Street Violence and Media Empowerment: Evidence from Guatemala

A first component focuses on individual behavior among at-risk youth. In this case, the source of variation is randomized. A sample of 1200 participants will be recruited from at-risk youth from the three most violent municipalities in the Department of Guatemala: Ciudad de Guatemala, Mixco and Villa Nueva. Participants will be recruited for the participation in a nation-wide media campaign. Participants will be randomly allocated to one of the following groups:
1. Give voice (T1). These participants will be invited to attend a meeting in which an individual video will be recorded for the inclusion in the media campaign. The participant will be given the opportunity to record his vision for the future of Guatemala and what he would like to improve in the city. The video will not be included in the nation-wide media campaign.
2. Give voice and share the voice (T2). These participants will be invited to attend a meeting in which an individual video will be recorded for the inclusion in the media campaign. The participant will be given the opportunity to record his vision for the future of Guatemala and what he would like to improve in the city. The video will be edited and will be included in the nation-wide media campaign. A team of media experts from working with UNDP’s Ciudades Conectadas Foundation, will support the research team in editing the material into a media material that can be shared on different channels (including electronic billboards in the city of Guatemala, metro-bus lines, radio, tv and social media). The media campaign will last for one year, and its focus aims at being impactful and safe to share. The media team will frame the T1 and T2 interventions in order to avoid messages that could be in direct confrontation with gangs.
3. Control group (T0). Participants in this group will not be selected for any activity.
Comparing T0, T1 and T2 will allow researchers to disentangle the effects of media exposure to a pure empowerment effect. In addition, within group T1 and T2, we will cross-randomize two types of content in the expression of the participant. One content related to self-presentation, which highlights the personal side of the participant, and one content related to community vision, which highlights the public side of the message. This will provide additional evidence on which content is more effective in reducing violent attitudes and increase individuals’ sense of belonging.

A second component focuses on overall violence by exploiting exposition to the media campaign created in the first component. The campaign is designed to reach a large number of individuals through broadcast on radio, electronic billboards, and social media. This component makes use of administrative data about homicides and exploits two sources of variation to identify the effect of exposure to the media campaign on violence. A first source of variation exploits semi-random exposure to the media campaign from radio and television using topography-corrected coverage as Enikolopov and Petrova (2011), Yanagizawa-Drott (2014), Olken (2009), Adena, et al. (2015), Della Vigna, et al. (2014), and Armand, Atwell and Gomes (2019).

Researchers will make use of the micro-targeting feature of Facebook paid content to randomly target different geographical areas of the country, therefore providing exogenous variation in exposure to the media campaign. In Guatemala, this social media platform is used by most youths. Using Guatemala’s age structure from the Central Intelligence Agency (2019) and Facebook (2019) data on users’ demographic profile, researchers estimate penetration in the 18-24 demographic between 73 and 85 percent.

2020-04-01

2021-04-01

Violent behavior, sense of belonging, aspirations, homicide rates

The evaluation design of the first component is a standard individual-level randomized controlled trial.

The evaluation design of the first component is a standard individual-level randomised controlled trial. The component aims to estimate the impact of empowering at-risk youth the impact of media in changing violent attitudes and sense of belonging. The first component of the project aims to estimate the impact of empowering at-risk youth the impact of media in changing violent attitudes and sense of belonging. It will focus on the following outcomes:
1. A survey of participants including thorough questions on their sense of belonging in their community, and verbal and physical altercations during the past two weeks, and aspirations. Demographic and socioeconomic data will also be collected allowing for a full description of the main characteristics of these individuals, their households, their network and their perceptions. Researchers will create indices of participants’ sense of belonging and propensity to engage in violent behavior from self-reported data. Information will be collected at three points in times: at the time of the registration as participant, 2 months following the intervention and 1 year following the intervention.
2. Behavioral measurements with at-risk youth measuring violent attitudes which will be piloted at first to test their informational value.
The second component of the project will estimate the impact of a national media campaign on administrative data about violence. Administrative data on homicides occurring across Guatemala which is to be provided by the Guatemalan Statistical Office, INE. See attached letter of support.
Target population: For the first component, the unit of analysis for individual experiments are 1200 at-risk youths between 16 and 21 years old from three municipalities: Guatemala, Mixco and Villa Nueva. These are the three municipalities in Guatemala with the highest murder rate. Researchers choose Mixco and Villa Nueva as they have seen recent growth in numbers of mareros due to its lack of basic social services and low-income background of several households (Clavel, 2019). The presence of mareros in Guatemala City is well documented and a driving force of the largest homicide rate in the country (Dudley, 2016; Muggah & Aguirre Tobón, 2018). For the second component, the unit of analysis is the administrative unit. Depending on the degree of granularity provided the INE, we will perform analysis at the level of the municipality (340 in total) or below this level (e.g. zonas).

Stata code

Individual

No

N/A

1200 at-risk youths between 16 and 21 year

400 individuals per treatment arm

This study proposes to have 1200 at-risk youth between the ages of 16 and 21 to be randomly allocated into one of three groups (T0, T1, T2). With 80 percent of statistical power and an alpha of 0.05, we could capture a minimum detectable effect of 0.2 of a standard deviation (as standard in randomized experiments) with 394 individuals per group. We therefore decide to have 400 individuals in each comparison group, which provides a power of 81 percent (alpha = 0.05). With cross-randomization, statistical testing could extend to 5 groups (T0, T1, T2 and cross-randomization within T1 and T2). This is in addition to test cross-randomized groups, for which previous calculations also apply. In this case, we would have 240 individuals in each comparison group. With 80 percent of statistical power and an alpha of 0.05, this allows capturing a minimum detectable effect of 0.26 of a standard deviation.