Estimating Inequality of Opportunity in Ghana Across Cohorts: A Machine Learning Approach

In the last decades, the inequality of opportunity has captured more and more of the attention of researchers and policymakers. While extensive theoretical and empirical work has been carried out in this field, only a small fraction has been conducted in developing countries, particularly in Africa, due to a lack of data. The paper presents new estimates of inequality of opportunity in Ghana across cohorts, showing the pattern of such measures over time and generations. The current approaches to estimating inequality of opportunity are often hindered by the ad-hoc model selection, which can lead researchers to either overestimate or underestimate the actual amount of inequality of opportunity. Hence, we implement a machine learning approach (Regression Tree and Forests) that is able to overcome the discretionary factors in the model and circumstances selection. We have shown that IOp follows an increasing pattern for the middle generation and then decreases for the last generation; hence, the generations born in more recent cohorts experience more equality in terms of opportunity than the older ones. The trend seems to follow the political events that affect the country: the more instability there is, the higher the Iop. We record the effect of the education policies (which opened up women’s enrollment) on the mother’s education as one of the main drivers of the inequality of opportunity for the last cohort. Geography and ethnicity matter considerably, and they seem to matter less and less than in the past.