The impact of happiness on household asset
allocation: the moderating effects of the big five
personality traits
Student nr: s3560287
Name: Guo Xinyuan
Supervisor: dr. R.O.S. Zaal
Study Program: MSc Finance
Abstract
This research studies the relationship between happiness and household asset allocation from the perspective of the moderation effects of the big five personality traits. This is conducted using Dutch survey data derived from the LISS Panel. The results confirm the negative relations between happiness and risky asset market participation as well as the share of financial wealth held in risky assets. The findings are significant for all three measures of happiness (subjective well-being, general happiness and current life satisfaction). Moreover, the paper finds that this relationship is more evident for people scoring high on neuroticism.
Introduction
Happiness is defined by Diener (2006), pp 400: as: “An umbrella term for different valuations that people make regarding their lives, the events happening to them, their bodies and minds, and the circumstances in which they live”. The definition is broad, ranging from individual’s general feelings to valuations towards general life experience (Diener, 1984; Medvedev and Landguis, 2018). Happiness is considered to playing a
significant role in financial decisions. For instance, Hermalin and Isen (2007) find that
a positive emotion, particularly happiness, increases the marginal utility of
consumption which implies that happiness can be used as a predictor for consumption.
Happier people also hold better expectancy for the future, and they are prone to having
more savings for future investment (Guven, 2012). Additionally, Shleifer et. al (1990)
examine the impact of happiness on stock market return and found a significant positive
correlation.
Although happiness has received growing attention in financial studies, there is
only limited literature on the relationship between happiness and household investment
decisions. Related literature reveals that people with higher levels of happiness are
showing unusual risk-taking tendencies when making decisions in the financial market
(see, e.g. Isen and Hermalin, 2008; Delis and Mylonidis, 2015; Forgas, 1995). Two
theories, the Affect-Infusion-Model and the Mood-Maintenance Hypothesis, are widely
used to explain the relationship between happiness and financial market decisions.
However, there is no consensus on whether a higher level of happiness reinforces or
constrains the risk-taking. On the one hand, the Affect-Infusion-Model by Forgas (1995)
indicates that happier people tend to take on higher risk in financial investments due to
their optimistic judgement towards the uncertainty. On the other hand, the
a high level of happiness avoid risk-taking behavior because they are inclined to
maintain positive affective states.
Specifically, in psychology, it is stated that the role of emotions in decision making
depends on the individual difference in personality traits (Zelenski, 2008). The
interaction effects between personality traits and the affective state can be used as an
explanation for the inconsistent findings regarding the relationship between happiness
and household asset allocations. We assume that people with different personality traits
react differently in decisions of risky financial assets investments when experiencing
happiness. Thus, we make the hypothesis that personality traits moderate the
relationship between happiness and asset allocations.
Following the paper by Delis and Mylonidis (2015), this study first seeks to
explore the relationship between happiness and risky decisions in household asset
allocation based on the Mood-Maintenance-Hypothesis. Happiness1is measured by
subjective well-being 2 in the main analysis. Then, we further examine the moderation
effects of the big five personality traits to clarify the conflicting results regarding the
impact of happiness on household asset allocation.
Data are collected from the LISS panel of CentERdata. The LISS-panel (Longitudinal Internet Study in the Social Sciences) is based on a true probability sample in Dutch individuals consists of almost 8000 people above the age of 16. In this paper, datasets of the big five personality traits, household asset allocation and happiness are collected for the years 2009, 2011, 2013 and 2015 and are combined into
1 Following Delis and Mylonidis (2015), happiness can be considered as a kind of affective state, an
emotion and a mood. We do not distinguish between these three conceptions since this paper mainly analyze the general impact of happiness on asset allocation decisions from a perspective of behavioral finance.
2 Following the paper by Delis and Mylonidis (2015), we use happiness, life satisfaction and subjective
cross-section data. This paper confirms a significantly negative relationship between happiness and risky assets market participation as well as the share of financial wealth held as risky assets. Moreover, the results also provide evidence that the neuroticism trait negatively moderates the relationship mentioned above.
Literature review
As early as 1985, Mehra and Prescott coined the term ‘equity premium puzzle’, which describes the abnormal situation where the observed historical equity premium is too large to be explained by the economic models. They find that the average equity returns are much higher than the average returns of the short-term risk-free debt, and the risk premium also fails to adequately explain the return spread. Besides, subsequent studies show that stock market participation rates around the world are relatively low.
Specifically, the stock market participation rates for most of the European countries are less than 25% and the rate for the United States is also less than 30% in the year 2005 (Guiso et al.,2008). Mankiw and Zeldes (1991) term this phenomenon as the ‘stockholding puzzle’, which is: why do individuals choose not to invest in stocks if the returns on risky assets are much higher than those on risk-free assets? To date, a series of theoretical and mathematical explanations are proposed by the economists, there’s no general rationale for this phenomenon (Campbell, 2006). In the field of behavioral finance, economists focus on the role of individual features on stock market participation. For instance, Guiso et. al (2008) find that the lack of trust can explain why people are reluctant to participate in the stock market. Additionally, financial literacy is also considered as an essential determinant of stock market participation (van Rooijn et al., 2011). In this paper, we follow Delis and Mylonidis (2015) to provide
2.1. The effect of happiness on asset allocation decisions
This paper intends to provide evidence for the impacts of happiness on
participation in the market of risky financial assets, as well as on the share of the households’ financial wealth held in risky assets. Due to the fact that investment decisions are fundamentally based on the risk-return tradeoff, it is important to consider
the relationship between happiness and risk-taking tendency. This is in line with
Kaplanski et. al (2015), Delis and Mylonidis (2015), and Rao et. al (2016), who identify
subjective risk attitudes and risk expectations as mediators of the association between
happiness and the exposure to risky assets.
Specifically, two alternative theories are commonly used in behavioral finance to
analyze this relationship. First, the Affect-Infusion-Model developed by Forgas (1995) suggests that happiness, as a kind of positive affect, exerts an influence on one’s judgements and eventually lead to a higher tendency to take risks. This theory is
complemented by Johnson and Tversky (1983), Jorgensen (1998), and Nygren et. al
(1996). Johnson and Tversky (1983) conduct comparable experiments to test the
relationship between affect and the risk perceptions. Their findings show that positive
affect and the frequency of risk judgements are negatively correlated. Additionally,
Jorgensen (1998) argues that people with negative affects perceive the uncertainty as
less pleasant, and thus they assess the situation thoroughly before making risky
decisions. Thereby, people with negative affects are more risk-averse when making
risky decisions. Furthermore, Nygren et. al (1996) propose that happier people
overestimate the chance of winning and underestimate the chance of losing in gambles.
This is in line with Wright and Bower (1992), who indicate that people in positive mood
optimistically estimate the probability of good outcome compared to those with
arguments, confirming that positive affect induces optimistic judgements towards
positive outcomes. Applying the Affect-Infusion-Model to financial decisions,
Kaplanski et. al (2015) propose that happier people show tendencies to optimistically
estimate returns and underestimate risks in investment plans. The evidence confirms
that the happier people are, the more likely they are to invest in risky assets. These
findings together support that people with higher levels of happiness display tendencies
to hold optimistic return expectations and lower levels of risk estimations.
Consequently, they are more likely to possess risky assets and hold a higher proportion
of risky investments compared to people with lower levels of happiness.
Taking the opposite stance, Isen and Patrick (1983) propose the
Mood-Maintenance Hypothesis, which states that happier people are inclined to maintain in
good mental states and hence they display a tendency to avoid risk-taking behavior.
Specifically, the theory suggests that compared to the small marginal return obtained
after taking the risk, losing what people have at the moment will cause them a much
higher utility loss (Drichoutis and Nayga,2013). Consequently, happier people are not
willing to risk the current positive state. This is also in line with Raghunathan and Pham
(1999), who argue that affective states transmit distinct messages to the individuals
when making decisions. Specifically, people have the desire to be compensated when
experiencing sadness, and thus they will choose options with higher risks. When
experiencing happiness, the desire diminishes, and thus they are less likely to take risks.
Applying these to financial decisions, evidence provided by Grable and Roszkowski
(2008) confirms that a higher level of happiness is indeed associated with lower risk
propensities. Similar evidence is provided by Delis and Mylonidis (2015), who
investigate the relationship between happiness and household financial decisions within
to possess risky financial assets compared to others. In general, although abundant
studies have examined the relationship between happiness and risk-taking decisions, no
consistent conclusion can be drawn.
This paper intends to extend the work of Delis and Mylonidis (2015) by examining
the moderation effects of the big five personality traits. Delis and Mylonidis (2015) find
a negative relationship between happiness and the probability of investing risky assets
by using a sample of the Dutch population. Using data extracted from the same database,
we make the assumptions based on the second theory, the Mood-maintenance
hypothesis, which is in line with findings of Delis and Mylonidis (2015). This leads to
the following hypotheses:
H1a: There is a significant and negative relationship between happiness and the risky assets market participation.
H1b: There is a significant and negative relationship between happiness and the share of wealth held in risky assets.
2.2. The moderation effect of the big five personality traits
The interaction effects between personality traits and mood in psychology studies
can be used as a guide to interpret the moderation effect of the big five personality traits
on the relationship between happiness and asset allocation decisions. As stated by Zelenski (2008, para.2): “To the extent that emotions help or hurt decision making, they likely do so more or less, depending on personality, because personality contributes to the experience of these emotion states”, we assume that personality traits moderate the relationship between happy emotion and risky asset investment decisions.
In psychology, the mood-congruent processing provides the theoretical basis for
the Affect-infusion model, indicating that people tend to recall memory (information)
‘mood-incongruent processing of emotional information’ is in line with the Mood-Maintenance Hypothesis (Rusting, 1998). The moderation approach which considers
the interaction effects between the personality traits and moods provides a valid
solution to the inconsistency (Rusting, 1998). Related studies indicate that in addition
to mood, personality and the combined interaction term also exert influence on the
emotional information processing and the subsequent decision-making (see e.g.,
Richards et. al,1992; Gotlib and Cane, 1987). Specifically, they find that the impacts of
mood on cognitive reactions are more evident when considering the individual
differences in personality traits (Teasdale and Taylor, 1981). In other words, personality
traits moderate the relationship between mood states and cognitive reactions.
Motivated by these studies and the argument stated by Gutnik et. al (2006), pp 726: “Emotions can have indirect effects on our behavior through implicitly shaping our attitudes and judgments (cognitive representations of the world)”, we can further
speculate that the personality traits have the potential moderation effects on the
relationship between and financial risk-taking decisions. If this were the case, it would
also provide evidence for the inconsistent relationships between both found in current
studies. In this study, we use the big five personality traits taxonomy (extraversion,
agreeableness, conscientiousness, neuroticism and openness to experience) advanced
by Costa and McCrea (1992) to classify individual personality traits. Next, we will
separately discuss the moderation effect of each personality trait.
Extraversion
Extraverted people are considered to be sociable, energetic, assertive and they
show more tendency to seek sensation (Heckman, 2011, McCrae and Costa, 1995).
Specifically, sensation seeking is a strong predictor of investment behavior, and people
According to Larsen and Ketelaar (1991), people who score high on extraversion
are likely to induce positive emotion. This is supported by Gross et. al (1998), who
argue that for extraverted people, a lower level of stimulation is enough to induce
happiness. Moreover, since extraverted people are sociable and energetic, they are
likely to engage in more social activities, and consequently receive more stimulation
from the outside world and experience positive emotions from a variety of sources
(Cooper et. al,1992). So, people scoring high on extraversion are likely to perceive the
frequent positive emotions as being normal, take them for granted, and they may react
insensitively to the state of happiness compared to less extraverted people. Furthermore,
as they do not value happiness as much, they may be willing to give much less effort to
maintain it and this will be pictured as they are less afraid of losing it by taking the risks.
Additionally, the often-accompanied assertiveness trait will lead them to focus more on
the potential rewards (Zuckerman and Kuhlman, 2000) and hold optimistic views about
the investment prospects when experiencing happiness. Consequently, happy
extraverted people may try to possess or increase their exposures to investment in risky
assets. In comparison, people scoring low on extraversion may show a tendency to be
anxious and indecisive (Stafford et. al, 2010). They may not experience the same level
happiness in general, and therefore they would not choose to risk the status quo when
experiencing happiness.
Considering those arguments together, we propose that for extraverted people with
higher levels of happiness, the mood-maintenance effect on investment in risky assets
is less evident. This yields the hypotheses:
H2b: The negative significant effect of happiness on the share of wealth held in risky assets is weakened by extraversion.
Agreeableness
Agreeable people are considered to be warm, socially adaptable and unselfish
(Heckman, 2011; Coker et.al, 2002). According to DeNeve (1998), people scoring high
on agreeableness are likely to induce positive emotions. Therefore, agreeable people
may also show the tendency to be insensitive to the state of happiness since they easily
experience positive emotions compared to less agreeable people.
In addition, agreeable people exhibit high levels of sociability and social
adaptability (see, e.g. Coker et.al, 2002; John and Srivastava, 1999). They are likely to
expose themselves in social situations. In combination with the facts that they are extremely sensitive to others’ views (Bierman, 2003) and that they easily trust others (Heckman, 2011), agreeable people are thus expected to take advice from others and
are not influenced as much by their own emotions in decision-making.
Considering those arguments together, we propose that happy agreeable people
may show less tendency to maintain the state of happiness and they are not afraid of
losing it by taking more investment risks since they are insensitive to happiness and rely much on others’ advice in decision-making. In other words, for agreeable people with higher levels of happiness, the mood-maintenance effect on the investment in risky
assets is less evident. This yields the hypotheses:
H3a: The negative significant effect of happiness on risky assets market participation is weakened by agreeableness.
Conscientiousness
Conscientious people are considered to be self-disciplined, goal-oriented,
responsible, and they are hard to be satisfied with themselves (see, e.g. Hackman, 2011;
McCrae and Costa, 1987; Woo and Ahn, 2015).
Conscientious people exhibit a strong sense of responsibility and they show the
tendency to achieve the goal of perfection (Woo and Ahn, 2015). Compared to others,
it takes more efforts for conscientious people to be satisfied with themselves. Therefore,
happiness is hard-earned and experienced infrequently by conscientious people.
Happiness will remind them the hardship of the process thus they will be seriously
affected by the scarce happy emotions and react sensitively to it. In general,
conscientious people will cherish the status quo, be prone to maintaining it, and not
taking more risks in investments when experiencing happiness. This is in line with Woo
and Ahn (2015), who suggest that conscientious people may perceive happiness as a
kind of calm without much excitement.
Combining these arguments, we can speculate that compared to others, people
who high on conscientiousness are less willing to possess risky assets or increase their
exposures to investments in risky assets when experiencing happiness. In other words,
for conscientious people with higher levels of happiness, the mood-maintenance effect
on the investments in risky assets is more evident. This yields the hypotheses:
H4a: The negative significant effect of happiness on risky assets market participation is strengthened by conscientiousness.
Neuroticism
People scoring high on neuroticism are considered to be stressful, insecure and
moody (see, e.g. Watson, 2001; Borghans et al., 2008; Heckman, 2011). According to
McCrae and John (1992), neurotic people are seldom satisfied with themselves or their
surroundings, and they show the tendency to be in negative emotions. This is supported
by Emma et. al (2014), who find that neurotic people induce more negative judgements
toward incidents, and they are trapped in negative emotions more frequently. As they
frequently experience negative emotions, suffer from anxiety, and feel despair, they
will cherish the infrequent happiness emotions more highly, and this will be further
reflected in a higher tendency to maintain the happiness.
Additionally, neurotic people are characterized by risk prevention and high levels
of sensitivity towards crisis (see, e.g. Rustichini et. al, 2016, Tanji et. al, 2015). These
intrinsic motivations towards anxiety avoidance force them to work harder (Nettle,
2007). Thus, compared to others, it takes more efforts for them to be happy or satisfied
with the current situation. The state of happiness will remind them of the hard work and
the perseverance during the process, and hence, they are not willing to risk the
hard-won happiness for uncertain outcomes.
Combining those arguments together, we propose that neurotic people with higher
levels of happiness show higher tendency to maintain the status quo, and they become
more unwilling to possess risky assets or increase their exposures to investments in
risky assets when experiencing happiness. This yields the hypotheses:
H5b: The negative significant effect of happiness on the share of wealth held in risky assets is strengthened by neuroticism.
Openness to experience
People scoring high on openness to experience are considered to be thrill-seeking,
curious, unconventional, and generally they hold wide interests (see, e.g. Woo and Ahn,
2015; Gough and Heilbrun, 1983; Almlund et. al, 2011). Openness is characterized by
engaging in new activities, having the tendency to change and not being afraid of losing
(McCrea and Costa, 1997; Becker, 2005).
Moreover, people who are open to experience are likely to seek more knowledge
in daily life (McCrae, 1994). Thus, when making investment decisions, they will have
more professional financial judgements to rely on, and consequently, they will not be
greatly affected by subjective evaluations like happiness. Additionally, people scoring
high on openness often expose themselves to stimulating events and enjoy dealing with
intense stimuli (Weisberg et. al, 2011). Therefore, when facing investment decisions in
happy emotions, they choose to seek more excitement and changes rather than maintain
the stable mental state of happiness.
Combining these arguments, we speculate that people scoring high on openness to
experience are likely to possess risky assets or increase their exposures to investment
in risky assets when experiencing happiness. In other words, for open people with
higher levels of happiness, the mood-maintenance effect on the investment in risky
assets is less evident. This yields the hypotheses:
H6b: The negative significant effect of happiness on the share of wealth held in risky assets is weakened by openness to experience.
3. Methodology 3.1. Database
The dataset used in this research is drawn from the Longitudinal Internet Studies
for the Social Sciences panel (LISS). LISS Panel is the core element of the MESS
project (Measurement and Experimentation in the Social Sciences) which is generated
by CentERdata (Tilburg University, The Netherlands). It is based on a true probability
sample in Dutch individuals consisting of almost 8,000 people above the age of 16. The
LISS panel collects information from disciplines including health, religion, social
connection, housing, work, education, economic situation, politic views, and
personality. Respondents are required to complete the questionnaires online every
month and get paid for each questionnaire.
3.2. Sample selection
Information on the dependent variable and explanatory variable is obtained
through the Surveys from the Core Studies in LISS Panel. Information on the control
variables is obtained through the Background Survey in LISS Panel. After accessing all
the data from the year 2009, 2011, 2013 and 2015, we match the data on household
asset allocations with information on happiness with the same number of the household
member for each year. During this time span, people gradually built awareness of the
financial crisis and the European debt crisis, which might be further reflected in the
data of household asset allocation. Moreover, it is hard to provide time-matched data
for the main variables after the year 2015. Four years of observations are merged into
cross-section data. The main point investigated in this paper is the general mechanism
decisions, so it is relatively less important to explore the time variation for individuals.
To guarantee the share of financial wealth invested in risky assets is between zero and
one, savings and risky investments accounts with a negative balance are excluded.
Filtering out the samples with missing data on the main variables yields 13,042
observations are left.
3.3. Variable construction 3.3.1. Household asset allocation
This paper intends to provide evidence for the impacts of happiness on
participation in the market for risky financial assets, as well as on the share of the households’ financial wealth held in risky assets. These two variables are commonly used in the related literature as they reflect the tendency to take financial risks (see, e.g.
Delis and Mylonidis, 2015; Rao et. al, 2016).
The first dependent variable representing household asset allocation is
participation in the market for risky financial assets. Risky financial assets in this study
contain growth funds, share funds, bonds, debentures, stocks, options and warrants as
included in the questionnaire constructed by LISS Panel. The variable is derived from
the question: “Did you possess one or more of the following assets? - Investments (growth funds, share funds, bonds, debentures, stocks, options, warrants, and so on)” and the answer “yes” indicates that the respondent is a participant.
The second dependent variable representing household asset allocation is the share of the households’ financial wealth invested in risky assets. Financial wealth contains the risky investment described as above and the relatively risk-free assets. The
relatively risk-free assets in this study contain current accounts, savings accounts, term
deposit accounts, savings bonds or savings certificates as involved in the questionnaire
assets holding by households, so it’s hard to use total assets or wealth as a denominator. The share of the households’ financial wealth invested in risky assets is defined as:
𝑆ℎ𝑎𝑟𝑒 𝑜𝑓 𝑡ℎ𝑒 ℎ𝑜𝑢𝑠𝑒ℎ𝑜𝑙𝑑𝑠’ 𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑤𝑒𝑎𝑙𝑡ℎ 𝑖𝑛𝑣𝑒𝑠𝑡𝑒𝑑 𝑖𝑛 𝑟𝑖𝑠𝑘𝑦 𝑎𝑠𝑠𝑒𝑡𝑠 =𝑟𝑖𝑠𝑘𝑦 𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑤𝑒𝑎𝑙𝑡ℎ
3.3.2. Happiness
This paper takes three measures for the independent variable ‘happiness’. All
questions on these measures are drawn from the questionnaire of the ‘Personality’ in
the LISS Panel. The first variable ‘subjective well-being’ is used as the main measure
of happiness in this paper and is estimated by merging proxies of positive affect,
negative affect together with a proxy of life satisfaction in accordance with the
theoretical basis (Diener, 1984). The proxies are based on questions related to affect
and life satisfaction.
Questions on the affect are drawn from the Positive-Negative Affect Scale
(PANAS) developed by Watson (1988), and questions on life satisfaction are drawn
from the Satisfaction with Life Scale (SWLS) developed by Diener et. al (1985). The
score of the subjective well-being is calculated by aggregating the score of life
satisfaction and the net affect (Libran, 2006). Before numeration, the values of those two variables are transformed into the form of standardized z-score since the corresponding questions are answered on different scales and scored differently.
General happiness and current life satisfaction are the two other measures of happiness used in this paper and they are used as alternative measures to provide
robustness for the main test. From a psychological view, subjective well-being is a
combination use of general happiness (affective evaluation) and current life satisfaction
3.3.3. The Big Five Personality Traits
In this study, we use the big five personality traits taxonomy (extraversion,
agreeableness, conscientiousness, neuroticism, and openness to experience) advanced
by Costa and McCrea (1992) to classify individual personality traits. The taxonomy
summarized stable dispositions for a long time in one’s life (Costa and McCrea, 1992).
It is considered to be the broadest and most comprehensive representation of the
personality traits (Almlund et. al, 2011) and is wildly used in economic and
psychological studies (Deck et al., 2008). There is plenty of evidence showing that the
big five personality traits play significant roles in financial behavior. Specifically,
Brown and Taylor (2011) discover that the stable dispositions represented by the big
five personality traits are significantly correlated with unsecured debt.
In this paper, the questions on the big five personality traits are drawn from the ‘International Personality Item Pool’ (IPIP) developed by Goldberg (1999). The questionnaire of the big five personality traits contains 50 questions. And each personality trait is described by 10 questions with the answers scored on a 5-point Likert scale (1932). An example question for the personality traits ‘openness to experience’ is: “I have difficulty understanding abstract ideas.”
3.3.4. Control variables
To accurately estimate the relationship between main variables, a number of
factors are considered as control variables. We take the factors which are considered to
have an essential influence on the investment in risky assets as control variables in this
research.
Firstly, basic demographic and socio-economic information as age, gender,
marital status, whether having children, net household income and work status are used
For instance, Charness and Gneezy (2007) collect data from 15 experiments to explore
the gender difference in the willingness to take risks in investment decisions. The
findings indicate that males are more willing to take financial risks and thereby invest
more in risky assets. Furthermore, we use net household income as a control. It
measures the monthly net income (in Euros) of all household members combined as
stated in LISS Panel. To correct the skewness of this data, the inverse hyperbolic sine
transformation is used. It should be mentioned that it would be better to use the net
amount of household asset as a control variable in this research since it is intuitive to
estimate the proportion of risky assets within their net assets. However, there is a high
shortage of data on loans, mortgage, and insurance to create this variable.
Secondly, education level, self-reported health status and taking care of
household financial matter are also considered to be control variables. Specifically, we
can expect that people with higher levels of education have the ability to learn about
the financial market. Campbell (2006) uses data of 2001 to study investment behavior
in the United States. He finds a significant positive effect of the education level on
equity participation. The variable of education level in this study is a 6-scale category
variable based on the CBS (Statistics Netherlands) classifications as involved in LISS
panel.
Similarly, people who consider themselves healthy would not worry much about
potential medical expenditure. Rosen and Wu (2004) examine the role of self-reported
health status in financial decisions. The finding demonstrates that the self-reported health status is an essential determinant for risky assets participation: people who consider themselves as healthy are likely to possess risky assets and to hold a large share of risky assets as well. In this study, self-reported health status is a 5-scale
3.3.5. Descriptive statistics
Table 1 represents the descriptive statistics of the variables for the total dataset.
As shown in the table, only 14.3% of the whole sample possess risky assets as part of
their financial assets. Whereas for the index of portfolio proportion, the mean value is
around 66.89%, accounting for more than half of their financial assets. It can be
concluded that, although the households rarely possess risky assets, the risky assets
account for a relatively high proportion of the household financial assets for the
households with risky assets.
The education level and self-reported health status are classified by category.
Whereas gender, marital status, children living at home, work status, whether taking
care of the household financial matter and whether possessing real estates are treated
as dummy variables. The data collect 13,041 observations and contain family members
Table 1. Descriptive statistics
This table presents the descriptive statistics of the initial database. The sample period runs from 2009-2015. Inverse hyperbolic sine transformation is used to process the net household income data. The category variables are presented by the number of observations and the corresponding
percentage.
Dependent variables N Mean Std.Dev. Min Max
Possessing risky assets 13,041 0.1432 0.3503 0 1
Share of risky assets 5,625 0.6689 0.4389 0 1
Control variables N Mean Std.Dev. Min Max
Age 13,041 49,9656 17.2458 16 96
Net household income (ln) 11,960 7.9643 0.8520 0 13
Male (dummy) 13,041 0.4674 0.4990 0 1
Married (dummy) 13,041 0.5852 0.4927 0 1
Having children (dummy) 13,041 0.5858 0.4926 0 1
Employed work (dummy) 13,041 0.4484 0.4974 0 1
Taking care of financial matter (dummy) 12,861 0.5806 0.7553 0 1
Possessing real estate (dummy) 13,041 0.0580 0.4935 0 1
Category variables N Percentage
Education level:
primary school 1,279 9.83%
intermediate secondary education 3,407 26.18%
higher secondary education 1,454 11.17%
intermediate vocational education 2,932 22.53%
higher vocational education 2,906 22.33%
3.4. Baseline models
The baseline models are constructed as follow:
First, we investigate the relationship between participation in the market for risky
financial assets and the household member’s happiness. According to Brooks (2014), ordinary
least squares regressions are not adequate to estimate the binary dependent variables, and the
Probit model which forms an S-shape probability line can be used for estimations of binary
dependent variables since it suits the case in real life. To implement a Probit model, we assume
that the endogenous variable is continuous. The interaction terms between happiness and
personality traits are then introduced in the model to test for the moderation effect of the big
five personality traits. 𝑌∗= 𝛼 + 𝛽 1𝑋1+ 𝛾1𝑋3+ 𝑒 (1) 𝑌1= {1 𝑌∗> 0 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 (2) 𝑌1= 𝛼 + 𝛽1𝑋1+ 𝛾1𝑋3+ 𝜀 , (3) 𝑌1= 𝛼′+ 𝛽 1′𝑋1+ 𝛽2′𝑋2+ 𝛽3′𝑋1𝑋2+ 𝛾1′𝑋3+ 𝜀′ , (4)
where 𝑌∗ represents the actual amount of risky assets held by each household, Y represents whether the household possesses any risky assets, 𝑋1 measures the happiness level of the household member, 𝑋2 is the vector of the household member’s big five personality traits, 𝑋3 is the matrix of control variables, and 𝜀 is the error term.
Alternatively, we explore the quantitative relation between the risky asset shares relative to the total financial asset holdings and house member’s happiness through the Tobit censored regression. According to Brooks (2014), the Tobit analysis can accurately make the estimation
of models with dependent variable censored at a certain zone. We use this approach in this
study to limit the short selling behavior, and the censoring share of risky assets ranges from 0
to 1. The interaction terms between happiness and personality traits are introduced in the
𝑌2 = { 𝑌2∗ 0 < 𝑌2∗ < 1 0 𝑌2∗≤ 0 1 𝑌2∗≥ 1 (5) 𝑌2= 𝛼 + 𝛽4𝑋1+ 𝛾2𝑋3+ 𝜀 , (6) 𝑌2= 𝛼′+ 𝛽 4′𝑋1+ 𝛽5′𝑋2+ 𝛽6′𝑋1𝑋2+ 𝛾2′𝑋3+ 𝜀′ , (7)
where 𝑌2∗ represents the proportion of risky assets in household financial wealth and 𝑌2 is used as the dependent variable in the regression models. 𝑋1 measures the happiness level of the household member, 𝑋2 is the vector of the household member’s personality traits, and 𝑋3 is the matrix of control variables, and 𝜀′ is the error term.
3.4.1 Endogeneity and Instrument variable
In this section, the potential problem of endogeneity is discussed. Endogeneity is a common issue occurring in economic empirical studies. Generally, endogeneity describes the situation that there exist correlations between the error terms and the independent variables, and thus the estimations will be inconsistent (Wooldridge, 2009). According to Guven and Hoxha (2015), the study of the impact of happiness on risk-taking behavior is often accompanied by the endogeneity caused by the omitted variables and reverse causality. Therefore, in this study, we mainly discuss the potential reverse causality and the omitted variable which are generally considered to be the sources of the endogeneity.
The omitted variable bias
In economic models, there can be factors which are correlated with the independent
variable and have an influence on the dependent variable but are not included in the model. If factors like those are ignored when constructing the model, the model might suffer from an omitted variable bias which implies that there is a correlation between the independent variable and the error terms, and this results in inconsistent estimators (Brooks, 2014).
For instance, in this study, the optimistic expectations for the economy may lead
be happy and satisfied about the future. If this were the case, happiness in this situation is both
the result of the increasing optimistic expectations and one determinant for the investment in
risky assets, and the optimistic expectations for the economy may be the omitted variable.
The reverse casualty bias
In studies of the causal relationship, reverse casualty often occurs to bias the estimations
(Katz, 2006). In regression models, we assume that the independent variable exerts influence
on the dependent variable, not the other way around. However, due to the existence of the
reversing impact, the estimations are always biased (Brooks, 2014).
As for this study, it is obvious to notice that it may not be a definite one-way relationship between happiness and investments in risky assets. For one thing, the level of happiness may affect one’s risk tendency towards investment in risky assets. And for another, the financial returns on the risky investment also influence their degree of happiness. Thus, the estimations of the impacts of happiness on risky investment may be biased by the reverse impacts. Combing the arguments together, we preliminary suspect that there is an endogeneity problem from the theoretical perspective.
Instrumental variable
In order to further check and avoid the endogeneity, we need to first introduce an instrumental variable for ‘happiness’ (Brooks, 2014). Studies in the literature use “social contact” factors as instrumental variables for ‘happiness’. Specifically,Sabatini (2011) uses “quality of friendship” and “social trust” as instrumental variables to analyze the relationship between happiness and health. The study finds that both of the variables are strongly correlated with happiness. Interactions with people are also found to be significantly associated with subjective well-being (Helliwell and Putnam, 2005).
In this paper, we adopted the instrumental variable ‘family relation’ following Delis and
Mylonidis (2015). The existence of the endogeneity problem and validity of this instrument
separate the independent variable (happiness) into an endogenous and an exogenous part, and
thus the exogenous part can be further used to estimate the one-way relationship. This
approach is technically termed as the “Two-Stage Least Squares Regression Analysis” (Rivers
and Vuong, 1988). Combined with the Probit regression and Tobit regression used in the
baseline models, this process can be executed in Stata.
4. Analysis and Results
4.1. Frequency table for risky asset market participation
Table 2 presents the two-way frequency table with the percentage of risky asset market
participation on each level of happiness. The total number of observations is 13,041. Samples
are selected by dropping the missing data on the main variables. The sample size for the year
2015 is relatively small compared to those for other years, which is caused by filtering out the
samples with missing data on the personality traits in that year. Due to the small sample of the
observations in the year 2015, the results can lead to a bias. The numbers shown in Table 2
represents the percentage of people possessing risky assets among all the respondents within
a certain level of happiness for the particular year. Happiness is measured by subjective
well-being which is a numeric variable that is calculated by summing positive affect, negative affect,
and life satisfaction after standardizing and it is coded into the category variable as follow: ‘extremely unhappy’ (1), ‘unhappy’ (2), ‘neutral’ (3), ‘happy’ (4) and ‘extremely happy’ (5). The table compares the risky asset market participation rate in terms of happiness level in the
year 2009, 2011, 2013 and 2015. Overall, the probability of holding risky assets is located
below 19%. Generally speaking, the risky asset market participation rate increases with a
higher level of happiness. However, it is notable that for the year 2011 and 2015 the people
with the highest level of happiness do not display the highest participation rate. This table
Table 2. Cross-tabulation of happiness on risky asset market participation (frequency)
This table presents the rates of risky assets market participation in different levels of happiness. In the first column, the numbers represent levels of happiness: extremely unhappy-1, unhappy-2, neutral-3, happy-4, and extremely happy-5. For each year, the first column presents the percentage of people possessing risky assets and the second column presents the corresponding numbers of observations. The sample contains a total of 13,041 observations which are based on Wave 2, 4, 6 and 8 of the Personality database from the LISS panel.
2009 2011 2013 2015 Total
Possess Observation Possess Observation Possess Observation Possess Observation Possess Observation
4.2. Validity of the big five personality traits - Cronbach’s Alpha
Each personality trait in International Personality Item Pool is based on 10
descriptive items as involved in the LISS panel. To test the consistency and validity of the items, we use Cronbach’s alpha. The Cronbach’s alpha developed by Lee Cronbach can examine the reliability of the phycological test by calculating the degree under
which the items are measuring the same topic (Cortina,1993). The higher the results,
the higher internal consistency can be perceived. Normally, an alpha no less than 0.7 is
considered to be acceptable for the internal consistency and a value between 0.8 and
0.9 is thought to be good (Griethuijsen et. al, 2014). Table 3 represents the results of the Cronbach’s alpha, the mean value and standard deviation for each personality. As shown in the table, all values are above 0.7, indicating that all the indexes are reliable.
Table 3 also presents a high mean for conscientiousness and a low mean for
extraversion, showing that the individuals in the sample are less outgoing, self-absorbed
and self-disciplined overall.
Table 3. Cronbach’s alpha for the big five personality traits
Big five personality traits Cronbach’s alpha Mean Std.
Extraversion 0.86 2.7961 0.6949
Agreeableness 0.81 2.9095 0.6413
Conscientiousness 0.79 3.2619 0.5960
Neuroticism 0.88 3.0089 0.7363
Openness to experience 0.76 3.0148 0.5579
4.3. The endogeneity problem and the validity of instrument variable
As mentioned in part 3, an endogeneity problem may exist. Before conducting the
main modeling, we need to confirm the presence of the endogeneity and test the validity
variable ‘family relation’ on subjective well-being as the first-stage of the least square regressions. Results are shown in Table 4.
Table 4. First-stage results of the 2SLS regression (Presence of the endogeneity)
The table presents the results of the first-stage regression which tests the endogeneity of happiness. The sample contains a total of 8,728 household member observations from the year of 2009, 2011, 2013 and 2015. Happiness is regressed on all control variables as well as the instrument variable “family relation”. Happiness here is measured by the main proxy “subjective well-being”. Coefficients, the corresponding significant level and standard errors are reported. P-values smaller than 0.01, 0.05 and 0.10 are indicated by ***, **, and *, respectively.
Subjective well-being First-stage results of the 2SLS
Constant 2.2672*** Age 0.0185*** Male -0.0210 Married 0.3929*** Having children 0.1225** Employed work 0.1859*** Education 0.0351***
Net household income (ln) 0.1037***
Possessing real estate -0.0103
Taking care of the financial matter -0.0874**
Self-reported health status 0.6440***
Family relation 0.3672***
Observations 8,728
F-statistic 183.18***
It is noticeable that happiness is strongly correlated with most of the control
variables. Specifically, married status and self-reported health status exhibit a high level
of correlations with subjective well-being. The coefficient on family relation is 0.3707,
which is significantly positive as supposed. Meanwhile, the F-statistic value at the
bottom is qualified for the rule of thumb that F should be bigger than 10. The results
confirm the presence of endogeneity and validate the reliability of the instrument
4.4. Empirical results
This section presents the results of the main analysis. The first part presents the
results of the relationship between happiness and household risky assets allocation. The
second part further reveals the moderation effects of the big five personality traits based
on the relationship between happiness and household risky assets allocation.
4.4.1 The effect of happiness on risky asset allocation decisions.
In this part, hypothesis 1a and hypothesis 1b are empirically tested. Table 5 reports
the Probit regression and Tobit regression for risky asset allocations using the instrument variable “family relation”. Controlling for the demographic and socio-economic factors, Panel A provides an overview of the impact of happiness on risky
asset market participation, and Panel B reveals the impact of happiness on the share of
financial wealth held in risky assets.
Firstly, at the bottom of Table 5, the statistics of the Wald test for the instrument
variable are significant. Thus, the null hypothesis of no endogeneity is rejected at 5%
significant level. The results further confirm that the instrument variable has the
explanatory power and that a weak instrument problem does not exist. Secondly, as can
be seen from table 5, all the control variables, except for married status, are showing
significant effects on the household risky assets allocation.
Hypothesis 1a predicts a negative relation between happiness and risky assets
market participation. Hypothesis 1b predicts a negative relation between happiness and
the share of financial wealth held in risky asset. As shown in Table 5, all coefficients
for the proxies of happiness are negative and statistically significant at the 5% level.
Table 5 Probit and Tobit regression results of happiness on the risky asset allocation
Panel A shows the IVProbit regressions which test the hypotheses 1a on the relationship between happiness and the risky assets market participation. Panel B shows the IVTobit regressions which test the hypotheses 1b on the relationship between happiness and the relative share of risky assets. All regressions are controlled for demographic and eco-social factors. The sample contains a total of 8,728 household member observations from the year of 2009, 2011, 2013 and 2015. The first, second and third column in each panel respectively uses subjective well-being, general happiness and current life satisfaction as the proxy for happiness. Wald test is used to check the endogeneity. Coefficients and the corresponding significant level are reported. P-values smaller than 0.01, 0.05 and 0.10 are indicated by ***, **, and *, respectively.
Panel A—Risky assets market participation Panel B — Relative share of the risky assets
IVProbit (1) IVProbit (2) IVProbit (3) IVTobit (1) IVTobit (2) IVTobit (3)
Subjective well-being -0.134** -0.264**
General happiness -0.133** -0.277**
Current life satisfaction -0.135** -0.262**
Constant -3.706*** -3.640*** -3.623*** 3.523*** 3.797*** 3.617*** Age 0.015*** 0.014*** 0.014*** 0.006* 0.002 0.003 Male 0.254*** 0.263*** 0.257*** -0.137** -0.146** -0.148** Married 0.036 0.035 0.041 -0.037 -0.034 -0.024 Having children -0.073* -0.063 -0.062 0.095*** 0.123*** 0.129*** Employed work 0.102** 0.084** 0.090** 0.084 0.037 0.044 Education level 0.180*** 0.178*** 0.175*** -0.114*** -0.125*** -0.125*** Net income 0.162*** 0.161*** 0.160*** -0.033 -0.032 -0.031
Possessing real estate 0.631*** 0.639*** 0.641*** -0.425*** -0.394*** -0.390***
Taking care of financial matter 0.286*** 0.294*** 0.285*** -0.120* -0.135** -0.137**
Self-reported health status 0.184*** 0.167*** 0.172*** 0.0932 0.071 0.069
Observations 8,728 8,597 8,630 3,984 3,949 3,968
Wald test of exogeneity 6.80*** 6.28** 8.70*** 3.34* 3.27* 3.28*
First-stage results
happiness and current life satisfaction, the probability of possessing risky assets will
respectively decrease by 13.4%, 13.3% and 13.5%, and the probability of buying more
risky assets will respectively decrease 26.4%, 27.7% and 26.2% for individuals. The
findings for all three proxies of happiness confirm hypothesis 1a and hypothesis 1b
which are consistent with the Mood-Maintenance Theory.
4.4.2 Moderation effects of the big five personality traits
As for the second part of this section, we further explore whether the negative
relationship between happiness and household risky assets allocation is moderated by
the big five personality traits. Table 6 to 10 present the results of model (4) and model
(7) which include personality traits as well as the interaction terms between the
personality traits and happiness. The tables report the Probit regression and Tobit
regression for the moderation effect of the five personality traits on the relationship
between happiness and household risky assets allocation. The moderation effects for
extraversion is presented in Table 6, agreeableness in Table 7, conscientiousness in
Table 8, neuroticism in Table 9 and openness to experience in Table 10.
As shown in Panel A (Table 6), the extraversion trait shows no significant
moderation effect on the relationship between happiness and risky asset market
participation. The coefficients for the interaction terms between each proxy of
happiness and the extraversion trait are all positive but insignificant. Thus, no evidence
is provided for hypothesis 2a.
Panel B shows the relationship between happiness and the relative share of risky
assets. In this panel it can be seen that the significant negative coefficients exist for the
interaction terms between extraversion and the general happiness as well as the current
life satisfaction. At the same time, the main effects of both proxies for happiness
the coefficient for current life satisfaction is -0.090 and is statistically significant at the
1% level. The coefficient for extraversion is 0.024 and is statistically significant at the
10% level. The interaction between them has a coefficient of -0.019 which is significant
at the 10% level. Similarly, the coefficient for general happiness is -0.097 and is
statistically significant at the 1% level. The coefficient for extraversion is 0.028 and is
statistically significant at the 10% level. The interaction term between them has a
coefficient of -0.031 which is significant at 1% level. From a mathematical view, with
one unit increase in life satisfaction, there will be a change of
(-0.090-0.019*extraversion) % in the probability of willingness to purchase more risky assets.
With one unit increase in general happiness, there will be a change of
(-0.097-0.031*extraversion) % in the probability of willingness to purchase more risky assets.
This is in contradiction to the hypothesis 2b, that the extraversion trait weakened the
negative impact of happiness on the decision to increase the share of risky assets.
Results indicate that for extraverted people with higher levels of general happiness or
current life satisfaction, the negative effect of happiness on the share of risky assets is
more evident. Nevertheless, as for the main proxy “subjective well-being”, no evidence
is found for hypothesis 2b. The results are inconclusive.
Table 7 presents the results for the moderation effect of the moderator “agreeableness” as mentioned in hypotheses 3a and 3b. All coefficients for the interaction terms are negative, however, no significance is found. This is similar to
Table 10, which shows the findings for the moderation effect of the moderator “openness to experience trait” as hypothesized in hypotheses 6a and 6b. No significant moderation effect of the interaction term was shown on either one of the investment
Table 6 Regression results of the moderation effects of the extraversion trait on the relationship between happiness (subjective well-being, happiness and life satisfaction) and the risky assets allocation
Panel A shows the IVProbit regressions which test the hypotheses 2a: the moderation effect of extraversion on the relationship between happiness and the risky assets market participation. Panel B shows the IVTobit regressions which test the hypotheses 2b: the moderation effect of extraversion on the relationship between happiness and the relative share of risky assets. All regressions are controlled for demographic and eco-social factors. The sample contains the household member observations from the year of 2009, 2011, 2013 and 2015. Column (1a) and (1b) uses subjective well-being as the proxy for happiness, column (2a) and (2b) uses the general happiness as the proxy, and column (3a) and (3b) uses current life satisfaction as the proxy. Subjective well-being is abbreviated to SWB. The interaction term between the indicators of happiness and extraversion traits are labeled as Mod.E. Coefficients and the corresponding significant level are reported. P-values smaller than 0.01, 0.05 and 0.10 are indicated by ***, **, and *, respectively.
Panel A —Risky assets market participation Panel B — Relative share of the risky assets
(1) (1a) (1b) (2a) (2b) (3a) (3b) (1) (1a) (1b) (2a) (2b) (3a) (3b)
Table 7 Regression results of the moderation effects of the agreeableness trait on the relationship between happiness (subjective well-being, happiness and life satisfaction) and the risky assets allocation
Panel A shows the IVProbit regressions which test the hypotheses 3a: the moderation effect of agreeableness on the relationship between happiness and the risky assets market participation. Panel B shows the IVTobit regressions which test the hypotheses 2b: the moderation effect of agreeableness on the relationship between happiness and the relative share of risky assets. All regressions are controlled for demographic and eco-social factors. The sample contains the household member observations from the year of 2009, 2011, 2013 and 2015. Column (1a) and (1b) uses subjective well-being as the proxy for happiness, column (2a) and (2b) uses the general happiness as the proxy, and column (3a) and (3b) uses current life satisfaction as the proxy. Subjective well-being is abbreviated to SWB. The interaction term between the indicators of happiness and agreeableness traits are labeled as Mod.A. Coefficients and the corresponding significant level are reported. P-values smaller than 0.01, 0.05 and 0.10 are indicated by ***, **, and *, respectively.
Panel A — Risky assets market participation Panel B — Relative share of the risky assets
Table 8 Regression results of the moderation effects of the conscientiousness trait on the relationship between happiness (subjective well-being, happiness and life satisfaction) and the risky assets allocation
Panel A shows the IVProbit regressions which test the hypotheses 4a: the moderation effect of conscientiousness on the relationship between happiness and the risky assets market participation. Panel B shows the IVTobit regressions which test the hypotheses 3b: the moderation effect of conscientiousness on the relationship between happiness and the relative share of risky assets. All regressions are controlled for demographic and eco-social factors. The sample contains the household member observations from the year of 2009, 2011, 2013 and 2015. Column (1a) and (1b) uses subjective well-being as the proxy for happiness, column (2a) and (2b) uses the general happiness as the proxy, and column (3a) and (3b) uses current life satisfaction as the proxy. Subjective well-being is abbreviated to SWB. The interaction term between the indicators of happiness and conscientiousness traits are labeled as Mod.C. Coefficients and the corresponding significant level are reported. P-values smaller than 0.01, 0.05 and 0.10 are indicated by ***, **, and *, respectively.
Panel A — Risky assets market participation Panel B — Relative share of the risky assets
Table 9 Regression results of the moderation effects of the neuroticism trait on the relationship between happiness (subjective well-being, happiness and life satisfaction) and the risky assets allocation
Panel A shows the IVProbit regressions which test the hypotheses 5a: the moderation effect of neuroticism on the relationship between happiness and the risky assets market participation. Panel B shows the IVTobit regressions which test the hypotheses 5b: the moderation effect of neuroticism on the relationship between happiness and the relative share of risky assets. All regressions are controlled for demographic and eco-social factors. The sample contains the household member observations from the year of 2009, 2011, 2013 and 2015. Column (1a) and (1b) uses subjective well-being as the proxy for happiness, column (2a) and (2b) uses the general happiness as the proxy, and column (3a) and (3b) uses current life satisfaction as the proxy. Subjective well-being is abbreviated to SWB. The interaction term between the indicators of happiness and neuroticism traits are labeled as Mod.N. Coefficients and the corresponding significant level are reported. P-values smaller than 0.01, 0.05 and 0.10 are indicated by ***, **, and *, respectively.
Panel A — Risky assets market participation Panel B — Relative share of the risky assets
Table 10 Regression results of the moderation effects of the openness to experience trait on the relationship between happiness (subjective well-being, happiness and life satisfaction) and the risky assets allocation
Panel A shows the IVProbit regressions which test the hypotheses 6a: the moderation effect of openness to experience on the relationship between happiness and the risky assets market participation. Panel B shows the IVTobit regressions which test the hypotheses 6b: the moderation effect of openness to experience on the relationship between happiness and the relative share of risky assets. All regressions are controlled for demographic and eco-social factors. The sample contains the household member observations from the year of 2009, 2011, 2013 and 2015. Column (1a) and (1b) uses subjective well-being as the proxy for happiness, column (2a) and (2b) uses the general happiness as the proxy, and column (3a) and (3b) uses current life satisfaction as the proxy. Subjective well-being is abbreviated to SWB. The interaction term between the indicators of happiness and openness to experience traits are labeled as Mod.O. Coefficients and the corresponding significant level are reported. P-values smaller than 0.01, 0.05 and 0.10 are indicated by ***, **, and *, respectively.
Panel A — Risky assets market participation Panel B — Relative share of the risky assets
Table 8 provides the findings for the moderation effect of “conscientiousness” as
mentioned in hypotheses 4a and 4b. As shown in table 8, all coefficients for the
interaction terms are insignificant except for the interaction term between subjective
well-being and conscientiousness towards the relative share of risky assets. More
Specific, the coefficient for subjective well-being is -0.095 and is statistically
significant at the 1% level. The coefficient for conscientiousness is 0.071 and is
statistically significant at the 5% level. The interaction between them has a coefficient
of 0.013 which is significant at the 10% level. From a mathematical view, with one unit
increase in subjective well-being, there will be a change of (-0.095 + 0.019 *
conscientiousness) % in the probability of willingness to purchase more risky assets.
For conscientious people with higher levels of subjective well-being, the negative effect
of happiness on the share of risky assets is less evident. This provides evidence that
conscientiousness trait interacts with an individual’s subjective well-being to weaken
the negative impact of happiness on the decision to increase the share of risky assets,
which is in contradiction to hypothesis 4b. However, the alternative measures for
happiness do not exhibit a similar result. Thus, the results are inconsistent.
Table 9 provides the estimation results for the moderation effect of “neuroticism”
as mentioned in hypotheses 5a and 5b. It is obvious to note that all coefficients for
interaction terms are significantly negative. At the meantime, the coefficients for all
proxies of happiness are significantly negative. Coefficients for neuroticism are all
positive at the 1% significance level. From a mathematical view, a unit increase in
subjective well-being will lead to a change of (-0.276-0.093*neuroticism) % in the
probability of willingness to possess risky assets and a change of (- 0.149 - 0.023 *
neuroticism) % in the probability of willingness to purchase more risky assets. A unit
neuroticism) % in the probability of willingness to possess risky assets and a change of
(- 0.111 -0.025 * neuroticism) % in the probability of willingness to purchase more
risky assets. A unit increase in current general happiness will lead to a change of (-
0.234 -0.124 * neuroticism) % in the probability of willingness to possess risky assets
and a change of (- 0.117 - 0.029 * neuroticism) % in the probability of willingness to
purchase more risky assets. In general, the findings indicate that for neurotic people
with higher levels of happiness, the negative effect of happiness on the decisions to
possess risky assets and increase the share of risky assets are strengthened. These
findings support hypotheses 5a and 5b.
5. Conclusion and discussion
Based on the Affect-Infusion-Model and Mood-Maintenance Hypothesis, this
study aims to address the impact of happiness on household asset allocations. Moreover,
we explore the potential moderation effects of the big five personality traits on the
relationship between happiness and household asset allocations to further clarify the
inconsistent findings. The results indicate that higher levels of happiness inhibit the
investment in risky assets for individuals in the Netherlands and the neuroticism trait
strengthen this effect.
We use a sample of 11,210 Dutch household observations from the year 2009,
2011, 2013 and 2013 to conduct the empirical models. Firstly, we find a significant
negative impact of happiness on risky investments. Specifically, happiness is mainly
measured by subjective well-being and alternatively, it is measured by general
happiness and current life satisfaction to provide robustness. The findings provide
evidence for the Mood-Maintenance Hypothesis (Isen and Patrick, 1983), which
impact of happiness on risk-taking behavior are in line with Raghunathan and Pham
(1999), Grable and Roszkowski (2008), and Delis and Mylonidis (2015).
Secondly, we further explore the moderation effects of the big five personality
traits on the relationship between happiness and investments in risky assets, which is
also the main contribution of this study. The results indicate that the neuroticism trait
strengthens the negative relationship between happiness and investments in risky assets.
In other words, neurotic happy people show less tendency to possess or invest more in
risky assets. This finding applies to all three measures of happiness. Additionally, the
conscientiousness trait also weakens the negative relations between subjective
well-being and the share of risky assets. However, this does not work for the other measures
of happiness, and thus the results are inconclusive. The alternative measures, general
happiness and current life satisfaction respectively interact with the extraversion trait,
and the extraversion trait strengthens the negative relations between happiness and the
share of risky assets. However, since this is not true for subjective well-being, the
results are also inconclusive. The moderation effect of the conscientiousness and the
extraversion can be further explored in a future analysis by refining the composition of
happiness.
This study uses subjective well-being as a composite index of the affective and
cognitive evaluations and calculates a general single score for the index as suggested
by Libran (2006). Although this approach can measure all the components in a single
score, it might also be inadequate since the commonality between affective evaluations
and cognitive evaluation components might get neglected (Busseri, 2015). Further
studies can focus on developing a more complete empirical approach of the overall
subjective well-being to estimate the potential relationships. Moreover, the lack of data