,880 8,291 6,053 229 247 517 333 (19.39) (2.983) (13.15) (9.603) (0.363) (0.392) (0.820) (0.528) Number of individuals 6,831 6,831 6,831 6,831 6,831 6,831 6,831 6,831 Total 63032 63032 63032 63032 63032 63032 63032 63032
- 14 -
3.3.2 Independent variable
The independent variable in our model is happiness. The DHS survey directly asks participants to what extent they consider themselves a happy person by means of the following question: “All in all, to what extent do you consider yourself a happy person?” The answers are on a one to five scale ranging from very happy to very unhappy. The following answers appear: 1. Very happy, 2. Happy, 3. Neither happy nor unhappy, 4. Unhappy, 5. Very unhappy and lastly -9. Don’t know. Table A2 in the appendix provides an overview of our independent variable question from the DHS survey. In order to deal with outliers, the -9 answers are dropped. This corresponds to a total of 316 droppings. Frey and Stutzer (2002) state that happiness can be captured and analysed because people can be asked how satisfied they are with their lives. They state that it is a sensible tradition in economics to rely on the judgement of the persons directly involved because individuals are the best judges of themselves. Frey and Stutzer (2002) measure happiness in two forms, evaluated happiness and experienced happiness. Since this paper uses an independent variable on happiness that is directly asked to participants themselves, we compare our measurement to the evaluated measurement form of Frey and Stutzer (2002). In order to evaluate the results of our analysis we transformed the data on happiness where a value of one corresponds to “very unhappy” and a value of five to “very happy”. Initially these values where vice versa. Table 2 presents the distribution of happiness and shows us that happiness is skewed where most of the individuals in our dataset consider themselves as “Happy” (64.07%).
Table 2. Distribution of Happiness All in all, to what extent do you consider yourself a
happy person?
Freq. Percent Cum.
very unhappy 139 0.24 0.24
unhappy 634 1.09 1.32
neither happy nor unhappy 8384 14.36 15.69
happy 37397 64.07 79.75
very happy 11819 20.25 100.00
3.3.3 Control variables
- 15 -
questions from the DHS survey are shown that are of interest for the construction of our set of control variables.
Agarwal et al. (2009) found that young and elderly persons make worse financial decisions compared to middle-aged people. Therefore, this study imposes a control variable on age. To account for outliers, the minimum age is set at 16 and the maximum age is set at 95. Furthermore, difference in gender also has effect on stock market participation. According to Almenberg and Dreber (2015), a gender gap arises when looking at stock market participation, where men are more likely to participate in the stock market then women. To account for this effect we create a dummy variable for gender where observations can have the value 1 for men and 2 for women.
Marital status is used as a control variable based on multiple research (Kumar, 2009; Grinblatt et al., 2011). Grinblatt et al. (2011) discovered a relationship between stock market participation and marital status. Therefore, a dummy variable is introduced where 0 indicates not married and 1 indicates married. To indicate whether someone is married we used the following two answers from the DHS question on marital status: “married or registered partnership (including separated), having community of property” and “married or registered partnership (including separated), with a marriage settlement”.
Furthermore, we add a dummy whether or not the participant has children, which is in line with previous research of van Rooij et al. (2011) and has strong theoretical and empirical backing in the literature. Based on research of Heaton and Lucas (2000) we add a dummy variable whether or not the participant is occupied or not. We create a dummy variable indicating 1 when occupied or 0 otherwise. We assess a participants’ work status as occupied (working) when one of the following answers is given: “employed on a contractual basis”, “works in own business” or “free profession, freelance or self employed”. A control variable based on income is added based on Guiso et al. (2008), where income is on an increasing scale and observations with negative income are dropped.
Another control variable is self-assessed health status. Health status is considered to be an important factor for the probability of owning different types of financial assets (Rosen and Wu, 2004). They found evidence that poor wealth is associated with lower stock market participation. To account for this relationship, self-assessed health status on a scale from 1 to 5 is used as a control variable. In order to ensure that higher self-assessed health status corresponds with a higher value, we transformed the data so in the end a higher health status is associated with a higher value.
- 16 -
different educational system in The Netherlands), which was asked on a scale of 1 to 13. So, to create a control variable for education we have to merge the two different questions into one. Table A5 in the appendix provides an overview in which can be seen which questions where merged together in order to arrive at our 0 to 7 scale. People who followed none or “other” education are given the value 0. In that case, education is an increasing control variable where a higher value corresponds to higher education followed.
3.4 Descriptive statistics
Table 3 provides descriptive statistics for all variables used in this paper, for the total dataset. Table B1 in the appendix provides descriptive statistics for all variables used in this paper, for the total dataset from 1993 until 2018, separated by year. A closer look to table B1 (appendix) shows us that happiness does not suffer from large deviations over time.
Taking into account our dependent variable, stock market participation, the descriptive statistics are given both for the created end variable as well as the different items that where used to construct this variable. Table 3 shows us that stock market participation has a mean value of 0.194, indicating that 19.4 percent of the respondents participates in the stock market. Taking into account our independent variable, happiness, we see that happiness equals 4.030 measured on a 1 to 5 scale, for our total dataset. This indicates that on average people consider themselves as “Happy”. When combining the results from table 2 and table 3, these statistics are confirmed since most people consider themselves to be happy according to table 2 as well.
- 17 -
Table 3. Descriptive Statistics for dependent, independent and control variables.
(1) (2) (3) (4) (5) (6) (7)
VARIABLES N mean median sd min max skewness
Dependent variables MutualFunds 63,032 0.132 0.132 0.338 0 1 1 Bonds 63,032 0.0298 0.0298 0.170 0 1 1 Shares 63,032 0.0960 0.0960 0.295 0 1 1 put1 63,032 0.00363 0.00363 0.0602 0 1 1 put2 63,032 0.00392 0.00392 0.0625 0 1 1 call1 63,032 0.00820 0.00820 0.0902 0 1 1 call2 63,032 0.00528 0.00528 0.0725 0 1 1
Stock market participation 63,032 0.194 0.194 0.395 0 1 1
Independent variable
Happiness 58,373 4.030 4.030 0.640 1 5 5
Control variables
Gender 131,307 1.496 1.496 0.500 1 2 2
Health_Status 61,747 3.913 3.913 0.733 1 5 5
Income 30,399 48,297 48,297 63,467 0 6.800e+06 6.800e+06
Education 125,344 4.207 4.207 1.955 0 7 7 Work 131,438 0.345 0.345 0.476 0 1 1 Marital_Status 131,438 0.320 0.320 0.466 0 1 1 Children 131,438 0.597 0.597 0.490 0 1 1 Age 105,813 46.34 46.34 16.95 16 95 95 Number of individuals 6,831 6,831 6,831 6,831 6,831 6,831 6,831
- 18 -
Table 4. Cross-tabulation of happiness on stock market participation. All in all, to what extent do you
consider yourself a happy person?
Stock market participation
No (1) Yes (2) Total (3) very unhappy 97 (82.91) 20 (17.09) 117 (100) unhappy 445 (81.06) 104 (18.94) 549 (100)
neither happy nor unhappy 6032 (81.90) 1333 (18.10) 7365 (100)
happy 25992 (79.60) 6660 (20.40) 32652 (100)
very happy 8263 (80.31) 2026 (19.69) 10289 (100)
Total 40829 (80.10) 10143 (19.90) 50972 (100)
Note: this table presents frequencies and percentages in parentheses
4. Methodology
In this section, we elaborate on the methodology and regressions used in this study.
To test our hypothesis that happiness influences stock market participation we first run a baseline regression. The data in this study ranges from 1993 to 2018, a panel of 26 years. Therefore, our baseline regression is in the form of a Pooled Ordinary Least Square (POLS) regression method. The following regression is used:
𝑌𝑖𝑡 = 𝛼 + 𝛽𝐻𝑖𝑡 + 𝛿𝑋𝑖𝑡′ + 𝜀𝑖𝑡 (1)
Where Y is the dependent variable, stock market participation of individual i at time t. Stock market participation in our model is a binary variable which only can have a value 1 (stock market participation) or 0 (no stock market participation) otherwise. The regression coefficient is denoted by β and H denotes our independent variable happiness (on a scale from 1 to 5). X’it denotes the vector of control variables including: age, gender, marital status,
having children, having work, income, education and health status. The error term is denoted by εit.
By estimating a regression by means of a Pooled OLS, we assume there is no heterogeneity (the same relationship holds for all data). However, since we deal with longitudinal data (and an unbalanced panel) there is bound to be heterogeneity in the model, which might also be unobservable. To assess whether this forms a problem in our model we perform a Breusch and Pagan Lagrangian multiplier test. Thereafter we continue estimating a panel data model in order to control for fixed effects. The following regression is used:
𝑌𝑖𝑡 = 𝛼 + 𝛽𝐻𝑖𝑡 + 𝛿𝑋𝑖𝑡′ + 𝑢𝑖 + 𝑣𝑖𝑡 (2)
Where ui is the cross section fixed effect and vit equals the error component, all other
- 19 -
the influence of variables that influence stock market participation but do not variate over time. We use a Hausman test to assess whether or not it is better to use a Fixed or Random effect panel data model. Furthermore this study performs several robustness checks to validate or model, these will be discussed in the section Results.
In our first model, we use a Pooled OLS regression to estimate whether happiness influences stock market participation. Thereafter, we add our control variables to check if the validity of our model increases. After we ran the Pooled OLS regression, we check whether or not it is preferable to use a panel data model, either fixed effects or random effects. Lastly, we perform several robustness checks to strengthen the validity of our regression results.
5. Results
This section presents the results of our base line model. Furthermore, we provide the results of several robustness checks and discusses the limitations that our model may have.
5.1 Main results
- 20 -
Table 5. Regression Results using Pooled OLS with and without control variables.
(1) (2)
VARIABLES POLS excl. CV POLS incl. CV
Happiness 0.00630** -0.0174*** (0.00276) (0.00494) Age 0.00442*** (0.000285) Gender -0.109*** (0.00619) Marital_Status -0.00326 (0.00691) Health_Status 0.0166*** (0.00482) Education 0.0367*** (0.00209) Income 5.54e-07*** (5.87e-08) Children -0.0264*** (0.00647) Work -0.0225*** (0.00637) Constant 0.174*** -0.00207 (0.0113) (0.0319) Observations 50,972 19,422 R-squared 0.000 0.065
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
To test whether this model suffers from heteroscedasticity we perform a Breusch Pagan / Cook-Weisberg test for heteroscedasticity. Results of this test are shown in table D1 of the appendix. With a p-value of 0.000 we reject the null hypothesis of a constant variance and we conclude that our model suffers from heteroscedasticity. In order to correct for this heteroscedasticity we use robust standard errors. Results from our Pooled OLS regression using robust standard errors can be found in table D2 of the appendix. The coefficients of our model do not change and are still significant at a 1% level. Next, we perform a Ramsey Reset test using powers of the fitted values of stock market participation in order to test for the presence of omitted variables. Results of this test can be found in table D3 of the appendix. With a corresponding p-value of 0.000 we reject the null hypothesis that our model has no omitted variables.
- 21 -
with a Pooled OLS regression and the preferred method is a panel data model. In order to determine whether it is better to use a fixed effect or a random effect model we perform a Hausman test. With a p-value of 0.000 and a corresponding chi-square statistic of 208.32, we reject the null hypothesis that a random effect model is preferred and we proceed with a panel data model using fixed effects. The results for this Hausman test can be found in table D5 of the appendix.
- 22 -
Table 6. Fixed effect regression results with robust standard errors (1) VARIABLES Happiness -0.0168*** (0.00580) Age -0.00370*** (0.00126) Gender -1.003*** (0.0115) Marital_Status -0.0110 (0.0139) Health_Status -0.00743 (0.00613) Education 0.00607 (0.00813) Income 4.90e-08* (2.85e-08) Children 0.0149 (0.0169) Work -0.000275 (0.00819) Constant 1.852*** (0.0748) Observations 19,422 Number of individuals 6,831 R-squared 0.006
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table D7 in the appendix provides our fixed effect panel data model regression results with and without robust standard errors. For comparison purposes, the Pooled OLS regression results with robust standard errors are also included. There are no major differences in outcome between our FE panel data model with or without robust standard errors, the p-value increased from 0.001 in our model without robust standard errors to 0.004 in the model containing robust standard errors.
- 23 -
average, people in The Netherlands consider themselves happy, not sad, so this might be an explanation for happy people not to participate in the stock market. Furthermore, our results are also more in line with Delis and Mylondidis (2015), who found that happiness negatively correlates with investing in risky assets and insurance (in The Netherlands). However, there is no direct link between happiness and stock market participation.
5.2 Robustness checks
This section provides several robustness checks to validate and strengthen our baseline model regression results. We use a random effect model, we cluster our data on household level and we run a Probit regression. Table 7 reports the results for these robustness checks.
Firstly, we use a random effect panel data model in order to counterpart the drawbacks that arise from our fixed effect panel data model. Variables that do not variate over time will cancel out in our fixed effect model. When using a random effect model we allow our variables to be part of our regression even when they do not vary over time. Furthermore, differences on individuals might have an influence on stock market participation when they are random and uncorrelated with happiness. Table 7 column (1) reports the results from this random effect regression. We can observe that our model is still significant at a 1% level and the negative coefficient is still present (-0.0148). The results of our random effect panel data model confirm the results of our baseline regression results. Happiness is still significantly negatively correlated with stock market participation.
Secondly, we cluster our data on household level. The survey performed by De Nederlandsche Bank collects data on individual level and on household level. Therefore, we want to know whether our findings differ on household level compared to individual level. Table 7 column (2) presents the results of this fixed effect panel data model, when clustering the data on household level. The results show only minor changes in our outcome compared to our baseline model. We still observe a negative significant effect (-0.0168) of happiness on stock market participation.
- 24 -
However, this model has a much higher coefficient, indicating that stock market participation is even lower when people are happier compared to our base line regression model.
Table 7. Regression results of various robustness checks
(1) (2) (3)
VARIABLES RE Clustered Probit
Happiness -0.0148*** -0.0168*** -0.142*** (0.00435) (0.00579) (0.0431) Age 0.00223*** -0.00370*** 0.0163*** (0.000332) (0.00126) (0.00326) Gender -0.108*** -1.003*** -1.166*** (0.00886) (0.0115) (0.0898) Marital_Status 0.00192 -0.0110 0.00197 (0.00720) (0.0139) (0.0708) Health_Status 0.000299 -0.00743 -0.00905 (0.00437) (0.00614) (0.0434) Education 0.0245*** 0.00607 0.222*** (0.00259) (0.00806) (0.0260)
Income 2.45e-07*** 4.90e-08* 3.56e-06***
(4.46e-08) (2.83e-08) (5.53e-07)
Children -0.00657 0.0149 -0.0623 (0.00702) (0.0167) (0.0684) Work -0.0135*** -0.000275 -0.122** (0.00512) (0.00827) (0.0485) lnsig2u 2.141*** (0.0511) Constant 0.189*** 1.852*** -2.408*** (0.0348) (0.0742) (0.343) Observations 19,422 19,422 19,422 R-squared 0.006 Number of individuals 6,831 6,831 6,831
Note: reported are (1): Random effect panel data model, (2): Fixed effect panel data model when clustering data on household level, (3): Probit model. Standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1
5.3 Limitations
This section provides several limitations that our model might have. We highlight omitted variable bias, reversed causality bias, selection effects and measurement error.
- 25 -
provides multiple researches that find different relationships on happiness. For example, Delis and Mylondidis (2015) use family relations to find a relationship on happiness whereas Guven and Hoxha (2015) use the weather to explain happiness. These different relationships on happiness might explain a possible cause for omitted variable bias in any research, also in our own. To control for this concept we added several control variables. However, in our Pooled OLS regression, even after controlling with several control variables, omitted variable bias still have been present. This is important when interpreting our results.
Secondly, another limitation of our model appears when taking into account the concept of reversed causality bias. This endogeneity problem implies that we need to interpret our results differently than we currently do. In the current situation, our results confirm that happier people participate less in the stock market. The concept of reversed causality bias makes us question if people who participate less in the stock market are happier, so the concept changes our way of thought to the other side. Delis and Mylondidis (2015) state that using a panel structure to analyse the data should resolve this problem of reversed causality. Since this study uses a panel data structure, the problem of revered causality should therefore not appear.
Thirdly, our results might suffer from selection effects. Since we use a fixed effect panel data model, observations that do not vary over time cancel out (leading to fewer observations). Collewet and Koning (2011) studied happiness based on data from the Dutch Household Survey as well and argue that people who rate themselves happy are less likely to experience a change in their level of happiness over time. Since our study uses a panel data model (on the same dataset) as well, this problem might also occur in our results.
- 26 -
6. Conclusion
This paper studies the effect of happiness on stock market participation using data from the Dutch Household Survey from De Nederlandsche Bank in a period from 1993 to 2018. The main findings are that happiness negatively correlates with stock market participation on a one percent significance level. This implies that happier people are less active on the stock market than unhappy people. These findings are confirmed after several robustness checks and alternative regressions. Also in the case where we cluster our data on household level, we can conclude that happier households are less active on the stock market. The use of several well-founded control variables strengthened our model and confirmed the described negative relationship.
The results of this study contrast with existing literature. Happiness is said to have a positive relation on stock market participation by Rao et al. (2016), who studied the same relationship as this study but focussed on the Chinese market using data from the China Finance Household Survey. Furthermore, we can state that on average a happier country, such as The Netherlands, does not participate more in the stock market than a less happy country, such as China. Therefore, we conclude that there is a difference between the relationship of happiness and stock market participation in China and The Netherlands. In addition to possible cultural explanations, a possible answer might be given by Hurd et al. (2011), who state that on average the Dutch hold a rather pessimistic view about the stock market.
Our results are important for policy makers since happiness is an important aspect of the equity premium puzzle and subsequently drives asset prices itself (Mankiw and Zeldes, 1991). However, our contrasting results compared to Rao et al. (2016) suggests that happiness affects utility (stock market participation should increase utility, Beshears et al., 2018) in very different ways across countries. Policy makers should therefore consider the role of emotions differently across countries to increase stock market participation and subsequently increase people’s utility.
- 27 -
- 28 -
References
Agarwal, S., Driscoll, J. C., Gabaix, X., & Laibson, D., 2009. The age of reason: Financial decisions over the life cycle and implications for regulation. Brookings Papers on Economic Activity, 2009(2), 51-117.
Almenberg, J., & Dreber, A., 2015. Gender, stock market participation and financial literacy. Economics Letters, 137, 140-142.
Barrotta, P., 2008. Why economists should be unhappy with the economics of happiness. Economics & Philosophy, 24(2), 145-165.
Beshears, J., Choi, J. J., Laibson, D., & Madrian, B. C., 2018. Behavioural household finance (No. w24854). National Bureau of Economic Research.
Brickman, P., Coates, D. and Janoff-Bulman, R., 1978. Lottery winners and accident victims: Is happiness relative? Journal of Personality and Social Psychology, vol. 36, p. 917–27. Briggs, J. S., Cesarini, D., Lindqvist, E., & Östling, R., 2015. Windfall gains and stock market participation. National Bureau of Economic Research Working Paper 21673.
Campbell, J., 2006. Household finance. Journal of Finance 61, 1553-1604.
Chang, L., & Krosnick, J. A., 2009. National surveys via RDD telephone interviewing versus the Internet: Comparing sample representativeness and response quality. Public Opinion Quarterly, 73(4), 641-678.
Cole, S., Paulson, A., & Shastry, G. K., 2014. Smart Money? The effect of education on financial outcomes. Review of Financial Studies 27(7), 2022–2051.
Collewet, M., & de Koning, J., 2011. Working time and happiness in Dutch panel data. In EALE Conference.
Cryder, C. E., Lerner, J. S., Gross, J. J., & Dahl, R. E., 2008. Misery is not miserly: Sad and self-focused individuals spend more. Psychological Science, 19(6), 525-530.
Das, S., Kuhnen, C. M., & Nagel, S., 2017. Socioeconomic status and macroeconomic expectations (No. w24045). National Bureau of Economic Research.
Daykin, A. R., & Moffatt, P. G., 2002. Analyzing ordered responses: A review of the ordered probit model. Understanding Statistics: Statistical Issues in Psychology, Education, and the Social Sciences, 1(3), 157-166.
- 29 -
Diener, E., & Oishi, S., 2005. The nonobvious social psychology of happiness. Psychological Inquiry, 16(4), 162-167.
Gerdtham, U. G., & Johannesson, M., 2001. The relationship between happiness, health, and socio-economic factors: results based on Swedish microdata. Journal of Socio-Economics, 30(6), 553-557.
Grinblatt, M., Keloharju, M., & Linnainmaa, J. T., 2011. IQ and stock market participation. Journal of Finance 66(6), 2121–2164.
Guiso, L., Sapienza, P., & Zingales, L., 2008. Trusting the stock market. Journal of Finance, 63(6), 2557-2600.
Guiso, L., & Sodini, P., 2013. Household finance: An emerging field. In: G. M. Constantinides, M. Harris, R. M. Stulz (Eds.), Handbook of the Economics of Finance. North-Holland, Amsterdam, pp. 1397-1532.
Guven, C., & Hoxha, I., 2015. Rain or shine: Happiness and risk-taking. Quarterly Review of Economics and Finance, 57, 1-10.
Frey, B. S., & Stutzer, A., 2002. What can economists learn from happiness research?. Journal of Economic literature, 40(2), 402-435.
Hagerty, M. R., & Veenhoven, R., 2003. Wealth and happiness revisited–growing national income does go with greater happiness. Social indicators research, 64(1), 1-27.
Haliassos, M., & Bertaut, C. C., 1995. Why do so few hold stocks? Economic Journal 105(432), 1110–1129.
Heaton, J., & Lucas, D., 2000. Portfolio choice in the presence of background risk. Economic Journal, 110(460), 1-26.
Helliwell, J. F., Barrington-Leigh, C. P., Harris, A., & Huang, H., 2009. International evidence on the social context of well-being (No. w14720). National Bureau of Economic Research.
Hofstede, G., 2011. Dimensionalizing cultures: The Hofstede model in context. Online readings in psychology and culture, 2(1), 8.
Hong, H., Kubik, J. D., & Stein, J. C., 2004. Social interaction and stock‐market participation. Journal of finance, 59(1), 137-163.
- 30 -
Hsee, C. K., & Weber, E. U., 1999. Cross‐national differences in risk preference and lay predictions. Journal of Behavioural Decision Making, 12(2), 165-179.
Hsu, J. C., 2012. What drives equity market non-participation? North American Journal of Economics and Finance, 23(1), 86-114.
Kahneman, D., & Krueger, A. B., 2006. Developments in the measurement of subjective well-being. Journal of Economic perspectives, 20(1), 3-24.
Kézdi, G., & Willis, R., 2011. Household stock market beliefs and learning. National Bureau of Economic Research Working Paper 17614.
Kuhnen, C. M., & Miu, A. C., 2017. Socioeconomic status and learning from financial information. Journal of Financial Economics 124(2), 349–372.
Kumar, A., 2009. Who gambles in the stock market?. Journal of Finance, 64(4), 1889-1933. Mankiw, N. G. and Zeldes, S. P., 1991. The consumption of stockholders and non-stockholders. Journal of Financial Economics 29(1): 97-112.
Lindqvist, E., Östling, R., & Cesarini, D., 2018. Long-run effects of lottery wealth on psychological well-being (No. w24667). National Bureau of Economic Research.
Moskowitz, T. J., & Vissing-Jørgensen, A., 2002. The returns to entrepreneurial investment: A private equity premium puzzle? American Economic Review, 92(4), 745-778.
Rao, Y., Mei, L., & Zhu, R., 2016. Happiness and stock-market participation: Empirical evidence from China. Journal of Happiness Studies, 17(1), 271-293.
Rosen, H. S., & Wu, S., 2004. Portfolio choice and health status. Journal of Financial Economics, 72(3), 457-484.
Sachs, J. D., Layard, R., & Helliwell, J. F., 2019. World Happiness Report 2019. New York: Sustainable Development Solutions Network.
Van Rooij, M., Lusardi, A., & Alessie, R., 2011. Financial literacy and stock market participation. Journal of Financial Economics, 101(2), 449-472.
Van Winden, F., Krawczyk, M., & Hopfensitz, A., 2011. Investment, resolution of risk, and the role of affect. Journal of Economic Psychology, 32(6), 918-939.
Veenhoven, R., 1991. Is happiness relative? Social indicators research, 24(1), 1-34.
- 31 -
Vissing-Jorgensen, A., 2002. Towards an explanation of household portfolio choice heterogeneity: Non-financial income and participation cost structures. National Bureau of Economic Research.
Vissing-Jørgensen, A., & Attanasio, O. P., 2003. Stock-market participation, intertemporal substitution, and risk-aversion. American Economic Review, 93(2), 383-391.
- 32 -
Appendix
Appendix A – Variable construction
Table A1. DHS survey questions used to construct our dependent variable stock market participation
Question
code Label Question
Answer range
BZ12 Mutual Funds
Did you, on 31 December 2017, have investments with MUTUAL FUNDS? Do not include investments in growth funds, investments (shares, bonds) in companies, or ‘insured saving’ (i.e. saving through a life-insurance) here.
Yes/No
BZ13 Bonds
Did you, on 31 December 2017, have any BONDS and/or MORTGAGE
BONDS? Do not include bonds through mutual funds here.
Yes/No
BZ14 Shares
Did you, on 31 December 2017, own any SHARES? Do not include shares of your own private limited company here, nor bonds through MUTUAL
FUNDS.
Yes/No
BZ15 Put1 Did you, on 31 December 2017, have
one or more PUT-OPTIONS? Yes/No
BZ16 Put2
Did you have any written PUT-OPTIONS outstanding on 31
December 2017?
Yes/No
BZ17 Call1
Had you, on 31 December 2017, bought one or more
CALL-OPTIONS, FALCONS, WARRANTS, SPRINTERS OR
TRACKERS?
- 33 -
BZ18 Call2
Had you, on 31 December 2017, any written CALL-OPTIONS, FALCONS, WARRANTS, SPRINTERS OR TRACKERS
outstanding?
Yes/No
Note: questions in this table are based on the 2018 survey. Surveys used from 1993-2017 use the same questions/notations but have a different year (y-1) mentioned in the question.
Table A2. DHS survey question used to construct our independent variable happiness. Question
code Label Question Answer range
GELUK Happiness
All in all, to what extent do you consider yourself a
happy person?
1. Very happy 2. Happy
3. Neither happy nor unhappy 4. Unhappy
5. Very unhappy -9. Don’t know
- 34 -
Table A3. List of control variables supported by corresponding literature.
Control variable Based on following Literature
Age
Agarwal, S., Driscoll, J. C., Gabaix, X., & Laibson, D., 2009. The age of reason: Financial decisions over the life cycle and implications for regulation. Brookings Papers on Economic Activity, 2009(2), 51-117.
Gender
Almenberg, J., & Dreber, A., 2015. Gender, stock market participation and financial literacy. Economics Letters, 137, 140-142.
Marital Status
Kumar, A., 2009. Who gambles in the stock market? Journal of Finance, 64(4), 1889-1933.
Grinblatt, M., Keloharju, M., & Linnainmaa, J., 2011. IQ and stock market participation. Journal of Finance, 66(6), 2121-2164.
Having Children
Van Rooij, M., Lusardi, A., & Alessie, R., 2011. Financial literacy and stock market participation. Journal of Financial Economics, 101(2), 449-472.
Work status
Heaton, J., & Lucas, D., 2000. Portfolio choice in the presence of background risk. Economic Journal, 110(460), 1-26.
Education
Campbell, J., 2006. Household finance. Journal of Finance 61, 1553-1604.
Van Rooij, M., Lusardi, A., & Alessie, R., 2011. Financial literacy and stock market participation. Journal of Financial Economics, 101(2), 449-472.
Health
Rosen, H., Wu, S., 2004. Portfolio choice and health status. Journal of Financial Economics 72, 457–484.
Income
Guiso, L., Sapienza, P., Zingales, L., 2008. Trusting the stock market. Journal of Finance 63, 2557–2600.
- 35 -
Table A4. DHS survey questions used to construct our set of control variables. Question
code Label Question Answer range
GEBJAAR Age Year of birth of the
respondent Any answer
GESLACHT Gender Sex of the
respondent
1. Male 2. Female
BURGST Marital Status What is your marital status?
1. married or registered partnership, having community of property (including separation from bed and table)
2. married or registered partnership, with a marriage settlement
(including separation from bed and table)
3. divorced from spouse
4. living together with partner (not married)
5. widowed 6. never married
AANTALKI Children Number of children in the household 6. 0 none 7. 1 child 8. 2 children 9. 3 children 10. 4 children 11. 5 children 12. 6 children 13. 7 children 14. 8 children 15. 9 children or more
BEZIGHEI Work Status Primary occupation of the respondent
1. employed on a contractual basis 2. works in own business
3. free profession, freelance, self-employed
4. looking for work after having lost job
5. looking for first-time work 6. student
7. works in own household
8. retired [pre-retired, AOW, VUT] 9. (partly) disabled
10. unpaid work, keeping benefit payments
11. works as a volunteer 12. other occupation
- 36 - OPLZON Education Highest level of education attended (regardless of certificate/diploma)
1. (Voortgezet) speciaal onderwijs / (continued) special education 2. Kleuter-, lager- of basisonderwijs
/ kindergarten/primary education 3. Voorbereidend middelbaar
beroepsonderwijs (VMBO) / pre-vocational education
4. HAVO/VWO / pre-university education
5. MBO of het leerlingwezen / senior vocational training or training through apprentice system
6. HBO (eerste of tweede fase) / vocational colleges
7. Wetenschappelijk onderwijs WO / university education
8. Did not have education (yet) 9. other sort of education/training .
ONDERW Education* Highest level of education attended
1. kindergarten/primary education 2. continued primary education
[VGLO] or elementary secondary education [LAVO]
3. continued special (low-level) education [MLK , VSO, LOM], secondary education
[MAVO/MULO]
4. pre-university education [HAVO, VWO, Atheneum, Gymnasium, HBS, MMS, Lyceum]
5. junior vocational training [e.g. LTS, LEAO, Lagere Land- en Tuinbouwschool]
6. senior vocational training [e.g. MTS, MEAO, Middelbare Land- en Tuinbouwschool]
7. vocational colleges [e.g. HTS, HEAO, opleidingen MO-akten] 8. vocational colleges 2nd tier [e.g.
accountant NIVRA, actuaris, opleidingen MO-B-akten] 9. university education
10. special (low-level) education [speciaal onderwijs]
11. vocational training through apprentice system [leerlingwezen] 12. other sort of education/training 13. doesn’t attend any education yet
GEZ3 Health Status
In general, would you say your health
- 37 -
IJ161 Income
What was your total gross income over the year 2017 (according to the annual statement) received from [NAME EMPLOYER]? 1. Any answer -9. Don’t know
*Education had a different question from 1993 to 2001
- 38 - Table A5. Merging of our education control variable.
Given
value Answer in DHS survey 2002-2018 Answer in DHS survey 1993-2001
1
(Voortgezet) speciaal onderwijs / (continued) special education
special (low-level) education [speciaal onderwijs]
2
Kleuter-, lager- of basisonderwijs / kindergarten/primary education
kindergarten/primary education
3
Voorbereidend middelbaar beroepsonderwijs (VMBO) /
pre-vocational education
continued primary education [VGLO] or elementary secondary education [LAVO] continued special (low-level) education [MLK , VSO, LOM], secondary education [MAVO/MULO]
4 HAVO/VWO / pre-university education
pre-university education [HAVO, VWO, Atheneum, Gymnasium, HBS, MMS, Lyceum]
5
MBO of het leerlingwezen / senior vocational training or training through
apprentice system
junior vocational training [e.g. LTS, LEAO, Lagere Land- en Tuinbouwschool]
senior vocational training [e.g. MTS, MEAO, Middelbare Land- en Tuinbouwschool]
vocational training through apprentice system [leerlingwezen]
6
HBO (eerste of tweede fase) / vocational colleges
vocational colleges [e.g. HTS, HEAO, opleidingen MO-akten]
vocational colleges 2nd tier [e.g. accountant NIVRA, actuaris, opleidingen MO-B-akten]
7 Wetenschappelijk onderwijs WO / university education
university education
0 Did not have education (yet) doesn’t attend any education yet 0 other sort of education/training other sort of education/training
- 39 -
Appendix B – Detailed descriptive statistics
Table B1. Descriptive Statistics for dependent, independent and control variables sorted by year. year: 1993
N mean p50 iqr sd min max
- 49 -
Appendix C – Model
Table C1. Correlation matrix on the various control variables
- 50 -
Appendix D – Results
Table D1. Breusch-Pagan / Cook-Weisberg test for heteroscedasticity H0: Contstant variance
Variables: fitted values of stock market participation
Chi2(1) 1301.60
Prob > chi2 0.0000
Table D2. Pooled OLS using robust standard errors
(1)
VARIABLES Stock market participation
Happiness -0.0174*** (0.00497) Age 0.00442*** (0.000286) Gender -0.109*** (0.00685) Marital_Status -0.00326 (0.00689) Health_Status 0.0166*** (0.00491) Education 0.0367*** (0.00221) Income 5.54e-07*** (1.67e-07) Children -0.0264*** (0.00635) Work -0.0225*** (0.00688) Constant -0.00207 (0.0323) Observations 19,422 R-squared 0.065
- 51 -
Table D3. Ramsey Reset test using powers of the fitted values of stock market participation H0: model has no omitted variables
F(3,19409) 52.47
Prob > F 0.0000
Table D4. Breusch and Pagan Lagrangian multiplier test. Test var(u)=0
Chibar2(01) 26879.92
Prob > chibar2 0.0000
Table D5. Hausman (1978) specification test. H0: difference in coefficients not symatic
Coefficient
Chi-square test value 208.318
Prob 0.0000
Table D6. Modified Wald-test for group wise heteroscedasticity in fixed effect regression model. H0: sigma(i)^2 = sigma^2 for all i
Coefficient
Chi2 (6831) 3.0e+11
- 52 -
Table D7. Pooled OLS regression results and fixed effect regression results with and without robust standard errors.
(1) (2) (3)
VARIABLES POLS FE FE ROBUST
Happiness -0.0174*** -0.0168*** -0.0168*** (0.00494) (0.00510) (0.00580) Age 0.00442*** -0.00370*** -0.00370*** (0.000285) (0.000597) (0.00126) Gender -0.109*** -1.003*** -1.003*** (0.00619) (0.340) (0.0115) Marital_Status -0.00326 -0.0110 -0.0110 (0.00691) (0.00985) (0.0139) Health_Status 0.0166*** -0.00743 -0.00743 (0.00482) (0.00532) (0.00613) Education 0.0367*** 0.00607 0.00607 (0.00209) (0.00475) (0.00813)
Income 5.54e-07*** 4.90e-08 4.90e-08*
(5.87e-08) (4.82e-08) (2.85e-08)
Children -0.0264*** 0.0149 0.0149 (0.00647) (0.0101) (0.0169) Work -0.0225*** -0.000275 -0.000275 (0.00637) (0.00584) (0.00819) Constant -0.00207 1.852*** 1.852*** (0.0319) (0.476) (0.0748) Observations 19,422 19,422 19,422 R-squared 0.065 0.006 0.006 Number of individuals 6,831 6,831
