Does the Gender of Top Managers and Owners Matter for Firm Exports?

ABSTRACT

How are export propensity and intensity affected by gender? Data from the World Bank's Enterprise Surveys (waves 2006–07, 2009–10) are used in a cross-country analysis to investigate whether export propensity and intensity differ according to the gender of top managers and entrepreneurs. Exporting is riskier than selling domestically and women, on average, tend to be more risk averse than men. Exporting entails costs, and women may have reduced access to finance compared to men. Most firms managed or solely owned by women are young and small and may have more difficulty obtaining credit. Women may self-select into routine sectors with lower mean productivity. Unlike most previous research, here the gender effect only takes into account firms where women have decision-making power. Accounting for the endogeneity of firm productivity, firm self-selection into exporting, and several factors influencing export propensity and intensity, the gender effect operates indirectly via some of those factors.

KEYWORDS:

• Gender
• entrepreneurship
• export propensity
• export intensity
• selection
• development

NOTES ON CONTRIBUTOR

Helena Marques joined the Department of Applied Economics at the University of the Balearic Islands, Spain, as Associate Professor in 2009. She was previously Senior Lecturer at Manchester Business School (2007–09), Lecturer at the Department of Economics at Loughborough University (2003–07), and Teaching Assistant at Newcastle University, first at the former Department of Economics (2001–02) and then at the newly enlarged Newcastle Business School (2002–03), where she also obtained her PhD in 2004.

ACKNOWLEDGMENTS

This research was carried out using data from the Enterprise Surveys (http://www.enterprisesurveys.orghttp://www.enterprisesurveys.org) collected by the World Bank. Thanks are due to the World Bank for making the data freely available. Thanks are also due to the reviewers and editors for their thorough comments that allowed substantial improvement of the manuscript. The usual disclaimer applies.

Notes

1 For comprehensive surveys on the exporting and investment decisions of heterogeneous firms, see Elhanan Helpman (2006), David Greenaway and Richard Kneller (2007), Andrew B. Bernard, Stephen J. Redding and Peter K. Schott (2012), and Marc J. Melitz and Stephen J. Redding (2014).

2 A related issue is whether women are part of the top management team and actually have decision-making power. Unfortunately, the World Bank's Enterprise Surveys database does not contain information on who holds decision-making power. All that can be done is to infer such power through the figures of the top manager or of the sole owner. It is well known from agency theory that the involvement of the owners of the firm in management decisions diminishes with firm size. In two extreme cases, the owners in very large firms often fully delegate management onto a CEO, while owners in very small firms are often the managers as well. By considering the two cases of top management and of sole ownership, the paper covers the two extreme cases.

3 Due to space constraints, only a summary is presented here. A fuller discussion is available from the author on request.

4 Due to space constraints, only a summary is presented here. A fuller discussion is available from the author on request. The data itself is available from http://www.enterprisesurveys.orghttp://www.enterprisesurveys.org.

5 Manova (2013) provides a thorough review of the literature, a theoretical analysis, and discussion of the empirical strategy, with a focus on developing countries but using product-level trade data. In the sample used in this paper, those firms with lower levels of (log) capital stock and (log) productivity do not export. In particular, all exporters in the sample have values of (log) capital stock and (log) productivity above -0.1278 and -3.1419, respectively; whereas, the minimum values of those variables for non-exporters are -3.8645 and -5.0685, respectively. Therefore -0.1278 and -3.1419 are the threshold levels at which firms start to export in this sample.

6 Due to missing data for many explanatory variables, the sample used is essentially a cross-section of twenty-three countries in Latin America, the Caribbean, Africa, and Asia (see Table A.1). The cross-sectional nature of the data prevents the extended use of lagged values, although the three-year lagged values of workers and productivity come directly out of survey questions. Productivity is defined as sales per worker instead of production per worker because production and stock data are not provided. However, this originates two caveats: (1) Given that total production equals total sales plus changes in stock, when those changes are not zero, productivity defined as sales per worker will be over or underestimated; and (2) It may include taxes, and the tax structure varies by sector as well as by country. Also, capital's book value is used instead of its market value, as the former was defined prior to the latter in time. See Table A.2 for more details on the definition of variables.

7 The variation in sample countries and sectors is due to missing data and is indicated by an asterisk in Table A.1. It consists in dropping African countries from the women-owners regressions and dropping chemicals and pharmaceuticals, and textiles from the propensity to export regressions. Note that in Table 2 with nine sectors, the sector fixed effects and selection effects are significant for women managers, but not for women owners. In order to check how the missing sectors and countries may affect the results, two more sets of regressions were run (results available upon request). The first reproduced Table 2 for the seven sectors of Table 1. The second reproduced the women-managers regressions for the seventeen countries of the women-owners regressions. Removing chemicals and pharmaceuticals and textiles from the export intensity regression introduced a negative autonomous gender effect for firms owned by women, but it also introduced a significant selection effect and significant sector fixed effects. Therefore, it could be argued that the gender effects and determinants of export intensity are dependent on the sector, which does not contradict with this paper's argument that the sector of activity matters. On the other hand, excluding the six countries missing from the women-owners regressions (mostly African countries) does not change the main results. The main change is that selection effects are not significant if those six countries are excluded – which makes sense if taking into account that African countries are probably those where barriers to entry into exporting activity are the most binding constraints (for example, see Bardasi, Sabarwal, and Terrell [2011] on the issue of credit constraints). Moreover, many more observations are lost with the removal of those two sectors (around 500) than with the removal of those six countries (around 100), which naturally impacts the results.

8 The regressions are carried out separately for firms whose top manager is a woman or whose sole owner is a woman. The firms both owned and managed by women (woman manager dummy equal to 1 and woman sole owner dummy equal to 1) are less than 4 percent of the sample and are the smallest ones (65 percent of them have less than twenty employees, against 49 percent for the whole sample). Unfortunately, in the sample there are only seven firms both managed and owned by women that do not get dropped from the regressions due to missing data. These also have a patent or formally trained workforce. With this important caveat in mind, when running the regressions with the owner and manager gender dummy in the same regression, the autonomous gender effect is negative for the propensity to export (given equal productivity by gender) but positive for export intensity; that is, women are less likely to export (even though their firms are as productive as those of men), but once they do they actually export a higher share of their sales compared to men (results available upon request). Selection effects are significant in the export intensity regression, so those women that select into exporting are not taken randomly from the sample. This reinforces the role of patents and workforce training for exporting. But it also reinforces the main point of the paper, which is that the negative gender effect at the means hides several indirect effects that can be picked up when a number of relevant determinants of exporting are taken into account. Given the reduced number of firms for which the GENDER dummy would be 1 in this case, the suspiciously high value of the remaining GENDER coefficients makes these results likely candidates to suffer from high collinearity.

9 The graphical representation of the firm distribution of the estimated export probabilities is available upon request.

10 Additional gender effects that operate through interactions between foreign ownership, technology, and imports were introduced as a further check (results available upon request). While most of the previous results are robust to these interactions, the selection effect disappears even for women managers, showing the importance of interactions between foreign ownership, technology, and imports in determining the propensity to export.

11 This effect is related to competition, but the cluster variables require a count of firms in the same city area. Meanwhile, for the number of competitors there is no geographical restriction.

12 To be more precise, there are two instances where there is a negative effect of the number of competitors on productivity for women: with more than five competitors for firms managed by women, and also with one competitor for firms owned by women.