The Big Five, financial market participation and risk-taking behaviour.
Abstract
This paper analyses whether intrinsic personality traits influence financial market participation and risk-taking behaviour. With the use of the Big Five personality traits (openness, conscientiousness, extraversion, agreeableness and neuroticism) and eleven sub-traits the relationship with financial market participation and risk-taking behaviour is examined. Data from the Survey of Health, Ageing,
and Retirement in Europe (SHARE) is used. The results show that neuroticism negatively influences the likelihood of financial market participation. The opposite holds for the personality trait openness.
Regarding risk-taking behaviour, individuals with an open personality have a preference for stocks. Several sub-traits also influence the likelihood of financial market participation, risk-taking behaviour
and a preference for bonds, stocks or mutual funds.
Keywords:
Financial market participation Big Five personality traits
Household finance Behavioural finance
JEL Classification: D14, G11, G41
Introduction
Why do some individuals invest in financial markets and why do some individuals not invest in financial markets? The equity premium puzzle (Mehra & Prescott, 1985.) shows us that it would be wiser to invest (more) money in financial markets, however, a lot of people restrain from doing so, henceforth the puzzle. The equity premium puzzle especially relates to the large differences in returns when comparing equity portfolios and bonds.
This study tackles the equity premium puzzle from a behavioural point of view. Humans have different intrinsic personality characteristics and this may influence whether they are inclined to participate in the financial markets. Are certain behavioural traits linked to the avoidance/entrapment of the equity premium puzzle? With the use of the Big Five personality traits model, this study explores whether behavioural traits can explain the equity premium puzzle.¹ The Big Five personality traits are openness, conscientiousness, extraversion, agreeableness and neuroticism. This paper addresses two questions regarding the equity premium puzzle. How does the Big Five influence the likelihood of financial market participation and does it influence the distribution of financial instruments that are held (i.e. risk-taking)?
Benartzi & Thaler (1995) developed the concept of myopic loss-aversion to explain the equity premium puzzle. Myopic loss-aversion entails that individuals are highly sensitive to losses and have a tendency to frequently control their wealth, henceforth leading to non-participation and a high equity premium. A potential shortcoming in their study is the assumption that all individuals suffer from myopic loss-aversion, although this may not be the case. Especially regarding the current market and its (negative) interest rate, myopic loss-aversion comes to be a strange concept. If individuals are highly sensitive to losses, they would not deposit their money at a bank, as this guarantees a loss of money with even the tiniest bit of inflation. The same holds for bonds, as the current interest rate on secure bonds is negative. This study adds value by linking the Big Five to investor behaviour. Particular personality traits may lead to financial market participation and the amount of risk an individual is willing to take.
Multiple studies have looked at personality traits and financial market participation (Bucciol & Zarri (2011); Brown and Taylor (2014); Conlin et alli. (2015) and Oehler et alli. (2018)). Only Brown & Taylor (2014) have used the Big Five as independent variables, for stock market participation in the British market. With the use of the vast cross-national SHARE dataset, this study is the first to look at other countries than the United Kingdom. Furthermore, all individuals in the dataset are aged over 50, when personality traits are considered to be in their most stable state (Bucciol & Zarri, 2011.). This study builds upon the study of Brown and Taylor (2014), as it not only checks the validity of their sample by extending it to continental Europe, a deeper analysis of risk-taking behaviour will also be performed.
Literature review
Mehra and Prescott (1985) investigated the large differences in returns between equity and bonds and developed the now famous equity premium puzzle. The puzzle, as the name implies, refers to the significant difference in average yields between equity and bonds, with equity having the larger yield, the equity premium. With the use of several models, Mehra and Prescott (1985) cannot explain the large difference in yields nor can the standard neoclassical models. They conclude that stocks should have, at most, a one percentage point higher return than bonds, while historically this figure has been around six percentage points. The equity premium puzzle is closely related to people investing in the financial markets in the first place. Hong et alli. (2004) states: “ The participation rate can have a direct effect on the equity premium; thus an understanding of what drives participation can help shed light on the equity-premium puzzle of Mehra and Prescott (1985).”
The social and psychological aspects of financial market participation have also been extensively researched. Hong et alli (2004)., Brown et alli (2008)., Guiso et alli. (2008) and Angelina & Cavadozzi (2017) provide useful insights. Hong et alli (2004) and Brown et alli. investigate peer effects. Peer effects refer to people close to you and how one could easily compare their own situation to theirs. To give an example, if everybody in your neighbourhood is prospering, it is unlikely that you want to stay behind compared to their welfare. Households who (closely) interact with their neighbours have an increased likelihood to invest in the stock market. The same holds for households who attend church (Hong et alli. (2004)). Whether the increased stock market participation for church-goers is solely due to the social effects remains up for debate. Religious, cultural, personal or psychological factors may also have an influence here. Another finding by Hong et alli. (2004) is that the sociability factor appears to be stronger in states where the stock market participation rates are higher. This may be attributed to word-of-mouth or a general interest in talking about the stock market with friends, as one would also talk about sports, films or other subjects of interest. Brown et alli. (2008) add to the findings by Hong et alli. (2004) by investigating communities. Their findings are in line with Hong et alli. (2004) and also find that in sociable communities the effects are larger due to word-of-mouth.
One of the possible psychological explanations is provided by Guiso et alli. (2008). An analysis and investigation into the amount of trust that individuals have, shows that less-trusting people are more unlikely to participate in the financial markets than trusting people. The amount of trust that one has largely originates from their family background, which is of course subject to demographics, culture, social standing and many other factors(Please see, Banfield (1958)). One may argue that there are missing pieces to the puzzle here. Since family background is incredibly complicated and has many (unknown) factors, it is impossible to safely conclude that every factor has been accounted for. However, the expectation for this research is that trust will positively influence the likelihood of financial market participation. Angelina and Cavapozzi (2017) look into dispositional optimism, the tendency to expect positive outcomes. When it comes to risk-averse investors, dispositional optimism makes no difference regarding financial market participation or the percentage of wealth invested. For risk-tolerant investors, being extremely optimistic instead of extremely pessimistic, a 16 percentage points increase in the probability of stock ownership is observed.
The Big Five can add to the current literature regarding the equity premium puzzle. Are there intrinsic fundamental behavioural traits that influence the decision to participate in financial markets and whether to invest in stocks or bonds? Fundamental behavioural traits might be responsible for the large premium, as individuals might not be willing to invest in equity due to these traits, henceforth driving up the premium. Furthermore, this study provides a deeper insight regarding myopic loss-aversion, because fundamental behavioural traits can be linked to this phenomenon.
linked to the probability of holding shares. For single individuals, agreeableness is also inversely related to the probability of owning shares. Important to note is information about the viability of using the Big Five as measurement. Brown and Taylor (2014) state: “In the psychology literature, it has been argued that the personality traits included in the Big Five taxonomy are stable over the life cycle (see, for example, Caspi, Roberts, & Shiner, 2005 and Borghans et al., 2008). There is however still some debate in the literature. For example, Almlund et al. (2011) conclude that personality traits do change over the life cycle.”
Conlin et alli. (2015) show that certain traits and sub-traits are indicators of stock market participation for Finnish investors. They use Cloninger’s (1993) Temperament and Character Inventory for traits. Novelty seeking, in general, has a positive effect on stock market participation. However, certain sub-traits of novelty seeking have large negative effects on stock market participation (e.g. extravagance). Furthermore, they find that higher sentimentality scores reduce the chance of participation and higher dependence increases the chance of participation. Following these studies, the first hypothesis is formed, where financial market participation is specified as the holding of stocks, bonds or investments in mutual funds.
Hypothesis 1
H0: The Big Five have no influence on financial market participation. H1: The Big Five have an influence on financial market participation.
Although the question of financial market participation is interesting on its own, this study will also examine the distribution of financial assets. Are there differences between certain types of individuals in the sense that they invest riskier (i.e. stocks) or safer (i.e. bonds).
Bucciol and Zarri (2017) find that agreeableness, cynical hostility and anxiety have significant influences on portfolio decisions. Agreeableness negatively affects the chance of holding a portfolio, anxiety negatively affects the proportion of shares in the portfolio and cynical hostility negatively affects both. Noteworthy is that the effects are quite large, going from the bottom to the top of the scale of cynical hostility decreases the chance of holding a portfolio by 6,7% and decreases the amount of shares in the portfolio by 6,6%.
The latest addition to the current literature comes from Oehler et alli. (2018). An experiment with undergraduate business students in fictional asset markets shows that extraversion and neuroticism have an influence on business decisions. More extravert students pay higher prices for assets and buy more assets when these are overpriced. Relatively neurotic students hold less risky assets in their portfolio. In order to examine whether individual with certain traits behave differently than others on the market, the following hypothesis is formed.
Hypothesis 2
Data
In order to correctly estimate the influence of the Big Five on financial market participation and risk-taking behaviour a large panel survey database is necessary. The SHARE (Survey of Health, Aging and Retirement in Europe) database is identified as a suitable database for this research. The database includes 140000 individuals from 27 European countries and Israel. All interviews are conducted with the use of a laptop and are face-to-face. Due to the wide scale of this database, the sample can be categorized as representative of individuals.
Most of the individuals in the SHARE database are over 50 years old, with a mean of approximately 70 years. One could argue that this poses a problem, however the opposite holds. Psychological literature argues that personality traits become more stable over time, with stability peaking for many traits when the individual is aged over 50 (Bucciol and Zarri, 2017). Henceforth, unstable personalities are minimized with the use of the SHARE dataset. The SHARE dataset consists of different waves, for this research, wave 7 is used. Wave 7 was released in April 2019 and is an interesting cross-sectional dataset due to its novelty.²
The data consists of multiple dependent variables: FMP, RiskS, RiskSMF, B/NW, S/NW and MF/NW. FMP (financial market participation) is an aggregation of Bonds, Stocks and Mutualfunds. FMP takes a value of 1 if the interviewed has a position in bonds, stocks and/or mutual funds. If the interviewed does not own any bonds, stocks or mutual funds the value of FMP is 0. RiskS is measured as 𝐵𝑜𝑛𝑑𝑠+𝑆𝑡𝑜𝑐𝑘𝑠+𝑀𝑢𝑡𝑢𝑎𝑙𝑓𝑢𝑛𝑑𝑠𝑆𝑡𝑜𝑐𝑘𝑠 . Henceforth, all values are between zero and one. For RiskSMF, the mutual funds are treated as risky assets. The following formula
for RiskSMF is applied 𝑆𝑡𝑜𝑐𝑘𝑠+𝑀𝑢𝑡𝑢𝑎𝑙𝑓𝑢𝑛𝑑𝑠
𝐵𝑜𝑛𝑑𝑠+𝑆𝑡𝑜𝑐𝑘𝑠+𝑀𝑢𝑡𝑢𝑎𝑙𝑓𝑢𝑛𝑑𝑠. The reason RiskS and RiskSMF are used is
because the risk of the chosen mutual funds are unknown. The mutual funds may be investing in emerging markets or only in equity, which make it quite a risky investment. On the other hand, the mutual fund may be a fixed income fund which tend to be less risky. Henceforth, this application is chosen.
B/NW, S/NW and MF/NW refer to the ratio of bonds (B), stocks (S) and mutual funds (MF) with respect to the net worth (NW). All amounts used for calculations have been converted to euro amounts using the respective exchange rates. Net worth is measured as real assets plus net financial assets, where net financial assets is measured as the value of the bank accounts, bonds, stocks, mutual funds and savings minus the liabilities. Real assets is calculated as the value of the home, their own business (if applicable), the cars and other real estate minus the mortgages.
are all measured using a questionnaire on a scale from one to five. The means of these personality traits create the Big Five. One additional personality trait is measured for Agreeableness, kind. However, kind is not aggregated in the score for Agreeableness.
Gender, age, education, education of the partner, country of residence net income and real assets are used as control variables. Gender is included because females generally make less risky investments (Almenberg & Dreber, 2015). The variable age, although the mean of the sample is around seventy, is still included as a control. The contrast in financial market participation between a 65 or 80 year old might be substantial, due to potential differences in the knowledge of online banking and trading. Education is included because it increases the likelihood of holding financial instruments (Cole & Shastry (2009)., Grinblatt et alli. (2009)). Education of the partner is included for potential spill-over effects. Both education and education of the partner are measured in years. The net income and value of real assets and are included for proxies of disposable income and wealth. Multiple wealth/income variables are included, because one can have a lot of wealth, but be relatively cash poor. The calculation of the value of real assets is explained in the calculation of net worth. Net income is calculated as earnings of employment, plus all pensions, social assistance, benefits, regular transfers, interest, income from rent and income of other household members. Country of residence is included as a proxy for cultural effects, wealth and many more.
Table 1.
Summary Statistics
Variables Observations Mean Standard Dev. Min Max
Thorough 43,204 4.174 0.891 1 5 Nervous 43,282 2.693 1.249 1 5 Imagination 43,129 3.338 1.186 1 5 Kind 43,249 4.148 0.804 1 5 Control/Other Variables Gender 44,384 0.582 0.493 0 1 Age 44,384 69.94 9.766 31 105 Education 44,383 10.95 4.390 0 35 EducationPartner 44,384 7.062 6.376 0 35 NetIncome 44,377 22.768 37.64 0 784.135 RealAssets 44,107 153.27 240.052 -100 1991 NW 42,898 150,320 195,992 -198,400 999,655 Bonds 2,468 28,685 41,599 260 250,000 Stocks 4,221 22,135 39,958 250 250,000 Mutualfunds 4,817 29,856 45,447 250 250,000
Table 1 shows the summary statistics of the dependent and independent variables. All data has been collected with the use of the SHARE database. The dependent variables are FMP (financial market participation), RiskS (ratio of stocks to financial investments), RiskSMF(ratio of stock and mutual funds to financial investments), B/NW (ratio of bonds to net worth), S/NW (ratio of stocks to net worth) and MF/NW (ratio of mutual funds to net worth). The independent variables of interest, the big five and personality traits are measured on a scale from
one to five. Gender takes a value of 1 if the person is a female, 0 otherwise. Education and EducationPartner is measured in years. NetIncome and RealAssets are measured in thousands of Euro’s. Net worth (NW), bonds, stocks and mutual funds are measured in Euro’s.
Research Methodology
Following in the footsteps of Brown and Taylor (2014), Conlin et alli. (2015) and Bucciol and Zarri (2017) this study will use a probit model for the testing of hypothesis 1. Furthermore, a linear probability model will be estimated. These regressions will be made twice, one time with the Big Five as the most important independent variables and once with the sub-traits. This will allow the analysis of whether certain sub-traits of the Big Five have a larger influence on financial market participation than their counterpart (as the Big Five dimensions are made up of two variables). In the analysis with sub-traits, the sub-trait ‘kind’ will also be included. A probit model is estimated with a dummy for financial market participation as the dependent variable. The following probit regression is estimated
𝐹𝑀𝑃𝑖 = 𝛼 + 𝛽𝑖𝐵𝐹𝑖 + 𝛾𝑖𝐶𝑖+ 𝑑𝑖𝑍𝑖 + 𝜀𝑖 (1)
Financial market participation (𝐹𝑀𝑃𝑖) is defined as holding bonds, stocks and/or mutual
funds. 𝐵𝐹𝑖 refers to the scores of the Big Five personality dimensions. 𝐶𝑖 refers to the control
variables gender, age, education, education of the partner, net income and real assets. Net worth is not included as a control variable, because the holding of stocks, bonds and mutual funds is used in the calculation of the net worth. Henceforth, net worth could cause problems with reverse causality and multicollinearity. One might argue that there are problems with collinearity between the variables real assets and net income, but a correlation coefficient of 0,4288 and a VIF (variance inflation factor) of 1,66 are well within all reasonable boundaries.³ 𝑍𝑖 is a country dummy that relates to the country of residence and 𝛼 and 𝜀𝑖 are the intercept
and error term, respectively. The variables for the Big Five are measured on a scale from 1 to 5 and are the means of their respective sub-traits.
The same regression will be used for the linear probability model (LPM), although obviously with the use of ordinary least squares instead of maximum likelihood estimation. The LPM will only be used to compare with the probit model. Since a LPM is not bound between 0 and 1, odds can become negative or greater than 100%, which obviously does not make sense. It is however interesting to see whether the effects of the probit model are also visible in the LPM. The coefficients of the probit model will be transformed into average marginal effects, for the sake of interpretation.
For the testing of hypothesis 2, multiple interpretations of risk will be assessed. Firstly, two ratio’s, the RiskS and RiskSMF will be used as the dependent variables. As previously stated in the data section, a limitation of these dependent variables is the unknown risk that mutual funds carry. However, as mutual funds will be analysed as carrying full risk and no risk, a proper interpretation of the results is still possible.
preference for bonds, stocks and mutual funds. Net worth is chosen as the proxy for wealth, since this includes both (passive) assets (e.g. real estate, cars) and liquid assets (e.g. cash, bank accounts). Because it includes both assets, this seems to give the best interpretation of once wealth.
Endogeneity in the form of self-selection (the entire sample participates in financial markets and is therefore not representable of the general population), will be solved by applying a Heckman twostep procedure. The twostep approach corrects for non-random samples. The first step in the two-stage process is a probit model to estimate the probability of an observation scoring a 1 and thus entering the sample. The second stage is an ordinary least squares to predict the dependent variable, in this case the multiple interpretations of risk. In order to account for biases, the Heckman procedure uses the OLS together with the probit to create the inverse Mills ratio, a selection parameter. The Mills ratio is computed by multiplying the standard deviation of the residuals in the second stage equation times the correlation between the error terms in the first stage and second stage equations (Certo et alli., 2016.). The Mills ratio is then added to the first stage equation to deal with sample selection bias (For all mathematical proof, please see Heckman, 1979).
Furthermore, by using the Heckman twostep, it is possible to see partial coefficients. A simple example of the benefits of the Heckman are as follows: Let us have a look at the relationship between smokers and wage. Assuming that people who smoke, want to smoke a lot, the wage one earns positively influences the amounts of cigarette one smokes. However, people with a high wage might not smoke (due to better education of the health hazards), henceforth wage negatively effects whether one smokes. By using the Heckman procedure, it is possible to see the partial effects of wage. In this case, the partial effects of the Big Five and sub-traits will also be visible.
Wooldridge (2010) provides three assumptions that influence the effectiveness of the Heckman procedure. Firstly, the independent variables are available for the entire sample, while the dependent variable is only available in the selected sample. Secondly, an omitted variable causes correlation between the two error terms in the selection equation and the equation of interest. The correlation is visible in the significance of the inverse Mills ratio. Thirdly, the independent variables are not correlated with the error terms from the first and second equation.
For the other three dependent variables, bonds/net worth (B/NW), stocks/net worth (S/NW) and mutual funds/net worth (MF/NW) the same exclusion restriction is applied. The dependent variables are ratios of the net worth in order to prevent problems with wealth/disposable income. Wealth can easily influence the amount of bonds, stocks or mutual funds one owns (as the individual has the capability to buy more). By correcting for net wealth however, this problem is avoided. Once again, wealth and disposable income influence the decision to participate in financial markets, but why would it influence the degree of net wealth that individuals invest in certain financial instruments?
The following selection equation is expressed :
𝑧𝑖∗ = 𝛼1+ 𝐴𝑖𝐵𝐹𝑖 + 𝐵𝑖𝐶𝑖+ 𝐶𝑖𝑍𝑖 + 𝑢𝑖, 𝑖 = 1, … , 𝑁 (2)
Where 𝑧𝑖∗ is a latent variable that cannot be observed, but an indicator variable is observed. 𝐵𝐹𝑖 refers to the scores of the Big Five, 𝐶𝑖 are the control variables and 𝑍𝑖 are the dummy
variables for the country of residence.
𝑧𝑖 = {1, 𝑧𝑖 ∗ > 0
0, otherwise (3) The equation of interest is:
𝑅𝑖𝑠𝑘𝑆𝑖 = 𝛼2+ 𝑎𝑖𝐵𝐹𝑖 + 𝑏𝑖𝐶𝑖 + 𝑐𝑖𝑍𝑖+ 𝜀𝑖 𝑖 = 1, … , 𝑛, 𝑁 > 𝑛 (4) 𝑅𝑖𝑠𝑘𝑆𝑖 is expressed as
𝑆𝑡𝑜𝑐𝑘𝑠
𝐵𝑜𝑛𝑑𝑠+𝑆𝑡𝑜𝑐𝑘𝑠+𝑀𝑢𝑡𝑢𝑎𝑙𝑓𝑢𝑛𝑑𝑠. The independent variables are the same as
for the selection equation, with the inclusion of real assets and net income.
A selectivity problem occurs because 𝑅𝑖𝑠𝑘𝑆𝑖 only occurs when FMP is 1. This implies that an
ordinary least squares estimator of the coefficients are biased an inconsistent. A conditional regression function is needed.
𝐸(𝑅𝑖𝑠𝑘𝑆𝑖|𝑧𝑖∗ > 0) = 𝛼2+ 𝑎𝑖𝐵𝐹𝑖 + 𝑏𝑖𝐶𝑖 + 𝑐𝑖𝑍𝑖+ 𝛼𝜆𝜆𝑖 𝑖 = 1, … , 𝑛 (5)
The additional variable 𝜆𝑖 is the inverse Mills ratio, it equals
𝜆𝑖 = 𝜙(𝛼1+ 𝐴𝑖𝐵𝐹𝑖+ 𝐵𝑖𝐶𝑖 + 𝐶𝑖𝑍𝑖)
𝛷(𝛼1+ 𝐴𝑖𝐵𝐹𝑖 + 𝐵𝑖𝐶𝑖+ 𝐶𝑖𝑍𝑖) (6) Where 𝜙 stands for the standard normal probability density function and 𝛷 denotes the cumulative distribution function for a standard normal random variable. However, because the value of 𝜆𝑖 is unknown, the parameters can be estimated with the use of a probit model.
𝜆̃𝑖 = 𝜙(𝛼̃1+ 𝐴̃𝑖𝐵𝐹𝑖+ 𝐵̃𝑖𝐶𝑖 + 𝐶̃𝑖𝑍𝑖) 𝛷(𝛼̃1+ 𝐴̃𝑖𝐵𝐹𝑖 + 𝐵̃𝑖𝐶𝑖+ 𝐶̃𝑖𝑍𝑖)
(7)
The estimated inverse is inserted into the regression of interest and leads to
Results
The full results of testing the first hypothesis, whether the Big Five influences the likelihood of financial market participation can be found in Table 2. (1) shows the results from the linear probability model with financial market participation (FMP) as the dependent variable. The variables of interest are the Big Five. (2) shows the results of the estimated probit model. (3) and (4) display another linear probability model and probit model, respectively. The variables of interest are now the personality sub-traits. (5) and (6) show the average marginal effects (AME) from the probit models in (2) and (4). These can be interpreted as follows: for openness, a one point increment for this Big Five trait increases the likelihood of financial market participation by 0,795%. Important to note is that the country dummies are not shown in the tables in order to keep the tables clear and comprehensible.
From the table it is visible that the dimensions Neuroticism, Openness and Conscientiousness significantly influence the likelihood of financial market participation (by looking at the average marginal effects).A highly neurotic person (scoring 5 on the scale), has a decreased likelihood of financial market participation by 4%. Very open individuals see their likelihood of participation increase by 4%. Highly Conscientiousness individuals see their likelihood of financial market participation decreased by 1,8%, although this is at a significance level of ten percent.
Since Neuroticism, Openness and Conscientiousness significantly influence the likelihood of financial market participation, H1 is rejected and the Big Five influence the likelihood of financial market participation.
income as a proxy for disposable income also significantly influence the probability of participation. Lastly, being a female decreases the likelihood of FMP by 3,5%.
The results from this research can be used by savings programs, pensions funds and others. For example, neurotic people have a smaller probability of financial market participation. Savings programs could make participants take the simple questionnaire and pay extra attention to neurotic individuals. The same holds for all sub-traits.
Table 3 shows the results from the Heckman twostep with RiskS and RiskSMF as the
dependent variables, where RiskS is defined as 𝑆𝑡𝑜𝑐𝑘𝑠
𝐵𝑜𝑛𝑑𝑠+𝑆𝑡𝑜𝑐𝑘𝑠+𝑀𝑢𝑡𝑢𝑎𝑙𝑓𝑢𝑛𝑑𝑠 and RiskSMF as 𝑆𝑡𝑜𝑐𝑘𝑠+𝑀𝑢𝑡𝑢𝑎𝑙𝑓𝑢𝑛𝑑𝑠
𝐵𝑜𝑛𝑑𝑠+𝑆𝑡𝑜𝑐𝑘𝑠+𝑀𝑢𝑡𝑢𝑎𝑙𝑓𝑢𝑛𝑑𝑠. Column 3 and 6 show the selection procedures, important to note
is that the inverse mills ratios are insignificant. This implies that the error terms in the selection equation and equation of interest are not correlated and henceforth there is no sample selection bias. Since all coefficients of the Big Five are not significant at a 5% level, they do not influence the amount of risk-taking behaviour. This holds for RiskS and RiskSMF. Noteworthy is Openness, which increases risk-taking behaviour when mutual funds are being treated as risky assets, albeit at a significance level of 10%. When mutual funds are being treated as bearing no risk, the sub-trait Reserved decreases risk-taking behaviour. Lazy, Imagination and Relaxed influence risk-taking behaviour, at a significance level of 10%.
Table 4 presents the findings of the Heckman twostep procedure with bonds divided by net worth as the dependent variable. The selection procedures are also visible in the table. Neurotic individuals see their likelihood of holding bonds decreased, while the ratio of bonds to net worth for extravert people is increased. In the sub-traits section, it shows that someone with few interests also has increased bonds to net worth ratio. If one has few interests, it can be considered that a sizeable amount of bonds are acquired for the long term and nothing else is done with them. Interesting to note is that women have a decreased likelihood of owning bonds, but when they do own bonds the ratio of bonds to net worth is increased. This is in line with the general risk-averseness of women. The significant negative inverse Mills ratio in (3) implies that without the selection procedure, the estimates would have been downward biased.
The mutual funds to net worth ratio is the dependent variable in Table 6. Similarly to the results of table 4 and 5, Neuroticism decreases the probability of owning financial instruments. However, Neuroticism seems to increase the ratio of mutual funds to net worth (at a 10% significance level). Unfortunately, a possible explanation for this phenomenon cannot be provided. The significant negative inverse mills ratios show a selection effect. Without the Heckman correction, the estimates would have been significantly downward biased.
Table 2. LPM Probit LPM Probit AME AME Variables (1) (2) (3) (4) (5) (6) Big Five Extraversion -0.00385** -0.0120 -0.00254 (0.00193) (0.00880) (0.00186) Agreeableness -0.00344 -0.0149 -0.00315 (0.00220) (0.0103) (0.00217) Conscientiousness -0.00362* -0.0168* -0.00356* (0.00217) (0.0102) (0.00215) Neuroticism -0.00753*** -0.0383*** -0.00809*** (0.00177) (0.00837) (0.00177) Openness 0.00748*** 0.0376*** 0.00795*** (0.00187) (0.00828) (0.00175) Personality Traits Reserved -0.000402 -0.00360 -0.000771 (0.00143) (0.00670) (0.00144) Trust 0.00531*** 0.0221*** 0.00473*** (0.00167) (0.00819) (0.00175) Lazy 0.00331** 0.0180** 0.00385** (0.00168) (0.00737) (0.00158) Relaxed 0.00336** 0.0155** 0.00333** (0.00169) (0.00777) (0.00166) FewInterests -0.00640*** -0.0316*** -0.00678*** (0.00135) (0.00611) (0.00131) Outgoing -0.00531*** -0.0187** -0.00401** (0.00188) (0.00864) (0.00185) FindFault 0.00655*** 0.0262*** 0.00562*** (0.00165) (0.00742) (0.00159) Thorough 0.00343* 0.0199* 0.00426* (0.00204) (0.0103) (0.00220) Nervous -0.00504*** -0.0247*** -0.00530*** (0.00155) (0.00725) (0.00155) Imagination 0.000575 0.00271 0.000581 (0.00158) (0.00709) (0.00152) Kind -0.00766*** -0.0331*** -0.00709*** (0.00232) (0.0111) (0.00239) Control Variables Age 0.000881*** 0.00228*** 0.00103*** 0.00312*** 0.000481*** 0.000669*** (0.000178) (0.000842) (0.000186) (0.000868) (0.000178) (0.000186) Gender -0.0357*** -0.165*** -0.0355*** -0.162*** -0.0349*** -0.0347*** (0.00359) (0.0160) (0.00369) (0.0164) (0.00338) (0.00350) Education 0.00797*** 0.0387*** 0.00767*** 0.0366*** 0.00818*** 0.00784*** (0.000472) (0.00207) (0.000483) (0.00210) (0.000435) (0.000447) EducationPartner 0.00322*** 0.0153*** 0.00300*** 0.0143*** 0.00323*** 0.00306*** (0.000324) (0.00136) (0.000329) (0.00138) (0.000287) (0.000294) NetIncome 0.000774*** 0.00209*** 0.000769*** 0.00208*** 0.000442*** 0.000446***
RealAssets 0.000282*** 0.000925*** 0.000284*** 0.000926*** 0.000195*** 0.000199*** (1.09e-05) (3.65e-05) (1.11e-05) (3.69e-05) (7.56e-06) (7.74e-06) Constant -0.0192 -1.701*** -0.0466** -1.828***
(0.0221) (0.103) (0.0237) (0.114)
Observations 44,099 44,099 42,543 42,543 44,099 42,543
R-squared 0.218 0.219
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 2 shows the output of a linear probability model, a probit and the marginal average effects of the probit model with financial market participation as the dependent variable. (1) and (3) are the linear probability models, with (1) showing the output of the Big Five and (3) the output of the personality traits. (2) and (4) show the probit outputs of the Big Five and the personality traits. (5) and (6) show the average
marginal effects (AME) of the probit model. Country dummies are not shown in the table.
Inverse Mills Ratio -0.0402 -0.0512
(0.0400) (0.0322)
Constant 0.483*** 1.062*** -1.644*** 0.577*** 1.160*** -1.958*** (0.0964) (0.0760) (0.104) (0.109) (0.0859) (0.118)
Observations 44,099 44,099 44,099 42,543 42,543 42,543
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 3 shows the results from the Heckman twostep procedure with RiskS and RiskSMF as the dependent variables. RiskS is defined as stocks/total investment and RiskSMF as (stocks+mutual funds)/total investment. (1) and (2) show the Big Five as most important independent variables and (4) and (5) the personality traits. (3) and (6) are the selection models with NetIncome and RealAssets as the exclusion variables. (3) is the selection model for (1) and (2), henceforth (6) the selection model for (4) and (5) The inverse mills ratio can
Table 4. B/NW Selection B/NW Selection Variables (1) (2) (3) (4) Big Five Extraversion 0.0139** -0.00654 (0.00697) (0.0127) Agreeableness -0.0105 -0.00558 (0.00828) (0.0150) Conscientiousness 0.00263 -0.00854 (0.00835) (0.0149) Neuroticism 0.0151 -0.0467*** (0.0109) (0.0124) Openness -0.0102 0.0136 (0.00705) (0.0120) Personality Traits Reserved -0.00688 0.00484 (0.00626) (0.00995) Trust -0.0180* 0.0355*** (0.0105) (0.0123) Lazy 0.00659 -0.00386 (0.00666) (0.0106) Relaxed -0.0140 0.0326*** (0.00967) (0.0114) FewInterests 0.0125** -0.0117 (0.00594) (0.00891) Outgoing 0.00685 -0.00283 (0.00803) (0.0129) FindFault -0.00371 0.0306*** (0.00887) (0.0107) Thorough 0.0105 -0.00713 (0.00956) (0.0151) Nervous 0.00892 -0.0161 (0.00745) (0.0107) Imagination 0.000792 0.000600 (0.00653) (0.0105) Kind 0.0173 -0.0231 (0.0117) (0.0169) Control Variables Gender 0.0530** -0.0921*** 0.0657*** -0.0860*** (0.0212) (0.0234) (0.0231) (0.0240) Age 0.000783 0.0135*** -0.00107 0.0143*** (0.00243) (0.00125) (0.00287) (0.00128) Education -0.00986* 0.0287*** -0.0130** 0.0274*** (0.00546) (0.00291) (0.00584) (0.00296) EducationPartner -0.00490*** 0.00452** -0.00502*** 0.00336* (0.00138) (0.00195) (0.00142) (0.00197)
RealAssets -3.88e-05 -5.14e-05
(5.16e-05) (5.21e-05)
NetIncome 0.000996*** 0.00100***
(0.000304) (0.000306)
(0.214) (0.238)
Constant 0.979 -2.853*** 1.419* -3.152***
(0.653) (0.152) (0.787) (0.173)
Observations 44,099 44,099 42,543 42,543
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 5. S/NW Selection S/NW Selection Variables (1) (2) (3) (4) Big Five Extraversion 0.00179 -0.000670 (0.00351) (0.0105) Agreeableness -8.11e-05 -0.00129 (0.00429) (0.0126) Conscientiousness -0.00405 -0.0216* (0.00434) (0.0124) Neuroticism -0.00227 -0.0365*** (0.00398) (0.0103) Openness 0.0101*** 0.0301*** (0.00368) (0.0101) Personality Traits Reserved -0.00282 -0.0120 (0.00290) (0.00836) Trust -0.00314 0.0319*** (0.00393) (0.0103) Lazy 0.00224 0.00479 (0.00295) (0.00895) Relaxed 0.00232 0.0236** (0.00341) (0.00950) FewInterests -0.00267 -0.0164** (0.00257) (0.00743) Outgoing -0.00255 -0.0237** (0.00380) (0.0107) FindFault -0.00175 0.0262*** (0.00321) (0.00900) Thorough -0.00327 -0.0107 (0.00427) (0.0124) Nervous 8.50e-05 -0.0185** (0.00317) (0.00898) Imagination 0.00739*** 0.0131 (0.00281) (0.00855) Kind 0.00574 0.00154 (0.00486) (0.0141) Control Variables Gender -0.0161 -0.261*** -0.00904 -0.259*** (0.0146) (0.0197) (0.0142) (0.0201) Age 0.00365*** 0.00378*** 0.00359*** 0.00459*** (0.000412) (0.00105) (0.000427) (0.00108) Education 0.00138 0.0358*** 0.000647 0.0352*** (0.00199) (0.00246) (0.00192) (0.00250) EducationPartner -0.00209*** 0.0105*** -0.00213*** 0.00973*** (0.000803) (0.00162) (0.000759) (0.00163)
RealAssets 2.65e-05 1.42e-05
(4.40e-05) (4.45e-05)
NetIncome 0.00169*** 0.00171***
Inverse Mills Ratio 0.0699 0.0455
(0.0620) (0.0608)
Constant -0.215 -1.942*** -0.150 -2.127***
(0.144) (0.126) (0.151) (0.142)
Observations 44,099 44,099 42,543 42,543
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 5 shows the Heckman procedure with stocks/net worth (S/NW) as the dependent variable. (1) and (3) are the regressions of interest with (1) showing the Big Five and (3) the personality traits. (2) is the selection procedure for (1) and (4) is the selection procedure for (3).
Table 6. MF/NW Selection MF/NW Selection Variables (1) (2) (3) (4) Big Five Extraversion -0.00160 -0.00276 (0.00399) (0.0102) Agreeableness -0.00460 0.0170 (0.00492) (0.0121) Conscientiousness -0.00379 -0.0419*** (0.00514) (0.0118) Neuroticism 0.00743* -0.0375*** (0.00438) (0.00981) Openness -0.00539 0.0212** (0.00392) (0.00968) Personality Traits Reserved 0.00145 -0.00227 (0.00325) (0.00802) Trust -0.0144*** 0.0464*** (0.00467) (0.00984) Lazy 0.00487 0.0265*** (0.00357) (0.00847) Relaxed -0.00675* 0.00952 (0.00367) (0.00906) FewInterests 0.00443 -0.0219*** (0.00303) (0.00719) Outgoing 0.000988 -0.00936 (0.00429) (0.0103) FindFault -0.00526 0.0177** (0.00357) (0.00868) Thorough 4.00e-05 -0.00249 (0.00483) (0.0121) Nervous 0.00219 -0.0307*** (0.00378) (0.00858) Imagination -0.00105 -0.00597 (0.00326) (0.00828) Kind 0.00795 -0.0179 (0.00557) (0.0135) Control Variables Gender 0.0371*** -0.177*** 0.0377*** -0.180*** (0.0121) (0.0190) (0.0122) (0.0194) Age 0.00391*** 0.00491*** 0.00367*** 0.00604*** (0.000483) (0.00101) (0.000519) (0.00104) Education -0.00464** 0.0332*** -0.00499** 0.0315*** (0.00203) (0.00239) (0.00195) (0.00242) EducationPartner -0.00470*** 0.00714*** -0.00446*** 0.00627*** (0.000741) (0.00155) (0.000718) (0.00157)
RealAssets -7.34e-05* -8.48e-05*
(4.31e-05) (4.36e-05)
NetIncome 0.00178*** 0.00181***
Inverse Mills Ratio -0.171** -0.185***
(0.0665) (0.0661)
Constant 0.366** -1.804*** 0.395** -2.013***
(0.147) (0.120) (0.160) (0.136)
Observations 44,099 44,099 42,543 42,543
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 6 displays the Heckman twostep for the dependent variable mutual funds/net worth (MF/NW). (1) and (3) are once more the equations of interest with (1) showing the regression with the Big Five and (3) with the personality traits. (2) and (4) are the selection
Conclusion
With the use of data from SHARE wave 7, an analysis between intrinsic personality traits and financial market participation was performed. Furthermore, the degree of risk-taking behaviour was also examined with respect to these personality traits. With the use of the Big Five (openness, conscientiousness, extraversion, agreeableness and neuroticism) and eleven sub-traits, a relationship between the likelihood of holding bonds, stocks and or mutual funds and the Big Five is visible.
Openness and neuroticism influence the probability of holding a financial instrument. Neuroticism has a negative effect on the likelihood of participating in the financial markets, decreasing the probability by 4% for highly neurotic individuals. Openness on the other hand increases the probability by 4% for very open persons. The results regarding Neuroticism can be linked to myopic loss-aversion. The tendency to frequently monitor once wealth and volatility of markets can be unpleasant for neurotic individuals. Findings for sub-traits are in agreement with the current literature, except for the findings in the social environment. Having social skills was generally thought of to increase the likelihood of owning financial instrument (Hong et al. (2004)., Brown et al. (2008)). The findings in this paper contradict the general consensus. A possible explanation could be the difference in social structures between Europe and the United States.
Regarding risk-taking behaviour, individuals who score high on Openness tend to have a higher percentage of their wealth invested in stocks and have an increased probability of owning stocks. Henceforth, it can be concluded that people with an open personality prefer riskier assets. Several sub-traits also have significant influence on the distribution of financial instruments. Interestingly, trusting individuals see the probability of owning a financial instrument increased, but this goes paired with an decrease of net worth invested. A possible explanation for this is unknown.
References
Almenberg, J., & Dreber, A. (2015). Gender, stock market participation and financial literacy. Economics Letters, 137, 140-142.
Angelini, V., & Cavapozzi, D. (2017). Dispositional optimism and stock investments. Journal of economic psychology, 59, 113-128.
Benartzi, S., & Thaler, R. H. (1995). Myopic loss aversion and the equity premium puzzle. The quarterly journal of Economics, 110(1), 73-92.
Bergmann, M., Scherpenzeel, A., & Börsch-Supan, A. (2019). SHARE Wave 7 Methodology: Panel innovations and life histories. Munich: MEA, Max Planck Institute for Social Law and Social Policy Brown, J. R., Ivković, Z., Smith, P. A., & Weisbenner, S. (2008). Neighbors matter: Causal community effects and stock market participation. The Journal of Finance, 63(3), 1509-1531.
Brown, S., & Taylor, K. (2014). Household finances and the ‘Big Five’personality traits. Journal of Economic Psychology, 45, 197-212.
Bucciol, A., & Miniaci, R. (2011). Household portfolios and implicit risk preference. Review of Economics and Statistics, 93(4), 1235-1250.
Bucciol, A., & Zarri, L. (2017). Do personality traits influence investors’ portfolios?. Journal of Behavioral and Experimental Economics, 68, 1-12.
Certo, S. T., Busenbark, J. R., Woo, H. S., & Semadeni, M. (2016). Sample selection bias and Heckman models in strategic management research. Strategic Management Journal, 37(13), 2639-2657. Cole, S. A., & Shastry, G. K. (2009). Smart money: The effect of education, cognitive ability, and financial literacy on financial market participation (pp. 09-071). Boston, MA: Harvard Business School.
Conlin, A., Kyröläinen, P., Kaakinen, M., Järvelin, M. R., Perttunen, J., & Svento, R. (2015). Personality traits and stock market participation. Journal of Empirical Finance, 33, 34-50.
Durand, R. B., Newby, R., & Sanghani, J. (2008). An intimate portrait of the individual investor. The Journal of Behavioral Finance, 9(4), 193-208.
Durand, R., Newby, R., Tant, K., & Trepongkaruna, S. (2013). Overconfidence, overreaction and personality. Review of Behavioral Finance, 5(2), 104-133.
Gaudecker, H. M. V. (2015). How does household portfolio diversification vary with financial literacy and financial advice?. The Journal of Finance, 70(2), 489-507.
Grinblatt, M., & Keloharju, M. (2009). Sensation seeking, overconfidence, and trading activity. The Journal of Finance, 64(2), 549-578.
Grinblatt, M., Keloharju, M., & Linnainmaa, J. (2011). IQ and stock market participation. The Journal of Finance, 66(6), 2121-2164.
Guiso, L., Sapienza, P., & Zingales, L. (2008). Trusting the stock market. The Journal of Finance, 63(6), 2557-2600.
Hong, H., Kubik, J. D., & Stein, J. C. (2004). Social interaction and stock‐market participation. The Journal of Finance, 59(1), 137-163.
Korniotis, G. M., & Kumar, A. (2011). Do older investors make better investment decisions? The Review of Economics and Statistics, 93(1), 244-265.
Mehra, R., & Prescott, E. C. (1985). The equity premium: A puzzle. Journal of monetary Economics, 15(2), 145-161.
Mehra, R. (2003). The equity premium: Why is it a puzzle? (corrected). Financial Analysts
Journal, 59(1), 54-69.
Oehler, A., Wendt, S., Wedlich, F., & Horn, M. (2018). Investors' personality influences investment decisions: Experimental evidence on extraversion and neuroticism. Journal of Behavioral Finance, 19(1), 30-48.
Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of choice. Science, 211(4481), 453-458.
Van Rooij, M., Lusardi, A., & Alessie, R. (2011). Financial literacy and stock market participation. Journal of Financial Economics, 101(2), 449-472.
Weil, P. (1989). The equity premium puzzle and the risk-free rate puzzle. Journal of monetary
Appendix 1.
Thurstone’s (1934) research let to the development of the “Big Five” model. The Big Five refers to five important personality traits within humans:
- Extraversion (outgoing/energetic vs. solitary/reserved) - Openness (inventive/curious vs. consistent/cautious) - Neuroticism (sensitive/nervous vs. secure/confident)
- Agreeableness (friendly/compassionate vs. challenging/detached) - Conscientiousness (efficient/organized vs. easy-going/careless)
One can argue that these personality traits are influenced by social norms, history and culture. However, individual differences across cultures are consistent and uninfluenced by culture (Mccrae & Costa, 2003).
Appendix 2.
The sample started with approximately 140000 respondents from Wave 7 of the SHARE database. Wave 7 is the first to include the Big Five personality test and henceforth the only Wave that is usable for the research conducted in this thesis. The target for the SHARE database are people aged over 50 at the time of sampling. For Wave 7 the target audience was 1966 and earlier. In contrast with Wave 1, where only household members with an age greater or equal to 50 were interviewed, later waves (2-7) also included household members with a lesser age. Partners of the interviewed were also interviewed, regardless of age.
The sample from Wave 7 had to be merged with all earlier waves, except Wave 3 (No necessary information was included in Wave 3). This is because modules that have previously been answered are not provided in the data that Wave 7 provides. After the merge, all the people that were interviewed but weren’t the financial respondent of the household were dropped from the sample. This leaves a sample of the financial respondents of the households/individuals and therefore the most trustworthy sample one can achieve when using interviews.
SHARE uses the BFI-10 as a measure for the Big Five. This short interview, that just contains ten questions which are measured on a scale of one to five, is quite reliable though. The correlation of the BFI-10 with the BFI-44 is r=.85 and the overall mean correlation of retest reliability across six weeks is r=.75 (Bergmann et alli. (2019)). The questions are asked in the following manner (the interviewer reads out the options):
I see myself as someone who is reserved. Do you...
1. Disagree strongly 2. Disagree a little 3. Neither agree nor disagree 4. Agree a little 5. Agree strongly In contrast with the other dimensions, Agreeableness loses some reliability in the short
questionnaire. Using a recommendation from Rammstedt and John (2007), SHARE adds a third question for Agreeableness in order to increase the reliability of the dimension. The third question for Agreeableness is whether one sees him/herself as considerate and kind. However, SHARE did not include kind in the Agreeableness score since the original BFI-10 provided the best result and
henceforth recommend these for use (Bergmann et alli. (2019)).
The BFI-10 is susceptible to acquiescence, which refers to the tendency to agree to a question by choosing the “yes” or “true” response, disregarding the content of the question along the way (Bergmann et alli. (2019)). The acquiescence bias is controlled by asking one question positively worded and the other negatively worded. Because acquiescence biases towards a “yes” or “true” response, the effect is automatically corrected for.
When handling data from surveys, it is always vulnerable to biases. An individual may not be willing to give the accurate answer. For example, the mean to the question whether the individual sees themselves as kind is 4.1 on a scale of 5. This might be true, nevertheless it is possible that the values are tainted because people want to be perceived as being kind.
Appendix 3
One might think that real assets and net income are heavily correlated and could cause multicollinearity. Table 7 shows that the correlation coefficient is below <.5 and henceforth
acceptable. The variance inflation factor (VIF) is smaller than 3. Therefore, multicollinearity is not an issue.
Table 7
Variables RealAssets NetIncome
RealAssets 1
NetIncome 0.4288 1
Table 7 is a correlogram that shows the correlation between RealAssets and NetIncome.
Table 8
Variable VIF 1/VIF
Age 1.16 0.861 Gender 1.08 0.922 Education 1.41 0.71 EducationPartner 1.31 0.764 NetIncome 1.63 0.615 RealAssets 1.5 0.665 Extraversion 1.14 0.874 Agreeableness 1.15 0.868 Conscientiousness 1.08 0.925 Neuroticism 1.19 0.837 Openness 1.13 0.885 Mean VIF 1.66
