Do returns to education depend on how and whom you ask?. (October 2017)

• Record Type: Journal Article
• Authors: Serneels, Pieter; Beegle, Kathleen; Dillon, Andrew
• Is Part Of: Economics of education review. Volume 60(2017:Oct.)
• Page Start: 5
• Page End: 19
• Publication Date: 2017-10
• Subjects: Returns to education -- Survey design -- Field experiment -- Development -- Tanzania; J24 -- J31 -- C83; Education -- Economic aspects -- Periodicals; 370
• DOI: 10.1016/j.econedurev.2017.07.010
• Languages: English
• ISSNs: 0272-7757

Abstract

Highlights: Estimates of the returns to education vary according to questionnaire design, but not according to who reports the information. The differences are substantial: 5 percentage points higher among the most highly educated men and 16 percentage points higher among the least well educated women. These differences are similar in magnitude to the bias in ordinary least squares estimation and to the differences by gender, educational attainment, and sectors. Abstract: Returns to education remain an important parameter of interest in economic analysis. A large literature estimates these returns, often carefully addressing issues such as selection into wage employment and endogeneity in terms of completed schooling. There has been much less exploration of whether the estimates of Mincerian returns depend on how information about wage work is collected. Relying on a survey experiment in Tanzania, this paper finds that estimates of the returns to education vary by questionnaire design, but not by whether the information on employment and wages is self-reported or collected by a proxy respondent. The differences derived from questionnaire type are substantial, varying from higher returns of 5 percentage points among the most well educated men to 16 percentage points among the least well educated women. These differences are at magnitudes similar to the bias in ordinary least squares estimation, which receives considerable attention in the literature. The findings demonstrate that survey design matters in the estimation of returns to schooling and that care is needed in comparing across contexts and over time, particularly if the data are generated through different surveys.