Skip to content

Gender Differentials in Returns to Education in Developing Countries

August 1, 2017

Editor’s note: This post is part of a series showcasing Barcelona GSE master projects by students in the Class of 2017. The project is a required component of every master program.


Authors:

Ignatius Barnardt, Golschan Khun Jush, Thies Wollesen, Samuel Hayden and Eva Sotosek

Master’s Program:

Economics and Finance

Paper Abstract:

We investigate a possible gender gap in returns to education using data from the World Bank’ STEP program for seven developing and emerging countries. We control for cognitive skills, non-cognitive skills and parental education – previously unobserved due to unavailability of data – to investigate how this heterogeneity is playing a role in estimating the gender differential in educational returns. We also model selection using the Heckman two-step estimation procedure to examine whether selection may be driving this phenomenon. Our findings suggest that gender gaps in returns to education are not as prominent in the countries in our sample as previously suggested. We also find that controlling for unobserved heterogeneity on the one hand, and selection on the other, has different effects in different countries, highlighting the importance of understanding individual countries’ labour markets in detail before drawing conclusions regarding the existence of a gender gap in returns to education.

Conclusions:

This paper explores gender gaps in returns to education for seven developing and emerging countries. First, we investigate the existence of such a gap in a standard Mincerian framework. We find a significant returns gap in only two countries, namely Ukraine and Ghana, while the estimates for the other countries are centred relatively tightly around statistically insignificant point estimates close to zero. Using quantile regressions to dig deeper does not materially affect our findings, although it does allow us to specify that the returns gaps estimated for Ghana and Ukraine are significant at two out of three quartiles of the wage distribution, and that in Vietnam there is a small but significant returns gap at the upper two quartiles of the distribution. These findings are important in providing context for the existing literature, showing that returns premiums in favour of females are not universally prevalent in developing countries for urban wage workers. This suggests that where large, significant returns gaps have been found in the literature, this seems to be driven to a large extent by other segments of the labour market.

Second, we use our novel dataset to analyse the extent to which controlling for previously unobserved heterogeneity, namely cognitive skills, personality traits and family background, affect OLS estimates of the returns gap. We find that controlling for these STEP variables does not materially affect our baseline estimates for Bolivia, Colombia, Georgia, Kenya and Vietnam (where the estimated gap remains insignificant and close to zero), or for Ukraine, where the estimated gap is of similar magnitude and remains significant. Only in Ghana we find that adding the STEP controls has a material effect, reducing the point estimate of the gap substantially and rendering it insignificant. The results of the quantile regressions qualify this finding somewhat, showing that controlling for the STEP variables does make a difference for estimates of the gap at certain quantiles of the distribution in Ukraine and Vietnam. Overall, our finding regarding the importance of these sources of previously unobserved heterogeneity is cautiously negative: although they do appear to make a small difference for the level estimates and have an important effect in Ghana, they do not appear to be universal sources of endogeneity in estimating the returns gap for urban wage workers.

Third, we examine the importance of controlling for selection in estimating the returns differential using the Heckman two-step procedure, dropping Kenya from our sample due to missing data. Here we find that after controlling for selection, our point estimates of the returns gap remain insignificant in Ghana, Georgia and Vietnam, albeit with a relatively high point estimate in Georgia. Similarly, our estimate of the returns gap in Ukraine does not change considerably and remains significant. In contrast, we obtain higher and significant point estimates of the returns gap in Bolivia and Colombia. As explained above, this somewhat counterintuitive result is due to positive selection of females into employment in Bolivia and Colombia, and the positive relationship between education levels and probability of employment. Interestingly, in the two countries where selection appears to be important, we found earlier that controlling for the STEP variables did not have an observable effect. Our findings therefore suggest that it is likely to be important to control for selection when estimating returns gaps in developing countries, even if only to exclude the possibility of selection bias. In addition, our approach suggests that selection is likely to operate through channels other than cognitive or non-cognitive abilities, or parental background.

Taken together, our findings show that, at least for urban wage workers in the countries in our sample, a returns premium for females may not be as prevalent as previously suggested. We also find that controlling for potential sources of endogeneity, such as unobserved heterogeneity and selection, substantially changes the estimates of the gender returns gap in three out of seven of the countries in our sample. This highlights the importance of considering these channels to avoid the risk of biased estimation. This paper therefore represents a starting point for more detailed research, which could zoom in on the existence and drivers of returns differentials in individual countries, and overcome some of the limitations of this paper by extending it to rural areas and using samples with a larger number of clusters. These findings are also relevant to policy makers, since they demonstrate the importance of understanding the characteristics and dynamics of each country’s individual labour market prior to making policy proposals.

Advertisements

Comments are closed.