An Experimental Evaluation of Hot Spot Policing in Brazil

Last registered on May 17, 2023

Pre-Trial

Trial Information

General Information

Title
An Experimental Evaluation of Hot Spot Policing in Brazil
RCT ID
AEARCTR-0011108
Initial registration date
May 12, 2023

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
May 17, 2023, 2:45 PM EDT

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

Locations

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

Request Information

Primary Investigator

Affiliation
Getulio Vargas Foundation

Other Primary Investigator(s)

PI Affiliation
IADB
PI Affiliation
Universidad Austral
PI Affiliation
Getulio Vargas Foundation

Additional Trial Information

Status
On going
Start date
2023-03-27
End date
2024-03-31
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
This project is the first experimental assessment of preventive patrolling in Brazil. We partnered with Parana Military Police to design and test the effects of hot spots policing on crime in Curitiba, Brazil. We identified 557 hot spot street segments and randomly assigned 211 to receive more police visits over the intervention period and will compare this treatment effect against business-as-usual patrols at hotspots. It is hypothesized that more time spent at hotspots (i.e., 30 minutes) with two or more visits per day will cause more reduction in robberies and other crimes that take place on streets. We also use non-experimental streets to test for spillovers onto non-hot spots and examine aggregate effects citywide. This study will complement the scant literature on hotspot policing in Latin America which so far has shown limited impact. The intervention has started on March 27th 2023 and is planned to last until July 2023.
External Link(s)

Registration Citation

Citation
Fagundes, Eduardo et al. 2023. "An Experimental Evaluation of Hot Spot Policing in Brazil." AEA RCT Registry. May 17. https://doi.org/10.1257/rct.11108-1.0
Experimental Details

Interventions

Intervention(s)
We will run a hot spot policing program in the city of Curitiba, in Paraná state, Brazil, as part of a hotspot initiative promoted by IADB.

We will test both the effectiveness and efficacy of saturated policing plus POP interventions at hotspots: a randomized controlled trial in which treatment conditions will receive more police visits over the intervention period – while comparing this treatment effect against business-as-usual patrols at hotspots. It is hypothesized that more time spent at hotspots (i.e., 30 minutes) with two or more visits per day would cause more reduction in robberies: a dose–response relationship between police presence and the level of crime at the hotspot.

The officers were instructed to be visible and outside of their cars during patrol time and were encouraged to perform their law enforcement duties as normal (e.g. only to perform stops and frisks if there was reasonable suspicion to do so) and to talk to people while they were on patrol.
Intervention Start Date
2023-03-27
Intervention End Date
2023-07-31

Primary Outcomes

Primary Outcomes (end points)
Number of robberies
Primary Outcomes (explanation)
Number of robberies defined according to Brazil penal code: the action of taking property unlawfully from a person or place by force or threat of force). It does include all registers, regardless the location and objects that were stolen.

Secondary Outcomes

Secondary Outcomes (end points)
- Index of any criminal events that take place in public space and subdivide this index into property and violent crimes
- Motor vehicle thefts and robberies
- Illegal possession of firearms
- Assaults in public space
- Robberies inside buildings (commercial or residences)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
Hotspots segments were identified as the one in which at least three robberies (top 1% of segments distribution) were registered in a twelve-month period and meet one of the following conditions:

• It is part of a statistically significant spatial clusters, calculated through Getis-Ord Gi* statistic
• It is part of an emerging hotspot, according to Space Time Pattern Mining ArCGIS tool
• It is part of a cluster, defined using Nearest Neighbor Hierarchical Clustering

This leads to the identification of 557 hot segments.

We used a two-stage randomization to maximize similarity between treated and control segments. We first divided segments into five categories using Jenks natural breaks to determine five different levels of crime density (number of crimes divided by street length). Second, we assigned segments to treatment and control groups by category. We start by randomly selecting one segment to treatment group and then select a second one to control group. Street segments within 200 m of a treatment unit were excluded from the pool of potential controls to minimize contamination effects from treatment areas.

This randomization procedure assigned 214 hot segments to treatment and the same number to control group. These 428 segments account for 27.7% of the city’s reported robberies. To examine spatial spillovers, we also analyze the 7646 street segments that lie within 200 meters of the 428 segments.

We then aggregated treatment segments into groups in order to create clusters to receive patrol. This procedure led to the creation of 27 hot spot patrol routes. Often, these hot segments were not coterminous but were located close to each other. Patrol routes were designed by the police and avoid control segments.
Experimental Design Details
Not available
Randomization Method
Randomization was done in Python
Randomization Unit
Street segment
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
Total sample size is comprised by 7346 street segments. The treatment is not clustered.
Sample size: planned number of observations
We have two samples. An experimental sample of 422 hot segments and a larger sample of 6924 street segments that we will use to assess spatial spillovers.
Sample size (or number of clusters) by treatment arms
The randomization procedure assigned 211 hot segments to treatment and the same number to control group. These 422 segments account for 27.7% of the city’s reported robberies. To examine spatial spillovers, we also analyze the 6924 street segments that lie within 200 meters of the 422 segments.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Using specification detailed in the analysis plan and baseline data, we are powered to detect ITTs of approximately 0.20 standard deviations relative to the control group and spillover effects of approximately 0.10 standard deviations.We also calculate the MDE using ordinary least squares and find similar magnitures, MDEs of respectively 0.18 and 0.11 standard deviations. Importantly, we do not adjust for covariates when simulating MDEs, although we do believe we will be able to gather neighborhood and segment-level data at the time of the evaluation. Therefore, we interpret these MDE as upper-bound effects that we will be able to detect.
IRB

Institutional Review Boards (IRBs)

IRB Name
IRB Approval Date
IRB Approval Number
Analysis Plan

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

Request Information