Street Police Patrols and Gender-Based Violence in Public Spaces: Experimental Evidence from Urban India

Last registered on December 02, 2019


Trial Information

General Information

Street Police Patrols and Gender-Based Violence in Public Spaces: Experimental Evidence from Urban India
Initial registration date
December 02, 2019
Last updated
December 02, 2019, 2:52 PM EST



Primary Investigator

ifo Institute

Other Primary Investigator(s)

PI Affiliation
Hyderabad City Police
PI Affiliation
Princeton University
PI Affiliation
University of Connecticut
PI Affiliation
University of Connecticut
PI Affiliation
World Bank

Additional Trial Information

On going
Start date
End date
Secondary IDs
How can gender-based violence in the public sphere (GBV) be prevented? Can improved police presence help curb street harassment? What works in improving victim’s engagement with police services? This project aims to answer these questions through a novel policing program in Hyderabad, India. The Safety, Health, and Environment (SHE Teams) Program is a hotspots street police patrolling intervention targeting GBV in public spaces. The researchers and Hyderabad City Police have jointly developed research that aims to test the role of increased police presence through patrolling and policing visibility (i.e., uniformed vs. undercover officers). Our research will address fundamental questions in the economics of crime and gender.
External Link(s)

Registration Citation

Amaral, Sofia et al. 2019. "Street Police Patrols and Gender-Based Violence in Public Spaces: Experimental Evidence from Urban India." AEA RCT Registry. December 02.
Sponsors & Partners

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

Request Information
Experimental Details


This study seeks to answer the following two main research questions:

1. What is the effect of GBV-targeted street patrolling on the frequency and type of street harassment incidents?
2. What drives these changes? Visible state presence and quantity of focused task- police force?

To address these research questions, we will make use of a clustered experiment to evaluate the efficacy of different forms of policing in curbing GBV crimes. First, we evaluate the effect of GBV-targeted street patrolling on the type and frequency of street sexual harassment (SH) and women’s proactive response through increased police presence. Second, we evaluate the effects of police visibility in reducing the incidence of GBV crimes. The intervention involves two treatment arms – Visible, patrolling by SHE Teams in uniform and Undercover, patrolling by SHE Teams in civvies. The research design will allow us to disentangle the direct deterrence effects generated by police presence separately from an additional impact of visible specialized police presence. In the context of GBV per se, the “police visibility” component has not been studied, and it becomes even more relevant given the low access women have to police services. Lastly, we also seek to evaluate what forms of policing shift women’s beliefs and choices regarding access to police. There is limited evidence on how improving policing towards GBV crimes may impact citizens’ engagement with the police, victimization, and detection of such crimes - key accountability outcomes for the police. Understanding such issues are a crucial steppingstone to addressing GBV, improving the accountability of the police, and changing citizen’s engagement with one of the essential components of governance, and this is the focus of our research project.
Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
• Incidence of street sexual harassment: We capture data on the incidence of street sexual harassment through an Enumerator Observation Survey. We ask enumerators to observe a hotspot for 15 to 20 minutes every day for a duration of six months and record all instances of street sexual harassment observed. Given the social stigma attached with victimization of sexual harassment, we may expect women to underreport their experience of street sexual harassment. Therefore, to circumvent this problem, we will use the ‘observed’ measure of the incidence of sexual harassment as captured by the Enumerator Observation Survey (EOS). The EOS provides us with a hotspot-level incidence of sexual harassment.
• Safety Perception: Lastly, we seek to study the effect of street patrolling on safety perception of women. We will obtain this measure from the victimization survey. Women are asked the following questions to arrive at a comprehensive measure of safety perception:
– How safe they feel at the hotspot?
– How safe they feel about traveling anywhere in the city between (i) 6 AM and 12 PM; (ii) 12 and 4 PM; (iii) 4 and 10 PM
Responses are collected on a four-point scale: ‘Completely Safe’, ‘Somewhat Safe’, ‘Somewhat Unsafe’ and ‘Completely Unsafe’. Through this variable, we will address whether women report feeling safer in areas patrolled by the police.
• Mobility: As a part of women’s proactive responses, we would like to see if increased police presence leads to increased mobility of women. Here, we define mobility as the footfall of women at a hotspot at any given point of time during the day. To measure the footfall of women, we will leverage the dense network of CCTV camera of the Hyderabad City Police. Making use of artificial intelligence and machine learning, we would arrive at a measure of footfall of women at the hotspot as captured by the CCTV cameras.
• Routes taken to work: In the victimization survey, women are asked about the routes they take to reach their workplace/colleges from their homes. Safety concerns may lead women to take inefficient routes (in terms of time/ distance and cost) to reach their workplace or colleges. Through this data, we can analyze whether women alter their routes to workplace/colleges in response to increased patrolling.
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Apart from the type and frequency of street sexual harassment, street patrolling may also affect other aspects of women like reporting behavior, participation in social activities or travel routes taken by altering their safety perception about locations. Therefore, we will look at the following variables.
• Change in type of harassment: We hypothesize that increased street patrolling will lead to a decrease in the incidence of street sexual harassment. However, we would also like to see if increased patrolling has differential effects on different forms of harassment experienced by women. Data from the EOS will be used to analyze the change in the type of harassment observed at the hotspot, if any.
• Reporting of incidence to police or other changes in pro-active behaviors: Findings from the baseline victimization survey indicate that the reporting rate of street sexual harassment is only 7.5 percent. Increased police patrolling, especially in the visible arm will increase the visibility of police. We thus hypothesize that reporting rates of GBV-related crimes may increase with increased police visibility as women may find it easier to approach police. We intend to test this hypothesis using data on reporting rates from the victimization survey. To measure other proactive behaviors, we will construct - through the observational data - a measure as to whether victimized women react more proactively. This will be done through question answers (1/0) to whether ”Does the victim...” did any of the following: i) ”called someone over the phone to help”, ii) ”called out the perpetrator directly”, , iii) ”informed the person accompanying her at the location”, iv) ”use any form of self-defense”, or v) ”ask for help from bystanders”.
• Incidence of harassment while traveling: As a part of the EOS, enumerators also make observations on sexual harassment while traveling between two hotspots, i.e., in public transport like buses, shared autos, trains, etc. This enables us to analyze whether the effect of hotspot-level police patrolling on incidence of sexual harassment is local, i.e., it reduces sexual harassment only at the hotspot-level, or whether it reduces sexual harassment faced by women in public transport as well.
• Labor Market Choices: To measure if increased police presence leads women to modify their labor market choices, we will analyze the ‘hours’ spent at workplace. This variable will be obtained from the victimization survey where women are asked to report their time of arrival/departure to/from the workplace.
• Participation in social activities: An improvement in safety perception of public places may induce women to step out more frequently and may also alter the time at which they are present in public places. For example, street police patrolling after in late evening may enable women to step out in late evening because a sense of safety induced by police presence. To record this, as a part of the Victimization survey we ask women whether they engage in any of the following social activities:
– Going out for a movie with female friends – Shopping for rations
– Attend recreational classes
– Visit relative’s home
– Got out to explore the city
Further, we also ask them about the time at which they carry out each of these activities.
Through the baseline and endline victimization survey, we will capture the change in the participation in activities as well as the time at which these are conducted.
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
We use a clustered experiment to identify the effects of different approaches to policing. The unit of intervention for this study is a hotspot. Hotspots are public places like schools, colleges, bus stops or marketplaces where the incidence of sexual harassment is relatively high. The research team in collaboration with the HCP identified 350 hotspots across Hyderabad which were randomly distributed across three groups. The first group – Treatment arm 1 or the Undercover arm consists of 100 hotspots, which receive undercover patrolling, i.e., patrolling by SHE Teams in civilian clothes. The second group – Treatment arm 2 or the Uniform arm consists of 100 hotspots, which receive patrolling by SHE Teams in their police uniform. The third group is the control arm which consists of 150 hotspots which receive no patrolling through the course of the intervention i.e. receive the typical police patrol any other area would receive. Under the two treatment arms, each SHE Team is comprised of three officers, with at least one female officer.
Experimental Design Details
Randomization Method
A stratified clustered randomization was done for the 350 hotspots across the city of Hyderabad. The stratification was done on the following basis:
• Nature of hotspot: Hotspots were categorized into four types – Educational (schools and colleges), Commute (bus stops and railway stations), General public places (markets and temples) and Residential – as the hotspots would differ in their composition of population (age and gender) basis their nature; and
• Population density: Hotspots were categorized into four categories – low, medium, large and very large – on the basis of the footfall at the hotspot. This stratification parameter was used because frequency of sexual harassment may be affected by how densely or sparsely occupied a location is. A footfall of less than 30 was termed as low; a footfall between 30 and 150 was termed as medium; between 150 and 400 as large and anything above 400 as very large.
A total of 2,000 iterations were done on a total of 57 key variables using data collected at baseline to arrive at a successful randomization. The three groups were balanced across observable characteristics of women like age, education, occupation, marital status, general safety perception and safety perceptions at various parts of the day. The randomization was implemented on Stata.
Randomization Unit
Randomization at the hotspot-level (cluster).
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
350 hotspots.
Sample size: planned number of observations
The estimated sample size contains information from a survey of 13,000 women.
Sample size (or number of clusters) by treatment arms
The estimated sample size contains information from a survey of 13,000 women.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Supporting Documents and Materials


Document Name
PAP_Amaral, Borker, Fiala, Kumar and Sviatschi
Document Type
Document Description
PAP_Amaral, Borker, Fiala, Kumar and Sviatschi

MD5: e503ac48d1a9da25b856d3ba42473fcd

SHA1: a8282c07884d92671d26dfea9795c363b41005c9

Uploaded At: December 02, 2019


Institutional Review Boards (IRBs)

IRB Name
University of Princeton
IRB Approval Date
IRB Approval Number


Post Trial Information

Study Withdrawal

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

Request Information


Is the intervention completed?
Data Collection Complete
Data Publication

Data Publication

Is public data available?

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials