Reforming the Gender-Related Development Index and the Gender Empowerment Measure: Implementing Some Specific Proposals

Abstract

Since their inception in 1995, the United Nations Development Programme (UNDP)'s Gender-Related Development Index (GDI) and Gender Empowerment Measure (GEM) have been criticized on conceptual and empirical grounds. In 2005–6, the UNDP's Human Development Report Office undertook a review of these indicators and suggested some modifications. This study extends this work by adjusting the recommendations, making concrete proposals for two gender-related indicators, and presenting illustrative results for these proposed measures. These new measures include the calculation of a male and female Human Development Index (HDI), as well as a gender gap measure (GGM) to replace the GDI as a measure of gender inequality. The study also proposes and implements several modifications and simplifications to the GEM. With these adjustments, a number of Sub-Saharan countries now rank much higher, countries in the Middle East have lower scores in both measures, and some European countries fare notably worse in the revised GEM.

Keywords:

• Gender inequality measures
• GDI
• GEM
• UNDP
• JEL Codes: I31, J16

Acknowledgments

The authors thank Claes Johannssen, Geske Dijkstra, Kevin Watkins, Allison Kennedy, the three anonymous referees and the editors of this journal, and participants at workshops in New York and Göttingen for helpful comments and discussion. Funding from the UNDP in support of this work and excellent research assistance from Ramona Rischke is gratefully acknowledged. We also thank Karina Trommlerova for pointing out another mathematical problem associated with UNDP's GEM, which we have briefly discussed here.

Notes

While it is likely that women with low earned incomes, relative to men, might suffer from inequalities in access to resources within the household, it is patently false to assume that a woman who earns no income at all therefore has no access to resources for human development such as food, clothing, and housing. See Klasen (2006a) and Stephan Klasen (2007) for a fuller discussion of these issues.

Another conceptual problem is that the procedure to adjust the HDI for gender inequality compounds penalties for gender inequality in different dimensions, even if the inequality hurts women in one dimension and men in another. Thus a country with gaps harming women in all three dimensions is treated the same as a country where equal-sized gaps impact women negatively in some dimensions and men in others, which seems problematic. As is shown in Klasen (2006b), this affects the results for many countries where women are advantaged in the life expectancy component but disadvantaged in the education and earned income component.

The only difference is that the GDI uses the log of income levels, while the GEM uses the income levels themselves.

There is an additional computational issue that needs to be raised. Due to the way the GEM is calculated, which considers the population shares of men and women relative to their shares in employment and parliamentary representation, another mathematical problem appears. If the female share of parliamentary representation or economic participation is above 50 percent but below the female population share (and thus women continue to be underrepresented), the component for parliamentary representation or economic participation is actually above 1, clearly an undesirable feature (which occurs, however, only rarely and does not greatly affect the rankings). Our alternatives proposed below deal with this issue. We want to thank Karina Trommlerova for pointing out this inconsistency to us.

As done in UNDP's GDI, we also assume in all of our measures that women have a five year life expectancy advantage; only deviations from that constitute inequality.

As the indicators included in the GDI, GEM, and GGM change relatively slowly over time, the rankings shown here would hardly change if more recent or somewhat older data were used.

These gaps are much larger than those between the HDI and the GDI, which are only about 1 percent on average. See Klasen (2006b) for a discussion.

It might be worth considering capping the female–male HDI ratio at 1 for each component. See our discussion on capping and the GGM.

Remember that in the GDI, gaps going in opposite directions are compounded just as much as gaps going in the same direction. With the ratio of the female–male HDI ratio, this is not the case, because it only compounds countries with gaps in the same direction, which is another reason for the low rank of these countries.

They are not identical, though. For example, Lesotho, Rwanda, and Madagascar each appear very low in the GDI, but fare quite differently in their female–male HDI ratios and the GGM. Lesotho increases its rank the most in the female–male HDI ratio and falls back in the GGM, while Madagascar and Rwanda rank higher in the GGM. Brazil and Botswana are two more countries whose ranks increase substantially in the female–male HDI ratio (compared with the GDI) but fall back again in the GGM; in these cases, as in Lesotho's, the relatively high inequality in labor force participation now weighs more heavily in the GGM than in the female–male HDI ratio, leading to this change. This is also the reason for further substantial rank losses in the GGM for many Middle Eastern countries and countries such as Italy, Greece, and Malta.

Particularly noticeable is the relatively low position of Luxembourg in the GGM, which only occupies rank fifty-six, despite faring much better in the GDI, the female HDI, and female–male HDI ratio. This difference is due to a particularity in Luxembourg's case. Because of its very high prosperity, male and female earned incomes reach the maximum of US$40,000, so the earned-income index is capped at 1 for both – (erroneously) suggesting perfect equality between the sexes. The GGM, however, considers existing gaps in labor force participation, and thus Luxembourg drops considerably in rank.

It is true that the empirical problems with the income component remain, but these are not much worse than using labor force participation rates.

This is also possible for the GGM, but does not actually happen.