The gender effect in risky asset holdings

https://doi.org/10.1016/j.jebo.2011.06.011Get rights and content

Abstract

We study the relation between gender and stock holdings in Finland, a particularly gender equal country. We show that it is crucial to control for risk aversion using a measure of subjective risk-taking, rather than measures derived from abstract gambles. Controls related to financial knowledge and resources also explain the gender difference. The residual effect of the male gender on the conditional equity share, after all appropriate controls, is 3 percentage points and statistically significant. The effect on stock market participation on the other hand is close to zero or negative, so men contribute more to the nonparticipation puzzle conditional on covariates. The gender difference mainly works through women's higher risk aversion, which we find extends to finance professionals and wealthy private banking customers.

Highlights

► We study the relation between gender and portfolio holdings. ► We compare several measures of risk attitude using respondents that are familiar with financial risk taking and show that gender is still a strong predictor of risk taking. ► The gender effect on conditional risky share is about 3 percentage points. ► The gender effect on the stock market participation is close to zero or negative.

Introduction

Several studies find that women are less likely to participate in the stock market, and conditional on participation, take less risk (see e.g., Sunden and Surette, 1998, Barber and Odean, 2001, Dwyer et al., 2002, Agnew et al., 2003). Compared to men, women are more risk-averse (e.g., Barsky et al., 1997; for reviews see Eckel and Grossman, 2008, Croson and Gneezy, 2009) and score lower in financial literacy tests (Chen and Volpe, 2002, Lusardi and Mitchell, 2008). Gender differences in portfolio holdings could thus be artifacts of imperfect controls for risk aversion or financial knowledge, and may vary between samples of different financial sophistication. Appropriate control variables would then eliminate the gender effect. On the other hand, it is possible that some gender difference remains, perhaps due to cultural or social norms.

In this paper we study the relation between gender and portfolio holdings in Finland, a country which ranks number 3 in the world in overall gender equality1 and number 2 in economic literacy in a sample of 55 countries studied by Jappelli (2010). To the extent that a possible cultural gender effect depends on the level of gender equality in the society, we should be able to establish a lower limit for it in Finland, given appropriate data. This idea is motivated by the observation that the gender gap in mathematical test scores in school disappears in countries with a more gender-equal culture (Guiso et al., 2008).

To begin answering this question, we first validate our method of eliciting risk-preference utilizing investment seminars in different populations that are familiar with financial risk-taking. We collect survey data on wealthy private banking investors (N = 177), investment advisers and managers (N = 81), as well as finance students (N = 77). With a 97 percent response rate, the sample is practically free of nonresponse bias. We measure risk-taking on an eleven-point scale first in general matters, and then in various domains, including financial matters. We also use more traditional methods, such as inputting risk aversion from a certainty equivalent to a hypothetical lottery, which have been widely used in the economics literature (Barsky et al., 1997, Donkers et al., 2001, Guiso et al., 2002, Guiso and Paiella, 2006, Guiso and Paiella, 2008, Nosic and Weber, 2010).

Our self-reported risk measures are based on the study by Dohmen et al. (2011b) where the authors analyze a large panel survey data and also validate the self-reported risk measures by conducting a smaller sample field experiment. They show that the general risk question more accurately predicts behavior in many contexts, for example portfolio choice, compared to the standard lottery measure. The same risk attitude measure is thereafter used in several studies (for example, Bonin et al., 2007, Dohmen et al., 2010, Dohmen and Falk, 2011, Dohmen et al., 2011a).

We find that women are significantly less willing to take risk than men in almost all domains of risk also in populations familiar with financial risks. Factor analysis of risk-taking in the different domains suggests the existence of two distinct factors in the tendency to take risk: a ‘cool factor’ comprising general and financial risk attitudes, and a hot factor comprising risk-taking in areas such as health, car driving, and sports. This result indicates that people who are more familiar with financial risks connect general risk taking especially with financial risk taking, and also separate financial risk taking from risk taking in other domains of life. Analyzing the relation between various risk measures and the subjects’ stock holdings, we find that the general and financial risk attitudes are very strong and robust predictors of portfolio choice. In contrast, methods traditionally used by economists, such as the certainty equivalent of a lottery, do a poor job. In particular, they have no incremental explanatory power on actual financial risk-taking when risk attitude is controlled for.

Having validated the performance of a self-reported financial risk attitude as a predictor of risky asset holdings, we then turn to data on retail bank clients’ risk profiling reports. These reports are the result of financial advisers’ client meetings in a large Finnish retail bank. The data offers several advantages. First, having 85,000 observations helps obtain precise estimates. Second, the data is more reliable than surveys in general as the data on investments comes directly from the bank's information system. Third, the customers have their own money at stake, and thus have incentives to carefully think about their risk attitude. The financial advisers have also been trained to emphasize the importance of correctly classifying their customers’ risk attitude.

The retail bank risk profiling data also shows a robust effect of male gender on financial risk attitude. The effect size decreases by about 30 percent when investment knowledge is controlled for. Education increases the willingness to take risk, but does not affect the magnitude of the gender difference by much. After all controls, the gender effect on risk taking is of similar magnitude as having a college education.

After controlling for risk attitude, financial knowledge, wealth, and other relevant factors, the coefficient on a male dummy in a regression explaining risky asset holdings brings out the residual gender effect. The results show that conditional on participation, men invest 2.8 percentage points more of their wealth in the stock market, and this difference is statistically highly significant. The difference in these conditional risky shares is 5.4 percentage points in the raw data, so control variables capture about half of the original gender effect. The magnitude of the residual gender effect corresponds to an increase in the willingness to take risk of about 0.3 points on a 5-point financial risk attitude scale. Investment knowledge provides another point of comparison: the impact of male gender is about the same as that of a dummy variable indicating some investment experience compared to the omitted group of no experience. The gender residual is not reduced if we limit the sample to highly experienced or highly educated investors.

It thus appears that even in a gender-equal country such as Finland, some measure of a gender effect on the risky share remains. We find that the residual gender effect in risky share increases with age2 while the difference in risk attitudes stays constant through the life cycle. Residual gender differences may be innate, or due to culture or social norms, as other effects should operate through the variables that we control for.

A different pattern nevertheless emerges when we look at the dichotomous stock market participation decision. The gender effect becomes insignificant when we control for risk attitude, financial knowledge and education. Furthermore, when we add income and wealth controls, the effect becomes significantly negative. It thus appears that the gender difference in stock market participation is completely explained by men having more favorable covariates on average, that is, higher wealth, lower risk aversion, and more investment knowledge.

In addition to the issue of gender and portfolio choice this paper also contributes to the literature on measuring risk aversion. Previous studies have used student subjects (e.g., Nosic and Weber, 2010), or respondents have been from the general population (e.g., Donkers et al., 2001, Dohmen et al., 2011b), mostly without any special familiarity with financial risks. In contrast to these studies, our subjects have considerable experience with financial risks. A recent experimental study by Gong et al. (2010) also uses real investors as subjects.

The rest of the paper is organized as follows. Section 2 describes the data. Section 3 presents results on risk attitude in different domains and studies gender differences. Section 4 investigates how the various risk measures are related to each other, as well as to risky asset holdings, and Section 5 estimates conditional gender differences in the best-performing risk measures as well as risky asset holdings. Section 6 concludes.

Section snippets

Survey data

We collected information on risk attitudes from 81 investment advisers and managers, 77 finance students, and 177 private banking customers (from here on referred to as investors). Investment adviser and manager responses were collected during three different investment seminars in the fall of 2007. Investors filled the questionnaire during an investment meeting organized by a private banking unit of a commercial bank in November 2007. In both cases the subjects arrived on the scene without

Basic results

The survey respondents rated their willingness to take risk, first in general, and then in six specific domains on an 11-point scale, from 0 (risk averse) to 10 (completely willing to take risk). On average, men rated their general willingness to take risk at significantly higher level than women (6.25 vs. 4.91; chi-square value 33.5, p-value < 0.01). The modal ratings are 7 for the men and 5 for women.3

Validating the risk measures

In this section we investigate the explanatory power of the various risk measures on actual portfolio choice. We use two outcome variables: a binary variable indicating stock market participation and a continuous variable indicating the percentage of wealth allocated in stocks, conditional on investing in stocks (risky share). We have survey data on these outcome variables from wealthy private banking clients (henceforth called investors) and finance students.

Stock market participation rates in

The gender effect in portfolio choice

The previous section shows that a simple financial risk attitude question dominates traditional risk measures based on abstract gambles. This gives us confidence for bringing in new data on retail bank clients’ financial risk attitude and portfolio choice for analysis. In addition to a large sample size (85,000 profiles, over 60,000 without any missing items), this data has other benefits as well. It is very accurate for most items. The data on investments comes directly from the bank's

Conclusions

In this paper we compare several measures of risk attitude using respondents that are familiar with financial risks. We find that a self-reported financial risk attitude (measured on an 11-point scale) is the strongest predictor of the proportion of wealth invested in stocks. Measures traditionally favored by economists, based on certainty equivalent or allocation to hypothetical investments, are dwarfed by the attitude variable. Factor analysis of risk attitudes in different domains suggests

References (34)

  • X. Caldú et al.

    Hormonal and genetic influences on processing reward and social information

    Annals of the New York Academy of Sciences

    (2007)
  • H. Chen et al.

    Gender differences in personal financial literacy among college students

    Financial Services Review

    (2002)
  • J.F. Cocco et al.

    Consumption and portfolio choice over the life cycle

    Review of Financial Studies

    (2005)
  • G.M. Constantinides et al.

    Junior can’t borrow: a new perspective on the equity premium puzzle

    Quarterly Journal of Economics

    (2002)
  • R. Croson et al.

    Gender differences in preferences

    Journal of Economic Literature

    (2009)
  • S. Curcuru et al.

    Heterogeneity and portfolio choice: theory and evidence

  • T. Dohmen et al.

    Performance pay and multidimensional sorting: productivity, preferences, and gender

    American Economic Review

    (2011)
  • Cited by (105)

    • Gender differences and measurement error in financial literacy

      2024, Journal of Behavioral and Experimental Finance
    • Beauty and stock market participation

      2023, Journal of Banking and Finance
    • What drives stock market participation? The role of institutional, traditional, and behavioral factors

      2023, Journal of Banking and Finance
      Citation Excerpt :

      Men are more likely to invest in stocks and the gender difference is larger in single households (Jianakoplos and Bernasek, 1998; Sundén and Surette, 1998; Barber and Odean, 2001). However, Halko et al. (2012) show that this effect vanishes after controlling for other important determinants in Finland, a country often considered the most gender equal in the world. Social interaction affects households’ investment decisions.

    • Broadband internet and stock market participation

      2023, International Review of Financial Analysis
    View all citing articles on Scopus

    We thank two anonymous referees, the editor (Rachel Croson), Markku Lanne, Elias Rantapuska, as well as the participants in the MOVE (Markets, Organizations and Votes in Economics) Workshop on Gender Differences in Competitiveness and Risk Taking in Barcelona, 2010, for comments, and Wei Li and Niklas Jahnsson for research assistance. This research was supported by the Academy of Finland, aivoAALTO project, OP Bank Group Research Foundation, and the Finnish Foundation for Advancement of Securities Markets.

    View full text