Child Labor and Illegal Chemical Pesticide Use in Jalal-Abad, Kyrgyzstan: A List Experiment with Randomized Survey Modes

2025-06-16

2025-06-29

Each respondent is randomly assigned to either a control group (with 4 non-sensitive items) or a treatment group (with the same 4 items plus 1 sensitive item).

Each respondent is independently and randomly assigned to a treatment or control group for two separate list experiments: one on illegal pesticide use and another on child labor within the farming household. Each list experiment consists of a control list (4 neutral items) and a treatment list (the same 4 items plus 1 sensitive item).

Our goal is to estimate the difference in the mean number of items reported between treatment and control groups in each list experiment. This allows us to estimate the prevalence of each sensitive behavior while reducing social desirability bias in responses.

Direct-question responses to each sensitive item (Illegal pesticide use; Child labor) among control-group respondents, recorded as Yes / No / I’m not sure.

After each list experiment, control-group respondents answer the sensitive question directly and choose one of three categories: Yes / No / I’m not sure. The “I’m not sure” option (i) captures genuine uncertainty and (ii) reduces non-response or drop-outs.

In collaboration with the Jalal-Abad regional office of the Ministry of Water Resources, Agriculture and Processing Industry of the Kyrgyz Republic, we obtained a list of 1,016 farming households (HHs) in Jalal-Abad. A two-week randomized survey experiment will be conducted with three treatments:

- Treatment 1: Survey mode (Phone vs. In-person, where In-person is the control)
- Treatment 2: Sensitive questions about child labor use (via list experiment)
- Treatment 3: Sensitive questions about illegal chemical pesticide use (via list experiment)

In-person surveys are planned to be conducted in the morning and early afternoon, while phone surveys are scheduled for after 3 PM; however, actual times may vary depending on field conditions.

Treatments 2 and 3 will be randomized independently; however, the order of the two list experiments will not be randomized in order to avoid confusion and reduce the risk of implementation errors among local interviewers, given the complexity of multiple overlapping randomizations.

For phone surveys, interviewers (2 male, 1 female) are randomly assigned to reduce interviewer-related bias, and call order is randomized to reduce timing-related bias. Re-calls are also randomized. Data collection will take place during the experiment period (June 16–29, 2025), as much as possible within logistical constraints.

Note: Households in the in-person survey will be selected purposively (not randomly), based on feasibility and input from local collaborators. As a result, comparisons between the phone and in-person groups may be subject to selection bias and will not allow us to draw conclusions about causality.

Randomization was conducted using computer-generated assignments in Excel and included the following elements:
- Survey mode assignment (Phone vs In-person)
- Treatment assignment for the child labor list experiment
- Treatment assignment for the illegal pesticide use list experiment
- Interviewer assignment (2 male, 1 female)
- Call order for the phone survey

Farming households (HHs) in Jalal-Abad, Kyrgyz Republic

No

In-person surveys are clustered by village or cooperative units, selected based on feasibility and input from local collaborators. These clusters were not randomly assigned and were chosen purposively, which may limit the ability to estimate causal effects when comparing in-person and phone survey responses. Phone surveys are not clustered.

The target sample is 1,016 farming HHs (1 respondent per HH). The final number of observations will depend on interviewer capacity during the survey week, aiming to reach up to 1,016.

Treatment 1 (Survey Mode):
- Phone: 716HHs
- In-person (Control): 300HHs (subject to local adjustments based on feasibility)

- Note: Households in the in-person survey will be selected purposively (not randomly), based on feasibility and input from local collaborators. As a result, comparisons between the phone and in-person groups may be subject to selection bias and will not allow us to draw conclusions about causality.

Treatment 2 (Illegal Pesticide List Experiment):
- Phone: Treatment 343HHs, Control 373HHs
- In-person: Treatment 147HHs, Control 153HHs

Treatment 3 (Child Labor List Experiment):
- Phone: Treatment 335HHs, Control 381HHs
- In-person: Treatment 157HHs, Control 143HHs