Got it
We use Google Analytics in compliance with German Data Protection Law. The site gathers data for the sole purpose of improving its services. You're able to decline now or later. By using our services, you agree to our use of cookies. You'll find more information here.

Data Set Description

This study developed and implemented a mobile-phone-based survey of garment workers in Bangladesh, providing insights about COVID-19 responses in their workplace and the effects on workers.

The study asked about factory shutdowns, the rate of safety measures like masks and increased distance between workspaces, late payment and earnings, working conditions, time out of the labor force, beliefs and aspirations, and physical and mental health, both during the pandemic and retrospectively in January 2020.

Because of the possibility that the pandemic caused separations (either voluntary or involuntary) from the garment sector, the project defined a garment worker as someone who was working in the garment industry in January 2020, even if they were no longer doing so at the time of the survey. For those workers no longer in the industry, the survey also asked about their reasons for separation and any alternate economic activity they were engaged in.

The survey was designed to elicit as close to representativeness as possible, even though it was conducted during a time when households could not be visited in person and there were no pre-existing sampling frames (such as official lists of garment workers) from which the research team could draw samples.

Accordingly, the survey started out with two seed samples and solicited referrals from them. Because of the potential that directly recruiting referrals from respondents (i.e., the well-known respondent-driven sampling technique) may result in non-representative samples at the population level, the research team utilized a technique called randomized network recruitment. Specifically, respondents were asked to tell the names of all the contacts in their phones (and gave a 10-taka incentive per name and phone number) so that it was possible to randomly follow the contacts.

To compare randomized network recruitment to standard respondent driven sampling, the researchers asked some respondents whom they “preferred” for the project to be contacted. In a small, randomly selected segment of the initial seeds, these preferred contacts were followed, and continued to be followed in referrals given by those contacts. The binary variable “random” in the final data equals 1 if referrals of the initial seed were recruited using randomized network recruitment, and 0 if preferred referrals of the initial seed were followed.

Date Created: 2022-05-11

Scope of Data Set


Time Periods: November 2020 - June 2021


G2LM|LIC - The Effects of Employer Responses to COVID-19 on Female Garment Workers in Bangladesh. Research Data Center of IZA (IDSC). Version 1.0. doi:10.15185/glmlic.710.1.

Researchers working with the “G2LM|LIC - The Effects of Employer Responses to COVID-19 on Female Garment Workers in Bangladesh” are obligated to acknowledge the data base and its documentation within their publications, including the DOI, by using this reference.




  1. Heath, Rachel M.
  2. Boudreau, Laura
  3. Matin, Imran
  4. Ahmed, Shakil

Cross section survey data


Telephone interview


Access to the data is provided to non-for-profit research, replication and teaching purposes. The data is available from the Research Data Center of IZA (IDSC).
Please contact IDSC for any access requests.

Geographic Coverage: