Job referrals for minority workers: Impacts on a manufacturing firm in India

Last registered on March 15, 2024


Trial Information

General Information

Job referrals for minority workers: Impacts on a manufacturing firm in India
Initial registration date
March 02, 2024

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
March 15, 2024, 2:34 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.



Primary Investigator

Harvard University

Other Primary Investigator(s)

Additional Trial Information

On going
Start date
End date
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Most manufacturing firms in the developing world rely on informally sourced referrals to recruit workers for entry-level jobs. Given homophilous networks, this practice may undermine diversity, and may contribute to occupational segregation. Does seeding more referrals amongst under-represented workers improve diversity, firm productivity, social cohesion within teams and worker retention? I study this in the context of India, where job search networks are concentrated by caste. I partner with a large manufacturing firm to experimentally vary their referral allocation process. In treatment teams, referrals are seeded disproportionately among incumbent lower caste workers. The study will focus on productivity, labor turnover, social cohesion and bargaining power as the key outcomes of interest.
External Link(s)

Registration Citation

Srivastava, Kartik. 2024. "Job referrals for minority workers: Impacts on a manufacturing firm in India." AEA RCT Registry. March 15.
Experimental Details


This study focuses on referral-based hiring at a manufacturing firm in India. The intervention disproportionately diverts referral opportunities to underrepresented incumbent workers belonging to lower caste groups, relative to a control group where referral opportunities are given to randomly picked workers, regardless of their caste. This intervention is conducted at the level of production teams inside the firm.
Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
- Demographic composition of teams (share of lower caste workers, team size)
- Team level productivity data
- Team level retention, attrition and turnover data
- Self-reported social cohesion, contact outside the firm, job satisfaction, worker bargaining power (team level, both average and extreme values)
- Job search intensity and worker outside options
- Beliefs about ability of, diligence of and cohesion with out-groups

In addition, I will explore lab-in-field games to measure beliefs towards out-groups. These will be specified prior to endline, where I expect to implement these. I will also attempt to get incentivized measures of cohesion, through workers' choices of sharing time with others at the firm during breaks and at optional training activities.
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
- Team leaders' beliefs about workers' attitudes, ability, cohesion, and referral quality
- Team leaders' reversion to status-quo referral allocation after the end of the intervention
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
The experiment involves randomizing teams at a manufacturing firm in India into one of two groups. In status quo, workers at this firm are primarily hired through referrals that are allocated to incumbent workers by team leaders based on their discretion. As part of this experiment, teams in the treatment group will have referral opportunities be given randomly within the set of lower caste incumbent workers, while teams in the control group will have referral opportunities be given to a randomly selected worker regardless of their caste. These teams are organized by task type within lines and shifts, and treatment status will be stratified by task type, team size, and baseline share of lower caste workers.
Experimental Design Details
Not available
Randomization Method
Randomization conducted through a computer algorithm. The implementation is controlled through the firm's HR department, who are given the names of workers who will receive referral opportunities for each team in advance.
Randomization Unit
The main intervention is delivered at the team level, of which there are 132 at the start of the experiment. Regressions on finer outcomes will have standard errors clustered at the team level.
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
Sample size: planned number of observations
132 teams, approximately 800-1000 workers at baseline, and approximately 2500-3000 referral candidates.
Sample size (or number of clusters) by treatment arms
61 treatment teams, 61 control teams.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Institutional Review Boards (IRBs)

IRB Name
Harvard University Institutional Review Board
IRB Approval Date
IRB Approval Number