CREDIT 21/01: Does the pay period matter in estimating returns to schooling? Evidence from East Africa

Abstract

This paper investigates whether returns to schooling differ according to the choice of the measure of earnings and the different periods in which workers are paid (daily, weekly, and monthly). Using comparable data from the Living Standards and Measurement Study (LSMS) for Malawi, Tanzania and Uganda, accounting for endogeneity using Gaussian Copula and for selection with the Heckman method, we show that converting earnings to common measures and pooling respondents produces different estimates of returns to education. Depending on the common measure chosen, estimates of returns for level of education can differ by up to 100% for Tanzania, up to 50% for Malawi and up to 20% for Uganda. Estimating separately for each pay period, returns also differ significantly. Returns to primary education are 40-70% in Uganda and 20-30% in Malawi and Tanzania. Returns to secondary education are about 80% in Malawi and Tanzania but vary between 50% and 90% in Uganda. Returns to higher education are 130% in Tanzania, 100-150% in Uganda and 120-165% in Malawi. Returns to increase with the level of education completed but estimating separately for different periods is more reliable than pooling.

Download the paper in PDF format

Authors

Livini Donath, Oliver Morrissey and Trudy Owens

View all CREDIT discussion papers