Why Would She? Polygyny and Women's Welfare in Ghana

ABSTRACT

This study examines whether polygynous marriages are beneficial to women in Ghana. While some scholars claim that women benefit from such marriages in terms of higher consumption or leisure time, others believe that such relationships can be oppressive for women, as compared to monogamous relationships. Using household data from the 2005/6 Ghanaian Living Standards Measurement Survey V and the 2008 Ghanaian Demographic Health Survey, this study finds little evidence to support the view that women experience economic benefits from these unions. Polygynous women in Ghana tend to be more accepting of and experience more domestic violence, and they have less decision-making power within the household than women in monogamous marriages. Thus, there seems to be more evidence to support the view of polygyny as an oppressive institution rather than the outcome of a woman's rational choice.

KEYWORDS:

• Marriage
• family
• women

JEL Code:

• J12

NOTES ON CONTRIBUTORS

Amy Ickowitz is a development economist working at the Center for International Forestry Research in Indonesia. Her work focuses on issues of health, nutrition, and gender dynamics within the household.

Lisa Mohanty is Program Director of General Education at Trident University International. Her current research focuses on demographic economics, with an emphasis on issues of race, gender, and marriage.

ACKNOWLEDGMENTS

We thank the three anonymous reviewers for their suggestions and comments.

Notes

¹ Polygamy is the general term used for a marriage in which either a man or a woman has more than one spouse. Polygyny is the specific form of polygamy in which a man has more than one wife.

² Jacoby (1995) uses cluster-level crop shares as an instrument for wealth in a two-stage least-squares regression.

³ We restrict the sample for the GLSS dataset so that the ages of women match those in the DHS dataset.

⁴ In the GLSS sample women with primary education are less likely to be in polygynous marriages, while in the DHS sample it is secondary education that appears to make the difference.

⁵ The method consists of running a probit regression with the endogenous dummy as the dependent variable on the instrumental variable and other independent variables and then using the predicted value from that regression as the instrument in the second stage.

⁶ The treatment effects model estimates two regressions simultaneously using maximum likelihood; the first regression is a probit for selection into treatment, and the second (in this case) is a linear model for expenditures. It assumes that the error terms of both equations are jointly normally distributed.

⁷ This is the procedure recommended by Wooldridge (2002).

⁸ We also matched using nearest neighbor and radius matching and found qualitatively similar results. These are available from the authors upon request.

⁹ For example, perhaps a woman is known to be particularly hardworking and thus is considered an attractive marriage partner in her community. If she prefers to be in a monogamous (polygynous) marriage, she may be more likely to end up in such a relationship.

¹⁰ Results from using radius matching were identical to the kernel results, but results using nearest neighbors were not statistically significant, although also positive.

¹¹ There is an extensive literature on household bargaining and many variants of household bargaining models. The basic structure of such models is that members of the households arrive at outcomes either through a process of cooperative or non-cooperative bargaining (Marjorie McElroy and Mary Horney 1981; Duncan Thomas 1990; Shelly Lundberg and Robert A. Pollak 1996; Zhiqi Chen and Frances R. Woolley 2001; Cheryl Doss 2006; Imran Rasul 2008). Bargaining power within the household in these models is determined by each member's fallback position. Thus if women who enter into polygynous marriages are doing so because they have little choice, then their bargaining power will be weak, leading to worse outcomes compared to less oppressed women.

¹² This logic also applies to the earlier regressions in the paper, but our identification strategy and sample size do not allow us to test on only junior wives.

¹³ Although regional dummies are included in the regressions, they are not reported in the tables.

¹⁴ It seems reasonable that an individual should have the final say on their own health and how their individual earnings are spent, but that a couple would make decisions on household expenditures together.

¹⁵ There are fewer observations for the earnings variable than for the others, so it is possible that the difference in result is driven by a difference in sample. We ran the other three regressions restricting the sample to those with earnings and still found that women in polygynous marriages had significantly less say on their health decisions and on big purchases, but more say on daily purchases. Results available from authors upon request.

¹⁶ We thank an anonymous reviewer for this point.