Measuring Response Bias in Surveys

How to cite

Iriberri, Nagore; Hospido, Laura. 2025. "Measuring Response Bias in Surveys." AEA RCT Registry. AEARCTR-0017405. https://doi.org/10.1257/rct.17405-1.0

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

Nagore Iriberri* , Laura Hospido

* Primary investigator

Registry ID
AEARCTR-0017405
DOI
https://doi.org/10.1257/rct.17405-1.0
First registered
Last updated
Status
In development
Trial dates
2025-12-15 – 2026-02-23
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/17405

Abstract

Surveys to measure perceptions and experiences are frequent nowadays in many professional settings. These surveys are administered via email that invite individuals to participate in them and therefore response rates are overall low. More worrying, as their results rely on answers by those individuals who have self-selected into participation, response bias is a serious concern. This field experiment proposes randomizing individuals into two main groups, such that a control group receives an email inviting individuals to participate in them, as it is usually the case, while the treatment group finds the survey embedded within a professional task. The control group is at the same time randomized into having a reminder email and no having the reminder email. The survey is always the same. Comparing the demographics and answers across these three groups allow us to measure the existence of response bias and its consequences.

Intervention

Individuals who self-select in a professional activity are randomized into two groups:

1. Treatment (survey embeded within the profesional task): the survey is embeded in the professional task such that all individuals who decide to do the task are directed to the survey as part of the profesional task.
2. Control group (survey received by email after the professional task is completed). Within this control group, there is also randomization between receiving a reminder email and not receiving it.

Experimental design

Individuals who self-select in a professional activity are randomized into two groups:

1. Treatment (survey embeded within the profesional task): the survey is embeded in the professional task such that all individuals who decide to do the task are directed to the survey as part of the profesional task.
2. Control group (survey received by email after the professional task is completed). Within this control group, there is also randomization between receiving a reminder email and not receiving it.

Design details

Randomization method
Randomization is done in office by a computer.
Randomization unit
Individuals who self-select into a specific professional activity.
Clusters
Approximately 2000 individuals.
Observations
Approximately 2000 individuals.
Arms
About 1000 in the control group and about 1000 in the treatment group.
Primary outcomes
All questions included in the short survey.

Cited by

No citation data available for this registration yet.

Keywords