Local Intelligence for Targeting High-Risk Populations

2025-01-28

2026-03-31

Economic, social, and psycological risks

We construct measures of risk factors for the following aspects: (i) labor market conditions, (ii) lack of human capital, (iii) use of drugs and substances, (iv) observed risk measures, (v) prior crime-related experiences, (vi) the IAT measure, (vii) preference for crime reduction, (viii) cognitive skills, (ix) approving attitudes towards crime, gangs, and violence, and (x) proximity to gang groups.

We will also construct a generalized risk measure, and the construction of each measure is described in more detail in the uploaded pre-analysis plan.

We conduct experiments with youth in fragile communities in Nairobi to find individuals at high ex-ante risk of being involved in violent crimes related to gang or extremist organizations. We randomly assign one out of following three different referral incentives to youth between 18 to 30 years old, who are randomly selected from the street and females recruited from randomly chosen households.

- T1 [Monetary Incentive]: Respondents who can identify individuals who match spefic targeting conditions will receive monetary compensation.
- T2 [Externality]: Referrers are informed about the positive externality for them from building a more peaceful community in that the number of vulnerable youths is reduced with improved economic opportunities by the training program, that it becomes more difficult for criminal groups to recruit local populations, and that the occurrence of crimes and violent events may eventually be reduced.
- T3 = T1+T2 [Interaction]: Combination of T1 and T2.
- C [Control]: No monetary compensation or information about positive externality.

After referral occurs, we additionally randomly assign one out of following three different referral incentives to those who are found through the referral process:

- T2 [Externality]: (Same as the above T2)
- T4 [Friend]: Respondents are asked to introduce a friend without any requirement based on targeting conditions.
- C [Control]: (Same as the above C)

Pre-Analysis Plan specifies more detailed experimental design.

Not available

We ask each randomly sampled individual to refer others under a randomized referral mechanism as first-order referrers. The randomized assignment of referral mechanisms is done by Qualtrics. We further ask each of the randomly selected referred individuals to refer others as second-order referrers under a referral mechanism that is randomly assigned by Qualtrics.

First-order referrer (randomly sampled individual) is the unit of randomization for our primary intervention. We also randomize treatment at the level of the second-order referrer (those referred by first-order referrers).

No

880 individuals (first-order referrers).

A total of 2,700 individuals, including 880 first-order referrers, with the goal of 1,200 first-order targets (those referred by first-order referrers) and 620 second-order targets (those referred by first-order targets as second-order referrers).

220 first-order referrers are assigned to the control (C), 220 first-order referrers to the monetary incentive (T1), 220 first-order referrers to the non-monetary incentive (T2), 220 first-order referrers to the combination of monetary and non-monetary incentives (T3). For the experiment on the second-order referrers, 300 individuals are assinged to the control (C), 300 individuals to the non-monetary incentive (T2), and 300 individuals to the friend group (T4). Although the exact sizes of the second-order referrer groups may be subject to change due to the uncertainty of observed referral behaviors, we will assign C, T2, and T4 with equal probability.