Tags: Harassment, whistleblowing, garbling, gender, garments, Bangladesh
Organizations often struggle to address harassment effectively due to challenges in gathering information from those involved. Reporting harassment is a daunting task for victims and witnesses alike, as it carries risks of retaliation and damage to reputations. This hinders organizations' ability to not only address individual cases but also understand the overall prevalence and nature of harassment.
In this study, we implemented a phone-based survey experiment at a major Bangladeshi apparel manufacturer. Our goal was to examine the impact of survey methods designed to provide plausible deniability, build trust in the surveyors, and reduce fears of information leaks on the willingness to report harassment. Additionally, we demonstrate how this reporting data can be leveraged to answer crucial policy questions about the extent of harassment, the proportion of managers responsible for the harm, and the isolation experienced by victims.
This project has several phases, including multiple surveys. It started with a pilot recruitment survey that ran from May 31, 2023, to June 16, 2023. After the pilot, the main recruitment survey began on August 1, 2024, and ended in June 2024. Using the data from the Recruitment survey, we launched the main phone call survey on September 25, 2024 with a new method called 'Hard Garbling'. Alongside these, there was a referral survey, which was a second version of the recruitment survey, running from June 7, 2024, to September 30, 2024. Once that ended, we tested 'Hard Garbling' in an updated referral survey, which included some extra questions. This test ran from October 7, 2024, to October 26, 2024. After testing, we rolled out another update to the referral survey with the 'Hard Garbling' method, which started on December 17, 2024, and continued until January 26, 2025. Additionally, we have been running a manual referral collection survey since September 11, 2024, which is still ongoing.
The sampling process follows three steps. First, areas for recruiting workers are randomly selected. Each of the 89 factories has a 3 km radius around it, called the commuting zone. These zones are grouped into recruitment areas in places like Savar, Gazipur, Narayanganj, and Chittagong, excluding areas like water bodies or industrial zones. The recruitment areas are divided into 2 km-wide sections or hexagons, and a random selection of these hexagons is made. The GPS coordinates of the selected hexagons are given to enumerators to begin recruitment.
Second, workers are chosen from the selected neighborhoods using a right-hand sampling method. Enumerators are assigned to specific areas and use tablets to help navigate. They start at the center of the neighborhood and move in a pattern, turning right at each intersection. At each household, they look for garment workers and recruit those who are eligible. If they recruit one worker, they skip the next house and continue until the entire area is covered. If they don't finish in one day, they continue the next day from where they left off.
Third, if more than one eligible worker is found in a household, one is randomly chosen. Enumerators check all available workers for eligibility and select one at random. If the chosen worker agrees to participate, the survey continues. If not, another worker is randomly selected. This ensures fairness and avoids bias. The process also accounts for different housing types, such as walled compounds with multiple buildings. The goal is to recruit workers fairly and systematically, following clear rules and procedures.
Hard Garbling: One important feature of the "Bangladesh Garments Whistleblowing Escrow" (BGWE) project is Hard Garbling, which helps protect workers' identities and ensures fairness. For every five garment workers' responses, the system randomly marks one answer as "yes" to create plausible deniability. This way, if a worker complains about harassment, they can later deny their answer, claiming the system made it "yes." The system sends these altered responses to a separate server via an API to track the hard garbling rate at the factory level without revealing any personal information. To ensure data protection, store survey responses offline and upload them in batches at the end of the day to prevent this issue, ensuring anonymity and privacy.
Api-store SCTO Plug-In: This plugin is designed to secure sensitive data without leaving any trace, ensuring that even individuals with full access to the data cannot identify respondents. Its primary objective is to facilitate the collection of sensitive information in the research sector, such as garbled data, while maintaining the highest level of anonymity and confidentiality.
Research Paper: Monitoring Harassment in Organizations , Authors: Laura Boudreau, Sylvain Chassang, Ada González-Torres, and Rachel Heath. This paper examines innovative approaches for monitoring and addressing harassment in workplace environments, ensuring that data collection methods are both effective and ethical.