Leaders in Action (LIA) is a norms-centered intervention that aims to reduce the acceptance and prevent the incidence of Intimate Partner Violence (IPV) in Peru by shifting social norms. This project takes advantage of the randomization of LIA across 250 villages.

LIA has two delivery models: a household-based module (HT), consisting of household training sessions by Community Health Volunteers, and a group-based module (GT) with education sessions in small gender-segregated groups organized by trained facilitators. The investigators will cross-randomize each approach to assess efficiency in reducing domestic violence and changing social norms about tolerance toward violence and gender roles. The study disentangles the impact of the two modules separately, as well as the interaction of the modules, while explicitly addressing methodological concerns of previous studies: reporting bias from self-reported domestic violence, limited statistical power and lack of long-term effects measures.

Potential and actual victims of IPV may profit from the intimate atmosphere of household visits, and that on the side of women, the transmission of information about IPV and services for victims may be facilitated in more private settings. At the same time, group-level workshops about harmful gender stereotypes and gender norms for women should, through social interactions and norm change, reinforce the effects of household-level treatments for women. The experiment will shed light on the potential mechanisms at play and the theoretical framework underlying IPV through extensive data collection and the calculation of heterogeneous effects. The goal of this project is to deliver new rigorous evidence to the scientific and policy community by experimentally evaluating the impact of a state-run IPV intervention and its main components. It provides insights into the effectiveness of distinct program components, assesses cost-effectiveness as well as potential to scale, and evaluates the mechanisms leading to the reduction of IPV.

Leaders in Action (LIA) is a norms-centered intervention run by the Peruvian Ministry for Women and Vulnerable Populations (MIMP) that aims to reduce the acceptance and incidence of Intimate Partner Violence (IPV) in Peru by shifting social norms. LIA has two main implementation phases: first, the recruitment of Community Health Volunteers (CHVs) and second, the delivery of content against IPV. The delivery of the intervention is assisted by MIMP professionals working in local "Women Emergency Centers" (Centro de Emergencia Mujer, or CEM), a nationwide network of hubs run by the MIMP that offer wide-ranging support services to IPV victims. The CEMs also implement other IPV prevention interventions at the local level.

CEM's Promotors (MIMP's professionals working at local CEM) recruit leaders of local social organizations and train them to be Facilitators of LIA. Facilitators are the principal CHVs located in each district center who conduct LIA activities. Another type of CHVs are leaders at the village-level who are trained by Facilitators to become Community Agents (CAs). Both CAs and Facilitators work together to conduct the second phase of LIA: the delivery of the intervention through its two delivery models, the Household Treatment (HT) and the Group Treatment (GT).

Originally, the study was designed to only evaluate the HT. Thus, a first randomization was conducted on our study sample with an even split of 125 control villages and 125 treatment villages. The research team and Innovations for Poverty Action (IPA) Peru deployed a census to identify the population of women in sample villages who met the eligibility criteria for the study, and conducted an in-depth baseline survey on women who reported IPV in the census. The implementation of the HT began in 2020, and was interrupted by the Covid-19 pandemic. The MIMP paused the program in March 2020 until national lockdowns were rescinded in February 2022. Given the intense migration and other demographic changes that took place during the 23-month interruption, as well as the potential for new IPV exposure over that period, new census and baseline surveys were conducted in 2022, which were used to define the final targeted sample for the intervention.

Our study tests two different delivery modes of the program, the HT and the GT, as well as their combination, in a cross-randomized design. The 250 villages in our study sample were randomized into 4 treatment arms following a factorial randomization model. This resulted in 62 villages receiving only the HT (HT only), 62 villages receiving only the GT (GT only), 63 villages receiving both the HT and GT (HT+GT), and 63 villages in the control group. The randomization procedure was stratified by CEM, the level of IPV risk, and the 2018 treatment status (villages that had begun implementation of the HT were restricted to remain in the HT only or HT+GT treatment arms).

The MIMP's original HT design followed a door-to-door delivery approach: CHVs, in coordination with the local CEM, offered 8 treatment sessions to households at risk of IPV. These HT sessions took place over a period of 1 to 2 months and were conducted in participants' residences. While all adult household members were invited to participate, the recipients of the HT were, in practice, primarily women. As a result, this delivery mode may have been ineffective at influencing norms and beliefs of male participants. The household-centered approach is also expensive and difficult to scale up. Thus, together with the MIMP, the research team developed modifications to the HT which resulted in the innovations described below.

First, in collaboration with the MIMP, the research team and IPA Peru developed a telenovela-style series as a key component of the LIA program (edutainment component). The series consists of four videos that reinforce themes of relationship conflict, masculinity, and social norms around gender. The edutainment component was randomized for the HT only treatment arm, so that in 31 villages, all households targeted for HT watched the edutainment component as part of the program, and in other 31 villages, only 50% of the households targeted for HT watched the edutainment component. In this second group, the households selected to watch the edutainment videos were randomly selected, which will allow the research team to experimentally identify the effect of the edutainment videos. In the other treatment arms (GT only and HT+GT), the edutainment component is implemented for all participants.

Second, in order to strengthen the effect of LIA and to reach potential perpetrators, which are predominantly men, as well as victims of IPV, we developed a different delivery format with the MIMP: the GT. The GT follows a gender-segregated yet community-wide delivery approach in a group discussion format. Existing evidence suggests that interventions which target both perpetrators and victims can have a substantial impact on attitudes towards and the incidence of IPV, and that simultaneously addressing IPV from different angles can be particularly effective (Chakraborty, P., Osrin, D. and Daruwalla, N., 2020; Abramsky, T. et al., 2014). These approaches commonly stress the importance of collaborative learning in support-group-style sessions to confront ideas of masculinity and gender norms. Moreover, norms change is more likely to happen when individuals are aware of relevant peers who are also changing their beliefs (Bursztyn, L., González, A. L., Yanagizawa-Drott, D. (2018)). As a result, the investigators proposed separating groups by gender to provide tailored settings during intervention.

• This group-based approach entailed 4 GT workshops that integrated various activities and group discussions focused on the same content as the HT. The group workshops directed at men had a slightly different curriculum than those directed at women, and always had at least one male CHV in the room facilitating the workshop. In order to meet this criteria during implementation, the MIMP required a larger number of male CHVs than those who were initially recruited through social organizations; men are usually recruited in disproportionately low numbers for this style of volunteer work, and the initial number of male CHVs was not sufficient for implementation of the GT. As a result, male CHVs were recruited as program Facilitators from local universities, communal associations, and other frequently-transited areas. GT sessions took place over the course of one month in village community centers, schools, churches, or other communal spaces where village residents would often gather, and lasted for approximately two hours. Larger villages held multiple GT sessions per workshop to cover all potential participants, and some villages were re-visited some months after the expected implementation time to ensure a high uptake.
• Participant targeting during the GT was the same as in other treatment arms: MIMP CHVs approached the houses of women in the targeted sample, and extended an invitation for them and their male partners to participate in the intervention. However, given that the GT is designed to involve the village community at large, invitations for the GT were extended to all adults in the village. Households in the targeted sample were approached door-to-door multiple times in the week before the GT session, while the rest of the village households were only approached once per workshop. Aside from door-to-door recruiting, the whole village received a notice an hour before each GT session through a loudspeaker usually used for other village-wide activities.
• The topics covered during both the HT and GT were equivalent to ensure comparability. These topics were: i) gender roles, beliefs and stereotypes; ii) violence, cultural patterns and human rights; iii) healthy relationships within the family; iv) respect between family members and self-care; v) assertive communication; vi) resolution and conflict management; vii) resources for domestic violence cases; and viii) leadership and women's agency.

Third, 50% of the villages assigned for either the HT only, GT only, or HT+GT treatment arms (94 villages total) were randomly selected to target one village leader (e.g. village president) to receive treatment given the village's treatment assignment. This will evaluate if community leaders are impactful in shifting norms at the village level.

Fourth, we randomized 50 percent of the treated villages to target a random selection of 30% of women who were interviewed in the census but did not report ever experiencing any form of IPV by their current partner. This will allow the research team to measure community-level impacts of the intervention through treatment density.

After the conclusion of both the HT and the GT, a set of qualitative interviews were conducted with the goal of informing the research team on the channels through which the treatments could be reducing IPV, as well as capturing information relevant to the design of the endline survey.

The research team plans to conduct an endline survey 6-months after the MIMP's implementation of the HT and GT to evaluate the impact of the LIA program and its different innovations on outcome measures such as: IPV, physical health, mental health, tolerance towards IPV, female agency, and social norms.

Eligibility to receive the initial census: Women between the ages of 18-59 in 2022, who lived with a male partner, and who were permanent residents in villages from our sample (defined as living in the village for at least 4 months of the year).

Eligibility to be part of the targeted sample:

• Women identified to be at risk of suffering from IPV in the census, who as a result, also received a baseline survey.
• In 50 percent of the treatment villages chosen at random, an additional 30% of women randomly selected from the census.
• Male partners of women in the targeted sample.
• Across a sub-sample of 94 treatment villages chosen at random, 1 village leader. Specifically, we identified two village leaders, one male and one female, and randomly selected which of the genders was selected for targeting. This resulted in 50 women leaders and 44 men leaders, and their partners.

Eligibility for the endline survey sample:

• We plan to interview all men, women, and village leaders in our targeted sample.
• We will also interview 4 additional women per village who were eligible to receive the initial census but were not part of the targeted sample, as well as their male partners, to measure spillover effects.