Attribution and Job-Information Sharing in Informal Labor Markets: Evidence from India

The study consists of a real paid task, a randomized source-attribution intervention around the transmission of that opportunity to another worker, and several additional exercises designed to measure the relational consequences of sharing job information.
1. Paid task
Workers receive a standardized description of a real one-day paid task located near the labor stand. They report their beliefs about the task and whether they are willing to undertake it. Conditional on accepting, each worker has an independent probability of being selected for the paid task.

2. Nomination decision
Workers are told that additional workers may be needed for the same paid task under the same terms on subsequent days. They are told that, if a worker whose name they provide is selected for contact, the research team will approach that worker, provide information about the paid task, and ask whether he would like to undertake it.

Workers are then asked whether they would like to provide the name of another worker whom they would send for the paid task.

3. Source-attribution treatment
Immediately before the nomination decision, respondents are independently randomized with equal probability to one of two conditions:
Anonymous: if the nominated worker is contacted, the research team will not tell him who provided his name.
Attributed: if the nominated worker is contacted, the research team will tell him who provided his name.
The paid task, the information conveyed to the nominated worker, the nominating worker's own probability of selection, and the nominating worker's formal responsibility for the nominated worker's performance are held fixed across conditions.

4. Follow-up with nominated workers
A randomly selected subset of nominated workers is subsequently approached. Each receives the source-disclosure information corresponding to the nominating worker's assigned condition, receives information about the same paid task, and is asked whether he is willing and available to undertake it.

5. Additional randomized components
The study includes additional exercises designed to measure the consequences of transmitting job information and workers' willingness to be identified as the source of such information.

First, respondents complete a randomized vignette in which one worker tells another about a job that subsequently turns out badly. The vignette varies whether the two workers are close friends or workers who know each other from the labor stand but are not close.

Second, nominated workers complete a randomized vignette that holds the adverse job outcome fixed but varies whether the information source had previously worked for the same employer without an adverse experience on the job.

Third, a separate sample of workers participates in an incentivized costly-punishment exercise. Respondents may give up part of their own experimental endowment to reduce the payoff of a worker who passed along a job opportunity in good faith. The randomized condition varies whether the subsequent job experience is normal or adverse.

Finally, workers who complete the paid task are asked whether information about their own experience may be shared with future workers under two formats: anonymously as part of a pooled summary or individually together with their name. Each participant makes a decision under both formats, with the order randomized.

2026-08-17

2026-12-31

We define four outcome domains. The first contains the study's primary outcome; the remaining domains contain secondary outcomes.
(i) Willingness to nominate. Whether the respondent is willing to provide the name of another worker whom he would send for the paid-task opportunity and the number of workers the respondent reports being willing to name for the opportunity.
(ii) Nominated-worker outcomes. Characteristics of the worker nominated, including a relational-strength index combining relationship category, interaction frequency, reciprocal support history, network embeddedness, and reciprocal job-information exchange, and a nominated-worker-quality index combining perceived ability and reliability. We also measure whether nominated workers who are selected for follow-up are willing and available to undertake the paid task.
(iII) Relational consequences directed toward an information source. Outcomes from three separate exercises: a relational-deterioration index and expected out-of-pocket compensation from the Relationship-Strength Vignette; a source-responsibility index and perceived source fault from the Source Prior-Experience Vignette; and the amount of the respondent's own experimental endowment sacrificed to reduce the information source's payoff in the Costly-Punishment Experiment.
(Iv) Willingness to be identified as a source. Whether workers who completed the paid task permit information about their own experience to be shared with future workers under an anonymous pooled format versus an individually named format.

The main Source-Attribution Experiment uses individual-level random assignment. Approximately 600 workers from approximately 30 labor stands will be independently randomized with equal probability to the Anonymous or Attributed condition. In the Anonymous condition, if a nominated worker is contacted, the research team will not disclose that the respondent provided his name. In the Attributed condition, the research team will disclose that the respondent provided his name.

Approximately half of nomination records will be randomly selected for nominated-worker follow-up within cells defined by labor stand and the nominating worker's treatment assignment.

The study also contains separate randomized exercises. Respondents in the main worker survey are independently randomized between close- and less-close-relationship versions of the Relationship-Strength Vignette. Respondents completing the nominated-worker follow-up are independently randomized between the two versions of the Source Prior-Experience Vignette. A separate sample of approximately 200-250 workers is independently randomized to a normal or adverse job experience in the Costly-Punishment Experiment. Among workers who complete the paid task, the order of the anonymous pooled and individually named information-sharing formats is randomized.

Not available

All randomizations will be pre-programmed and implemented electronically through SurveyCTO during survey administration.

All treatment assignments in the main Source-Attribution Experiment and the additional randomized exercises are randomized at the individual level.

No

Approximately 40 urban labor stands (there is no clustering of treatment).

Approximately 950 daily-wage workers

Main Source-Attribution Experiment: approximately 300 workers in the Anonymous condition and 300 workers in the Attributed condition across approximately 30 labor stands.

Costly-Punishment Experiment: approximately 100-125 workers in the normal-outcome condition and 100-125 workers in the adverse-outcome condition across approximately 10 labor stands.

Nominated-worker follow-up: approximately 100 nominated workers may be followed up, depending on realized nomination rates and random selection for follow-up.