A Framework for Reconsidering the Lake Wobegon Effect

Abstract

The "Lake Wobegon Effect'' describes the potential bias introduced into survey-based analyses of education issues, because students systematically over-report academic achievements such as grade-point average. While the use of official-records data negates this effect, many researchers can only access student-reported data. In this paper we examine whether student-survey data in place of official-records data meaningfully biases regression estimates. We investigate this by capitalizing on a subtle but extremely useful statistical feature of over-reporting on bounded variables like grade-point average. Specifically, the misreports will be negatively correlated with the true grade-point average - a form of non-classical measurement error - which actually counteracts the bias. Around this insight we build a simple testing framework that permits a researcher with access to both data sets to assess the bias's size and significance. We demonstrate our method using two data sets, and find that the bias does not meaningfully impact the signs, significance, or magnitude of regression slope coefficients. Our results, and that the special econometric feature is likely present in all student-reported grade data, lead us to believe the Lake Wobegon Effect bias is inconsequential.

Keywords: Education Practice, Student Survey Data, Grade Bias, Measurement Error, Aptitude Data

JEL Classification: A22, C13, C52