Social Norms and Firm Productivity: Evidence from Bangladeshi Knitwear Factories

Last registered on November 04, 2022

Pre-Trial

Trial Information

General Information

Title
Social Norms and Firm Productivity: Evidence from Bangladeshi Knitwear Factories
RCT ID
AEARCTR-0009548
Initial registration date
September 25, 2022

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
September 27, 2022, 11:56 AM EDT

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

Last updated
November 04, 2022, 2:40 PM EDT

Last updated is the most recent time when changes to the trial's registration were published.

Locations

Region

Primary Investigator

Affiliation
Columbia University

Other Primary Investigator(s)

PI Affiliation
BRAC Institute of Governance and Development
PI Affiliation
Washington University in St. Louis

Additional Trial Information

Status
In development
Start date
2022-09-23
End date
2024-01-31
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
In this project, we aim to test whether social norms, in particular norms around communication, hinder information transmission inside the firm. Particularly, we argue that in certain settings, higher social status individuals are unwilling to receive productivity improving information from lower social status individuals, that this dynamic is costly to the firm, and that it limits low-status individuals’ career advancement. To test these hypotheses, we collaborate with Bangladeshi knitwear factories to conduct an incentivized survey experiment to examine workers’ willingness to participate in information sharing sessions with workers of the same versus the opposite status. We also implement a field experiment in which we inject information about productivity-enhancing practices into the firm through selected high- and low-status individuals. We then study how social status affects information diffusion across workers and quantify the downstream effects on trained and untrained workers’ productivity. Finally, we explore methods to alleviate the impact of social norms on information transmission in the factory.
External Link(s)

Registration Citation

Citation
Boudreau, Laura, Sakib Mahmood and Oren Reshef. 2022. "Social Norms and Firm Productivity: Evidence from Bangladeshi Knitwear Factories." AEA RCT Registry. November 04. https://doi.org/10.1257/rct.9548-1.2
Sponsors & Partners

Sponsors

Experimental Details

Interventions

Intervention(s)
This research consists of two field experiments with workers employed in the linking section of sweater factories in Bangladesh. In experiment 1, we will conduct an incentivized survey experiment with the objectives of eliciting workers’ willingness to pay to receive and to share productivity-enhancing information and of identifying how this willingness changes when social norms are breached. The information being exchanged will be provided by the research team through technical skills training for workers. In experiment 2, after conducting the training, we will introduce the productivity-enhancing information into the factory by training selected workers and studying the downstream effects on their coworkers’ knowledge of the technical skills, adoption of the technical skills, and labor productivity.
Intervention Start Date
2022-09-23
Intervention End Date
2023-12-01

Primary Outcomes

Primary Outcomes (end points)
Experiment 1 (survey experiment): Willingness-to-pay/to-accept to participate in teaching session.

Experiment 2 (training experiment):
• Awareness of technical skills among coworkers;
• Adoption of technical skills among coworkers;
• Productivity of coworkers.

In experiment 2, we will test for heterogeneous treatment effects by coworkers’ social status. We hypothesize that the positive effect of low-status workers (instead of high-status workers) being trained, will be smaller for high status coworkers compared to low status coworkers. We will also allow for heterogenous effects by proximity. For example, workers on the same, comparted to adjacent, lines to trained workers.

Finally, we note that there is some risk that we will not be able to measure labor productivity for all factories, which may require us to change whether we include it as a primary or a secondary outcome. This is because we have found factories to maintain very different production records that sometimes allow for calculation of individual-level labor productivity and sometimes do not. As of posting, the research team has recruited three factories to participate, and we are able to calculate labor productivity for all three. Consequently, we are hopeful that we will be able to measure it for all factories. If we are not able to measure it in factories recruited in the future, however, then we will update our pre-registration with this information.
Primary Outcomes (explanation)
Experiment 2 (training experiment):

1. Awareness of technical skills training among coworkers: We will construct an index of awareness using the following questions from the endline survey:

10. Can you please tell me which skills were covered in the skill training session?
Instructions: Enumerator does not read skills aloud. Enumerator waits for respondent to list skills learned during the training.
a. Skill 1: Bending body less (Other acceptable names for this skill: curve the body; motion improvement)  Listed by participant  Not listed
b. Skill 2: Stopping machine less (Other acceptable names for this skill: not pausing machine; machine stoppage reduce)  Listed by participant  Not listed
c. Skill 3: Keeping panels nearby (Other acceptable names for this skill: keep panels on the lap)  Listed by participant  Not listed
d. Skill 4: 5S (Other acceptable names for this skill: housekeeping; clean and tidy)  Listed by participant  Not listed
e. The participant provided a skill that was not covered in training: (open)


27. For each of the following statements, can you please tell me whether you agree, disagree, or neither agree nor disagree with the statement?
a. When linking, workers should minimize the number of machine stops to improve their production.
b. When linking, workers should bend their body to improve their production.
c. When linking, workers should keep panels nearby to where they are working.

Response options: Agree, disagree, neither agree nor disagree.

28. Can you please tell me each of the 5S practices? If you can tell me them in order, please do.
FOR ENUMERATOR ONLY: CORRECT ORDER: Sort, Set in order, Shine, Standardize, Sustain.
Instructions: Select the appropriate response option.
a. Respondent lists all 5 practices, in correct order.
b. Respondent lists all 5 practices, not in correct order.
c. Respondent lists between 1-4 practices (order does not matter).
d. Respondent lists no practices.


2. Adoption of technical skills among coworkers: We will construct an index of adoption using the following measures from the in-factory skill assessment (measured pre-, immediately post, and delayed post) and the voucher redemption:
• Body Bending
• Machine Stop
• Location of Panels
• 5S (Wastebasket available; Wastebasket being used to keep excess yarns and/or empty cones; NO excess yarn, empty cones, or other material on floor)
• Wastebasket redeemed (Source: Voucher redemption form)

We will construct index variables using Anderson (2008)’s methodology.

3. Productivity of coworkers: Our approach will ultimately depend on the productivity measurement format of the participating factories. Tentatively, we plan to use an efficiency measure calculated based on the number of operations completed in a day relative to the total number of operations that could be completed in the duration of hours worked in that day given the operations’ standard allowable minute (SAM). SAM is defined as the number of minutes that should be required for a single, particular operation of a particular style being worked on.

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
Experiment 1:

We will conduct an incentivized survey experiment with the objectives of indirectly eliciting individuals’ willingness to pay to receive and to share productivity-enhancing information and of identifying how this willingness changes when social norms are breached.

We will first stratify based on factory, production unit, high vs. low productivity, sex, tenure, education level, and age. We may not be able to use all variables in all factories if the strata that result are too small. Within strata, we will randomly assign workers to the “Sender” condition or to the “Receiver” condition. Senders are the workers who are randomly assigned to imagine that they receive training. Their prompt will entail an invitation to participate in information-sharing sessions in which they teach workers who have not participated in training. Receivers are the workers who are randomly to imagine that they are not selected to participate in training. Their prompt will entail an invitation to participate in information-sharing sessions in which they are taught new skills by workers who received training. The survey experiment will occur prior to announcement of which workers have been selected to participate in the training.

Within strata and condition (Sender/receiver), for the information sharing sessions, workers will be randomly assigned to be matched with another worker who is (1) anonymous; (2) high status; or (3) low status. We consider social status related to sex (men vs women), age (older vs younger), and religion (Muslim vs non-Muslim); where the higher status group is always listed first. We may not be able to use all dimensions of social status in all factories if the number of particular groups is too low. In the experiment, workers will not receive information about their partner’s identity, only their type. The anonymous condition identifies baseline take-up rates. Within strata and condition (sender/receiver x match type), workers will be randomly assigned to be informed that their decision will be made public or will be kept private.

In the experiment, workers will be read the experimental prompt, which is tailored to their treatment assignment. In the prompt, workers will be invited to participate in the information sharing session and informed about the type of worker whom they are matched with and whether their decision to participate will be public (private). The survey enumerator will then elicit workers’ willingness to pay (receivers) or to accept (senders) to participate in the sessions using the Becker-DeGroot-Marschak (BDM) method.
During experiment 1, workers will also complete other survey modules.

Experiment 2:

We will conduct a skills training experiment in which we randomly assign a subset of either high or low status workers in a linking production unit to 2 days of technical skills training. We will then use this exogenous variation in worker training to study to what extent diffusion of new knowledge in the workplace depends on workers’ social status.

We will first stratify based on factory, the production unit’s quantile in the productivity distribution, and the production unit’s quantile in the distribution of sub-sections by the share of low types among those eligible for training. We may not be able to use all variables in all factories if the resulting stratas are too small. Within strata, we will randomly assign production units to either have a random subset of high types or a random subset of low types trained. We anticipate training 20-30% of workers in each subsection, but ultimately, the share will depend on the numbers of eligible workers and take-up of the training. The training entails two days of technical skills training. The technical skills training curriculum will be tailored to each factory’s needs based on a diagnostic conducted by a team of industrial engineers. The training will be conducted on two consecutive Fridays. Incentivized skills tests will be administered at the beginning and the end of each day of training.

In the week before the first training, in the week after the second training, and approximately three weeks after the second training, a trained assessor will visit the factory and will observe adoption of the skills taught during the training by all workers in the linking section. If relevant, they will also observe adoption of housekeeping tools.

In the third week after the second skills training session, an endline survey will be conducted with workers in the linking section. Individual-level production data will be collected for several months prior to the training intervention and for multiple weeks following it. The research team will measure coworkers’ awareness of the technical skills taught during the training sessions, coworkers’ adoption of these skills, and coworkers’ productivity.

At factories where industrial engineers deem poor housekeeping to hinder production quality, the training will include housekeeping techniques. On the second day of training, workers will be provided with tools needed to maintain good housekeeping (e.g., small baskets to store certain materials). Workers will be encouraged but not required to use these materials at their workstations. They will also be endowed with 4 vouchers for baskets that they can provide to their coworkers. Their coworkers can use the vouchers to collect these materials from a member of the management team who will make them available at their office. Given the staggered nature of the project’s role-out to factories, it is difficult to assess how many factories will be eligible for this intervention ahead of time.
Experimental Design Details
Not available
Randomization Method
The randomization is done by a computer.
Randomization Unit
Experiment 1: The unit of randomization is a worker, stratified by factory, production unit, high vs. low productivity, sex, education, tenure, and age. We may not be able to use all variables in all factories if the strata that result are too small.

Experiment 2: The unit of randomization is a production sub-section of the linking section, stratified by factory, the production sub-section’s quantile in the productivity distribution, and the production unit’s quantile in the distribution of sub-sections by the share of low types among those eligible for training. We may not be able to use all variables in all factories if the strata that result are too small.
Was the treatment clustered?
Yes

Experiment Characteristics

Sample size: planned number of clusters
The treatment is clustered for experiment 2. The treatment is not clustered for experiment 1.

Experiment 2: Tentatively, we are targeting a total of 60 clusters. Ultimately, the number of clusters will depend on the organization of the linking sections in the factories that we partner with, how many production units there are per factory, and the duration of the worker panels that we construct. Consequently, this number may decrease or increase somewhat.
Sample size: planned number of observations
Approximately 3,200 workers (we assume ~40 workers per production unit). Ultimately, the number of workers per cluster will depend on the organization of the linking section and how many production units there are per factory. If needed, we will increase the number of factories to achieve our target number of observations.
Sample size (or number of clusters) by treatment arms
Experiment 1: Within strata, 50% of participants will be assigned to the receiver condition, and 50% allocated to the sender. 50% of participants will be assigned to the private condition, and 50% to the public condition. In terms of the match between sender and receiver types, ~33% of respondents will be assigned to same social type (high with high, low with low), 25% will be assigned to a compatible type (high to low), 25% to an incompatible type (low to high), and ~16% will be assigned to an anonymous type.

Experiment 2: 50% of clusters will be allocated to the high-status arm and 50% to the low-status arm.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB

Institutional Review Boards (IRBs)

IRB Name
Columbia University Human Research Protection Office
IRB Approval Date
2021-12-14
IRB Approval Number
IRB-AAAS9618
IRB Name
Institutional Review Board (IRB) of the BRAC James P Grant School of Public Health, BRAC University
IRB Approval Date
2022-01-19
IRB Approval Number
IRB-13 December'21-040
IRB Name
The Washington University in St. Louis Institutional Review Board
IRB Approval Date
2022-02-08
IRB Approval Number
202201154