Working Paper
How far does the apple really fall from the tree?

Despite the importance attributed to intergenerational educational mobility in the process of development, there remains little consensus on how mobility should be measured. We present analytical and empirical evidence regarding the sensitivity of alternative estimators to different forms of measurement error and data transformations.

To do so, we develop a flexible simulation model, allowing us to quantify the bias associated with different empirical choices. Our evidence suggests that use of an upwards mobility estimator, complemented by an out-performance metric, based on a reference distribution transformation is comparatively most robust.

Applying these recommendations to the case of Mozambique, using complete census records from 1997, 2007, and 2017, we find a high degree of provincial heterogeneity, with a marked north-south gradient in mobility. Moreover, we identify a clear slowdown in educational mobility between 2007 and 2017, especially in the northern region, suggesting the education system is not delivering consistent gains to children across the country.