Hungry for Equality: A Longitudinal Analysis of Women’s Legal Rights and Food Security in Developing Countries

ABSTRACT

Access to food is one of the most basic human needs, yet undernourishment remains one of the greatest barriers to health and development in the Global South. The entitlement approach to hunger maintains that the real issue is not the overall availability of food, but access to it. This paper specifically highlights the access failures of women by focusing on how gender-based stratification limits command over food, even in light of existing food supply. Going beyond traditionally used indicators of gender inequality, we examine how women’s legal rights affect the depth of hunger in developing countries. In a first-time longitudinal study of its kind, we use the 50 Years of Women’s Legal Rights historical dataset from the World Bank to measure changes in women’s property rights and constitutional rights from 1990 to 2010. A two-way fixed effects regression analysis of 42 low- and middle-income nations demonstrates that improving women’s rights is associated with lower levels of hunger over time, even controlling for the food supply, economic development, and other relevant variables.

KEYWORDS:

• Development
• Sociology of Global and Transnational Sociology Sex and Gender

Notes

1. For a more detailed explanation of the rights and the methodology for data collection, see (Hallward-Driemeier, Hasan, and Iqbal 2013).

2. In additional analyses not shown but available upon request, we consider each right by itself. For two rights, there is not enough variance in the data for the model to converge (i.e., “default marital property regime protects wife’s interests” and “religious law is invalid if it violates non-discrimination clause”). For the rest of the rights, the results are consistent with those presented here, suggesting that the indices are valid and reliable. We considered Item Response Theory (IRT) and factor analysis to create a weighted scale of the indices (Coppedge et al. 2016). However, there is not enough variability within individual rights over time. Once a country is coded as 1, it tends to stay at 1. IRT requires more variability and larger sample sizes than our dataset provides (Marvelde 2006). Furthermore, factor analysis is not suitable for dichotomous predictors as it assumes linear associations between observed predictors (Wirth and Edwards 2007). Since the individual rights are consistently statistically significant, this suggests that no single right is driving the significance of the indices in the final models.

3. We also tested women’s literacy rates and contraceptive prevalence as alternate measures of women’s status. Neither of these variables reaches statistical significance, and both reduce the goodness of fit of the model.

4. GDP growth is an alternative measure to GDP per capita. In additional analyses, we considered annual GDP growth, which measures the annual percentage growth rate of GDP at market prices based on constant local currency (World Bank 2015). This measure of GDP is not statistically significant, and results for the key independent variables of interest are robust to this alternate model specification. We choose GDP per capita because it is the most common indicator of economic development in previous literature and it improves the fit of the models over GDP growth.

5. We also considered indicators of foreign direct investment (FDI), since imports/exports and FDI both measure embeddedness in the international market. However, previous research shows that FDI significantly affects wellbeing only when measured by sector (Mihalache-O’keef and Quan 2011). Unfortunately, sectoral data are not available for the time points used here.

6. Polity IV provides another popular data source for democracy indicators. However, we choose Freedom House for theoretical and statistical reasons. First, Freedom House measures political rights and civil liberties enjoyed by a state’s populace, whereas Polity IV measures levels of autocracy versus democracy (Casper & Tufis 2003). Given that the analysis is anchored around Sen’s (1999) concept of entitlements, a measure of democracy that assesses the rights and liberties of the population is better aligned with our key independent variables that measure women’s rights/entitlements. Second, even though the two common measures of democracy are highly correlated, empirical evidence shows that they do not always yield consistent results and are not interchangeable in statistical analyses (Casper & Tufis 2003; Högström 2013). In this case, the Polity IV measure fails to reach levels of statistical significance, whereas the Freedom House measure is consistently significant. Finally, the Freedom House measure yields a higher R-square (within) value, suggesting that it improves the goodness of fit of the model.

7. Arable land per hectare provides another measure of land. Arable land per hectare is not statistically significant and the key findings regarding the main independent variables are robust to this alternative model specification. We choose arable land equipped for irrigation because it is a more precise measure of agricultural activity. That is, it tells us how much of the available land is able to facilitate irrigation technologies, without which agricultural productivity is stunted due to reliance on rainfall.

8. The 42 countries in the analytic sample are as follows: Bangladesh, Benin, Bolivia, Burkina Faso, Cambodia, Cameroon, Central African Republic, Ethiopia, Egypt, Arab Rep., Georgia, Ghana, Guatemala, Honduras, India, Indonesia, Kenya, Kyrgyz Republic, Lesotho, Madagascar, Malawi, Mali, Mauritania, Morocco, Mozambique, Nepal, Nicaragua, Niger, Nigeria, Pakistan, Philippines, Rwanda, Senegal, Sri Lanka, Swaziland, Tajikistan, Tanzania, Togo, Uganda, Vietnam, Yemen Republic, Zambia, Zimbabwe.

9. The four excluded countries are: Sierra Leone, Uzbekistan, Lao, and Liberia. These countries are dropped because data on food imports/exports are unavailable for the time period. As a robustness check, we ran all models without measures of food imports and exports, thereby increasing the sample to 46 countries. The major findings do not change substantially.