Country level differences in hardship, risk taking propensity and
stock market participation across Europe
Lloyd Tijhaar S1887068 MSc IFM/MSc Finance Faculty of Business and Economics
University of Groningen Supervisor: Dr. M. Kramer
Acknowledgement
"This paper uses data from SHARE Waves 1, 2, 3 (SHARELIFE), 4 and 5 (DOIs: 10.6103/SHARE.w1.500, 10.6103/SHARE.w2.500, 10.6103/SHARE.w3.500, 10.6103/SHARE.w4.500, 10.6103/SHARE.w5.500), see Börsch-Supan et al. (2013) for methodological details.*
Abstract
I find that country level differences in hardship explain differences in stock market participation across Europe. Hardship and risk aversion negatively affect stock market participation rates. The negative association between hardship and stock market participation is more pronounced for high wealth individuals. Results are based on the fifth wave of the Survey of Health, Ageing and Retirement in Europe (SHARE). Logistic regression is used to estimate the relationship. The results imply that country specific characteristics are relevant in explaining the stock market participation puzzle. Governments and policy makers should take this into consideration when shifting the responsibility for someone’s financial future more towards the individual.
1. Introduction
Individuals have become more and more active in financial markets. This is becoming increasingly important, as governments shift the responsibility for someone’s financial future more towards the individual. Still, in the Netherlands stock-market participation is approximately only 14% (Giannetti and Koskinen, 2010). In most other European countries participation rates are not higher. It is, therefore, paramount to increase our understanding of why individuals engage in this non-normative behaviour.
Classical theories assume that investors only differ in wealth and risk-aversion (Markowizt, 1952). Contemporary behavioural finance research however, has shown a wide array of other factors affecting investment behaviour. Behavioural finance is progressing in its exploration and explanation of investor heterogeneity (Conlin et al., 2015), yet we only have an imperfect understanding of the factors affecting stock market participation. Explanations range from individual variables such as gender, overconfidence (Barber and Odean, 2001), financial literacy (van Rooij et al., 2011) and IQ (Grinblatt et al., 2011) to country-level factors such as institutions (Asgharian et al., 2013) and shareholder protection (Giannetti and Koskinen, 2010). Traditional models in financial economics cannot explain why participation is low, as participation would result in significant gains triggered by the risk premium and diversification (Guiso et al., 2008). This phenomenon has been named the participation puzzle.
Prior research discovered a collection of factors that affect stock market participation, but there are still more factors to uncover. This research will expand our current knowledge and advance the field by studying the effects of country level differences in hardship and risk aversion on stock market participation. Additionally, well-known household level determinants and country level variables explaining stock market participation are included in the analysis. My contribution to the literature consists of introducing hardship as new explanatory variable in the participation puzzle.
hardship affects risky behaviour and financial risk taking differently. Considering stock market participation, I expect that hardship will not be related to increased financial risk taking as people who encounter more hardship might be less future oriented. The question that this study aims to answer is ‘how do country level differences in hardship and individual risk aversion affect stock market participation across Europe?’
Stock market participation differs moderately between European countries, however in general participation rates are very low (see figure 1). Stock market participation in the sample is highest in Sweden (40%) and the lowest in Estonia (2%) with a sample average of 12%. Figure 1 indicates that country level characteristics could be hugely relevant as explanation for stock market participation. The question of stock market participation itself is fascinating because stock market participation is the first decision regarding investing that an individual makes, as he simply chooses whether to invest or not (Conlin et al., 2015). Studying the determinants of stock market participation is a relevant topic as it causes welfare loss and
affects policy (Cocco et al., 2005; Hong et al., 2004). Cocco et al. (2005) calculated that the welfare loss associated with not investing in stocks can be quite substantial. The loss can amount up to two percent of annual consumption. Individuals participating in the stock market are able to accumulate more wealth than those who do not participate, therefore the participation puzzle is vastly relevant (Mehra and Prescott, 1985). Besides missed gains for individuals, certain policy debates also depend fundamentally on the explanations for
0% 5% 10% 15% 20% 25% 30% 35% 40% 45%
households not to participate in the stock market (Hong et al., 2004). Hong et al. (2004) explain these dilemmas using government proposals that result in investing social security tax proceeds in the stock market. The results can be twofold, if people do not invest because they judge the risk-return trade-off to be unappealing, no gains can be acquired when the government invests on their behalf. On the other hand, if the motivation for not-participating is related to a lack of information or other factors, then these proposals make some sense.
To study stock market participation, I will use the Survey of Health, Ageing and Retirement in Europe (SHARE). SHARE is a multidisciplinary survey including a wide range of topics. The database includes individuals aged 50 and above in 14 European countries and Israel. The fifth wave of the SHARE survey will be used; this wave was completed in 2013 and is the most recent at this moment. The SHARE dataset is a perfect fit for this study as it has many observations in many countries, and it contains questions on all the needed variables. To my knowledge the fifth wave of SHARE is not yet used in similar household finance studies before, therefore the results of this study will add new verification to the existing literature.
Empirical analysis shows a clear negative relationship between hardship in a country and stock market participation among its people. Harsh and unpredictable circumstances in a country decrease the rate of stock market participation even when other country level variables are accounted for.
The remainder of this paper is structured as follows. Section two first discusses the current known factors affecting stock market participation. After that it discusses hardship, risk aversion and their separate and combined effects on participation. Section three presents the method and data. Section four presents the empirical results. Section five presents the results of the robustness tests. Finally, section six summarizes and discusses the results.
2. Literature review
2.1 Factors affecting stock market participation
In this section I will discuss factors related to stock market participation. First, I will discuss individual and household level factors and after that I will discuss country level factors affecting participation.
A large share of the early empirical literature on the topic of stock market participation explains limited participation as a result of transaction and information cost (Vissing-Jorgensen, 2003). Wealth and income are the determinants of stock market participation that occur in transaction cost based explanation. Hong et al. (2004) show that participation increases strongly with wealth. The effect of income on stock market participation is similar to the effect of wealth (Cocco et al., 2005). Cocco et al. (2005) state that the presence of labour income increases the demand for stocks, this effect is especially present early in life. The rationale behind the effects of wealth and income on participation is that stock market participation involves fixed costs, such as entry and transaction cost (Vissing-Jorgensen, 2003). Since wealthier households have more to invest, the fixed costs are less of a barrier to them (Hong et al., 2004). Empirical studies have shown that relatively small fixed entry cost can already explain the low stock market participation rates that are observed in practice (Paiella, 2001; Vissing-jørgensen, 2002).
Just as Mankiw and Zeldes (1991), Campbell (2006) states that income and wealth cannot fully explain whether a household participates in the stock market. The author suggests that there must be other demographic factors affecting with stock market participation. These other demographic factors that affect stock market participation are: age, marital status and gender.
In general, age decreases the probability that an individual is willing to take risk across all domains. Dohmen et al. (2011) state that the effect is relatively small in financial matters though. The lifecycle model predicts an inverted U-shape relationship in the willingness to take risk as a function of age. This inverted U-shape does not always occur in practice though. Cocco et al. (2005) find that stock allocation declines as age increases. The authors explain this decline as being driven by the fact that the labour income profile is downward sloping at older ages. With the increase of age, labour becomes less important and thus so do the implicit risk-free asset holdings represented by it. Investors react optimally by shifting their portfolio more toward the risk-free asset. Campbell (2006b) focuses on participation instead of allocation, and finds a negative age effect on participation in public equity markets. Increased participation by younger households starting in the 1990s is stated as probable cause for this negative effect.
Being married is related to higher stock market participation, for example Grinblatt et al. (2011) show that participants are 1.24 times more likely to be married than non-participants. Chrisiansen et al. (2015) also find that the participation rate increases for both men and women after marriage. Their rationale behind this result is that marriage frees up economic resources, this facilitates payment of costs related to stock market participation. The same authors find similar effects for both married couples and cohabiting couples, therefore they conclude that it is not necessary to distinguish between the two.
Finally, gender is related to stock market participation. In general, women are significantly less willing to take risk in all domains, this also includes the financial domain (Dohmen et al., 2011). These results are corroborated by others stating that women are less likely to participate in the stock market and when they participate they take less risk than men (Barber and Odean, 2001; Halko et al., 2012).
2010), economic development, market transparency, and the legal environment (Georgarakos and Pasini, 2011).
Guiso et al. (2008) state that investors take the risk of being cheated into account when deciding whether to invest in stock or not. They conclude, based on both micro data and cross country data, that trust is an important factor in explaining the limited participation puzzle. Moreover, the authors find that both individual differences in trust as country level differences in trust explain why some individuals partake in the stock market while others do not. These country level differences are caused by objective characteristics of the financial system such as the quality of investor protection and its enforcement. Asgharian et al. (2013) continue this road by focusing on institutional quality and trust. Institutional quality is related to the protection of property rights and investor protection. They find that both institutional quality and trust are positively related to stock market participation, where institutional quality affects trust through learning. Low institutional quality leaves fraudulent behaviour unpunished, this results in less trust. As a result, people are less willing to engage in any type of financial contract. In better functioning markets with stronger institutions people tend to have more trust. Therefore, they are willing to enter financial contracts with other parties they do not have ties to (Asgharian et al., 2013). Finally, they conclude that stock market participation is significantly affected by the part of trust related to institutional quality. The same results regarding the effects of institutions on trust are found in the paper of Georgarakos and Pasini (2011), who claim that country dummies capture differences in average trust levels, as differences are mostly due to country level differences in the institutional environment. Related to institutional quality is investor protection, Giannetti and Koskinen (2010) state that weak investor protection decreases the motivation to participate in the stock market, as this is seen as an additional cost for participating. Their proxy of investor protection, the Anti Director Rights Index explains 48% of the variation in stock market participation rates. Aforementioned results again stress the relevance of country level characteristics as explanation for differences in stock market participation.
2.2 Hardship
both the economic and social environment inside a country. The authors state that their results suggest that hardship can be used as a cue to guide life-history strategies.
Life-history theory is related to evolutionary biology and behavioural ecology (Hill et al., 1997). Moreover, Hill et al. (1997) state that applying concepts from life-history theory is useful in increasing our understanding of human behaviour, which is what we are trying to do when explaining stock market participation. Life-history theory attempts to explain variation involving adaptive trade-offs in how resources are distributed across differing life functions (Brumbach et al., 2009; Roff, 2002; Stearns, 1992). Variables in life-history theory that affect risk-taking include gender, age, marital and parental status and environmental characteristics (Hill and Chow, 2002). The environmental characteristics include the amount and predictability of resources, and rates and sources of mortality.
Following results from the field of life-history theory Mata et al. (2016) also state that ecological circumstances can cause life-history strategies to change. They use hardship as a measure to reflect these ecological circumstances within different countries. Mata et al. (2016) focus on reproduction strategies as explanation for behaviour. In harsh environments, risky strategies appear to be more adaptive as mortality rates are high and competition for resources is fierce. In more predictable and rich environments, slower reproductive strategies are more appropriate. The logic behind hardship is that individuals gamble on shorter lifespans in harsh and unpredictable environments, and adjust their actions according to these expectations. This rationale is supported by Hill et al. (1997) who find increased frequency of risk taking for individuals with future unpredictability beliefs and shorter lifespan estimates.
Mata et al. (2016) constructed an index that is meant to capture both social and economic hardship within a country. Their index consists of homicide rate, GDP per capita, income inequality, gender equality, infant mortality and life expectancy at birth. Linear regression is used to model the effect of hardship on risk taking propensity. Their results are in line with the predictions of life-history theory. They find that variation in risk taking propensity between countries is affected by local characteristics measured by the hardship index. The authors control for age, gender, education, parental status, marital status, and occupation.
Mata et al. (2016) found that hardship affects risk taking propensity. The theory behind this suggests that when individuals live in countries with more hardship they tend to gamble on shorter life spans. Gambling on a short lifespan and an unpredictable environment do not only affect reproduction behaviour but also risky behaviours. Hill and Chow (2002) conclude that environmental instability increases risky alcohol consumption. While Rahkonen et al. (2005) found an association between financial hardship and smoking, even within high-income groups. Similarly to financial hardship, material hardship is also positively related to smoking (Gilmore et al., 2000). Gambling on a shorter life-span might thus mean that individuals in countries experiencing more hardship are less future oriented. This could also explain other risky behaviours such as smoking and drinking, since the negative effects of these behaviours are valued differently. Howlett et al. (2009) offer future orientation as potential explanation for why individuals do not take financial decisions that will be positive in the long-run. They find that more future oriented consumers are more likely to participate in a retirement plan. The same mechanism could be at play for stock market participation, as this also is a future oriented strategy. Therefore, through its effect on future orientation hardship could explain country level differences in stock market participation.
In this study, I will take a different approach than Mata et al. (2016). I expect that the effect of hardship operates through future orientation. Thus, I propose that hardship affects stock market participation differently than it affects other risky behaviours. I will test whether country-level differences in hardship can be used to explain variation in stock market participation across Europe. The expectation is that individuals in countries that are most exposed to hardship gamble on a shorter life span and will therefore be less inclined to have money invested in stocks. This results in the following hypothesis: There is a negative relationship between country level differences in hardship and stock market participation in Europe.
2.3 Risk taking propensity
measurements are complex topics, it is necessary to give risk taking propensity a prominent position in this research. Doing this will increase our understanding of the mechanisms through which hardship affects risk aversion and more importantly how it affects stock market participation.
When discussing the relationship between risk taking propensity and stock market participation a definition is needed that is suited for this context. In the remainder of this paper the definition of Sitkin and Pablo (1992) will be used. In their study, they define risk propensity as: “the tendency of a decision maker either to take or to avoid risks” (pp.12). An individual can be either risk averse or risk tolerant; a risk averse individual will be hesitant to take risk, while a risk tolerant individual is more willing to absorb risk. Thus, risk-taking propensity is an individual difference that affects the level of risk someone is willing to take.
2.3.1 Risk taking propensity in behavioural economics
Risk-taking is an often-studied topic in classic as well as behavioural economics. Classis theory assumes that investors only differ in risk taking propensity and wealth (Markowizt, 1952). “In classic decision theory, risk is considered a function of the variation in the distribution of possible outcomes, the associated outcome likelihoods, and their subjective values” (Steward and Roth, 2001, pp.145). Classic decision theory assumes that decision makers take rational decisions that will result in the best possible outcomes for everyone. Behavioural economists think investor rationality is unrealistic. They claim that risky decisions are not solely based on these rational calculations, but they are also affected by an individual’s predisposition towards risk in general and by the situation (Bromiley and Curley, 1992). Their within-trait approach focuses on how individuals with specific traits behave in different situations. Contrastingly, their within-situation approach focuses on how individuals with different personality characteristics behave in specific situations. Szrek et al. (2012) corroborate that risk taking is a universal part of our lives, but there are large differences between individuals in their risk-taking propensity. Moreover, in a classic paper addressing risk attitude Dohmen et al. (2005) conclude, based on a large sample of 22,000 German citizens, that “the distribution of the willingness to take risks exhibits heterogeneity across individuals” (pp 34). Risk taking propensity is thus affected by an individual’s predisposition as well as by the situation.
2011). Next to the principal component there are five more components explaining at least five percent of the variation. This indicates that there is some additional content captured by domain specific measures. To conclude, risk taking propensity varies across individuals and although there is a general risk attitude it differs between domains.
These statements concerning risk propensity are in agreement with the Big Five personality theory, as this theory suggests that risk propensity is a smaller part of the overarching trait extraversion (Mount and Barrick, 1995). Considering that risk-taking propensity is a personality trait might mean that risk taking can be seen as one of the factors that are trans-situational, as personality is generally considered to be, which means that risk propensity is more a characteristic of an individual than of their situation (Nicholson et al., 2002).
2.3.2 Measurements of risk taking propensity
There seems to be a consensus now that context specific risk measures are more accurate than general risk measures. Markiewicz and Weber (2013) state that self-reported risk-taking propensity as assessed with survey items concerning a specific domain are more successful in predicting real world risk taking than expected utility model derived risk attitude coefficients. The same was concluded by Dohmen et al. (2011) in their study as they state that the best predictor of financial risk taking is the domain specific question on so-called “financial matters”. The only measure to predict risky behaviours across all different domains is Dohmen’s general risk measure (Dohmen et al., 2011). This means that the domain specific risk question relating to financial matters does not necessarily predict an individual’s general risk-attitude reliably while the general risk question can predict a large fraction of an individuals’ financial risk taking attitude. Moreover, the authors conclude that their general risk question more accurately predicts risky behaviour in different contexts than the, at that time, more common lottery measure. Halko et al. (2012) corroborated these results in their study where they compared several measures of risk attitude in a sample group consisting of individuals that are familiar with financial risk. In this sample, they found that self-reported financial risk attitude is the strongest predictor of the proportion of wealth invested in stocks.
2.3.3 Risk taking propensity and stock market participation
there is a relationship between the risk taking propensity and the level of investment risk that is taken by individual investors (Masters, 1989). The author finds that people who are less willing to take risk are more conservative in their investments, while those more inclined to take risk invest in riskier portfolios. While developing the general risk measure Dohmen et al. (2011) also study risk taking behaviour in a financial context. In this paper the authors conclude that their risk measure is positively correlated with investment in stocks, which seems logical as stocks are a relatively risky financial investment. These results were corroborated by Halko et al. (2012) and Hamid and Rangel (2013), who found similar results in their study as risk-taking propensity was positively related with risk-risk-taking behaviour for individual investors in the stock market. Based on this prior empirical research, I expect to find a negative relationship between risk aversion and stock market participation as more risk taking is positively related to stock market participation. This results in the second hypothesis: There is a negative relationship between risk aversion and stock market participation in Europe.
2.4 Interaction hardship and risk taking propensity
In the previous two sections I discussed the separate effects of hardship and risk aversion on stock market participation. I proposed that life-history strategies affect future orientation. Moreover, I suggested that future orientation is the mechanism through which hardship might work. This explains how hardship affects other risky behaviours differently than stock market participation. This research explores new terrain and theoretical backing is scarce. Therefore, the final hypothesis proposing an interaction effect between hardship and risk aversion is purely explorative. An interaction effect might be present as risk taking propensity is affected by both an individual’s predisposition as the situation. Differences in hardship relate to an individual’s situation.
3. Methods and data
3.1 SHARE data
The main source of data for this study is the fifth wave of the Survey of Health, Ageing and Retirement in Europe (SHARE) (Börsch-Supan, 2016). This dataset is made available for free by the university of Tilburg. The fieldwork for the fifth wave of SHARE was completed in November 2013. SHARE is a multi-disciplinary, cross-national survey containing micro data that includes both individual and household characteristics and investment behaviour. The dataset covers a large variety of different questions including topics relevant for this research such as: stock market participation, risk-aversion, risky behaviours, wealth and income. The dataset contains data on 66,246 observations from 14 European countries and Israel. The other countries are: Austria, Belgium, Czech Republic, Denmark, Estonia, France, Germany, Italy, Luxembourg, Netherlands, Slovenia, Spain, Sweden and Switzerland. The SHARE survey is aimed at individuals aged 50 and above, it includes only 122 individuals younger than 50. This results in an old sample with a mean age of 69 (see panel A of table 1). The main advantage of using SHARE is that survey data enables me to do a cross-country comparison.
Based on previously discussed theory, I construct an empirical model that investigates how stock market participation is affected by differences in hardship and risk aversion. The main hypotheses for this study are a negative relationship between hardship and stock market participation and a negative relationship between risk aversion and stock market participation. The third hypothesis is more explorative and suggests a negative interaction effect of hardship and risk aversion. The dependent variable in the estimation is a binary variable representing stock market participation, where an individual does or does not participate in the stock market. The independent variables are hardship and risk aversion. Hardship is measured at the country level and is therefore the same for every individual living in the same country. Risk aversion is an individual attitude Since the dependent variable is binary and contains many more zeros than ones a logit model is the best approach, in choosing this approach I follow Asgharian et al. (2013).
education as the decision maker. Finally, if both income and education are not reported the man is picked as the decision maker.
3.2 construction of key variables 3.2.1 Stock market participation
Panel A of table 1 presents the summary statistics of the variables used in the estimation of the stock market participation regression. Respondents answered the following question relating to stock market participation:
“Do you currently have any money in stocks or shares (listed or unlisted on stock market)?”
Stock market participation is a binary variable which is equal to one if the households owns stocks and equal to zero if the household does not own stocks. Figure 1 in the introduction displays stocks market participation across the sample. Average stock market participation across the sample is 12%. Stock market participation is the highest in Scandinavian countries, in Sweden almost 40% holds stock and in Denmark 32%. Stock market participation is the lowest in Estonia and the Czech Republic with two and four percent, respectively. For only five countries the rate of stock ownership is above the average, this thus means that ten countries have a stock market participation rate below the average.
Before the analysis I dropped the observations relating to stock market participation and risk aversion where the respondents refused to answer the question or did not know the answer to the question. This resulted in dropping 643 and 1505 observations, respectively. Excluding these and the missing observations reduces the sample size to 42,698 observations. Since over a third of the observations of the dependent variable is missing I analysed whether the missing observations differ from the non-missing observations on the independent and control variables.
differ between missing and non-missing observations (see table 2 in appendix B for results). All control variables differ for the missing and non-missing dependent variable. The differences are most significant for wealth, income, and gender. Wealth and income are much larger for the variables where the dependent variable is not missing. Furthermore, individuals in the missing dependent variable group are on average lower educated and older, less individuals are married and more are male. Wealth appears to be the most troublesome variable; the robustness section will deal with this issue.
Table 1
Panel A N Mean Std. Dev. Min Max
Stock market participation 42698 0.12 0.33 0 1
Hardship (standardized) 42698 0.00 1.00 -1.78 2.46 Risk aversion 42698 3.70 0.58 1 4 Wealth 25319 24042.39 117885.40 0 7566512 Income 12534 23692.52 46684.26 0 2500000 Education 13958 5.32 2.60 1 15 Age 42693 69.24 10.28 28 106 Married 16004 0.63 0.48 0 1 Gender 42698 0.43 0.50 0 1 HDI (standardized) 42698 0.00 1.00 -1.48 1.80 Rule of Law 42698 1.47 0.49 0.30 2.00 ADRI 31953 3.14 0.99 1.50 5.00
Panel B N Mean Std. Dev. Min Max
Homicide rate 15 1.27 0.59 1.00 3.00 GDP per capita 15 42383.90 22891.51 17231.30 101450.00 Income inequality 15 0.30 0.04 0.25 0.36 Gender inequality 15 0.07 0.04 0.01 0.16 Infant mortality 15 2.87 0.64 2.00 4.00 Life expectancy 15 81.20 1.70 77.00 83.00 Hardship 15 0.00 0.65 -1.02 1.63
3.2.2 Hardship
Hardship is the first independent variable. To construct the hardship index for the cross-country comparison, I combined data from several sources. Following Mata et al. (2016), I use the same indicators that most likely capture adversity and social strife to measure the exposure to hardship in the countries included in the sample. The hardship index is meant to capture both economic and social hardship and consists of the homicide rate, GDP per capita, income inequality, gender equality, infant mortality and life expectancy at birth. In the robustness section I empirically test the different effects of economic and social hardship on stock market participation. The homicide rate was acquired from the World Bank and is from 2014, the number expresses the intentional homicides per 100.000 people. GDP per capita is also from the World Bank, this data is from 2015 and is measured in current US dollars. Income inequality data is acquired from the OECD; the data is from 2015 and is measured by the Gini coefficient which ranges from 0 in the case of perfect equality to 1 in the case of perfect inequality. For data on gender equality the World Bank is used, their measurement is the GII inequality index, data is from 2013, again like the Gini it ranges from 0 to 1 where higher values indicate higher levels of inequalities. Infant mortality data is also from the World Bank; this data is from 2015 and it shows the mortality rate of infants per 1,000 live births. Finally, the life expectancy at birth was acquired from the World bank, this data is from 2015 and consists of the life expectancy at birth in years. An overview of the components of hardship can be found in appendix A. Panel B of table 1 shows the unstandardized variables and summary statistics for the different hardship variables and for the total index.
countries the Estonian score is much higher than all the other scores. Figure 2 below graphically depicts the hardship scores of all the countries in the sample.
3.2.3 Risk aversion
Risk aversion is the second independent variable. SHARE measures risk aversion by the following question specifically aimed at financial risk taking:
“When people invest their savings they can choose between assets that give low return with little risk to lose money, for instance a bank account or a safe bond, or assets with a high return but also a higher risk of losing, for instance stocks and shares. Which of the statements on the card comes closest to the amount of financial risk that you are willing to take when you save or make investments?” The risk aversion question has the following four answer categories: no risk, average risk, above average risk, and substantial risk. In each country, most people are not willing to take any risk, a smaller part is willing to take average risk, and a very small group is willing to take above average and substantial risk. This question should be an accurate predictor of stock market participation, as the domain specific risk question has been shown to be the most accurate predictor of risk taking (Dohmen et al., 2011; Halko et al., 2012; Hamid and Rangel, 2013). In the Scandinavian countries, the willingness to take risk is higher than in the other countries. Figure 3 depicts the willingness’ to take risk across the countries in sample.
-2.00 -1.50 -1.00 -0.50 0.00 0.50 1.00 1.50 2.00 2.50
Figure 3. Risk aversion by country in the sample
3.2.4 Control variables
The empirical specification acknowledges that there are many determinants of stock market participation. Therefore, a considerable number of control variables is included in the regressions. Summary statistics of the control variables are presented in panel A of table 1, and descriptions and measurements of the variables are listed in appendix A. Following previous studies, I include controls for wealth and income. Wealth is measured as the financial wealth (amount in euros) at the time of the survey. Income is measured as household’s approximate income in euros after taxes and contributions in the year prior to the survey. For both wealth and income dummies are included for each quartile. The expectation is that the signs are positive for all quartiles and that the coefficients associated with higher quartiles are larger. Furthermore, control variables are included for demographic variables such as education, age, marital status and gender. Education is measured as the number of years someone has been in full-time education. The sign of education is expected to be positive, as higher education increases the probability of stock market participation. Age is expected to show an inverted U-shape relationship; therefore, I also include age squared. The expectation is that the sign of age is positive and the sign of age squared is negative since research has shown reduced stock market participation when age increases. Marital status and gender are both dummies, where being married and being male will have the value of one. Since married individuals and men tend to invest more in stock both signs are expected to be positive.
0% 20% 40% 60% 80% 100%
To filter out recognised country level effects that might possibly capture part of hardship three country-level control variables are included in the analysis. First, the Human Development Index (HDI) from the United Nations Development Programme was included. This index measures average achievement in key dimension of human development. These dimensions are a long and healthy life, being knowledgeable and a decent standard of living1. The United Nations uses life expectancy at birth, mean years of schooling and GNI per capita as indicators of these dimensions. Two of these three indicators are also included in the hardship index. The HDI only measures development though, and does not reflect security and inequalities, where hardship includes security and inequality. HDI is higher for more developed countries, therefore I expect a positive sign for this relationship.
The “Rule of law” index as collected by the World Bank is included, this index captures perceptions of the degree to which agents have confidence in and abide by the rules of their society. The main factors are the quality of contract enforcement, property rights, the police, the courts, and the likelihood of violence and crimes2. Institutional quality might affect stock market participation as it increases transaction cost. Moreover, it affects the risk attitude of people and limits their knowledge concerning alternative investments (Asgharian et al., 2013). The Rule of Law index is higher when institutional quality is higher, therefore I expect a positive sign for rule of law. Following Asgharian et al. (2013), I also used the “Index of legal system and security of property rights” from the Fraser Institute as measure of institutional quality to increase robustness. This does not affect the empirical results.
I also include the Anti Directors Rights Index (ADRI). This index is an often-used measure of shareholder protection. I use the revised version of Spamann (2010). The ADRI is not available for the Czech Republic, Estonia, Luxembourg and Slovenia; therefore, these countries are dropped from the sample once the ADRI is included. For the ADRI a higher figure indicates better shareholder protection, therefore I expect a positive sign for shareholder protection.
4. Results
4.1 hardship and risk aversion: univariate results
In this section I present some univariate results to get a first indication of the main relationships. Figure 4 graphically depicts stock market participation rates by hardship
and risk aversion. The sample is separated in three groups of 5 countries each, depicting high, average and low hardship countries. Additionally, the sample it split at the median to create a low and high risk averse group of individuals. The figure shows that participation rates are higher in countries with less hardship. Moreover, it shows that participation rates are higher for low risk averse individuals than for high risk averse individuals for every hardship category. Stock market participation is the lowest with 2.6% in high hardship countries for high risk averse individuals and highest with 44% in low hardship countries for low risk averse individuals. The figure appears to be in line with the first two hypotheses as lower levels of hardship and lower levels of risk aversion coincide with higher stock market participation rates.
Figure 4. Stock market participation over risk aversion by hardship category
0% 10% 20% 30% 40% St o ck ma rke t p a rt ici p a ti o n
High hardship Average hardship Low hardship
Figure 5. Stock market participation over risk aversion separated by hardship
Figure 5 displays how the relationship between hardship and stock market participation is affected by risk aversion. The relationship is negative for all three categories of hardship, which means that higher risk aversion results in lower stock market participation for each category of hardship. Moreover, stock market participation is higher in low hardship countries than in average and high hardship countries for every level of risk aversion. The difference between stock market participation rates decreases when risk aversion increases. The difference in participation between the low and high hardship group is much larger for substantial risk (39%) than for no risk (13%). This suggest that there might be an interaction effect.
4.2 Hardship and risk aversion: multivariate results
In this section I present the empirical results of the multivariate analyses to determine whether hardship and risk aversion explain differences in stock market participation across Europe. I use logistic regression to estimate the effects of hardship and risk aversion on stock market participation. Table 2 presents the results of the estimation. The regression confirms the most important results from the univariate analysis. Stock market participation is negatively related to the level of hardship and risk aversion. The first column of table 2 shows the
40% 20% 0% 60% 80% St o ck ma rke t p a rt ici p a ti o n
Substantial risk Above average risk Average risk No risk Risk aversion
95% Confidence interval High hardship
regression containing only hardship, the coefficient is negative and very significant. Living in harsh and unpredictable environments thus decreases the rate of stock market participation. The second column contains both hardship and risk aversion, both coefficients are negative and highly significant. Hence, hardship and risk aversion both explain differences in stock market participation. In column three of table 2 the individual and household level control variables are included in the regression; this does not change signs or significance of the independent variables. Finally, in the fourth column of table 2 the interaction effect of hardship and risk aversion is added. I did not find a significant interaction effect of country level differences in hardship and risk-aversion on stock market participation in Europe. Additionally, adding the interaction effect strongly decreases the significance of hardship itself as it is now only significant at the 10% level.
I confirm most of the results from prior studies on the determinants affecting stock market participation (Campbell, 2006b; Cocco et al., 2005; Cole and Shastry, 2009; Hong et al., 2004). Column three of table 2 includes the household level determinants related to stock market participation. The quartiles for wealth behave exactly as expected were each higher quartile results in a higher coefficient. All quartiles for wealth are highly significant. This confirms the positive relationship between wealth and stock market participation. It is noteworthy that the coefficients for income is negative and significant for the third quartile as compared to the first quartile base group. The coefficients for the other quartiles are not significant, income therefore does not explain stock market participation in this sample. One explanation for this is that most individuals are retired. Cocco et al. (2005) found that labour income becomes less relevant as age increases. Education and marital status are both positively and significantly related to the dependent variable. The coefficient for gender, being male, is slightly positive but not significant. Finally, both the coefficients for age and age squared are not significant. One explanation for this result may be that the sample was reasonably old, and respondents who are retired may be in the de-cumulation phase of their life cycle (van Rooij et al., 2011).
the results of the logistic regression including the country level control variables. This regression reveals some peculiar results. The coefficient for HDI is negative and highly significant. This suggest that individuals in higher developed countries are less likely to invest in the stock market. The coefficient for Rule of Law is not significant, this contradicts other studies showing that institutional quality affect stock market participation (Asgharian et al.,
Table 2
Hardship Independent Full model Interaction
Country controls Country controls + interaction -0.846*** -0.784*** -0.350*** -0.472* -1.655*** -1.309*** (0.021) (0.022) (0.066) (0.281) (0.209) (0.401) -1.114*** -0.883*** -0.860*** -0.804*** -0.868*** (0.022) (0.055) (0.076) (0.061) (0.087) 0.037 -0.115 (0.082) (0.114)
Wealth (base group Q1)
Wealth (Q2) 1.685*** 1.684*** 1.577*** 1.584*** (0.268) (0.268) (0.296) (0.296) Wealth (Q3) 2.297*** 2.299*** 2.423*** 2.422*** (0.261) (0.262) (0.286) (0.286) Wealth (Q4) 3.851*** 3.851*** 3.766*** 3.766*** (0.256) (0.256) (0.281) (0.281) Income (base group Q1)
Income (Q2) -0.008 -0.009 0.036 0.041 (0.143) (0.144) (0.159) (0.159) Income (Q3) -0.292** -0.292** -0.17 -0.170 (0.128) (0.128) (0.14) (1.14) Income (Q4) 0.049 0.049 0.117 0.118 (0.131) (0.131) (0.142) (1.142) Education 0.037*** 0.037*** 0.034*** 0.035*** (0.011) (0.011) (0.011) (0.011) Age -0.001 -0.002 -0.084 -0.084 (0.069) (0.069) (0.076) (0.076) Age squared 0.000 0.000 0.000 0.001 (0.001) (0.001) (0.001) (0.001) Married 0.188** 0.188** 0.242** 0.241** (0.092) (0.092) (0.089) (0.098) Gender 0.085 0.085 0.061 0.062 (0.083) (0.083) (0.089) (0.089) HDI (standardized) -0.510*** -0.540*** (0.171) (0.174) Rule of law 0.262 0.299 (0.295) (0.299) ADRI 0.308*** 0.302*** (0.109) (0.109) Pseudo R2 0.07 0.15 0.29 0.29 0.30 0.30 Log. L -14773.28 -13465.49 -2107.61 -2107.51 -1826.05 -1825.54 N 42614 42614 6160 6160 5207 5207
Hardship x risk aversion Risk aversion Hardship (standardized)
Coefficients
2013). Finally, the ADRI is positive and significant which is in line with expectations. These results confirm that hardship is a particularly relevant variable for as it explains more variation in stock-market participation than already captured by other country level variables.
Some results deviate from my expectations, therefore I tested whether this may have resulted from multicollinearity issues. The variance inflation factors for hardship, HDI, Rule of Law and ADRI all lie between six and eight. Although the variance inflation factors do not exceed the recommended maximum level of 10 (Hair et al., 1995), they are still quite large. To analyse the separate effects, I include HDI and Rule of Law in the regression with the full model one by one. This time the effect of HDI is positive and highly significant. These results are in line with my prediction. Again, I do not find significant results for Rule of Law. Finally, the variance inflation factors for these separate regressions do not exceed 2.1. This means that no multicollinearity issues arise between hardship and the country level control variables. Analysis excluding hardship and including the three controls gives variance inflation factors over six. I conclude that slight multicollinearity issues exist between the three control variables, but not between the control variables and hardship.
5. Robustness
In this section I will test the robustness of the results. First, I will change the measurement of stock market participation by replacing the binary variable by risky share. Secondly, I will test how observations with missing dependent variables affect the results. Thirdly, I regress hardship on different risky behaviour such as smoking. Finally, I analyse which components of the hardship index drive the results.
5.1 Risky share
before. Both hardship and risk aversion remain highly significant. The regression shows that a one standard deviation increase in hardship decreases the risky share by ten percent. In addition, a one-step increase in risk aversion decreases the fraction of wealth that is allocated to stocks by 29%. Again, the interaction effect for hardship and risk aversion was not significant, which is similar to the result from the logistic regression. Moreover, the effect of hardship becomes much stronger when the country level control variables are added to the regression. After adding these controls a one standard deviation increase in hardship decreases the risky share by 54%. Thus, the risky share tobit regression produces the same results as the logistic regression.
5.2 Missing versus non-missing data of the dependent variables
In the data section, I suggested when discussing the data that missing data might be a problem. Especially the large difference in mean wealth between observations with a missing versus a non-missing dependent variable. Moreover, the logistic regression indicated that
Table 2 Model excluding country controls Model including country controls Model and interaction + country countrols -0.01*** -0.535*** -0.454*** (0.022) (0.065) (0.121) Risk aversion -0.292*** -0.260*** -0.276*** (0.018) (0.019) (0.027)
Hardship x risk aversion -0.027
(0.034)
Individual controls yes yes yes
Country controls no yes yes
(Pseudo) R2 0.26 0.27 0.27
Log. L -2348.72 -2033.15 -2032.85
N 6158 5206 5206
Multivariate analysis of hardship and risk aversion on risky share. This table presents coefficient estimates of hardship and risk aversion measures on risky share using a two limit tobit regression with limits of 0 and 1. The dependent variable is a continuous variable that represents the fraction of wealth allocated to equity. Apart from hardship and risk aversion, I include wealth, income, years of
education, age and age-squared, a dummy of being married, a male dummy, the Human Development Index, Rule of Law and the Anti Directors Rights Index. These control variables are not presented below as the results are familiar. The first column presents the results of the regression excluding the country controls. The second column presents the results of the regression including the country controls. The third column present the regression including the interaction and country level controls. Standard errors are in parenthesis. *** , ** and * indicate significance at the 1%, 5% and 10% level respectively.
wealthy individuals are much more likely to have money invested in stocks. The coefficient for the fourth quartile is almost four times larger than the coefficient for the first quartile. The difference between means and the magnitude of the coefficient imply that results for the groups may differ. To test whether missing observations affect the results, I performed another regression where I split the sample in a below median wealth and above median wealth group. The rationale behind this analysis is to determine whether missing observations corrupt the
Figure 6. Stock market participation over hardship split by median wealth
results. A first impression of the difference between the low and high wealth groups is obtained from figure 6 and 7. Comparing the low and high wealth group in the first graph, it can be observed that the negative effect of hardship on stock market participation is much stronger for the high wealth group than for the low wealth group. The same is visible for risk aversion, where the negative relationship between risk aversion and stock market participation is much stronger for individuals with an above median wealth than for individuals with below median wealth.
A more formal test is needed to see whether the effects of hardship and risk aversion are significant for the low wealth group. The regression results can be found in table 3. The analysis regressing hardship and risk aversion on stock market participation confirms the impression that was obtained from the graphs. The estimation (see: column 1 and 2 of table 3) shows that hardship and risk aversion are still significant for both the low wealth and high wealth group. For the low wealth group the hardship coefficient becomes less significant though, as it is now only significant at the 5% level as compared to the 1% level. Finally, all other variables show the same signs in these regressions as in the estimation with the full sample. Moreover, for the high wealth group the dummy variable of being married is significant while it is not significant for the low wealth group. Finally, the country level control variables are not significant for the low wealth group, but they are significant for the high wealth group. Again, the two-sided tobit regression to test the risky-share shows the same results for the below and above median group as the logistic regressions (see table 3 in appendix
Table 3
Low wealth High wealth
Low wealth with interaction
High wealth with interaction
-2.645** -1.616*** -4.937* -1.181***
(1.346) (0.211) (2.928) (0.41)
Risk aversion -0.899*** -0.797*** -0.358 -0.877***
(0.258) (0.062) (0.664) (0.09)
Hardship x risk aversion 0.761 -0.144
(0.815) (0.116)
Individual controls yes yes yes yes
Country controls yes yes yes yes
(Pseudo) R2 0.19 -0.24 0.19 0.24
Log. L -88.59 -1728.01 -88.09 -1727.25
N 1175 4031 1175 4031
Multivariate analysis of hardship and risk aversion on stock market participation separated by wealth groups. This table presents coefficient estimates of hardship and risk aversion measures on stock market participation using a logistic regression model. The dependent variable is a binary variable taking the value of 1 when an individuals has invested in stocks and taking a value of zero when an individual does not own stocks. Apart from hardship and risk aversion, I include wealth, income, years of education, age and age-squared, a dummy of being married, a male dummy, the Human Development Index, Rule of Law and the Anti Directors Rights Index. These control variables are not presented below as the results are familiar. Standard errors are in parenthesis. *** , ** and * indicate significance at the 1%, 5% and 10% level respectively.
B). To conclude, hardship and risk aversion significantly explain variation in stock market participation between countries and individuals for both low and high wealth individuals. However, both hardship and the other country level control variables decrease in significance for the low wealth group as compared to the high wealth group and the full sample. This might imply that country level variables are less relevant in explaining stock market participation of low wealth individuals as compared to high wealth individuals.
5.3 Hardship and other risky behaviour
Theory predicted that hardship results in an increased probability of gambling on a short life span (Mata et al., 2016). From this statement, I deduced that hardship results in reduced future orientation. This results in less investing, as this is a future oriented strategy, and more other risky behaviours. Therefore, the first hypothesis proposed that the relationship between hardship and stock market participation would be negative. The results confirm the negative relationship between the two variables, even when I controlled for relevant household and country level variables. Investing in stocks generally considered riskier than investing in bonds. Mata et al. (2016) stated that hardship is linked to an increased willingness to take risk, but the results show that this is not the case when it comes to stock market participation. At this point, the question remains whether hardship does not affect an individual’s risky behaviours in general or whether hardship does not increase an individual’s financial risk taking. I deem the second explanation to be more plausible. To test validity of this statement I will again run a regression but this time I will substitute stock market participation by other risky behaviours as dependent variable.
drinks on one occasion during the last three months. For both questions the answers range from not at all (option 1) to daily or almost daily (option 7).
First, I performed a formal test to examine the impact of hardship on smoking. The regression included the same control variables that were used to explain stock market participation. Column 1 of table 4 presents the results of the regression. The sign of the coefficient for hardship is positive and highly significant. This means that when controlled for income, wealth, education, age, marital status, gender and country level variables, hardship has a positive relationship with smoking. Column 2 and 3 show that the coefficient for hardship is also positive for both dependent variables related to alcohol consumption. A positive coefficient means that when hardship increases alcohol consumption increases. The coefficient for hardship related to both drinking behaviours is highly significant.
Based on this analysis of the effects of hardship on risky behaviours I conclude that hardship does increase risky behaviour. Its effects on smoking and alcohol consumption are positive and significant even when controlled for other individual and country level controls. These results indicate that hardship increases risky behaviours (Mata et al., 2016), but that the relationship does not hold for financial risk-taking.
Table 4
Smoking Drinking (days) Drinking (>6)
Hardship (standardized) 0.923*** 0.319*** 0.686***
(0.214) (0.116) (0.175)
0.277*** 0.096*** 0.047
(0.067) (0.039) (0.049)
Individual controls yes yes yes
Country controls yes yes yes
(Pseudo) R2 0.075 0.032 0.056
Log. L -1694.44 -9159.4 -4424.48
N 2711 5207 4242
Risk aversion
5.4 Separating the effects of the components of the hardship index
The hardship index captures both social and economic hardship within a country. The index consists of homicide rate, GDP per capita, income inequality, gender equality, infant mortality and life expectancy at birth. I already showed that hardship significantly explains stock market participation across Europe. This final analysis tests which components of the index drive the results. I perform a similar logistic regression as before, however now I replace hardship by its six separate components. Table 5 below presents the empirical results of the regression. All separate components are standardized to ease interpretation. The first column presents the results of the analysis excluding the country level control variables. In this analysis, the coefficients for homicide, income inequality and life expectancy are negative and highly significant. Infant mortality is negative and significant at the 5% level. These results are in line with expectation. GDP per capita is highly significant but positive, which is remarkable. This suggest that stock market participation is higher in countries with a lower GDP per capita. Regressing GDP per capita on stock market participation without the control variables results in a strong negative relationship between the two. Further inspection indicates that the negative effect in the full regression is caused by the correlation between GDP per capita and income and wealth. When income and wealth are excluded the coefficient for GDP becomes negative and highly significant. Finally, gender equality is not significant.
Next, I included the country level control variables in the regression (column 2 of table 5). In this regression, only the negative coefficients for homicide and infant mortality are highly significant. The results of this regression are not very reliable as the variance inflation factor exceeds 10 for all components and country level control variables.
6. Discussion
The key finding of this paper is that hardship negatively affects stock market participation across Europe. Individuals living in countries with a harsher and more unpredictable environment are less likely have money invested in stocks. Moreover, an individual’s risk aversion also negatively affects stock market participation. Further analysis revealed a positive relationship between hardship and smoking and drinking. Individuals living in harsher environments are more likely to smoke, drink more often and drink more. This is in line with the results of Mata et al. (2016) that hardship increases risk taking propensity. One possible explanation might be that hardship is related to an increased willingness to take risk in specific domains but that the financial domain is not included in these domains. Another explanation may be that the effect of hardship operates through another underlying mechanism which effect is not captured by the other control variables. This underlying mechanism might
Table 5 M1 M2 -0.62*** -0.83** (0.089) (0.4) 0.415*** -0.28 (0.01) (1.191) -0.745*** -0.306 (0.106) (0.398) 0.133 0.205 (0.122) (0.375) -0.113*** -0.989*** (0.06) (0.252) -0.544*** 0.244 (0.134) (0.952)
Individual controls yes yes
yes
(Pseudo) R2 0.30 0.30
Log. L -2058.85 -1819.73
N 6160 5207
Gender inequality
Multivariate analysis of the separate components of hardship on stock market participation This table presents coefficient estimates of the six separate components of hardship on stock market participation using logistic regression. The dependent variable is a binary variable taking the value of 1 when an individuals has invested in stocks and taking a value of zero when an individual does not own stocks. Apart from hardship , I include risk aversion, wealth, income, years of education, age and age-squared, a dummy of being married, a male dummy, the Human Development Index, Rule of Law and the Anti Directors Rights Index. Standard errors are in parenthesis. *** , ** and * indicate significance at the 1%, 5% and 10% level respectively.
GDP per capita
Country controls Homicide rate
Income inequality
be future orientation. This explanation holds for all observed behaviours. First, hardship reduces the future orientation of individuals in a country and future orientation is linked to stock market participation. Secondly, smoking and drinking behaviours are in line with being less future oriented. Ultimately, the effect of hardship is more pronounced for high wealth individuals than for low wealth individuals. This suggest that country level factors are less important as explanation for stock market participation for low wealth individuals. This latter finding again stresses the importance of wealth.
Limited stock market participation has been a topic of interest for a couple of decades. This study exposed hardship as additional explanation for stock market participation. Still, the participation puzzle remains complex. The exact mechanisms through which hardship works are not well-defined yet. The results do stress the importance of the effects of country specific characteristics on stock market participation. I draw two main implications from these results. First, governments shift the responsibility of an individual’s financial future more toward the individual. This study increases our understanding of how harsh and unpredictable environments affect the individual’s investment behaviour. In countries with more hardship, individuals are perhaps less likely to take their own responsibility and invest in stocks. Governments should take this into consideration when leaving this responsibility to individuals. The second one is related to the first. These results indicate that other factors than an unappealing risk-return trade-off affect participation for individuals. In this situation, governments could invest money from social security proceeds for example in the stock market on behalf of their population. These investments would result in more wealth accumulation.
6.1 Suggestions for further research
Further research focusing on the relationship between hardship and risk taking propensity could study the effects of hardship on risk propensity in different domains. This could shed more light on the effect of hardship and help explain or support these findings. Additionally, a measure of future orientation could be included to determine if this is the mechanism through which hardship operates.
Another direction for further research is to replicate this on a more representative sample. One limitation of the SHARE dataset is that it included almost singularly individuals aged 50 and above. This is confirmed by the average age of 69 in the sample. Using a sample with a wider age group would result in more reliable and generalizable results.
approximately ten percent of that number. Analysis has shown that there are large differences between the response and non-response groups for the dependent variable. More extensive analysis and splitting the sample into two groups did not alter the results from the study though. This suggests that even though the two groups differ the effects of the independent variables do not affect the groups differently.
References
Asgharian, H., Liu, L., Lundtofte, F., 2013. Institutional quality and stock market participation : learning to forget (december 2015). Available SSRN
https//ssrn.com/abstract=2369732. doi:10.2139/ssrn.2369732
Barber, B.M., Odean, T., 2001. Boys will be boys : gender , overconfidence , and common stock investment The Quarte, 261–292.
Börsch-Supan, A., 2016. Survey of health, ageing and retirement in Europe (SHARE). Wave 5. Release version 5.0.0. SHARE-ERIC. Data set. doi:10.6103/SHARE.w5.500
Bromiley, P., Curley, S., 1992. Individual differences in risk taking. J. F. Yates (Ed.), Risk-taking Behav. 87–132.
Brumbach, B.H., Figueredo, A.J., Ellis, B.J., 2009. Effects of harsh and unpredictable environments in adolescence on development of life history strategies: A Longitudinal Test of an Evolutionary Model. Hum. Nat. 20, 25–51. doi:10.1007/s12110-009-9059-3 Campbell, J.Y., 2006a. Household Finance. J. Finance v61, 1553–1604.
Campbell, J.Y., 2006b. Household finance. J. Finance 61, 1553–1604.
Chrisiansen, C., Schroter, J., Rangvid, J., 2015. Understanding the effects of marriage and divorce on financial investment: The role of background risk sharing. Econ. Inq. 53, 431–447. doi:10.1111/ecin.12113
Cocco, J.F., Gomes, F.J., Maenhout, P.J., 2005. Consumption and portfolio choice over the life cycle. Rev. Financ. Stud. 18, 491–533. doi:10.1093/rfs/hhi017
Cole, S., Shastry, G.K., 2009. Smart money : The effect of education , cognitive ability , and financial literacy on financial market participation. Unpubl. Pap. from Harvard Bus. Sch.
Conlin, A., Kyröläinen, P., Kaakinen, M., Järvelin, M.-R., Perttunen, J., Svento, R., 2015. Personality traits and stock market participation. J. Empir. Financ. 33, 34–50.
doi:10.1007/s13398-014-0173-7.2
Dohmen, T., Falk, A., Huffman, D., Sunde, U., Schupp, J., Wagner, G.G., 2011. Individual risk attitudes: Measurement, determinants, and behavioral consequences. J. Eur. Econ. Assoc. 9, 522–550. doi:10.1111/j.1542-4774.2011.01015.x
Dohmen, T., Falk, A., Huffman, D., Sunde, U., Schupp, J., Wagner, G.G., 2005. Individual risk attitudes: new evidence from a large, representative, experimentally-validated survey. IZA Discuss. Pap. 1–56.
doi:10.1198/073500107000000287
Gali, J., 1994. Keeping up with the Joneses : Consumption Externalities , Portfolio Choice , and Asset Prices. J. Money , Credit Bank. 26, 1–8.
Georgarakos, D., Pasini, G., 2011. Trust, sociability, and stock market participation. Rev. Financ. 15, 693–725. doi:10.1093/rof/rfr028
Giannetti, M., Koskinen, Y., 2010. Investor protection, equity returns, and financial globalization. J. Financ. Quant. Anal. 45, 135. doi:10.1017/S0022109009990524 Gilmore, A.B., McKee, M., Telishevska, M., Rose, R., 2000. Epidemiology of smoking in
Ukraine. Prev. Med. (Baltim). 33, 453–461.
Grinblatt, M., Keloharju, M., Linnainmaa, J., 2011. American finance association IQ and stock market participation. J. Finance 66, 2121–2164.
Guiso, L., Sapienza, P., Zingales, L., Guiso, L., Sapienza, P., Zingales, L., 2008. Trusting the stock market. J. Finance 63, 2557–2600.
Hair, J.F.J., Anderson, R.E., Tatham, R.L., Black, W.C., 1995. Multivariate data analysis (3rd ed). New York: Macmillan.
Haliassos, M., Bertaut, C.C., 1995. Why do so few hold stocks? Econ. J. 105, 1110–1129. Halko, M.L., Kaustia, M., Alanko, E., 2012. The gender effect in risky asset holdings. J.
Econ. Behav. Organ. 83, 66–81. doi:10.1016/j.jebo.2011.06.011
Hamid, F.S., Rangel, G.J., 2013. The relationship between risk propensity , risk perception and risk-taking behaviour in an emerging market. Int. J. Bank. Financ. 10, 134–146. Hill, E.M., Chow, K., 2002. Life-history theory and risky drinking. Addiction 97, 401–413.
doi:10.1046/j.1360-0443.2002.00020.x
Hill, E.M., Ross, L.T., Low, B.S., 1997. The role of future unpredictability in human risk-taking. Hum. Nat. 8, 287–325. doi:10.1007/BF02913037
Hong, H., Kubik, J.D., Stein, J.C., 2004. Social interaction and stock-market participation. J. Financ. @Bullet LIX, 137–163. doi:10.1111/j.1540-6261.2004.00629.x
Howlett, E., Kees, J., Kemp, E., 2009. The role of self-regulation, future orientation, and financial knowledge in long-term financial decisions. J. Consum. Aff. 42, 223–242. Mankiw, N.G.G., Zeldes, S.P.S.P., 1991. The consumption of stockholders and
nonstockholders. J. financ. econ. 29, 97–112. doi:10.1016/0304-405X(91)90015-C Markiewicz, Ł., Weber, E.U., 2013. DOSPERT’s gambling risk-taking propensity scale
Markowizt, H., 1952. Portfolio selection. J. Finance 7, 77–91.
Masters, R., 1989. Study examines investors’ risk-taking propensities. J. Financ. Plan. 2, 151–156.
Mata, R., Josef, A.K., Hertwig, R., 2016. Propensity for risk taking across the life span and around the globe. Psychol. Sci. 27, 231–243. doi:10.1177/0956797615617811
Mehra, R., Prescott, E.C., 1985. The equity premium: A puzzle. J. Monet. Econ. 15, 145– 161. doi:10.1016/0304-3932(85)90061-3
Mount, M., Barrick, M., 1995. The Big Five personality dimensions: Implications for
research and practice in human resources management. Res. Pers. Hum. Resour. Manag. a Res. Annu. 13, 153–200.
Nicholson, N., Fenton-O’Creevy, M., Soane, E., Willman, P., 2002. Risk propensity and personality. . London. Edu/Docs/Risk. 1–33. doi:10.1080/1366987032000123856 Paiella, M., 2001. Limited financial market participation: A transaction cost-based
explanation. Inst. Fisc. Stud. WP 01/06.
Rahkonen, O., Laaksonen, M., Karvonen, S., 2005. The contribution of lone parenthood and economic difficulties to smoking. Soc. Sci. Med. 61, 211–216.
Roff, D., 2002. Life history evolution. Sunderland, MA: Sinauer.
Sitkin, S., Pablo, A., 1992. Reconceptualizing the determinants of risk behavior. Acad. Manag. Rev. 17, 9–38.
Spamann, H., 2010. The “antidirector rights index” revisited. Rev. Financ. Stud. 23, 467– 486. doi:10.1093/rfs/hhp067
Stearns, S.C., 1992. The evolution of life histories. Oxford: Oxford University Press. Steward, W.H., Roth, P.L.L., 2001. Risk propensity differences between entrepreneurs and
managers: A meta-analytic review. J. Appl. Psychol. 86, 145–153. doi:10.1037//0021-9010.86.1.145
Szrek, H., Chao, L.-W., Ramlagan, S., Peltzer, K., 2012. Predicting (un)healthy behavior: A comparison of risk-taking propensity measures. Judgm. Decis. Mak. 7, 716–727. van Rooij, M., Lusardi, A., Alessie, R., 2011. Financial literacy and stock market
participation. J. financ. econ. 101, 449–472. doi:10.1016/j.jfineco.2011.03.006 Vissing-Jorgensen, A., 2003. Perspectives on behavioral finance: Does “irrationality”
disappear with wealth? Evidence from expectations and actions. SSRN Electron. J. 18, 139–208. doi:10.2139/ssrn.417421
