"Violence-related experiences and opinions in Mali: a list-experiment based analysis"

How to cite

Bertelli, Olivia; Mercier, Marion; Mesple-Somps, Sandrine; Lavallée, Emmanuelle; Calvo, Thomas. 2021. ""Violence-related experiences and opinions in Mali: a list-experiment based analysis"." AEA RCT Registry. AEARCTR-0008275. https://doi.org/10.1257/rct.8275-1.2

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

Olivia Bertelli* , Marion Mercier , Sandrine Mesple-Somps , Emmanuelle Lavallée , Thomas Calvo

* Primary investigator

Registry ID
AEARCTR-0008275
DOI
https://doi.org/10.1257/rct.8275-1.2
First registered
Last updated
Status
Completed
Trial dates
2021-06-20 – 2021-07-19
Countries
Mali (Sub-saharan Africa)
JEL codes
C81, C83, D74, O12
Registry record
http://www.socialscienceregistry.org/trials/8275

Abstract

We aim to conduct two list experiments in Mali. The first one addresses response biases affecting the reporting of preferences and experiences related to conflict in Mali. The second one addresses response biases affecting the reporting of men's experiences and attitudes towards gender-based violence. In the first one, we plan to conduct a phone survey to interview 1400 individuals (men and women) across Mali. The second one will be conducted with face-to-face interviews surveying men living in Bamako. In both experiments we randomly assign half of the sample to a survey List Experiment and the other half to a Direct Questioning survey. The comparison between the prevalence rates measured with the List Experiment and with the Direct Questioning will provide the size of the reporting bias present in our sample.

Intervention

In a list experiment, respondents are given a list of statements and are asked to declare how many of these statements they experienced (or agree with), without revealing which particular statements they experienced (or agree with). Respondents are randomly allocated to two groups. Control-group respondents are asked about a list of J so-called "baseline" items, while treatment-group respondents are asked about a list of J + 1 items, the additional item being the sensitive item that the researcher is interested in. Given the random allocation of respondents into the two groups, the prevalence rates of the baseline items should be equal across groups. The difference between the average number of items declared in the treatment group and in the control group thus provides an estimate of the prevalence of the sensitive item in the population.
Our List Experiment (LE) survey contains five list experiments in link with violence-related experiences and opinions. The Direct Questioning (DQ) survey contains five symmetrical direct Yes or No questions. The sensitive items of the LE survey are 1) "Over the last 12 months, me or a member of my household was physically assaulted outside home." 2) "Me or a member of my household owns a firearm." 3) "Me or a member of my household is ready to take up arms to defend our community." 4) "Nowadays, the army should rule in Mali". In addition, we will ask participants whether they trust international army forces in Mali.
Respondents assigned to the DQ survey will serve as control group. They will be administrated the same lists as the LE group, but without the sensitive items. The comparison of the response rates between the LE and the DQ groups will give us the measure of the violence prevalence rate in the sample.

Experimental design

See intervention

Design details

Randomization method
Assignment to the Treatment (List Experiment) group and to the Control group is randomly determined by the research team using the Stata software. Among the one-thousand men initially selected, 500 are assigned to the List Experiment and 500 to the Control group. The same is done separately for women. Each randomization is stratified on regions. All individuals in the replacement list are also randomly assigned to either groups, so that, if necessary, replacements are done within the same group. The replacement list is also stratified on gender and regions, as for the primary list.
Randomization unit
Individuals
Clusters
There are 12 regions. However, as described above, randomization is conducted at the individual level.
Observations
1400 individuals.
Arms
700 individuals assigned to the treatment group (List experiment) and 700 to the Control group. Equal proportion of men and women in both groups.
Primary outcomes
The main outcome is the response given by respondents to one of the five lists of items. The outcome variable will range between 0 and 3 for members of the DQ group and between 0 and 4 for members of the LE group.

Cited by

No citation data available for this registration yet.

Keywords