Factory safety information and work decisions by garment sector workers in Bangladesh

Treatment group: After participating in a survey (see survey attachment), the treatment group participants receive information about their factory and other factories’ nearby performance on structural, electrical, and fire safety audits conducted by a third-party auditor. Specifically, enumerators follow a script (see treatment group post-survey script attachment) in which they tell the participant how many times their factory does not comply with the “High Priority” safety rules for the audits; they also tell the participant how their factory ranks compared to 71 factories nearby for which the safety audits are available. Then the enumerator provides the participant with a flyer that includes information about the audits and a list of the best and relatively worse performing factories nearby (see attachment). It also includes a phone number that the participant can call for additional information about safety.

Control group: Control group participants also participate in a survey do not receive the information about safety at the end. They hear a short post-survey script and receive a flyer with a phone number that they can call for more information (see control group post-survey script and flyer attachments).

2015-08-28

2015-09-28

1. Calls to safety information hotline: (1) Whether a participant places a call to the safety hotline ({0,1}); and (2) the duration of the participant’s call to the safety information hotline. (Exploratory analysis: Which safety information options the participant selects and whether the participant leaves a voicemail (the safety information hotline is a touchtone system where the caller selects different options to receive different safety information).)

2. Employment: Whether participant is more or less likely to report: (1) Having changed jobs in past three months; (2) having left the workforce in the past three months; (3) plans to look for a job with a different employer or to stop working in the near future. If participant responds positively to any of these, (4) whether they report safety as a factor in their decision. If the participant has changed jobs to another garment factory: (5) If the factory is one of the factories for which a safety audit is available, whether the factory performed as well or better than the participant’s former factory on the safety audits; (6) Change in participant’s position and wages from old to new factory; (7) Change in participant’s commuting distance (reported commuting time) from old to new factory; (8) Change in participant’s reported nonpecuniary benefits from old to new factory. If the participant has changed positions within the same factory: (9) Change in positions; (10) Change in wages.

3. Job referral practices: (1) Whether participant is more or less likely to refer family and friends to jobs at their factory and/or (2) other factories nearby compared to participant’s baseline. (3) Whether factories participants reports referring family/friends to are more likely to be top safety performers (less likely to be bottom safety performers). (4) Whether participant is more or less likely to report safety as a factor in their decisions to refer family and friends to jobs at their factory and/or other factories nearby compared to participant’s baseline.

4. Perceived stress and risk at work: Response to survey questions 31, 32, 38, and 39.

5. Prioritization of safety relative to other nonpecuniary benefits: Response to survey question 33.

6. Information sharing with family, friends, and coworkers.

7. Participation in workers’ organizations: Reported participation in trade unions, Worker Participation Committees, and/or Workers’ Welfare Associations (where relevant).

I randomly assign walking routes for enumerators into treatment (factory safety information provided) and control (no factory safety information provided) conditions. Enumerators obtain a random sample of garments along each walking route using a right-hand sampling rule (see attachments). All participants along one walking route are assigned to either the treatment or the control group.

Identification of walking routes: After selecting the location of the field experiment, I mapped the locations of all factories for which the third-party safety audits were publicly available (71 factories). Following a site visit to determine reasonable assumptions for how proximate garment workers’ residences are to their factories, I determined the geographic area to be included in the study. Then I determined the locations of 24 clusters of workers by aiming to maximize distance between walking routes subject to identifying housing units along each route (using Google satellite imagery).

Treatment and control group participants receive follow-up phone calls three and six months after the survey during which they are asked a series of follow-up questions.

Participants’ factories in this study can be grouped into three categories: Top safety performers, intermediate safety performers, and poor safety performers. Participants in the treatment group will learn which type of factory they work at, and therefore their response to the treatment is expected to be different depending on which type of factory they work at. When I test whether the treatment had an effect on the outcome variables of interest, I will specifically test for differential effects across these three groups. The three groups are pre-determined by the study design: 10 factories are listed as top performers, and 17 factories are listed as bottom performers, on the safety information sheet provided to participants. The remaining 44 factories are the intermediate group.

Randomization was done in the office using the statistical program R.

Walking routes through neighborhoods with dense populations of garment workers were the unit of randomization. Each walking route was randomly assigned to either treatment or control.

Yes

24 clusters of 13 workers per cluster. 12 treatment clusters and 12 control clusters.

312 participants.

156 participants in the treatment group and 156 participants in the control group.