Please fill out this short user survey of only 3 questions in order to help us improve the site. We appreciate your feedback!
Examining Poverty and Anti-Social Behavior in the Lab
Last registered on August 03, 2015


Trial Information
General Information
Examining Poverty and Anti-Social Behavior in the Lab
Initial registration date
August 03, 2015
Last updated
August 03, 2015 1:59 PM EDT
Primary Investigator
UC San Diego
Other Primary Investigator(s)
PI Affiliation
Busara Center for Behavioral Economics
PI Affiliation
Princeton University
Additional Trial Information
Start date
End date
Secondary IDs
This paper describes the analysis plan for a randomized experiment examining the effect of poverty on anti-sociality. This study aims to identify the extent to which poverty, through individual cognition and decision making, can influence behavior typically classified as anti-social (dishonesty, aggression, etc.). We simulate poverty in the lab by inducing a priming effect which makes salient psychological states associated with poverty. We observe anti-social behavior with a battery of tasks and questionnaires selected to obtain a broad measure of anti-sociality. We conducted this experiment with a sample of 200 respondents from the Kibera and Viwandani slums, two of Nairobi's largest informal settlements that suffer from both economic hardship and violent crime. This plan outlines our outcomes of interest and econometric approach.
External Link(s)
Registration Citation
Abraham, Justin, Johannes Haushofer and Jeremy Shapiro. 2015. "Examining Poverty and Anti-Social Behavior in the Lab." AEA RCT Registry. August 03. https://doi.org/10.1257/rct.791-1.0.
Former Citation
Abraham, Justin et al. 2015. "Examining Poverty and Anti-Social Behavior in the Lab." AEA RCT Registry. August 03. http://www.socialscienceregistry.org/trials/791/history/4893.
Experimental Details
Our study utilizes a methodology developed by Mani et al (2013) and adapted to our Kenyan sample to identify the psychological effect of poverty on anti-social behavior in the lab. We presented three hypothetical scenarios to the respondents, which describe a financial problem they might experience. The primes are described in detail in the appendix. Respondents are given 5 minutes per scenario to contemplate about how they might deal with these problems. These scenarios, by touching on financial issues, act as primes that trigger thoughts of the respondent's own economic situation.
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)

Poverty primes (randomly assigned treatment) Cantril Self-Anchoring Ladder
(a) Current life
(b) Life five years from now
Ring Task
Noise Aversion Task
Coin Toss Game
Maudsley Violence Questionnaire
Buss-Perry Aggression Questionnaire (Short)
Primary Outcomes (explanation)
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
Our main specification of interest is the OLS analogue to the analysis conducted in Mani et al. (2013):
yi =β0 +β1Ti +β2Richi +β3(Ti ×Richi)+εi (1)
We collect detailed asset ownership information from our respondents and use this information to construct an objective measure of wealth. We construct a weighted asset index and define the dummy variable Richi = 1 if the respondent is above the median of this index. We construct a weighted asset ownership index following the procedure in Anderson (2008). In addition, we will run a basic treatment effects specification to capture the impact of treatment relative to control:
yi = β0 + β1Ti + εi (2)
where yi is the outcome of interest for respondent i. Ti is a treatment indicator that takes the value 1 for respondents that received the “difficult” financial scenario and 0 for those with the “easy” scenario. εiht is the idiosynratic error term, which we assume is serially uncorrelated. Thus, β1 estimates the treatment effect of the poverty prime on each outcome. We are also interested in the treatment effect as it varies across gender. To examine heterogeneous effects, we estimate the following model:
yi = β0 + β1Ti + β2Femalei + β3(Ti × Femalei) + εi (3) Femalei is an indicator for respondent gender that takes the value 1 for females. Therefore,
β3 estimates the differential effect of the treatment for females compared to males.
Experimental Design Details
Randomization Method
Computerized randomization
Randomization Unit
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
Sample size: planned number of observations
Sample size (or number of clusters) by treatment arms
105 treatment, 105 control
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
To achieve a sample size of 200, we conducted several sessions of the experiment with an average of 20 respondents per session.
IRB Name
Princeton University Institutional Review Board
IRB Approval Date
IRB Approval Number
Analysis Plan
Analysis Plan Documents
Analysis Plan



Uploaded At: July 31, 2015

Post Trial Information
Study Withdrawal
Is the intervention completed?
Is data collection complete?
Data Publication
Data Publication
Is public data available?
Program Files
Program Files
Reports, Papers & Other Materials
Relevant Paper(s)