Lying on Surveys

• In Event: 4351 - Measurement Issues in Social Statistical Models

Abstract

This paper develops methods for analyzing misreporting of sensitive traits on surveys.
Methods are provided for surveys that combine direct questioning with one of two indirect
questioning techniques, the list experiment and randomized response technique. By modeling
the joint distribution of direct and indirect responses, the methods enable efficient estima-
tion of the relationships between misreporting and respondent characteristics and may improve
regression analysis of the underlying traits. Extensions of the methods address robustness,
incorporate auxiliary information and outside surveys, and mitigate measurement error arising
from respondent noncompliance. I demonstrate the methods by revisiting two list experiments
on anti-immigrant sentiment in the United States, which show that avowed though not latent
support for restricting immigration increased during the Great Recession (Creighton, Jamal and
Malancu, 2015).

Author

• Winston Chou, Princeton University