The Determinants of Domestic Violence in Bangladesh - A Randomized Control Trial

Last registered on September 08, 2022


Trial Information

General Information

The Determinants of Domestic Violence in Bangladesh - A Randomized Control Trial
Initial registration date
September 02, 2022

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
September 08, 2022, 12:40 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.


There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Primary Investigator

Stanford University

Other Primary Investigator(s)

Additional Trial Information

In development
Start date
End date
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
This study uses a combination of lab-in-the-field and vignette experiments to assess the relative importance of different motives for which husbands might use violence against their wives, as well as the reasons for which low-income husbands use more violence than high-income husbands. In the lab-in-the-field experiment, “choosers” choose whether or not to transgress a social norm (eat the better plate themselves) and “responders” choose whether and how much to respond using a harmless punishment proxy (seconds of annoying sound). We randomly vary conditions to test whether the responders’ decision varies with 1) whether the chooser has additional information about the responder, 2) whether it was made before or after learning the chooser’s choice, and 3) whether the chooser’s choice (and the responder’s response) is observable to others. In the vignette experiment, subjects respond to a vignette story and oral description of a hypothetical couple and state their first- and second-order beliefs about the prevalence and social acceptability of spousal “disciplining” in different situations in which the wife might transgress a social norm. We will correlate the stated beliefs in the vignettes with the incentivized choices in the lab-in-the-field experiment to establish the external relevance of the lab-in-the-field experiment for intimate partner violence.
External Link(s)

Registration Citation

Buchmann, Nina. 2022. "The Determinants of Domestic Violence in Bangladesh - A Randomized Control Trial." AEA RCT Registry. September 08.
Experimental Details


Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
Lab-in-the-field experiment: The primary outcome is seconds of annoying sound chosen by the responder.
Vignette experiment: The primary outcomes are i) perceived likelihood of disciplining (first-order beliefs about prevalence), ii) perceived likelihood of disciplining by 20 subjects surveyed in the village (second-order beliefs about prevalence), iii) perceived social acceptability of disciplining (first-order), iv) perceived social acceptability of disciplining by 20 subjects surveyed in the village (second-order).
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
Lab-in-the-field experiment:

We randomly match two couples from different villages virtually (and anonymously) to participate in the experiment. The possible matched pairs are: i) the two wives play against each other, ii) the two husbands play against each other, and iii) the wife from one couple plays against the husband from the other couple. Within matched pairs, we randomly assign one spouse to be the “chooser” and the other spouse to be the “responder”.

The chooser decides between two plates of flat rice, the “good” plate, and the “bad” plate, and the responder receives the other plate. The responder can respond to each choice in terms of seconds that the chooser needs to listen to an annoying sound. The decisions are made and implemented in real-time and online.

We cross-randomize subjects into three treatment assignments:
1) Image relevance: We randomly assign responders to either an image-relevance or non-image relevance treatment arm. In the image-relevance treatment arm, we inform the responder that the chooser, before selecting the plate, will find out whether the responder scored within the top 25% on a respectable man or soft charactered woman scale. In the non-image relevant treatment arm, we inform the responder that the chooser has no additional information about the responder.
2) Timing of sound assignment: Experiment pairs are assigned to one of three treatment arms: i) ex-ante, the responder allocates the sound for each choice before the chooser selects a plate (i.e., the responder can still influence the plate selection); ii) ex-post cold, the responder selects the sound for each outcome after the chooser selects a plate, but before he/she finds out about the plate selection (i.e., the responder cannot influence the plate selection anymore but also cannot respond in a hot anger state); iii) ex-post hot, the responder assigns the sound after the chooser selects a plate and after finding out about the plate selection (i.e., the responder cannot influence the plate selection and may respond in a hot anger state).
3) Observability of transgression and sound assignment: We vary whether i) both transgression and sound assignment are private; ii) transgression is public to onlookers, but the sound assignment is private; or iii) both transgression and sound assignment are public to onlookers.

We will further test for heterogeneity by the income of the player randomly assigned as the “responder”.

Vignette experiment:

The vignette experiment aims to test the external relevance of the sound experiment for intimate partner violence and elicit norms and social acceptability of violence by income. We show vignettes in which either a high- or low-income husband instructs his wife not to do a certain action, which she might or might not follow. We elicit respondents’ first- and second-order beliefs about the likelihood and social acceptability of disciplining in the case that the wife does or does not follow the husband’s instruction. This will allow us to measure community-level norms about violence.
Experimental Design Details
Not available
Randomization Method
Spouses pull colored pieces of paper from a sealed bag to determine the matched pairs as well as the roles in the experiment. Pairs are randomly pre-assigned to treatment in Stata. Participants know that they can only be matched with someone from the other household (never with their own spouse).

We will stratify across unions, an administrative grouping of roughly ten communities. We will not stratify by income, as we do not yet know the income of the respondents in our sample. However, in the middle of data collection, we will test whether treatment assignment has been balanced by income thus far, and then decide whether to stratify going forward, using the income collected to date.
Randomization Unit
Matched pair.
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
Sample size: planned number of observations
We plan to survey 8,800 male respondents and 8,800 female respondents. In particular, we will randomly select 440 communities, and, within each community, we will select 20 households and survey both husband and wife.
Sample size (or number of clusters) by treatment arms
We expect the following distribution of couple pairings:
- Standard sample: female chooser, male responder: 90% (7,920)
- Male chooser, female responder: 3.3% (293)
- Female chooser, female responder: 3.3% (293)
- Male chooser, male responder: 3.3% (294)

Within the standard sample, we will randomize 88% of respondents (6,966) to the main experiment, and the remaining 954 respondents to a series of diagnostic tests and mini interventions.

Within the main experiment, respondents will be cross-randomized across the image relevance, timing, and observability treatment arms.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Institutional Review Boards (IRBs)

IRB Name
Stanford IRB
IRB Approval Date
IRB Approval Number
IRB Name
IRB Approval Date
IRB Approval Number
IRB Name
IRB Approval Date
IRB Approval Number