ABSTRACT
This contribution evaluates whether affiliation with Islam is a theoretically and statistically robust proxy for patriarchal preferences when studying the relationship between gender inequality and economic growth. A cross-country endogenous growth analysis shows that direct measures of patriarchal institutions dominate a variety of religious affiliation variables and model specifications in explaining country growth rates, and that using religious affiliation, particularly Islam, as a control for culture produces misleading conclusions. This result is robust to the inclusion of measures of gender inequality in education and income, indicating that establishing and maintaining patriarchal institutions (a process this study calls “patriarchal rent-seeking”) exact economic growth costs over and above those measured by standard gender inequality variables. One of the key contributions of this study is to draw on unique institutional data from the Organisation for Economic Co-operation and Development's Gender, Institutions and Development (GID) database to better understand the gendered dynamics of growth.
NOTES ON CONTRIBUTOR
Elissa Braunstein is Associate Professor of Economics at Colorado State University in Fort Collins. In her work, she uses a feminist lens to better understand macroeconomic and international economic processes and outcomes, with particular emphasis on issues of economic development, growth, and gender equality.
ACKNOWLEDGMENTS
Many thanks to the anonymous reviewers and the guest editors for the time and care they took in giving constructive feedback on earlier versions of the article and to the participants in IAFFE and Feminist Economics’ Gender and Economics in Muslim Communities workshop. All mistakes are, of course, my own.
Notes
1 One important exception is Rafael La Porta, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny, who empirically link percent Muslim (and percent Catholic) to poorer government performance, though these associations “generally become insignificant” once they control for per capita income and latitude (Citation1999: 262).
2 Note that because the GID database (from which I draw the gendered institutional variables) does not include economies with fewer than one million people, I do not include them in my analysis either.
3 This specification stems from trying a number of geography variables as instruments. Area in the tropics consistently failed over-identification tests, so I came to include it in the growth regressions. I include the Frankel and Romer (1999) variable because it is widely accepted as a valid instrument for trade. I used location in the Southern hemisphere as an additional instrument because it was the only geography variable that had a marked impact on the 2SLS estimations, though primarily in terms of the overall fit of the first-stage regressions, as opposed to having much of a direct impact on the endogenous variables themselves.
4 For instance, the variance inflation factor for patriarchal dominance when run with the other gender inequality and religion variables is 2.41 and the condition number is 24.7.
5 First stage results are available from the author on request.
6 Because some scholars use a Muslim dummy variable (which equals one if 50 percent or more of the population is Muslim and zero otherwise) to proxy for patriarchal preferences, I also ran the model with just a Muslim dummy variable for religious affiliation, but failed to get any statistical or economic significance. However, in a prior version of the paper, with a smaller sample of primarily Muslim countries, the Muslim dummy coefficient estimate was larger and statistically significant.
7 Middle East and North Africa (MENA) does gain statistical significance, but this is primarily the result of a decline in standard errors rather than an increase in coefficients.
8 Furthermore, the first stage regressions (not presented) indicate that there is a negative and statistically significant association between PD and investment, a correlation that does not manifest for any of the other gendered variables nor for percent Muslim.