Cultural Differences in Risk Tolerance*Christoph S. Weber**

IWE Working Paper No. 01-2013

ISSN: 1862-0787

Erlangen, May 2013

Institut fur Wirtschaftswissenschaft(Institute of Economics)Kochstraße 4 (17), 91054 Erlangen, GermanyPhone: +49 9131 85-22376Fax: +49 9131 85-22060E-mail: christoph.weber@fau.deInternet: www.economics.phil.fau.de

*The data used in this publication were made available by the German Socio-EconomicPanel Study (GSOEP) at the German Institute of Economic Research (DIW Berlin), Berlin.**The author wishes to thank Jurgen Kahler for very useful comments and suggestions.

Abstract

Risk aversion is of great importance on a microeconomic and a macroeconomic level.First of all, risk aversion is an important factor in explaining many everyday decisions.Among these are decisions to invest money in different types of assets, the decisionto enter the labour market and others like the decision to move (Guiso and Paiella,2004). As risk tolerance is so important in life, one asks which determinants can explaindifferent attitudes towards risk. The question is then how risk averse people or specialgroups are. Several studies show different risk attitudes with respect to gender, age,income, and wealth (see e.g. Dohmen et al., 2011). While these findings are hardlycontroversial, there is still some uncertainty about the effect of culture on risk tolerance.Thus, the main issue of this survey is to elaborate possible differences in risk preferencesthat are caused by cultural background. The main question in this context is whetherreligion or nationality are of importance for explaining risk attitudes. For this purpose,this study employs the German Socio-Economic Panel (GSOEP) to figure out differencesin risk attitudes with respect to certain characteristics. The estimations show that thecultural background does indeed have some impact on risk taking behaviour. Thus, bothProtestants and atheists tend to be less risk averse than people with other religions ordenominations. Furthermore, foreigners are less likely to be risk prone. This applies alsofor specific nationalities like Turks, Italians, Greeks, and Ex-Yugoslavs.

JEL: D12, D80

Keywords: Risk Aversion, Risk Tolerance, Cultural Background, Generalized OrderedLogit Model

II

1 Introduction

Risk aversion is of great importance on a microeconomic and a macroeconomic level.First of all, risk aversion is an important factor in explaining many everyday decisions.Among these are decisions to invest money in different types of assets, the decision toenter the labour market and others like the decision to move (Guiso and Paiella, 2004).Therefore, risk preferences play a key role in economic decisions. Assuming that riskaverters invest less in risky assets, they will get a lower return in the long run. Thiscould result in greater inequalities especially when we think of richer people to be lessrisk averse. Furthermore, people who are more risk tolerant tend to earn more money.This is due to the fact that risk prone individuals work more often in risky sectors1 ofthe economy (see e.g. Shaw, 1996). Higher risk tolerance also increases the probabilitythat somebody works as entrepreneur2 (Cramer et al., 2002). Risk-taking behaviouris also an important factor for successful entrepreneurial skill (van Praag and Cramer,2001). This has macroeconomic consequences as well. Societies with lower uncertaintyavoidance might be more innovative (Shane, 1995). Taking this into account risk aversepeople tend to work less in risky sectors, be less willing to found a company or workas self-employed (Ekelund et al., 2005) and work less frequently in the research area3.Finally, this might affect economic growth in the long run (Klasing, 2008).

As risk tolerance is so important in life, one asks which determinants can explaindifferent attitudes towards risk. The question is then how risk averse people or specialgroups are. Several studies show different risk attitudes with respect to gender, age,income, and wealth (see e.g. Dohmen et al., 2011). Thus, women are more risk aversethan men. Risk aversion also increases also with age. Income and wealth mainly havethe same effect meaning that risk tolerance increases when income and wealth increases.While these findings are hardly controversial, there is still some doubt about the effectof culture on risk tolerance. Thus, the main issue of this survey is to elaborate possibledifferences in risk preferences that are caused by cultural background. The main ques-tion in this context is whether religion or nationality are of importance for explaining therisk attitudes. For this purpose, this study employs the German Socio-Economic Panel(GSOEP) to figure out differences in risk attitudes with respect to certain characteristics.The advantage of this survey is that it has a large number of observations which makesit easier to find specific effects.

This paper is structured as follows. Section 2 gives an overview of existing studiesabout attitudes towards risk. Then the data and the estimation procedure are briefly

1Risk averse individuals prefer to work in the public sector. The amount of risk aversion is also correlatedwith wage, but only in the private sector (Pfeifer, 2011).

2Xu and Ruef (2004) have a contrary point of view. According to them, entrepreneurs are more riskaverse than non-entrepreneurs and other managers.

3To be honest, in a special context of a contest with limited liabilities risk averse are more likely towin (Skaperdas and Gan, 1995). However, this seems to be only an exception to the rule that risk averseindividuals perform worse.

1

explained. In Section 4, estimation results are presented and discussed. Section 5 con-cludes.

2 Literature Review

Some aspects of risk preferences have already been covered. The following part will be aliterature review on differences in risk aversion with respect to gender, age, wealth andincome, and cultural background (religion, nationality, race).

2.1 Gender differences in risk preferences

The role of gender, in particular, has been analyzed quite a lot. Many studies haveshown that women are generally more risk averse than men (see e.g. Brooks et al.,2009; Cohen and Einav, 2007; Watson and McNaughton, 2007; Bajtelsmit et al., 1999)or, indeed, have different risk preferences (see e.g. Croson and Gneezy, 2009). Womenare also considered to be more loss averse (Brooks and Zank, 2005) and, to a largerextent, to exhibit loss aversion4 (Schmidt and Traub, 2002). Furthermore, women arealso not only less risk-seeking, they have different ways of making decisions (Powell andAnsic, 1997). To explain gender differences in risk behaviour, there are some differentexplanations5. Research by Amanatullah et al. (2010) explored the difference in riskaversion by gender in managerial decision-making and found it to be a result of the self-interest-orientation. Men are more selfish and, therefore, make riskier choices, which aremore in their interest than in the corporation’s interest. In contrast, women focus moreon the interest of the firm they work for. Another explanation for differences may beoverconfidence of men. According to this, men are more convinced about their decisionsthan women (Barber and Odean, 2001). This may indicate that men do not think thatthey are acting more risk prone than women, since they believe they can predict possibleoutcomes more accurately. Another explanation in line with the Prospect Theory6 is thatwomen have a different perception of probabilities. Thus, they think that gains are lesslikely than they actually are (Schubert, 2006).

Other studies expand the gender issue with studies about the role of marital status.In general, married people are risk averters (see e.g. Sunden and Surette, 1998). Inmarried couples, wives are more risk averse than husbands (Arano et. al., 2010). Inaddition, a man with a wife who is ready to take some financial risk is more risk-seekingthan a man with a very risk averse wife. This effect does not occur to the same extent for

4Tversky and Kahneman (1991) introduced the concept of loss aversion. According to this theory, peopleare more responsive to losses than to gains. Loss aversion is the key assumption of Prospect Theory. Hence,people are twice as sensitive to risks as to gains (Kahneman, Tversky, 1979).

5Other explanations to those presented here are given by Graham et al. (2002) and Bajtelsmit andBernasek (1996).

6Kahnemann and Tversky developed Prospect Theory as an alternative to expected utility theory in 1979.According to this theory, people’s behaviour depends on whether they act in the “loss area” or the “gainarea”. If they are below their target they are risk prone. If one is above his target he acts risk averse.

2

a woman and her husband7 (Grazier and Sloane, 2008). Besides, risk preferences alsodiffer among individuals who do not have children and those with children. Thus, mar-ried individuals with children are more risk averse than individuals without children8.Furthermore, single mothers are less risk-seeking than single people without children(Grazier and Sloane, 2008).

However, there are also several studies that show opposite results when reviewingthe role of gender. Thus, women are not more risk averse than men (Papke, 1998; Em-brey and Fox, 1997; Johnson and Powell, 1994). For example, Borghans et al. (2009)show that women’s preferences do not differ from men’s all the time when ambiguityis introduced in a game. For example, at greater levels of uncertainty, women are asrisk averse as men (Borghans et al., 2009). Moreover, the survey of Adhiraki (2009)did not display a significant dependency of risk aversion on gender or marital status.Following his interpretation, the difference in risk-taking between genders arises fromthe fact that women think that they know less about financial markets and assets. AlsoAtkinson et al. (2003) assume that possible gender differences may come from the factthat women are less informed about financial markets and that they are less wealthy.According to Booth and Nolen (2012), the environment plays an important role in gen-der differences. Women are less risk averse when educated in single-gender schoolsand acting in a solely female experimental group. Furthermore, Johnson and Gleason(2005) conclude in analyzing a game show that men act more cautiously than women.The reason why there are sometimes differences in risk preferences could be caused bydifferent investment opportunities (Schubert et al., 1999). To summarize, the usual as-sumption is that women are risk averters. The majority of studies find this correlationbetween gender and risk preferences. However, several works have also shown that thisstereotype does not always appear.

2.2 Risk preferences and age

Another approach to explain the amount of risk aversion is by considering the factor ofage. Arrow considered relative risk aversion9 to increase with respect to age. There isindeed some evidence for this hypothesis (see e.g. Morin and Suarez, 1983; Palsson,1996). According to Halek and Eisenhauer (2001) elderly individuals are more likely tobe risk averse than younger ones. Thus, risk aversion decreases with respect to age upto a certain point and then increases again (see Cohen and Einav, 2007). This can beexplained partly by the investment’s horizon. The shorter the duration of an investment,the more risk-free portfolios are structured (see Berkelaar et al., 2004). Age-dependent

7Though women are considered to be more influenced by their husbands, women who are recommendedby their husband to invest more in assets are still more risk averse than similar men who are told by theirwives to be less risk averse (Arano et al., 2010).

8This can be explained by the fact that risk aversion also influences the time of marriage. More risk-averse individuals marry earlier than risk inclined individuals (Spivey, 2010).

9The distinction between absolute and relative risk aversion is not of importance in this survey. Thus, itis not explicitly mentioned. A definition of absolute and relative risk aversion can be found in Pratt (1964).

3

risk aversion is also a result of the life-cycle models (see Gourinchas and Parker, 2002).An inquiry, however, by Bommier and Rochet (2006) found that risk aversion increasesif the investment period is longer. This might explain that older people hold more riskyassets.

In contrast, Wang and Hanna (1997) found relative risk aversion to decrease withrespect to age. Moreover, moderate decreasing relative risk aversion was observed alsoby Bellante and Green (2004). These surprising findings may be explained by myopicloss aversion (Benartzi and Thaler, 1995). Younger investors do not consider the longertime horizon for investments which they have. Their decisions are, thus, biased to con-sumption in the short-run. Nevertheless, age has an impact on risk preferences, even ifsome discrepancies still remain.

2.3 Risk preferences and education

Another factor which may influence risk attitudes is education. Several studies haveshown the effect of education on risk aversion. According to them, people with highereducation are usually less risk averse (see e.g. Bertaut, 1998). Thus, more educatedpeople tend be more risk prone than less educated people (Shaw, 1996). This effect isonly significant for speculative risk, whilst the aversion to pure risk even increases amongthe higher educated. This is due to the fact that one must first search for speculative risk(Halek and Eisenhauer, 2001). The question is where this effect may come from. Acommon assumption is that this effect is moderated by cognitive ability. In fact, peoplewith lower cognitive ability are greater risk averters (Dohmen et al., 2010). Surprisingly,financial literacy does not affect decisions10 (Cole and Shastry, 2012). Consequently, abetter education in finance would not decrease risk aversion. However, one has to becareful with those assumptions because the effect could also be the other way round.Since risk aversion is also a key determinant in the decision to enter higher education(Belzil and Leonardi, 2007), cause and consequence could be confused.

2.4 Risk preferences and income/ wealth

Along with gender, age, and education income also plays a major role in explainingrisk taking behaviour. The way wealth affects risk preferences has, for instance, alreadybeen discussed by Arrow (1970) who proposed that absolute risk aversion decreaseswith respect to wealth and relative risk aversion increases with respect to wealth. Whiledecreasing absolute risk aversion is usually supported (Levy, 1994), the hypothesis ofincreasing relative risk aversion is very controversial. There are several studies showingthat relative risk aversion decreases with respect to wealth (Riley and Chow, 1992; Cohnet al., 1975). However, there are also studies which find constant risk aversion (Friendand Blume, 1975). This is in line with the work of Siegel and Hoban (1982) who found

10In comparison, Grable (2000) draws the conclusion that people who are better informed about financialmarkets tend to be more risk tolerant.

4

evidence for increasing relative risk aversion. On the other hand Brunnermeier andNagel (2008) came up with the conclusion that wealth changes do not influence riskpropensity. Nonetheless, the usual assumption nowadays is still that risk tolerance in-creases as wealth increases. Apart from wealth, income plays a role in risk preferences.People with higher income tend to be less risk averse (see e.g. Tanaka et al., 2010). Notonly does the salary matter, however, but also tenure increases risk tolerance (Agnewet al., 2003). This may be due to higher job safety11. Moreover, income uncertain-ties negatively affect willingness to take risks (Grazier and Sloane, 2008). In contrast,Vereshchagina and Hopenhayd (2009) find poorer entrepreneurs to be more likely to berisk-inclined. The Prospect Theory may give an explanation for behaviour in this mannerand also for increasing relative risk aversion. Someone who is below his reference pointis willing to take on risks to improve his situation.

2.5 Risk preferences and cultural background

Previous studies have already partly revealed the role of nationality, ethnicity, and reli-gion on risk preferences. Sung and Hanna (1996) showed, for instance, that Blacks12

and Hispanics in the U.S. are less risk tolerant than whites. Similar to this conclusion, astudy by Brown (2007) says that Blacks and Hispanics tend to have less stocks, whereaseven low-income Whites are more risk tolerant than Blacks and Hispanics at any incomelevel. A study with newer data supports the greater risk aversion among Hispanics, butdoes not find a significant effect in Blacks (Coleman, 2003). However, Weber and Hsee(1998) conclude that culture is not the reason for that. According to them, the effectonly comes from the fact that people from different cultures perceive risk differently.The reason why differences between racial groups are often found could be that ethnic-ity is confused with other variables such as household size and the number of children(Gutter et al., 1999). Thus, blacks and Hispanics are only more risk averse, because theyhave larger families.

In addition, nationality also matters in the decision-making processes. Accordingto their self-assessment, the French, Indians, British, Chinese and US-Americans con-sider themselves to be highly risk tolerant. Germans take an average position, whilstBrazilians, Mexicans and the Spanish consider themselves to be risk averters. Only theJapanese assess themselves as highly risk averse (Salacuse, 1998). A study by Cummingset al. (1971) indicates that Americans are the most risk prone13. In this work individualsfrom Central Europe (in this instance: France, Belgium, Germany, and Great Britain) aremore risk averse than Scandinavians, the Spanish and the Greeks.

Furthermore, recent cross-country comparisons revealed the surprising14 fact that11Another explanation comes from the observation by Schooley and Worden (1996) that part-time work-

ers are less risk-seeking.12In addition, Benjamin et al. (2010) show that non-immigrant Blacks are more risk averse than immi-

grant Blacks.13Risk propensity and risk tolerance are used as synonyms in this article.14The surprise seems to diminish when taking into account a study by Weber et al. (1998). Hence, a

5

the Chinese are more risk tolerant than Americans15 (Weber and Hsee, 1998; Fan andXiao, 2005). This finding is also supported by an experimental analysis by Wang and Fis-chbeck (2008). In contrast, a survey on risk tolerance among German residents (Bartkeand Schwarze, 2008) concludes that someone’s nationality cannot explain his or her risktolerance16. The different risk preferences between people with different citizenshipscan be explained by some other factors. Among these factors are religiousness and reli-gion which have a significant effect on risk aversion. In addition to previous studies, thefindings are not only that religious people are more risk averse than non-religious peo-ple, but also that among religious people Christians are less risk averse than Muslims17.There are further studies which try to capture the effect of religion and religiousnesson risk attitudes. Several studies have already shown that religiousness is linked to riskaversion (see e.g. Hilary and Hui, 2009). In general, religious individuals seem to bemore risk-averse than atheists18. For instance, firms with a local base in a country withhigher religiousness are less willing to take on risk (Hilarya and Hui, 2009). Since beingirreligious was a risky behaviour in Western (Christian and Muslim) societies, atheistswere more risk prone. In contrast, being irreligious in Eastern (Hindu and Buddhist)societies did not reflect risk-taking behaviour. Therefore, being atheist in these countriesis not correlated with risk tolerance (Miller, 2000).

3 Data

After discussing previous studies about risk aversion, we will now briefly explain thedata set that is employed in this study. Then we will discuss the estimation procedure.

3.1 Data

In this paper, we use a panel data set provided by the German Institute for EconomicResearch (DIW). This German Socio-Economic Panel Study (SOEP) was first conductedin 1984. There are 27 waves of the SOEP up to now. This paper uses results from the2009 questionnaire as this contains some specific questions about risk aversion. Theadvantage of this data set is that it contains more than 22,000 individuals19.

The dependent variable is always a variable that covers attitudes towards risk. Par-ticipants are asked to self-assess their risk preferences on an 11-point scale with values

comparison of American and Chinese proverbs reveals that one would expect Chinese to be less risk averse.However, this is only true for financial, not social risks.

15The risk aversion of the Chinese is lower in the investment domain only. An explanation for the higherrisk tolerance is the “cushion hypothesis”, which says that people in a collectivist society expect to getfinancial help if they need it. Therefore, the Chinese are more risk prone (Hsee and Weber, 1999).

16This also partly contradicts the effect that migrants are more risk tolerant (Jaeger et al., 2007; Halekand Eisenhauer, 2001).

17Hence, Benjamin et al. (2010) find Catholics to be less risk prone than Protestants and Jews.18Since women are more risk averse than men, the higher level of religiousness among women can be

explained by their risk preferences (Miller and Hoffmann, 1995).19For further information about the GSOEP, see Wagner, Frick, and Schupp (2007).

6

Values General willingness totake risks

Willingness to take risksin financial matters

No Answer 0.41 1.540 9.21 37.691 7.55 14.032 14.48 15.273 16.16 11.34 11.39 5.755 20.06 7.96 9.03 2.537 7.36 1.968 3.41 1.249 0.55 0.3310 0.38 0.46

Table 1: Relative Frequencies of Answers to Questions about Risk Tolerance (DataSource: GSOEP 2009 )

between 0 and 10 . A value of 0 means that a person says that he or she is not risk proneat all. A value of 10 indicates that the individual thinks that he is very risk prone.

There are eight different variables covering risk preferences in the SOEP. The first isthe personal willingness to take risks. Then there are questions relating to risk takingbehaviour in specific situations. Among these is risk tolerance while driving, in financialmatters, in sports and leisure, and in occupation. Furthermore, there are questionsdealing with willingness to take health risks and risk in trusting other people. The lastone asks how much people would like to invest in a game after winning the lottery.We focus on the general willingness to take risks and the willingness to take risks onfinancial matters as these are the most important in an economic context. Table 1 showsthe frequencies of answers to the questions about risk tolerance.

This table shows the proportion of people answering the questions about general riskpreferences and risk tolerance in financial matters. Also the proportion of participantswho did not answer that question is shown. According to general risk tolerance, morethan 30 percent consider themselves as very risk averse (values of 0,1, or 2). Around27 percent are moderately risk averse (values of 4 or 5) and 20 percent take an averageposition (value of 5). 16 percent of the participants assess themselves as moderately riskprone (values of 6 or 7), while only 4 percent are relatively risk tolerant (values of 8, 9,or 10). The amount of people being highly risk averse is highest in the context of risktaking in financial matters. Thus, more than 66 percent consider themselves as very riskaverse in this context.

Table 2 shows the relative frequencies of answers to the question about willingness totake risks about winning the lottery. Participants were asked how much of a hypothetical

7

Answer Relative Frequency

100,000 Euro 1.8880,000 Euro 0.5560,000 Euro 2.4940,000 Euro 6.7220,000 Euro 13.01Nothing 74.78No Answer 0.57

Table 2: Relative Frequencies of Answers for Risk Taking after Winning the Lottery (DataSource: GSOEP 2009 )

lottery win of 100,000 euros they would like to invest into a game in which the betwill be either doubled or halved with a probability of 0.5 in each case. 0.57 per centdid not answer that question. The overall majority is not willing to spend any of thathypothetical lottery win. Only around 5 percent would like to risk more than half of that100,000 euro.

This analysis uses several different independent variables. Among these are gender,marital status, number of children, age, a dummy for a university degree, salary, wealth,nationality, and religion as these are routinely included in all estimations of determinantsof risk aversion. Furthermore, we use variables for political position, the degree ofworries about financial markets, the current health status, the degree of job worries, thedegree of satisfaction, and a dummy for unemployed. All these variables are taken fromthe 2009 sample with exception of the variables for religion and wealth which are takenfrom the 2007 sample because these variables were not included in the 2009 sample. Asdata sets had to be merged, the maximum number of observations decreased to roughly17,000.

3.2 Estimation Procedure

The dependent variable in this study is a variable with 11 different possible values.However, we cannot say anything about the differences between two outcomes. Theonly valid conclusion is to put people in order according to their self-assessment. Thismeans that the dependent variable is an ordinal one. Therefore, it is not useful to usea simple linear regression framework as this requires a metric measurement level. Theproblem is that OLS estimators are not efficient as they do not provide the minimumvariance estimator in this context. Furthermore, the standard errors would be biasedusing an OLS regression. Other papers, therefore, use different techniques like inter-val regressions (Dohmen et al., 2011). In this paper, we use an ordinal logit model.However, the problem is that a simple ordinal logit model uses the proportional oddsassumption. This means that the model has the same coefficients for all values of the

8

dependent variable. This assumption is often violated. A procedure for testing the pro-portional odds assumption is the Brant Test of parallel regression assumption (Brant,1990). Running this test reveals that the assumption is violated. A solution to overcomethis problem is to use a generalized ordered logit model. In this case, we use the gener-alized ordered logit model provided by R. Williams (2006). After estimating this model,average marginal effects are computed. However, every model was also estimated usingan ordinal logit model. The estimations show very similar results. The lottery game,however, was estimated using a simple ordered logit model.

4 Estimation Results

In this section, the results of the estimation of the variables that influence risk takingbehaviour are presented. As mentioned in the last section, we use a generalized orderedlogit model. For reasons of clarity, the tables are presented in the Appendix. Sinceall estimations contain a huge amount of information, we will not discuss every singlevalue. Thus, the first step is to present and discuss the estimation results with respect toeach dependent variable. Then the differences between the estimations are discussed.

First, we take a look at the determinants of the general attitudes towards risk. Resultsof the estimations are shown in Table 3 and 4 in the appendix. Gender has the expectedeffect on risk taking behaviour. Thus, women have a higher probability of being veryor moderately risk averse (values from 0 to 3). On the other hand, they have a lowerprobability of being risk neutral or risk prone.

The same accounts for being married. Married individuals are more probably beingrisk averse and less probably being risk prone. In contrast to previous studies, risktolerance is positively correlated with the number of children. However, this study uses avariable for children not living at home anymore while previous studies used the numberof children living at home. The next independent variable is age. As in previous studies,risk propensity decreases with age. On the other hand, elderly individuals are morelikely to be very risk averse and less likely to be risk neutral or risk prone. As previousstudies showed a correlation between education or cognitive ability and risk tolerance,a dummy variable for having a university degree was included. This variable did not,however, have explanatory power in this model. Therefore, it is not included in Tables3 and 4. On the contrary, income has the expected effect on attitudes towards risk.Thus, the higher the income, the lower the probability of being very risk averse. Salarysignificantly decreases the likelihood of being risk neutral or moderately risk prone. Inthe first model, however, it decreases the likelihood of being risk-neutral. Wealth has,to a large extent, the same effect as income. It decreases the probability of being veryrisk averse and increases the probability of being risk neutral. However, the connectionbetween being rich and being moderately or very risk prone is ambiguous.

Besides those variables that are routinely included in estimations of risk taking be-haviour there are several other variables that matter in this estimation. The first oneis a variable which shows how worried respondents are about their finances. The re-sponse is again measured on a 3-point scale. The higher the value of this variable is,

9

the smaller the worries are about finances. Therefore, people with large worries aboutfinances have a higher probability of being very risk averse and have a lower probabilityof being moderately risk averse and risk neutral. On the other hand, people with virtu-ally no worries about finances are not significantly more likely to be risk prone. In thefirst model they even are less likely to be very risk prone. This could be explained bythe fact that people who are not concerned about their future will not risk their futureby being too risk prone. The degree of life satisfaction has a similar effect. Participantswere asked how satisfied they are with their lives with respect to the last five years. Theestimation shows that more satisfied people tend to be less risk averse. Moreover, thoseindividuals with higher degrees of life satisfaction are more likely to be moderately orvery risk prone. This is in contrast to Prospect Theory which says that people tend tobe risk prone if they are below a subjective reference point (this could be income level).On the other hand, people tend to be risk averse if they are above this reference point.The results reveal exactly the opposite. Being highly satisfied increases the probabilityof being risk prone and vice versa.

Current health status also has some impact in explaining risk attitudes. Participantswere asked about their current health status. The higher the value is, the worse thehealth status is. Thus, a bad health status raises the probability of being risk averseand decreases the probability of being moderately or very risk prone. Furthermore, adummy for being unemployed is included. The unemployed show a higher probabilityof being moderately risk averse and a lower probability of being risk neutral or veryrisk prone. The overall result of all those variables is that being in an uncomfortablesituation leads people to be more risk averse as they are concerned about potentiallosses. An alternative explanation might be that people tried risky behaviour in the pastand lost and are therefore now both unsatisfied with their situation and also risk averse.

Moreover, planning to leave occupation in the following two years decreases theprobability of being completely risk averse. It also makes it more likely that somebodyis moderately risk prone. Furthermore, the political position can also partly explainattitudes towards risk. People were asked how they self-assess themselves on a left-rightscale according to their political attitudes. This variable is also measured on an 11-pointscale with a value of 1 meaning that an individual says about himself or herself that heor she is completely left-wing. According to their self-assessment, the overall majorityof people is neither left wing nor right wing as they have of a 4, 5 or 6 on this scale.Roughly ten percent are moderately right and roughly 18 percent are moderately left.According to the results in Table 3, being more left-wing increases the probability ofbeing very risk averse. In contrast, being less left increases the likelihood of being riskneutral, moderately risk prone and very risk prone. This may be explained by the factthat the left-wing tend to favour equality in contrast to liberty. Thus, they also dislikesituations in which people can be better or worse off. Consequently, they try to avoidrisky choices and favour situations with certain outcomes.

Lastly, we discuss the role of cultural background as this is the main purpose of thispaper. The effect of being a foreigner is ambiguous. It increases the probability of be-ing maximally risk averse and also of being very risk prone. This could be explained

10

by the fact that the group of foreigners in this data set is heterogeneous. It consists ofpeople living in Germany for many years and people who immigrated just a few yearsago. Some argue that immigration is a risky behaviour. Thus, immigrants should havea higher risk propensity. On the other hand, individuals might be risk averse after theymoved to another country. Therefore, we need to take a closer look at certain nationali-ties. The data set provides information about Germans, Turks, Italians, immigrants fromthe former Yugoslavia, Greeks, and Austrians. The other groups of nationalities of partic-ipants in the GSOEP are simply too small to use them in this analysis. Table 4 shows allvariables covering nationality have some explanatory power. Thus, Greeks have a lowerprobability of being very risk averse and Austrians and Italians have a lower probabilityof being moderately risk averse. Furthermore, Greeks and Yugoslavs are less likely to bemoderately risk prone. Turks, on the other hand, have a lower probability of being riskneutral but a higher probability of being moderately risk prone.

Furthermore, the role of religion was also of importance in previous studies. In thisestimation, dummy variables for being Protestant and being atheist are included. Thereference group consists of Catholics, Muslims and Christians with other denominationsin the first estimation. Protestants have a lower probability of being very risk averse.Similarly, being atheist decreases the likelihood of being very risk averse. However,atheists are also more likely to be risk neutral and moderately risk prone. Both effectscan partly be explained by German history. Being atheist was a risky behaviour in thepast as these people were prosecuted. Also Protestants had to fight for their faith andwere prosecuted in several regions in which Catholicism was the dominant denomina-tion. We also tested whether being Muslim plays a role in this estimation. It decreasedthe probability of being moderately risk prone. In the ordered logit model the dummyfor Muslims was not significant. This means that the higher risk aversion among Turkscannot solely be explained by their religion. Religiousness is not important in this esti-mation. Religiosity is measured by the frequency of attending religious services. Highervalues mean that people go less often to church. Hence, there is no evidence that morereligious people are more risk averse.

The data set does not only provide information about general risk attitudes but alsoabout risk aversion in financial matters (see Table 5). The effect of gender is somewhatdifferent as in the previous estimation. Being female significantly increases the proba-bility of being highly risk averse. In contrast to the general risk attitudes, women havea lower probability of being moderately risk averse. The influence of marriage is am-biguous. It decreases the likelihood of having a value of 2 in the generalized orderedlogit. As before, people with children tend to be less risk averse. Again, risk aversion iscorrelated with age. Thus, elderly are more likely to be completely risk averse. However,it decreases the probability of being moderately risk averse. Contrary to the previous ex-ample, educational background is a relevant factor in this context. Having a universitydegree lowers the likelihood of being completely risk averse. Salary has the same effectas in the previous estimation. Income is positively correlated with being moderately riskaverse. Wealth, on the whole, has mainly the same effect. Worries about finances makeit less probable to be very risk averse, moderately risk averse or risk neutral. Life and

11

income satisfaction largely have the same effect. More satisfied people tend to be lessrisk averse. In contrast to the previous estimation, current health status does not playa role in explaining the attitudes towards financial risk. Unemployment has a some-what different effect than before. It increases the likelihood of being completely riskaverse but decreases the probability of being very risk averse (value of 1). Planningto leave occupation in the next two years significantly increases the likelihood of beingmoderately risk averse. Future expectations mainly have the same effect. Thus, havingpositive expectations with respect to the future decreases risk aversion. The effect ofpolitical attitudes is the same as in the previous estimation. Being left-wing increasesthe probability of being very risk averse.

Foreigners have a higher probability of being completely risk averse. However, theyare also less likely to be very risk averse (value of 2) and more likely to be risk neu-tral. Again being very risk prone is positively correlated with non-German nationality.The next step is to look more closely at specific nationalities. Thus, both Italians andGreeks have a significant lower risk tolerance in the ordered logit model. In contrastto the previous estimation, the variable for Turks is not significant. According to thegeneralized ordered logit model, the effects are ambiguous for Yugoslavs, Greeks, andItalians. In this model, however, being Turkish is negatively correlated with being veryrisk averse. The next question is then whether religion matters in this context. Theeffect of being Protestant is ambiguous. It decreases the likelihood of having a valueof 1 but increases the likelihood of having a value of 2. In the ordered logit model thedummy for Protestants is not significant. Atheist, on the other hand, tend to be more riskprone. The dummy for Islamic faith is again not significant. In contrast to the previousestimation, religiousness is also of importance in this estimation. Thus, religiousnessis positively correlated with being completely risk averse but negatively correlated withbeing very risk averse (values of 1 or 2). Hence, this estimation supports the theorythat religious people are more risk averse. It is noteworthy that there are hardly anysignificant variables in the generalized ordered logit model for values above 5. This canmainly be explained by the fact that there are simply very few people in the sample whoare moderately or very risk averse in financial matters (see Table 1).

Lastly, we analyse the determinants of risk tolerance in a hypothetical situation (seeTable 6). Participants were asked how much of a lottery win of 100,000 euro theywould invest in a game. The possible answers were the whole amount of 100,000, anamount of 80,000, 60,000, 40,000, 20,000 or nothing. The estimation shows that thecoefficients have the expected signs as in the previously described contexts. However,only being atheist and being Italian matters out of the variables that capture the culturalbackground. While the dummy for Italians has the expected sign, being atheist increasesthe probability of investing a lower amount in assets. This is in contrast to previousestimations in which atheists had a higher likelihood of being risk prone .

Taking all estimations into account, we found that the variables that are routinelyincluded mainly have the expected effect. In addition, this study showed that also cul-tural background is of importance for explaining risk tolerance. Culture was definedin this study by religion and nationality. The overall conclusion is that Protestants and

12

atheists tend to be less risk averse than Muslims and Catholics. Furthermore, the studyalso revealed an effect of nationality. Turks, Italians, and Greeks showed lower risk tol-erance. In addition, Ex-Yugoslavs also tend to be more risk averse in some contexts. Thiscontradicts the hypothesis that migrants are more risk prone. However, this hypothesismight still hold. The main wave of migration into Germany was in the sixties. Thus,Turks, Italians, and Greeks immigrated several decades ago as guest workers. The par-ticipants in this survey are mainly of the second or third generation of migrants. Thispartly explains why those groups no longer have a higher risk tolerance. Furthermore,Austrians did not have in any of these estimations a significantly different risk toleranceto others with the exception of one case in the generalized ordered logit model. This canbe explained by the fact that, to a large extent, Germans and Austrians share the sameculture.

All in all, one has to consider the cultural background if one wants to explain de-terminants of attitudes towards risks. As risk aversion is of major relevance in everydaylife, it is necessary to keep this finding in mind when politicians or academics want toaddress these groups.

5 Conclusions

Risk aversion is an important factor not only in economic situations but also in mostlife decisions. Thus, the aim of this study was to figure out which determinants canexplain different attitudes towards risk. This study showed that women tend to berisk averters. Married individuals are also more likely to be risk averse. Age has theeffect that it decreases risk tolerance. Both income and wealth are positively correlatedwith risk tolerance. Worries about finances tend to increase risk aversion while lifeand income satisfaction increase risk propensity. Unemployed individuals have a higherprobability of being risk averters. Furthermore, people who plan to leave occupation inthe following two years are more risk prone. In addition to these variables, there areseveral factors that matter in specific contexts. Thus, current health status plays a rolein explaining general risk tolerance. Political attitudes are also of relevance. Being lessleft-wing is associated with a higher probability of being risk prone.

The main issue of this study was to figure out possible differences in risk tolerancewith respect to cultural background. The estimations revealed that both religion andnationality matter for risk aversion. Thus, both Protestants and atheists tend to be lessrisk averse than people with other religions or denominations. Furthermore, foreignersare less likely to be risk prone. This also applies for specific nationalities. Turks, Ital-ians, Greeks, and Ex-Yugoslavs tend to be more risk averse than others. Hence, culturalbackground does indeed have some impact on risk taking behaviour. Consequently, oneshould take these results into consideration if one thinks, for instance, about reforms ofthe national social pension programme.

13

References

Adhikari, B. K. (2009), Gender Differences in Risk Aversion: A Study of NepaleseBanking Sector Employees, Working Paper.

Agnew, J., Balduzzi, P. and A. Sunden (2003), Portfolio Choice and Trading in a Large401(k) Plan, The American Economic Review 93(1), 193-215.

Amanatullah, E. T., Shropshire, C., James, E. H. and P. Lee (2010), Risky Business ...For Whom? Gender, Self- vs. Other-Orientation and Risk in ManagerialDecision-Making, Darden Business School Working Paper No. 1633978.

Arano, K., Parker, C. And R. L. Terry (2010), Gender-Based Risk Aversion andRetirement Asset Allocation, Economic Inquiry 48(1), 147-155.

Arrow, K. J. (1970), Essays in the Theory of Risk-Bearing, Amsterdam.

Atkinson, S. M., Baird, S. B. and M. B. Frye (2003), Do Female Mutual Fund ManagersManage Differently?, The Journal of Financial Research 26(1), 1–18.

Bajtelsmit, V. L. and A. Bernasek (1996), Why Do Women Invest Differently ThanMen?, Financial Counseling and Planning 7, 1-10.

Bajtelsmit, V. L., Bernasek, A. and N. A. Jianakoplos (1999), Gender Differences inDefined Contribution Pension Decisions, Financial Services Review 8(1), 1–10.

Barber, B. M. and T. Odean (2001), Boys Will Be Boys: Gender, Overconfidence, andCommon Stock Investment, Quarterly Journal of Economics 116(1), 261–92.

Bartke, S. and R. Schwarze (2008), Risk-Averse by Nation or by Religion? SomeInsights on the Determinants of Individual Risk Attitudes, SOEPpapers No. 131.

Bellante, D. and C. A. Green (2004), Relative Risk Aversion among the Elderly, Reviewof Financial Economics 13(3), 269–281.

Belzil, C. and M. Leonardi (2007), Can Risk Aversion Explain Schooling Attainments?Evidence from Italy, Labour Economics 14(6), 957-970.

Benartzi, S. and R. H. Thaler (1995), Myopic Loss Aversion and the Equity PremiumPuzzle, The Quarterly Journal of Economics 110(1), 73-92.

Benjamin, D. J., Choi, J. J. and A. Strickland (2010), Social Identity and Preferences,The American Economic Review 100(4), 1913–1928.

Benjamin, D. J., Choi, J. J. and G. Fisher (2010), Religious Identity and EconomicBehavior, NBER Working Paper No. 15925.

Berkelaar, A. B., Kouwenberg, R. and T. Post (2004), Optimal Portfolio Choice underLoss Aversion, The Review of Economics and Statistics 86(4), 973-987.

Bertaut, C. C. (1998), Stockholding Behavior of U.S. Households: Evidence from the1983–1989 Survey of Consumer Finances, The Review of Economics and Statistics

14

80(2), 263-275.

Bommier, A. and J.-C. Rochet (2006), Risk Aversion and Planning Horizons, Journal ofthe European Economic Association 4(4), 708-734.

Booth, A. L. and P. Nolen (2012), Gender Differences in Risk Behaviour: Does NurtureMatter?, The Economic Journal 122(558), 56–78.

Borghans, L., Golsteyn, B. H. H., Heckman, J. J. and H. Meijers (2009), GenderDifferences in Risk Aversion and Ambiguity Aversion, Journal of the EuropeanEconomic Association 7(2-3), 649-658.

Brant, R. (1990), Assessing Proportionality in the Proportional Odds Model for OrdinalLogistic Regression, Biometrics 46(4), 1171-1178.

Brooks, P. and H. Zank (2005), Loss Averse Behavior, The Journal of Risk andUncertainty 31(3), 301–325.

Brooks, R., Faff, R. W., Mulino, D. and R. Scheelings (2009), Deal or No Deal, that isthe Question: The Impact of Increasing Stakes and Framing Effects onDecision-Making Under Risk, International Review of Finance 9(1-2), 27-50.

Brown, D. A. (2007), Pensions and Risk Aversion: The Influence or Race, Ethnicity, andClass on Investor Behavior, Lewis and Clark Law Review 11(2), 385-406.

Brunnermeier, M. K. and S. Nagel (2008), Do Wealth Fluctuations GenerateTime-Varying Risk Aversion? Micro-Evidence on Individuals’ Asset Allocation, TheAmerican Economic Review 98(3), 713-736.

Cohen, A. and L. Einav (2007), Estimating Risk Preferences from Deductible Choice,The American Economic Review 97(3), 745-788.

Cohn, R. A., Lewellen, W. G., Lease, R. C. and G. G. Schlarbaum (2012), IndividualInvestor Risk Aversion and Investment Portfolio Composition, The Journal of Finance30(2), 605-620.

Cole, S. A., Paulson, A. and G. K Shastry (2012), Smart Money: The Effect ofEducation, Cognitive Ability, and Financial Literacy on Financial MarketParticipation, Harvard Business School Finance Working Paper No. 09-071.

Coleman, S. (2003), Risk Tolerance and the Investment Behavior of Black and HispanicHeads of Household, Financial Counseling and Planning 14(2), 43-52.

Cramer, J., Hartog, J., Jonker, N. and C. M. van Praag (2001), Low Risk AversionEncourages the Choice for Entrepreneurs: An Empirical Test of Truism, Journal ofEconomic Behavior and Organization 48(1), 29-36.

Croson, R. and U. Gneezy (2009), Gender Differences in Preferences, Journal ofEconomic Literature 47(2), 448–474.

Cummings, L. L., Harnett, D. L. and O. J. Stevens (1971), Risk, Fate, Conciliation andTrust: An International Study of Attitudinal Differences among Executives, The

15

Academy of Management Journal 14(3), 285-304.

Dohmen, T., Falk, A., Huffman, D. and U. Sunde (2010), Are Risk Aversion andImpatience Related to Cognitive Ability?, The American Economic Review 100(3),1238-1260.

Dohmen, T., Falk, A., Huffman, D., Sunde, U., Schupp, J. and G. G. Wagner (2011),Individual Risk Attitudes: Measurement, determinants, and BehavioralConsequences, Journal of the European Economic Association 9(3), 522–550.

Ekelund, J., Johansson, E., Jarvelin, M.-R. and D. Lichtermann (2005),Self-Employment and Risk Aversion – Evidence from Psychological Test Data, LabourEconomics 12(5), 649-659.

Embrey, L. L. and J. J. Fox (1997), Gender Differences In The InvestmentDecision-Making Process, Financial Counseling and Planning 8(2), 33-39.

Fan, J. X. and J. J. Xiao (2006), Cross-Cultural Differences in Risk Tolerance: AComparison between Chinese and Americans, Journal of Personal Finance 5(3),54-75.

Friend, I. and M. E. Blume (1975), The Demand for Risky Assets, The AmericanEconomic Review 65(5), 900-922. Gourinchas, P.-O. and J. A. Parker (2002),Consumption over the Life Cycle, Econometrica 70(1), 47-89.

Grable, J. E. (2000), Financial Risk Tolerance and Additional Factors that Affect RiskTaking in Everyday Money Matters, Journal of Business and Psychology 14(4),625-630.

Graham, J. F., Stendardi, E. J., Myers, J. and M. F. Graham (2002), Gender Differencesin Investment Strategies: an Information Processing Perspective, InternationalJournal of Bank Marketing 20(1), 17-26.

Grazier, S. and P. J. Sloane (2008), Accident Risk, Gender, Family Status andOccupational Choice in the UK, Labour Economics 15(5), 938–957.

Guiso, L. and M. Paiella (2004), The Role of Risk Aversion in Predicting IndividualBehavior, CEPR Discussion Paper 4591.

Gutter, M. S., Fox, J. J. and C. P. Montalto (1999), Racial Differences in InvestorDecision Making, Financial Services Review 8(3), 149–162.

Halek, M. and J. G. Eisenhauer (2001), Demography of Risk Aversion, The Journal ofRisk and Insurance 68(1), 1-24.

Hilary, G. and K. W. Hui (2009), Does Religion Matter in Corporate Decision Making inAmerica?, Journal of Financial Economics 93(3), 455-473.

Hsee, C. K. and E. U. Weber (1999), Cross-National Differences in Risk Preference andLay Predictions, Journal of Behavioral Decision Making 12(2), 165-179.

Jaeger, D. A., Bonin, H., Dohmen, T., Falk, A., Huffman, D. and U. Sunde (2010), Direct

16

Evidence on Risk Attitudes and Migration, The Review of Economics and Statistics92(3), 684-689.

Johnson, D. K. N. and T. R. Gleason (2009), Who REALLY Wants to be a Millionaire?Gender Differences in Game Show Contestant Behavior under Risk, Social ScienceQuarterly 90(2), 243-261.

Johnson, J. E. V. and P. L. Powell (1994), Decision Making, Risk and Gender: AreManagers Different?, British Journal of Management 5(2), 123-138.

Kahneman, D. and A. Tversky (1979), Prospect Theory: An Analysis of Decision underRisk. Econometrica 47(2), 263-291.

Klasing, M. J. (2008), Culturally Risk-Averse? A Model of Economic Growth withEndogenous Culture, University of St. Gallen Department of Economics WorkingPaper Series 2008-23.

Levy, H. (1994), Absolute and Relative Risk Aversion: An Experimental Study, Journalof Risk and Uncertainty 8(3), 289-307.

Miller, A. S. (2000), Going to Hell in Asia: The Relationship between Risk and Religionin a Cross Cultural Setting, Review of Religious Research 42(1), 5-18.

Miller, A. S. and J. P. Hoffmann (1995), Risk and Religion: An Explanation of GenderDifferences in Religiosity, Journal for the Scientific Study of Religion 34(1), 63-75.

Morin, R. A. and A. Fernandez Suarez (1983), Risk Aversion Revisited, The Journal ofFinance 38(4), 1201-1216.

Palsson, A.-M. (1996), Does the Degree of Relative Risk Aversion Vary with HouseholdCharacteristics?, Journal of Economic Psychology 17(6), 771-787.

Papke, L. E. (1998), How Are Participants Investing Their Accounts in ParticipantDirected Individual Account Pension Plans?, The American Economic Review 88(2),212-216.

Pfeier, C. (2011), Risk Aversion and Sorting into the Public Sector, German EconomicReview 12(1), 85-99.

Powell, M. and D. Ansic (1997), Gender Differences in Risk Behaviour in FinancialDecision-Making: An Experimental Analysis, Journal of Economic Psychology 18(6),605-628.

Pratt, J. (1964), Risk Aversion in the Small and in the Large, Econometrica 32(1/2),122-136.

Riley, W. B. and K. V. Chow (1992), Asset Allocation and Individual Risk Aversion,Financial Analysts Journal 48(6), 32-37.

Salacuse, J. W. (1998), Ten Ways that Culture Affects Negotiating Style: Some SurveyResults, Negotiation Journal 14(3), 221–240.

17

Schmidt, U. and S. Traub (2002), An Experimental Test of Loss Aversion, The Journal ofRisk and Uncertainty 25(3), 233–249.

Schooley, D. K. and D. D. Worden (1996), Risk Aversion Measures: ComparingAttitudes and Asset Allocation, Financial Services Review 5(2), 87-99.

Schubert, R. (2006), Analyzing and Managing Risks – on the Importance of GenderDifferences in Risk Attitudes, Managerial Finance 32(9), 706-715.

Schubert, R., Brown, M., Gysler, M. and H. W. Brachinger (1999), FinancialDecision-Making: Are Women Really More Risk-Averse?, The American EconomicReview 89(2), 381-385.

Shane, S. (1992), Uncertainty Avoidance and the Preference for InnovationChampioning Roles, Journal of International Business Studies 26(1), 47-68.

Shaw, K. L. (1996), An Empirical Analysis of Risk Aversion and Income Growth, Journalof Labor Economics 14(4), 626-653.

Siegel, F. W. and J. P. Hoban Jr. (1982), Relative Risk Aversion Revisited, The Review ofEconomics and Statistics 64(3), 481-487.

Skaperdas, S. and L. Gan (1995), Risk Aversion in Contests, The Economic Journal105(431), 951-962.

Spivey, C. (2010), Desperation or Desire? The Role of Risk Aversion in Marriage,Economic Inquiry 48(2), 499-516.

Sunden, A. E. and J. Surette (1998), Gender Differences in the Allocation of Assets inRetirement Savings Plans, The American Economic Review 88(2), 207-211.

Sung, J. and S. Hanna (1996), Factors Related To Risk Tolerance, Financial Counselingand Planning 7, 11-19. Tanaka, T., Camerer, C. F. and Q. Nguyen (2010), Risk andTime Preferences: Linking Experimental and Household Survey Data from Vietnam,The American Economic Review 100(1), 557-571.

Tversky, A. and D. Kahneman (1991), Loss Aversion in Riskless Choice: A ReferenceDependent Model, Quarterly Journal of Economics 106(4), 1039-1061.

Van Praag, C. M. and J. Cramer (2001), The Roots of Entrepreneurship and LabourDemand: Individual Ability and Low Risk Aversion, Economica 68(269), 45-62.

Vereshchagina, G. and H. A. Hopenhayn (2009), Risk Taking by Entrepreneurs,American Economic Review 99(5), 1808-1830.

Wagner, G. G., Frick, J. R. and J. Schupp (2007), The German Socio-Economic PanelStudy (SOEP) – Scope, Evolution and Enhancements, Schmollers Jahrbuch 127(1),139-169.

Wang, H. and S. Hanna (1997), Does Risk Tolerance Decrease with Age?, FinancialCounseling and Planning 8(2), 27-31.

18

Wang, M. and P. S. Fischbeck (2008), Evaluating Lotteries, Risks, and Risk-mitigationPrograms, Journal of Risk Research 11(6), 775–795.

Watson, J. and M. McNaughton (2007), Gender Differences in Risk Aversion andExpected Retirement Benefits, Financial Analysts Journal 63(4), 52-62.

Weber, E. U. and C. Hsee (1998), Cross-Cultural Differences in Risk Perception, butCross-Cultural Similarities in Attitudes towards Perceived Risk, Management Science44(9), 1205-1217.

Weber, E. U., Hsee, C. K. and J. Sokolowska (1998), What Folklore Tells Us about Riskand Risk Taking: Cross-Cultural Comparisons of American, German, and ChineseProverbs, Organizational Behavior and Human Decision Process 75(2), 170–186.

Williams, R. (2006), Generalized Ordered Logit/Partial Proportional Odds Models forOrdinal Dependent Variables, The Stata Journal 6(1), 58–82.

Xu, H. and M. Ruef (2004), The Myth of the Risk-Tolerant Entrepreneur, StrategicOrganization 2(4), 331-355.

19

Appendix

General Willingness to Take Risks

Variable 0 1 2 3 4

Gender 36.9350*** 33.2770*** 52.0342*** 14.8373** -3.0121Married -4.6203 9.434 20.414** 13.5636 -3.871Age 2.2041*** 1.2605*** 1.564*** -0.3527 -0.6916***Salary -0.0087*** -0.0070*** 0.0032 0.0003 0.0045*Wealth -0.0001*** -0.0000 -0.0000 -0.0000 0.0001**Worries aboutFinances

-23.3283*** -3.1296 1.6961 15.1922*** -0.0502

Satisfaction -3.501*** -4.6734*** -1.9014 -3.8164** -1.0392Current Health 1.4393 4.239* 9.8045*** 10.4959*** -6.6072**Unemployed -6.3042 -6.2811 10.6653 31.9062** 19.277*Exit LabourForce

-0.574*** -0.1334 -0.2594 0.2423 -0.1094

Political Atti-tudes

0.3878 -4.7623*** -8.8553*** -6.0145*** 0.413

Foreigner 7.7774*** -0.0266 4.6887 -5.0903 -2.5468Protestant -7.9086* -8.6438** -7.5603 3.7268 2.6986Atheist -21.1461*** -17.5821*** -14.3073** -0.814 4.8762

Table 3: Determinants of the General Willingness to Take Risks 1

20

General Willingness to Take Risks

Variable 5 6 7 8 9 10

Gender -28.4391*** -38.5048*** -42.9307*** -19.1889*** -4.3602*** -0.5574Married -1.9967 -13.1010** -9.0509* -7.2026** -0.9127 -1.6939Age -1.9286*** -1.0498*** -0.9758*** -0.0916 -0.0269 0.0441Salary -0.0046** 0.0068*** 0.0035*** 0.0025*** 0.0023 0.0003Wealth 0.0001** -0.0000 -0.0000* 0.0000 0.0000** -0.0000*WorriesaboutFinances

10.8317 0.3855 2.8142 -1.3937 -1.9535*** 0.1238

Satisfaction -1.6119 3.5028*** 8.0611*** 4.6461*** 0.8019*** -0.0781CurrentHealth

-0.6351 -9.7644*** -3.5934* -3.7313*** 0.2872 -1.4274

Unemployed -31.082** 4.9782 4.238 -22.0752*** -2.0623 -3.5614***Exit LabourForce

0.0803 0.5313*** 0.1219 0.0586 0.026 0.0161

Political At-titudes

5.9928*** 6.2995*** 4.2501*** 1.6779*** 0.5432*** 0.3235**

Foreigner -3.5642 -2.2684 -0.2521 0.2123 0.8723* 0.6629***Protestant 6.4783 6.6081 3.8210 1.7771 -0.2776 -0.1724Atheist 32.1127*** 1.967 13.7502*** 1.4777 0.0815 -0.2369

Number of Observations: 17,230 LR chi2(154): 2840.38 Pseudo R2: 0.0384

Table 3: Determinants of the General Willingness to Take Risks 1 (continued)Note: The table shows estimation results of a generalized ordered logit model. Thevalues show the average marginal effects of the respective coefficients. All values aredivided by 1000. The asterisks indicate whether a coefficient is significantly differentfrom zero at the 10 % (one asterisk), 5 % (two asterisks) or 1 % (three asterisks)significance level.

21

General Willingness to Take Risks

Variable 0 1 2 3 4

Gender 36.6287*** 33.6265*** 51.7889*** 14.1364** -3.0360Married -4.4862 9.8416 20.7106** 14.8086 -3.1293Children -14.2300 -6.5482 -14.9890* -11.1870 -20.1939**Age 2.2694*** 1.3000*** 1.6338*** -0.3279 -0.6137***Salary -0.0000 -0.0068*** 0.0034 0.0003 0.0048*Wealth -0.0001 -0.0000 -0.0000 -0.0000 0.0001**Worries aboutFinances

-23.3274 -2.4813 2.2775 15.0574*** -0.0065

Satisfaction -3.5289 -4.6500*** -1.857 -3.7077** -0.9560Current Health 1.1397 4.3266** 9.6468*** 10.9088*** -6.6899**Unemployed -5.6465 -6.8454 10.6172 31.1442** 20.5811*Exit LabourForce

-0.5444 -0.1296 -0.2578 0.2181 -0.1127

Political Atti-tudes

0.4020 -4.7959*** -8.8499*** -6.1883*** 0.3111

Turkish 46.8124 19.6591 33.2373 -42.5201 -19.667Italian 12.818 7.5954 24.0723 -62.0352** -9.0327Yugoslav 49.0602 -39.6255 -31.6589 14.082 -17.9200Greek 54.0416 -50.3868** 65.7576 -5.2246 30.0135Austrian 0.2487 17.1885 -25.3033 84.2712 -103.658***Protestant -7.5149 -7.8281* -7.2088 002.0007 004.1206Atheist -20.1409*** -16.5295*** -13.4319* -2.3733 6.4332Muslim -0.8784 3.1925 -7.1698 -7.5583 10.7915

Table 4: Determinants of the General Willingness to Take Risks 2

22

General Willingness to Take Risks

Variable 5 6 7 8 9 10

Gender -28.5884*** -38.0846*** -42.6988*** -19.2099 -3.9143 -0.6357Married -2.7714 -13.8105** -9.6739* -8.0364 -1.2605 -1.2699Children 21.7013** 22.6407*** 9.9683 6.5423 2.8605 2.2598Age -2.0227*** -1.1539*** -1.0119*** -0.1079 -0.0218 0.0228Salary -0.0052** 0.0064*** 0.0033*** 0.0026 0.0002 0.0001Wealth 0.0001** -0.0000 -0.0000 0.0000 0.0000 -0.0000WorriesaboutFinances

10.0187** 0.0803 2.7537 -1.6345 -1.7658 0.0228

Satisfaction -1.6918 3.6056*** 7.9951*** 4.4944 0.6883 -0.0708CurrentHealth

-0.9067 -9.3678*** -3.6084* -3.9882 0.2037 -1.2147

Unemployed -31.121** 5.0149 3.5613 -22.5007 -1.9313 -3.0346Exit LabourForce

0.0896 0.5227*** 0.1366 0.0498 0.0213 0.0096

Political At-titudes

6.0454*** 6.4985*** 4.2536*** 1.776 0.4625 0.2705

Turkish -71.3463** 52.7181** -13.3866 -6.3593 -998.9183 996.9604Italian 29.3666 -18.9234 -9.8398 17.2643 6.6709 3.6914Yugoslav 9.5991 6.4817 -51.9039* 31.7527 15.4225 16.2079Greek -4.8824 -68.8976*** -3.9050 -12.3178 -6.5567 0 (omitted)Austrian -12.6057 33.059 12.5105 2.1035 -6.3298 0 (omitted)Protestant 6.4494 5.4677 2.7409 2.7421 -0.2273 -0.3023Atheist 31.8079*** -0.0833 12.5429*** 2.3083 0.0675 -0.4474Muslim 15.4903 -30.8557** 3.9797 8.7611 8.3408 -2.9248

Number of Observations: 17,230 LR chi2(154): 2942.38 Pseudo R2: 0.0398

Table 4: Determinants of the General Willingness to Take Risks 2 (continued)Note: The table shows estimation results of a generalized ordered logit model. Thevalues show the average marginal effects of the respective coefficients. All values aredivided by 1000. The asterisks indicate whether a coefficient is significantly differentfrom zero at the 10 % (one asterisk), 5 % (two asterisks) or 1 % (three asterisks)significance level.

23

Willingness to Take Financial Risks

Variable 0 1 2 3 4

Gender 111.5097*** 23.8152*** -0.833 -29.8376*** -24.1800***Married -20.7281 30.1510*** 2.0183 -3.4629 -2.3147Children -47.9488*** -0.842 3.2758 15.6452* 2.8639Age 4.9420*** -0.2238 -1.6235*** -1.1295*** -0.5431***Uni -57.7216** 5.1841 10.6571 2.942 0.0188Salary -0.0271 -0.0026 0.0112*** 0.0070*** 0.0030***Wealth -0.0008*** -0.0000 0.00038*** 0.0002*** 0.0002***Worries aboutFinances

-41.8039 2.1031 17.0057*** 12.5989*** 3.3596

Satisfaction -1.7245 -1.2014 -3.2173* 1.5013 0.9916Income satis-faction

-2.9627 -1.6674 0.294 0.8389 1.2499

Unemployed 59.5558*** -27.1381** -14.0045 -20.6924* -3.2836Exit LabourForce

-1.5129 -0.0623 0.1699 0.4841*** 0.2591***

Future -25.2404*** -4.0622 -0.9797 7.8503** 7.1771***Political Atti-tudes

-6.2672 -6.0187*** -0.9642 2.4691* 1.5106

Turkish 105.8463** -44.2596 -65.5218*** -44.3263 14.1316Italian 127.077** -51.3643** -28.0912 -36.3877 -33.6393Yugoslav 138.593 -75.2398 -54.8582 -87.9557** 23.4537Greek 83.3916 84.003 -41.7118 -83.3443*** -8.0792Austrian 73.3649 -33.2721 -11.7658 1.4212 20.3459Protestant -5.0535 -13.6563* 15.2233** 6.6747 3.0637Atheist -15.477 -7.3105 14.5607* 0.4157 0.5568Muslim 11.3717 -20.0713 -37.1326 15.0113 4.2098Service -20.2613** 6.5136* 9.8484*** 1.9988 2.482

Table 5: Determinants of the Willingness to Take Risks in Financial Matters

24

Willingness to Take Financial Risks

Variable 5 6 7 8 9 10

Gender -33.9031 -18.5817 -19.1916 -5.7968 -0.5032 -3.2654Married 3.3059 -2.1737 -4.5396 2.547 -0.0612 -1.6303Children 21.2823 6.0454 -1.3377 -0.7164 1.3409 -0.0965Age -1.0767 -0.3032 -0.2178 0.0175 -0.0291 0.1176Uni 25.3064 7.1121 3.2447 5.6183 -2.7239 -0.0209Salary 0.0060 0.0026 0.0011 0.0020 -0.0000 0.0002Wealth 0.0001 0.0000 0.0000 0.0000 0.0000 -0.0000WorriesaboutFinances

7.9094** -0.193 4.3347 -2.1476 -0.8285 -0.7836

Satisfaction 1.4585 1.1296 1.1353 0.431 -0.3036 0.2575Income sat-isfaction

0.8567 0.6359 0.5174 0.5805 0.1946 -0.3206

Unemployed -4.9445 -5.1374 2.1096 7.327 1.8993 4.0863Exit LabourForce

0.03452 0.1669 0.0708 0.097 0.033 0.0108

Future 6.0501 3.8825 0.3435 3.5843 1.027 0.6755Political At-titudes

5.7285 1.9644 1.6405 0.1809 0.2949 0.1272

Turkish 4.4613 8.2095 -2.3868 20.628 4.1591 0.1941Italian -14.9026 16.6177 -6.7851 9.1431 12.5342 3.1859Yugoslav 58.2505 -8.6622 28.9549 -1012.931** 997.0332*** -2.2742***Greek -13.8748 10.9255 -1017.394*** 989.8678*** -5.7869 0 (omitted)Austrian 6.9696 -1022.816*** 595.4719 380.2476 -005.6582 0 (omitted)Protestant -5.5498 0.8331 1.9709 -1.571 0.1817 -1.7545Atheist 4.6851 0.6533 4.7131 -0.9218 -0.0203 0.281Muslim 8.8984 -7.4781 7.9743 0.9953 5.1834 2.1687Service -0.4232 0.2161 2.627 -1.7854 -0.3158 -0.4216

Number of Observations: 17,230 LR chi2(154): 2619.40 Pseudo R2: 0.0408

Table 5: Determinants of the Willingness to Take Risks in Financial Matters (continued)Note: The table shows estimation results of a generalized ordered logit model. Thevalues show the average marginal effects of the respective coefficients. All values aredivided by 1000. The asterisks indicate whether a coefficient is significantly differentfrom zero at the 10 % (one asterisk), 5 % (two asterisks) or 1 % (three asterisks)significance level.

25

Variable Willingness to Take Risks AfterWinning the Lottery

Uni -0.4688*** -0.4689***Age 0.0185*** 0.0185***Married 0.1377*** 0.1386***Salary -0.0002*** -0.0002***Wealth 0.0000 0.0000Worries about Finances -0.1182*** -0.1187***Gender 0.3510*** 0.3504***Children -0.1175** -0.1176**Exit Labour Force -0.0025** -0.0025**Satisfaction -0.0529*** -0.0528***Protestant 0.0696 0.0676Atheist 0.1953*** 0.1930***Muslim -0.0498Political Attitudes -0.0022 -0.0023Height -0.0018 -0.0018Turkish -0.1691 -0.1272Italian 0.3848* 0.3825*Yugoslav 0.3705 0.3881Greek 0.3658 0.3681Austrian -0.3384 -0.3390

Number of oberservations 16,990 16,990Pseudo R2 0.0356 0.0356

Table 6: Determinants Willingness to Take Risks After Winning the LotteryNote: The table shows estimation results of an ordinal logit model. Values show theaverage marginal effects of the respective coefficients. The asterisks indicate whethera coefficient is significantly different from zero at the 10 % (one asterisk), 5 % (twoasterisks) or 1 % (three asterisks) significance level.

26