Covariate-Dependent Reporting Bias: Methods and Application to the LGBQ Earnings Gap

How to cite

Deal, Cameron. 2026. "Covariate-Dependent Reporting Bias: Methods and Application to the LGBQ Earnings Gap." AEA RCT Registry. AEARCTR-0018394. https://doi.org/10.1257/rct.18394-2.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Cameron Deal*

* Primary investigator

Registry ID
AEARCTR-0018394
DOI
https://doi.org/10.1257/rct.18394-2.0
First registered
Last updated
Status
Completed
Trial dates
2026-05-01 – 2026-05-15
Countries
United States of America ()
Registry record
http://www.socialscienceregistry.org/trials/18394

Abstract

This paper studies covariate-dependent reporting bias of binary traits. I develop a practical method that combines a list experiment with direct measurement to estimate both overall underreporting and the characteristics and outcomes of individuals who conceal under direct questioning. I apply the method using a question fielded in the Understanding American Survey to characterize the mental health and economic outcomes of non-heterosexual individuals who report their sexual identity indirectly. Additionally, I estimate the share and characteristics of misreporters--individuals who reveal their sexual identity indirectly (but not directly).

Intervention

A treatment group of approximately 4500 respondents will see a list of 5 items, including a question about sexual orientation, while a control group of roughly the same size will see a shortened list that omits the sensitive item. They will be asked how many of the items apply to them.

Experimental design

The study will randomly select roughly half of respondents to receive the veiled vs. direct report condition.

Design details

Randomization method
Randomization by Understanding America computer.
Randomization unit
Individual
Clusters
9000--this depends on how many individuals respond to the monthly survey for UAS.
Observations
9000
Arms
4500 in long list condition, 4500 in short list condition.
Primary outcomes
This study will use the outlined method to examine the level of misreporting in LGBQ identity using direct reports of sexual orientation and the implied prevalence from the list experiment above. Additionally, the study will use the proposed methods to characterize the following covariates for direct reporters, the full complier population, and misreporters: 1. LGBQ identity: Earnings, employment, wealth/assets, job characteristics, mental health measures, age, sex/gender, religious affiliation, political affiliation/ideology, vote choice, and race.

Cited by

No citation data available for this registration yet.

Keywords