The sensitivity of measured prevalence of street harassment to questionnaire design

How to cite

Beaman, Lori; Jayachandran, Seema; Sharma, Anisha. 2026. "The sensitivity of measured prevalence of street harassment to questionnaire design." AEA RCT Registry. AEARCTR-0019421. https://doi.org/10.1257/rct.19421-1.0

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

Lori Beaman* , Seema Jayachandran , Anisha Sharma

* Primary investigator

Registry ID
AEARCTR-0019421
DOI
https://doi.org/10.1257/rct.19421-1.0
First registered
Last updated
Status
Completed
Trial dates
2024-11-13 – 2025-04-22
Countries
India ()
JEL codes
83, J16, K42, O12
Registry record
http://www.socialscienceregistry.org/trials/19421

Abstract

Gender-based harassment in public spaces affects women directly through victimization and indirectly through fear. We survey women aged 19–40 recruited at randomly sampled public spaces (bus stops, metro stations, and markets) in New Delhi (N=2,249) and Bengaluru (N=2,290), measuring the prevalence of seven forms of harassment, the anxiety the risk of each form induces, and self-reported effects of harassment risk on work and mobility decisions. To quantify how questionnaire design affects measured prevalence, respondents were randomly assigned to either the main questionnaire (with 90% probability) which asks directly about specific experiences related to gender-based harassment or an alternative (with 10% probability) that used screener questions. The screener questionnaire asked two questions about women's experiences with either physical or non-physical sexual harassment before asking the same set of specific experiences as in the main questionnaire, conditional on a woman indicating she had been harassed.

Experimental design

In-person survey of women aged 19–40 recruited at sampled public venues (bus stops, metro stations, retail markets) in Delhi (Nov–Dec 2024) and Bengaluru (Mar–Apr 2025). Respondents were randomized in real time (via SurveyCTO) between two questionnaire versions: a behaviorally specific version with neutral framing (90%) and a screener-question version labeling behaviors as "sexual harassment" (10%). The screener-question version replicates the instrument previously fielded to the full sample in a companion (non-randomized) survey in Jaipur (May–Jun 2024, N=1,863), whose data provide a cross-city benchmark and are reported in the paper's appendix; the Jaipur survey is not part of the randomized trial registered here.

Design details

Randomization method
Real-time computer randomization within the SurveyCTO survey instrument (uniform random draws generated at runtime for each respondent).
Randomization unit
Individual respondent (questionnaire version).
Clusters
N/A
Observations
Completed surveys: 2,101 in Delhi; 2,024 in Bengaluru (4,125 total in the trial).
Arms
Delhi: 1,900 main questionnaire, 201 screener version. Bengaluru: 1,823 main questionnaire, 201 screener version.
Primary outcomes
Reported prevalence of gender-based harassment in public spaces: any form, less severe forms (unwanted staring, catcalls, lewd gestures, flashing), and severe forms (groping, stalking, sexual assault), over the past year and ever, by questionnaire version.
Secondary outcomes
Type-specific prevalence; locations of harassment; self-reported anxiety about each harassment type; self-reported influence of harassment risk on whether and where to work, social activities, and errands.

Cited by

No citation data available for this registration yet.

Keywords