Finite Response Windows for Self-Administered Surveys

How to cite

Glazerman, Steven; Tabassum, Tanzila. 2022. "Finite Response Windows for Self-Administered Surveys." AEA RCT Registry. AEARCTR-0009813. https://doi.org/10.1257/rct.9813-1.0

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

Steven Glazerman* , Tanzila Tabassum

* Primary investigator

Registry ID
AEARCTR-0009813
DOI
https://doi.org/10.1257/rct.9813-1.0
First registered
Last updated
Status
In development
Trial dates
2022-08-01 – 2022-08-31
Countries
Bangladesh ()
JEL codes
C83
Registry record
http://www.socialscienceregistry.org/trials/9813

Abstract

Empirical social sciences increasingly rely on data derived from sefl-administered web (SAW) surveys, but such survey modes suffer from low response rates. Understanding survey respondent behavior is critical to finding successful ways to increase response rates and improve data quality. There is a gap in the research literature in particular on the role of procrastination and commitment devices in respodnents' decisions to complete surveys in a timely fashion.

This study is a survey protocol experiment designed to test whether self-administered web surveys can achieve higher response rates by asking respondents to commit to a specific narrow window of time instead of the full data collection period. For example, would respondents be more likely to compete a survey if they were given a particular day and a three-hour time slot within which to complete (and an option to reschedule) instead of a one--week window?

Study subjects, 331 participants in an RCT on job training for survey enumerators, are randomly assigned to either be given a long window or a narrow window with a specific appointment time (3-hour period on a given day) to complete the survey by web. For the narrow window group, the link is only active for the specified period. For subjects in the other treatment arm, they are only given an end date and may access the survey at any time. The study's outcome is survey completion.

Intervention

The intervention is the size of window for completing the survey, small or large. This is an A/B test of two alternative survey protocols. The idea for a shorter window originally arose from a technical limitation on the number of respondents who could be on the server at a given time. In an earlier survey round, this method seemed to produce high response rates, so once that technical barrier was lifted and we did not need to assign short windows, we decided to implement the current experimental test, to see whether the artificial constraint would lead to better outcomes than an unconstrained/open-ended request to complete the survey.

Experimental design

Study subjects randomized within treatment arm to one of two groups:

Group A. Given a specific 3-hour appointment window to complete the survey. Respondents in this group are randomized to one of three possible windows.

Group B. Given the survey link and asked to complete any time during the one-week survey period.

Design details

Randomization method
Randomization built into SurveyCTO software.
Randomization unit
Individual respondents
Clusters
Not a clustered design
Observations
331 Respondents
Arms
165 in Group A 166 in Group B
Primary outcomes
Survey completion rate
Secondary outcomes
Respondent collusion: Another motivation for using the finite time slots is to reduce the possibility that study subjects interact with each other and influence each others' responses to survey questions, particularly knowledge questions. It is rare, but possible that survey respondents are members of the same family. We will analyze study outcomes within each window and compare across windows for the finite window treatment group to look for evidence of collusion on survey questions.

Cited by

No citation data available for this registration yet.

Keywords