Information provision, police beliefs and public trust: Experimental evidence from Colombia

We are testing two interventions: (i) intensified interactions with citizens plus procedural justice instructions; and (ii) intensified interactions with citizens plus procedural justice instructions, with the addition of an information campaign on procedural justice values. In our context, procedural justice refers to the idea of fairness in police decisions and interactions with citizens. Our units of analysis are police quadrants or patrolling areas, akin to police beats.

The first treatment arm consists of increasing police presence and the quality of interactions with citizens within the patrolling area. Police patrols will be instructed to intensify their presence and interactions, with specific instructions on the importance of police contact with citizens. Following the beginning of the intervention, patrols in this treatment arm will receive daily messages on procedural justice protocols, e.g., to be fair and to listen to citizens' complaints and questions.

The second treatment arm consists of exactly the same components as the first with the addition of an information campaign where police officers received daily reminders on the importance of procedural justice principles. These messages highlight, for instance, cordially greeting people before starting any dialogue or being empathetic with whoever they interact with.

Because patrolling areas are relatively large, we collect data on one street block within the quadrant, and ask police patrols to target it as a "pilot" block. We randomly sampled "pilot" street blocks within quadrants.

Note: We updated the intervention description after the intervention finished, but not as an ex-post change. The intervention evolved after we registered the PAP, and this is a more accurate description of what the intervention effectively was.

2022-03-21

2022-04-22

We have three main outcomes:

1. Public trust in the police
2. Demand for policing services
3. Police beliefs

We have three main outcomes:

1. Public trust in the police, measured through one single public trust question in the citizens survey

2. Demand for policing services, measured through a costly request (we asked residents whether they would support a new tax directed at funding the police)

3. Police beliefs, measured as their trust in citizens and their second-order beliefs on public trust

We expect the intervention will increase public trust and police beliefs.

Note: We updated the description of the police measures after the intervention finished, but not as an ex-post change. The instrument aimed to understand how effective procedural justice principles were by asking directly to officers how much they trust the public (a measure of reciprocity) and their second-order beliefs on public trust (the main aim of the intervention).

Our secondary outcomes are:

1. A first stage measure of police activity
2. Four indices to better understand the mechanisms that change public trust in the police: procedural justice, perceived effectiveness, perceived accountability and transparency, and convergence of values

We will measure impacts on two secondary outcomes:

1. A first stage measure of police activity, measured as an additive index of a set of questions on police presence and behavior, asked in the citizens survey

2. Four indices to better understand the mechanisms that change public trust in the police, all measured in the citizens survey
2.1. Procedural justice, measured as an additive index of a set of questions on police behavior
2.2. Perceived effectiveness, measured as an additive index of a set of questions on perceived effectiveness of the police
2.3. Perceived accountability and transparency, measured as an additive index of a set of questions on perceived accountability, corruption and other transparency outcomes
2.4. Convergence of values, measured as an additive index of a set of questions on values and the idea of police legitimate use of force

We do not have priors on the directions of these secondary outcomes.

The study focuses on five Colombian cities: Barranquilla, Bucaramanga, Cali, Cartagena and Medellín---these are the largest Colombian cities excluding the capital Bogotá.

For our experimental sample, we randomly chose 351 quadrants out of the universe of police quadrants across the five cities---we excluded from the sampling frame only exceptional cases such as quadrants in a Caribbean island nearby Cartagena. From the experimental sample, quadrants will be assigned to each treatment arm using stratified randomization. Probability of assignment to each treatment arm is 33%.

Randomization was done in an office using a computer. We assigned treatment as follows:

First, we stratified by city. Within each city, we stratified by terciles of poverty scores using census data. Within each tercile of poverty score, we ranked quadrants based on the baseline public trust measure. Within each triplet in the ranking, we randomized quadrants. Probability of treatment assignment is 33% for all treatment arms.

Our units of analysis are police quadrants or patrolling areas, akin to police beats. These are discussed above.

No

351 police quadrants.

About 2,100 surveys, 6 from each police quadrant. We will cluster observations by quadrant in our analysis.

~351/3 police quadrants assigned to intensified interactions with citizens
~351/3 police quadrants assigned to intensified interactions with citizens plus procedural justice instructions
~351/3 police quadrants assigned to control

We used our baseline measure of public trust and ran 10,000 simulations to estimate the minimum detectable effects. For each randomization, we estimated treatment effects using a regression of a z-score of the public trust measure on indicators for both treatments, and fixed effects for city, income tercile and trust triplet. Our minimum detectable effects are of roughly 0.1 standard deviations for both treatments.