The CEDAW Effect: International Law's Impact on Women's Rights

Abstract

Evidence of demonstrable, positive effects for the United Nations’ international human rights treaties has generally eluded researchers. However, the Convention for the Elimination of All Forms of Discrimination Against Women (CEDAW) has a statistically significant and positive effect on women's rights, even when other key factors are controlled. This result is counterintuitive, given that CEDAW's enforcement mechanisms were initially weaker than other human rights treaties. Furthermore, women's rights are implicated in deeply ingrained cultural systems that are difficult to change. This article confirms CEDAW's positive effects, finding that they are robust but not uniform. They are most pronounced for women's political rights, somewhat less pronounced for women's social rights, and absent for women's economic rights. Several possible counterexplanations for CEDAW's effects are tested but garner little support. We conclude by offering alternative hypotheses for CEDAW's effects that focus on the internal dynamics ratification may trigger within individual countries.

Notes

There are a few exceptions to this apparent consensus. Landman (2005) finds some significant, positive relationships based on data covering multiple treaties that includes treaty reservations. Even so, these relationships are relatively weak. Simmons (2009) and Hafner-Burton (2009) find that human rights treaties can be effective indirectly by altering domestic dynamics and providing local activists with resources to mobilize to protect rights.

For a good example of a study that reports change on the ground within individual countries, see Heyns and Viljoen (2001, 2002). Their comprehensive work on the impact of six UN human rights treaties in 20 countries finds that individual countries do take tangible steps to incorporate treaty norms in their domestic legal structures and cultures. They do not, however, measure the effects of these changes on actual human rights conditions within each country.

Included in the CEDAW-specific analysis are Australia, Brazil, Canada, Colombia, Czech Republic, Egypt, Estonia, Finland, India, Jamaica, Japan, Mexico, Philippines, Romania, Russia, Senegal, South Africa, Spain, and Zambia. Iran is included among the 20 countries studied by Heyns and Viljoen, but it has not ratified CEDAW.

The Committee is designed to represent participating states in terms of geography, legal structure, and “form of civilization” (see CEDAW 1979, Part V, Article 17).

Details about the dataset can be found at http://ciri.binghamton.edu/.

Arbitrary cut-points may influence the results. We also ran each model with a dichotomized dependent variable obtained by collapsing CIRI categories 0 and 1 into one category and 2 and 3 into another. This yielded similar results to those reported in Table 1.

Hendrix (2010) provides a thorough theoretical and empirical review of 15 different measures of state capacity and finds that survey measures of bureaucratic quality and tax capacity are best employed for both theoretical and empirical reasons. We employ the latter and, following Hendrix, measure tax capacity as tax as a proportion of GDP. For a fuller discussion of the merits of using this specific proxy for state capacity, see Englehart (2009).

For instance, Fein (1995) argues that there is “more murder in the middle” in that transitional and semi-democratic regimes have more human rights abuses than fully democratic or fully authoritarian political systems. By contrast, Davenport and Armstrong (2004) and Bueno de Mesquita et al. (2005) suggest a tipping point, with most of the effect of democracy coming only with the establishment of full multiparty competition. Englehart (2009) finds that the nonlinearity is not simply curvilinear, and it differs for different measures of rights.

Data obtained from the World Bank's (2011) World Development Indicators.

Data obtained from the World Bank's (2011) World Development Indicators.

Introducing a lagged dependent variable can create problems of endogeneity and bias that are likely to worsen rather than improve the quality of the model (Achen 2000). These problems are likely to be particularly pronounced in time-series analysis and in models with binary and ordinal dependent variables (Brandt, Williams Fordham, and Pollins 2000; Brandt and Williams 2001; Pang 2008). As a robustness check, we ran the models in Ordinary Least Squares regression with the lagged dependent variable and as ordered logits omitting the lagged dependent variable. We obtained similar results using these alternative specifications. Omitting the lagged dependent variable tends to increase the coefficients for the other variables, but the pattern of significance, including that of the CEDAW ratification variable, remains basically the same.

These calculations were performed using Clarify (King, Tomz, and Wittenberg 2000).

It is likely that international war deaths do not emerge as a strong predictor because international wars are relatively rare. Civil wars are far more common than international wars, which may explain why civil war deaths emerge as statistically significant.

In the model of women's political rights using five-year lagged results, state capacity and economic development have such a large effect that they shift almost all countries into Category 3 on women's political rights, thereby reducing the probability of being in Category 2.

Saudi Arabia is currently the only country in the world where women are denied suffrage. However, only local offices are elected there: as of January 2014, there was no national legislature and the head of state was a hereditary monarch. The government has promised that women will be allowed to participate in local elections in the future. In many states, however, the legal right to vote is infringed in practice due to cultural stereotypes and active discrimination.

States may weaken their commitments by lodging reservations against certain articles of the treaty with which they do not intend to comply. In the case of CEDAW, these reservations may be substantial, with some states making reservations even against Article 2, which arguably undercuts the central purpose of the treaty. We considered the effect of the “reservations” states parties file at the time of ratification. Landman (2005) constructed a variable measuring the seriousness of state reservations to CEDAW, as well as other human rights treaties, and generously shared it with us. Using this to weight the ratification variable does not, however, change the results in any significant way. Neither the pattern of significance nor the relative impact of the treaty changes very much from the results shown in Tables 1 and 2. The 2000 Optional Protocol (OP) to the treaty permits individual complaints against states parties that have ratified the OP, rather than waiting for the CEDAW Committee to assess compliance during the regular reporting process. As of June 2012 the Optional Protocol had 104 states parties and 29 complaints had been considered. For details, see http://www2.ohchr.org/english/law/jurisprudence.htm (accessed 29 January 2013). Using ratification of the Optional Protocol to indicate CEDAW commitment does not produce very different results from the ratification variable.

We are indebted to an anonymous reviewer for this suggestion. The reporting variable produces somewhat stronger results for women's political rights and weaker ones for social rights, but the direction of the coefficients and the pattern of significance remain the same.

Jamaica, for instance, apparently ratifies treaties only when it believes that it is already in compliance (Heyns and Viljoen 2001: 494).

For a good discussion, see Hill (2010).

The lower house of parliament is used rather than the upper house, or a combination of the two, because lower houses are more likely to be democratically elected. Women as a percent of the lower house is a relatively “sticky” indicator, because, once elected, a legislature will typically sit for a few years before a new election. For this reason, the proportion of female parliamentarians is shown in Figure 2 but was not employed as a dependent variable in Tables 1and 2. The proportion of female parliamentarians does not perform well as a dependent variable in time-series cross-section models since values for each country tend to remain the same for several years. This leads to a huge coefficient on the lagged dependent variable that essentially wipes out all other results.

These include Syria, San Marino, and Timor-Leste, all scoring 1 on the CIRI Women's Social Rights variable, and Sao Tome, with a score of 2. Afghanistan also ratified in 2003, but there is no CIRI data for Afghanistan in that year.

Since educational systems vary widely, we use this proportion rather than the raw percentage of female enrollment to compensate for the fact that in many countries enrollment for both boys and girls may be very low. The proportion of female-to-male enrollment specifically targets the concept we are interested in: whether girls are less likely to receive an education than boys.

The spike in 2008 is due to the fact that data are available for only three countries in that year: Qatar, Sudan, and the United States. Qatar lists an extraordinary ratio of 119% female over male enrollment in that year. In 2009, Qatar ratified CEDAW, causing a huge drop in the mean for the nonparties group. It is unclear how reliable the Qatar data are; if it is accurate, it registers a very large decline in male enrollment for unexplained reasons.

^. The first-year effect may actually be blunted by the fact that some countries also make changes in anticipation of ratification. Liechtenstein, for instance, passed a constitutional amendment on gender equality in 1992, in anticipation of ratifying CEDAW in 1995. See Government of the Principality of Liechtenstein (1997).

The year dummies are therefore dropped in the models reported in Table 3.

^. The lagged models for political and civil rights (including the number of ratifications variable) produce similar results as those that appear in Table 1 and thus are not shown here. In both lagged versions of the political rights model, the only difference is that GDP per capita loses significance when the ratifications variable is added, for reasons that are unclear. Meanwhile, the number of ratifications variable is not significant in either the one- or five-year lagged models of women's political rights. Both lagged versions of the social rights model (including the number of ratifications variable) show the same patterns of significance as those shown in Table 1. The number of ratifications variable is significant and negative when a one-year lag is employed and loses significance when a five-year lag is employed. As with the contemporaneous results displayed in Table 3, there is little evidence to suggest that a worldwide cultural trend toward improved women's rights conditions is explaining away the effect of CEDAW ratification.

See CEDAW (1995, para. 30).

See CEDAW (2004, para. 3).

See CEDAW (2004, para. 15). Passage is attributed to Ms. Hanna Beate Schöpp-Schilling (Germany), but “urgent legal review” may not be a direct quotation.

The Government of Morocco maintains one reservation to CEDAW. It pertains to the mechanism for arbitration of disputes brought by other states under the treaty.