Gender differences in entrepreneurial choice and risk aversion – a decomposition based on a microeconometric model

Abstract

Why are female entrepreneurs so rare? In Germany, women exhibit both a lower entry rate into and higher exit rate from self-employment. To explain this gender gap, this study estimates a structural microeconometric model of transition rates that includes a standard risk aversion parameter. Inputs into the model are the expected value and variance of earnings from self-employment and dependent employment, estimated separately by gender and accounting for nonrandom selection into self-employment. The gender differential in the transition rates is decomposed using a novel extension of the Blinder–Oaxaca technique for nonlinear models. Women's higher estimated risk aversion explains the largest part of their higher exit rate but only a small portion of their lower entry rate.

Keywords:

• entrepreneurship
• self-employment, risk aversion
• gender differential
• nonlinear Blinder–Oaxaca decomposition

JEL Classification::

• J23
• J16
• D81

Acknowledgements

I thank the anonymous referees, Viktor Steiner, and participants at various seminars for their helpful comments and suggestions. Financial support from the German Research Foundation (DFG) for the project ‘Tax Policy and Entrepreneurial Choice’ (STE 681/7-1) is gratefully acknowledged.

Notes

¹ Only in Mexico and Turkey are women's self-employment shares higher than men's, which may reflect the high number of small agricultural establishments in these countries. Blanchflower (2000) offers more detail about self-employment in OECD countries, including the lower self-employment rate among women.

² Credit constraints for entrepreneurs have been widely discussed (e.g. Hurst and Lusardi, 2004; Disney and Gathergood, 2009). It is plausible that they may be more severe for female entrepreneurs. Borjas and Bronars (1989) discuss consumer discrimination in the context of self-employed African–Americans.

³ Wagner (2007) also analyses gender differences among nascent entrepreneurs using a matching approach and controls for mentions of ‘fear of failure’ as a reason not to start a business. Though an interesting variable, fear of failure differs from risk aversion, in that it depends on the person's expected success probability.

⁴ A related stream of literature has analysed earnings differentials between self- and dependent employment, without considering differences in the variance of earnings. For example, Taylor (1996) and Fraser and Greene (2006) confirm that higher expected earnings from self-employment relative to paid employment significantly increase the probability of becoming self-employed; Hammarstedt (2006) establishes the same result for Swedish immigrants; and Rees and Shah (1986) and Dolton and Makepeace (1990) find a positive but insignificant effect. Hamilton (2000) instead concludes that factors other than earnings induce people to become self-employed.

⁵ For more details about this model, see Fossen (2009a). Fossen (2009b) also uses a similar model to study the effects of progressive income taxation on self-employment.

⁶ Similar analyses of entrepreneurial exit using hazard rate models appear in Evans and Leighton (1989), Audretsch and Mahmood (1995), Taylor (1999) and Haapanen and Tervo (2009), as well as Falck (2007), who uses German establishment data.

⁷ Alternatively one could assume Constant Absolute Risk Aversion (CARA). The advantage of the CARA utility is that a closed-form representation of expected utility exists if y is normally distributed, and no Taylor approximation is needed. However, prior research prefers CRRA as the more realistic specification (in the context of entrepreneurship, see Kanbur (1982), Rees and Shah (1986) and Pfeiffer and Pohlmeier (1992)).

⁸ The only difference is that the coefficient α of the risk-adjusted income differential (i.e. difference between self-employment and dependent employment in both models) is expected to be negative in the exit model. In the estimation of the parameters, α is left unconstrained, so a check to determine if α has the expected sign in all models serves as a test of the models’ consistency.

⁹ The 2005 wave is used to obtain retrospective income information for 2004 only.

¹⁰ The coefficient of the selectivity term λ is negative in all earnings regressions, which indicates that the error terms in the selection equation and the earnings equation are negatively correlated. The coefficient is significant in the models of dependent employment only. Insignificant and sometimes negative selection terms in regressions of earnings from self-employment are common (e.g. Rees and Shah, 1986; Borjas and Bronars, 1989; Evans and Leighton, 1989; Dolton and Makepeace, 1990), which suggests that there is no significant selection on unobservables; Taylor (1996), in contrast, reports positive and significant selection effects.

¹¹ In the earnings variance regression, the explanatory variables are jointly significant for both employment states and genders, which confirms that earnings are heteroscedastic (Breusch–Pagan test). This result shows that the variance of earnings not only differs between dependent and self-employment and by gender, but also between individual workers, depending on their characteristics and covariates.

¹² The results from the tax rate regression show that the individual average tax rate increases with gross income at diminishing rates, which reflects the progressive income tax code in Germany.

¹³ The full estimation results are available from the author upon request.