close× Call Us
close×
Title Excessive Alcohol Consumption and Violent Brawls: A Randomized Controlled Trial with Bartenders in Bogotá, Colombia
Post date 03/07/2019
C1 Background and Explanation of Rationale

Excessive alcohol consumption remains a pressing concern for public policy: it is associated with a host of physical maladies, psychological disorders, as well as the commission of crimes. A recent meta-analysis found that 3 million deaths in 2016 were attributable to alcohol, not including those related to alcohol-attributable violence. Global alcohol-attributable violence is estimated to account for 248,000 annual deaths, and observational studies frequently find alcohol consumption to be positively correlated with homicides. Despite existing evidence suggesting such a link, few studies use experimental methods to convincingly establish causal relationships between changes in alcohol consumption and crime.

In this Pre-Analysis Plan we describe the proposed evaluation of a randomized controlled trial that studies the impact of an intervention directed at bartenders in four neighborhoods or localities in Bogotá, Colombia. The goal of the study is to determine whether alcohol-induced violent brawls can be reduced via a low-cost, scalable intervention undertaken through a public-private partnership. The intervention has been designed by Bogotá's mayor's office, Colombia's largest brewery (Bavaria), and the team of researchers. Its objectives are to empower bartenders, provide didactic materials to bar patrons, and offer food and non-alcoholic drinks to bar patrons.

C2 What are the hypotheses to be tested?

H1: Treated street segments located within treated police quadrants will experience fewer violent brawls than either untreated street segments located within treated police quadrants or untreated street segments located within untreated police quadrants.

H2: The intervention will not displace violent brawls toward street segments geographically close to treated street segments.

H3: The intervention will produce the largest effects on treated street segments where sales of water and food in bars have increased the most.

H4: The intervention will not change alcohol sales.

C3 How will these hypotheses be tested? *

We will employ two sources of data for the evaluation: i) administrative records on brawls in and around bars and ii) survey data for bars collected after the intervention.

The administrative data consists of brawl reports from 2014 onward collected by the Security and Justice Secretariat of the city of Bogotá, which we have referred to as the NUSE database. We will also use records from the Statistical, Criminal, Contravention and Operational Information System (SIEDCO) from the National Police that measure other violent events in the city. Merged with client records from Bavaria, we can measure reported brawls in and around the street segments where the bar is located, given that both sources are geocoded. These data will provide measures for a number of primary outcomes: 1) violent brawls per police quadrant; 2) violent brawls per street segment; and 3) other violent events in both street segments and police quadrants in Bogotá.

Given the available administrative data, we will estimate the effects of the intervention on brawls using a difference-in-differences specification.
A survey instrument was designed to measure the mechanisms that could drive potential changes in reported brawls, including a) the value of all alcohol sold in different temporal windows, b) the value of water sold in different temporal windows; and c) the value of food sold in different windows. Via this endline survey instrument, we will gather data on the bar (e.g. years of operation; hours of operation; whether it has a bathroom for clients; how many tables and chairs; whether private security is provided; the kind of alcohol served; whether non-alcoholic beverage options are available; the kinds of food served; sales of alcohol in recent months; sales of water in recent months; and sales of food in recent months), the bartender (e.g. whether s/he is the owner, whether s/he lives in the neighborhood, education level, number of years as a bartender), and the characteristics of the clientele (regulars or not; gender; age; number of clients during the previous weekend; whether clients pay after each drink or accumulate a "tab"). We also will ask bartenders about the frequency of disagreements within their bars, how these disagreements are resolved, and pose a series of scenarios related to the potential causes of violent brawls and how bartenders would resolve them. Given that survey data will only be available after the treatment has taken place, we will study the effects of the intervention using a cross-section post specification.

C4 Country Colombia
C5 Scale (# of Units) 221 police quadrants, 2539 street segments
C6 Was a power analysis conducted prior to data collection? Yes
C7 Has this research received Insitutional Review Board (IRB) or ethics committee approval? Yes
C8 IRB Number This research was approved by the Comité de Ética de la Universidad de los Andes via Acto 970 de 2019.
C9 Date of IRB Approval March 7, 2019
C10 Will the intervention be implemented by the researcher or a third party? LOWE
C11 Did any of the research team receive remuneration from the implementing agency for taking part in this research? No
C12 If relevant, is there an advance agreement with the implementation group that all results can be published? No
C13 JEL Classification(s) C93, D74, K42, D90