Explaining female and male entrepreneurship at the country level

Abstract

Using Global Entrepreneurship Monitor data for 29 countries this study investigates the (differential) impact of several factors on female and male entrepreneurship at the country level. These factors are derived from three streams of literature, including that on entrepreneurship in general, on female labour force participation and on female entrepreneurship. The paper deals with the methodological aspects of investigating (female) entrepreneurship by distinguishing between two measures of female entrepreneurship: the number of female entrepreneurs and the share of women in the total number of entrepreneurs. The first measure is used to investigate whether variables have an impact on entrepreneurship in general (influencing both the number of female and male entrepreneurs). The second measure is used to investigate whether factors have a differential relative impact on female and male entrepreneurship, i.e. whether they influence the diversity or gender composition of entrepreneurship. Findings indicate that – by and large – female and male entrepreneurial activity rates are influenced by the same factors and in the same direction. However, for some factors (e.g. unemployment, life satisfaction) we find a differential impact on female and male entrepreneurship. The present study also shows that the factors influencing the number of female entrepreneurs may be different from those influencing the share of female entrepreneurs. In this light it is important that governments are aware of what they want to accomplish (i.e. do they want to stimulate the number of female entrepreneurs or the gender composition of entrepreneurship) to be able to select appropriate policy measures.

Keywords:

• entrepreneurship
• gender
• determinants of entrepreneurship

Acknowledgement

The authors would like to thank Nancy Carter, Sander Wennekers and two anonymous referees for comments on an earlier version of the paper. Ingrid Verheul acknowledges financial support by the Fund Schiedam Vlaardingen e.o. and the Trust Fund Rotterdam. Early versions of the present paper have been read at the Babson Kauffman Entrepreneurship Research Conference at Babson College, Boston, 6–8 June 2003, at the RENT conference in Lodz, Poland, 20–21 November 2003, and at the First GEM Research Conference, Berlin, 1–3 April 2004.

Notes

Notes

1. See also Grilo and Thurik (2005a) for data of the European member states.

2. In 2002, there were 37 countries participating in GEM. For eight of these countries there was no information available for several of the explanatory variables that we use in this study. Therefore, the analysis in the current paper is restricted to 29 countries.

3. Indeed, for the 29 countries the Spearman rank correlation coefficient between the female and total entrepreneurial activity rate is 0.96, which is significant at the 0.01 level.

4. Although the Spearman rank correlation coefficient between the female entrepreneurial activity rate and the female share in entrepreneurship is significant, its value is only 0.53, confirming that the two concepts are indeed different.

5. A factor that has a positive impact on the absolute number of female entrepreneurs may at the same time have a negative impact on the female share in total entrepreneurship if its influence on the number of male entrepreneurs is relatively larger than that on the number of female entrepreneurs.

6. We use 2002 as in this year the number of countries participating in GEM was higher than in the more recent years 2003 and 2004.

7. See Verheul (2005), Blanchflower (2004) and Parker (2004) for surveys.

8. In this paper technological development is operationalized as R&D investments per capita.

9. This implies that we expect a negative sign for the linear income variable and a positive sign for the squared income variable.

10. Here it is proposed that there is no differential effect of income level on female and male entrepreneurship.

11. Note that this is a reversed causality effect as it refers to unemployment as an effect of entrepreneurship and not as a cause of entrepreneurship. For this study we do not have time series data at our disposal and therefore we are not able to test for reversed causality effects in our empirical analysis.

12. Kovalainen et al. (2002) find a negative association between female unemployment and business start-ups by women.

13. On the other hand, as women already occupy more than one-half of the employment in services, and men increasingly enter service jobs, the differential effect of growth in the number of service jobs on female and male entrepreneurship may be diminishing.

14. As, in principle, GEM measures entrepreneurial activity in the formal sector, it may be argued that the size of the informal sector negatively impacts entrepreneurial activity in the formal sector.

15. In the present paper we use the share of women in the labour force as an indicator of female labour force participation.

16. Note that the entrepreneurial activity rate of GEM is scaled on population and not on labour force (or total employment).

17. Grilo and Thurik (2005b) report on the differences of the entrepreneurial engagement levels between old and new member countries of the European Union.

18. See also OECD (2001) and Breedveld (2000).

19. For a discussion of these other demographic factors, we refer to Verheul (2005). It should be noted here that relatively few studies have been able to systematically link demographic factors to business start-ups at the macro-level Delmar and Davidsson (2000).

20. The possible loss of entitlements to social security upon becoming self-employed may constrain entrepreneurial activity (Henrekson and Johansson 1999).

21. Entrepreneurial aspects may include business qualities (e.g. management, financing, marketing knowledge) as well as more inherent entrepreneurial qualities (e.g. creativity, independence, perseverance). The latter qualities should be introduced and developed in an early phase of education (Van der Kuip and Verheul 2004).

22. We choose to formulate a hypothesis on entry regulation as this is likely to have an important impact on start-up and new venture activity (as measured by GEM).

23. Other studies do not find significant gender differences (Buttner and Rosen 1989, Riding and Swift 1990).

24. We focus on informal venture capital instead of formal venture capital as the bulk of the entrepreneurs measured by GEM run small businesses. Formal venture capital is often acquired by larger businesses.

25. Kovalainen et al. (2002) also find a negative relationship between the statutory support payment scheme as a percentage of wages and the new business rate for women.

26. Several studies have focused upon explaining entrepreneurship from a cultural perspective (McGrath and MacMillan 1992, McGrath et al. 1992, Shane 1992, 1993, Davidsson 1995, Busenitz et al. 2000, Mueller and Thomas 2000, Uhlaner et al. 2002, Hofstede et al. 2004, Noorderhaven et al. 2004).

27. Life satisfaction may be more likely to be related to job satisfaction for men than for women, in particular since employment often absorbs more time in the lives of men.

28. Based on the empirical evidence provided by Noorderhaven et al. (2004) we choose our hypothesis to be in line with the social legitimation perspective.

29. The values for the informal sector variable range from 3.8 to 4.8 for the four countries and this corresponds with an estimated size of the informal sector of approximately 20 to 35% of the economy (table 3). Note that ‘informal’ is not the same as ‘illegal’.

30. Note that the correlation between female share in entrepreneurial activity and size of the informal sector is positive and significant (r = 0.54, p < 0.01). See table 4. Remarkably, Chile scores low on both variables, contributing to the positive relationship. Chile combines a score of only 1.7 on the informal sector index, with a low share of women in total entrepreneurship (30.3%). See table 2. Apparently, informal entrepreneurship by women occurs less often in Chile as compared to other Latin American countries such as Argentina and Brazil.

31. Note that the negative Hypothesis 5 relates to official or formal entrepreneurial activity.

32. The questions asked in the GEM Adult Population Survey do not necessarily exclude owner-managers of unofficial businesses. In particular, respondents who indicate that they ‘sell any goods or services to others’ are included in the TEA index. The fact that the Adult Population Survey is a survey among randomly selected adults does also not give reason to assume that unofficial entrepreneurs are excluded from the TEA count.

33. Per capita income and squared per capita income are counted separately.

34. During the general-to-specific modelling procedure we applied two-tailed tests because removal of variables with high t-values but with a sign opposite to the predicted sign would give biased results. However, note that the effects of the selected variables in table 6 are interpreted in terms of one-tailed tests.

35. Note that not all of these variables are marked as significant in table 5 because we did one-sided tests. However, a high t-value does seem to indicate that there is an effect, even if the effect is not consistent with the sign of the effect in the corresponding hypothesis.

36. Wennekers et al. (2005) provide empirical support for a U-shaped relation between the ‘innovative capacity index’ (see Global Competitiveness Report) and the nascent entrepreneurship rate of the Global Entrepreneurship Monitor. The downward part of the curve reflects the Schumpeter II regime (creative accumulation), where the innovative advantage lies with large, established firms. The upward part of the curve reflects the Schumpeter I regime (creative destruction), where the technological regime is more favourable to innovative entry.

37. Although life satisfaction has a significant impact on female entrepreneurial activity in table 6. Note however that the sign is opposite to what was predicted in Hypothesis 12.

38. Again here we select variables based on two-tailed tests. Selection based on one-tailed tests would create a bias if we were to exclude variables with high t-values but with an unpredicted sign. See also note 34.

39. Reversed causality (i.e. a positive effect of self-employment on satisfaction of women) is not an issue here as entrepreneurial activity refers to only a part of the population (20% at most, see table 1), while the life satisfaction variable is an average country score (see table 3) referring to the whole population. Even if (female) entrepreneurs report to be more satisfied with their life as compared to (female) wage earners, it is unlikely that this has a large impact on the life satisfaction variable as this refers to the whole population of a country.

40. For women this was partly due to multicollinearity (table 6). Also note that the absolute effect of life satisfaction is larger for women (table 5). Thus, given the smaller number of female entrepreneurs the significant effect on the female share of entrepreneurs is not surprising.

41. Indeed, comparing tables 5, 6, 7 and 8, we see that, by and large, the ordering of variables based on the size of the effects is quite similar to the ordering based on the significance (t-values) of the effects. We feel that this increases the credibility of our estimation results.

42. Female and male unemployment rates for 2001 and data on the employment levels of women and men in the service sector for 2000 are obtained from the ILO LABORSTA database (http://laborsta.ilo.org). No gender-specific unemployment data are available for India. Instead, we make use of the general unemployment rate in India to enable comparisons between the analyses using either general or gender-specific variables. For the same reason, we use the general service employment rate for South Africa. (Fe)male employment in the service sector is scaled on total (fe)male employment. For employment definitions and measurement per country we refer to Verheul (2005), Chapter 2. Gender-specific data for importance of family and life satisfaction are obtained from the World and European Values Surveys. Average country scores for women and men are used.

43. Indeed, most of the research in the area of female entrepreneurship focuses upon Anglo-Saxon countries (Verheul 2005).

44. The possibility of including more countries is largely dependent on the number of countries participating in GEM. In 2003 and 2004 the number of participating countries in GEM was lower than that in 2002. To safeguard a sufficient number of observations in our analyses in this paper we used data of 2002. Also note that pooling of data of different years is no option as the effect of the business cycle would distort results (Reynolds et al. 2002).

45. Note that the variables ‘informal sector’ and ‘former communist country’ already account for some of these possible differences between different parts of the world.