Promoting regular labor supply among the urban poor

In this study, we hypothesize that workers’ preference for, and ability to maintain, long-term employment is malleable. We test this hypothesis by exploiting a habit-formation design, whereby we provide small financial incentives to attend daily to a randomly selected group of informal laborers for a limited period of time (6-8 weeks) and then observe labor market outcomes. The primary hypothesis we will test is that treated participants will attend the stand more often than Control participants not only in Phase 1 (incentives), but also in Phase 2 (follow-up phase, without incentives).

2022-05-09

2022-10-31

Our primary outcome is attendance at the labor stand during phase 2.

We will measure impacts on employment levels and earnings.

These two outcomes are considered secondary because: 1) they are self-reported leading to additional noise in the data, and 2) they are dependent on labor demand as well, which may vary across contexts. We will also collect supplementary data including: reasons for absences and a brief questionnaire on time use, habits, and identity.

Additional outcomes subject to feasibility. We are piloting to determine whether participants can be asked to wear pedometers for a portion of the study, in the hopes of providing an “objective” measure of the amount of work done. However, we will only analyze this data if we are able to incentivize participants to wear the devices regularly.

Similarly, we are attempting to partner with a construction firm to offer participants the choice of a “regular” job that is salaried and requires consistent attendance. Incentive compatible preferences for this type of work will be elicited and, if possible, a (small) random subset will be offered their choice at the end of the study. Take-up of the offer will be dependent on our being able to secure such a partner firm.
Additional outcomes subject to feasibility. We are piloting to determine whether participants can be asked to wear pedometers for a portion of the study, in the hopes of providing an “objective” measure of the amount of work done. However, we will only analyze this data if we are able to incentivize participants to wear the devices regularly.

Similarly, we are attempting to partner with a construction firm to offer participants the choice of a “regular” job that is salaried and requires consistent attendance. Incentive compatible preferences for this type of work will be elicited and, if possible, a (small) random subset will be offered their choice at the end of the study. Take-up of the offer will be dependent on our being able to secure such a partner firm.

The RCT will incentivize a subset of workers to increase their attendance at the stand, and then remove the incentives to examine the persistence of labor supply changes among treated participants.

The RCT consists of three phases:
Baseline (3 weeks): which includes screening, baseline daily data collection on attendance and labor outcomes, randomization, and announcement of treatment conditions. Participants also take a survey—prior to learning their treatment status—on demographics and baseline beliefs about job search. Participants are tracked for 11 days to gather baseline data on labor supply habits.

Phase 1 (6-8 weeks): participants enter Phase 1 and are assigned to a Treatment group (which receives incentives to increase their labor supply) or a Control group (which receives matched but unconditional payments to address income effects). Payments, described in more detail below, are made at the end of each week. This phase may be extended slightly for groups with disruptions to Phase 1 (e.g. holidays, weather shocks, election periods).

Phase 2 (6–8 weeks): During Phase 2, incentive payments will be removed for both Treatment and Control groups, but participants will continue to be surveyed about their labor market outcomes.

Treatment conditions
1) Treatment (financial incentives) participants receive a fixed incentive payment for each day they attend the labor stand and arrive before a predetermined cutoff time (computed during baseline based on worker flows at the labor stand) from Monday to Saturday during Phase 1. In order to be eligible for an incentive payment, treatment participants need to be observed at the stand by a surveyor before the predetermined cutoff time. For festival days – when the field staff is not working – no payments are provided to either treatment or control participants. These days are communicated to participants in advance.

2) Control participants receive a weekly lump sum payment, unconditional on attendance, during Phase 1. The payment amount is determined by randomly matching each control individual with a treatment individual within the same stand-stratum. This process ensures that both the mean and variance of payments is similar across arms to avoid differential income effects, but prevents collusion. This random match between a control and treatment participant is carried out each week. During informed consent, control participants are told that the payment amount is chosen randomly by the computer.

In Phase 2 (after the removal of incentives), the treatment participants will no longer receive payments conditional on attendance at the stand. They will receive payments for survey completion which are the same between treatment and control groups throughout the study.

The randomization is performed using Stata

Participants are randomly assigned at the individual level, stratified on stand, baseline attendance and baseline average daily wage earnings, to either the control arm or the treatment arm. Cut offs to determine the strata are computed using the median number of days of baseline attendance and the median baseline average self-reported daily wage from pilot stands.

No

We will enroll approximately 250-350 workers (depending on take-up at the stands) across approximately 15 stands in Chennai.

We will enroll approximately 350-500 workers (depending on take-up at the stands) across approximately 15 stands in Chennai.

We plan to have half of the sample enrolled in the treatment arm and half in the control arm. Some variation may occur because of the difference size of the strata.