Introducing Piece-rate Pay in Bangladeshi Factories: Does Managerial and Worker Training Prevent Pitfalls and Improve Worker and Factory Outcomes?

We are evaluating two training programs in Bangladeshi apparel and footwear factories. The first program is delivered to factory management and focuses on identifying and preventing sexual harassment, creating promotion opportunities for women, increasing supervisor accountability, and introducing pay incentives in a thoughtful way. The second program is delivered to the entire workforce via a train-the-trainer approach, and focuses on rehumanizing workers in the eyes of supervisors and managers.

The program is designed to first address the way factory managers process information about the relationship between humane workplace systems and firm performance. Empirical evidence indicates that supervisors are not likely to accurately process information about workers if they see them in dehumanized terms.
The training then helps factories develop the systems that allow them to introduce optimizing workplace systems without adversely impacting workers. The training first develops workplace systems related to accountability, sexual harassment norms and gender neutral promotion ladders. We then turn to workplace systems necessary to support optimizing pay practices.

2024-05-01

2025-03-10

We are anticipating the following outcomes.
Rehumanize workers in the minds of managers/supervisors so that they accurately process information related to the relationship between abuse of workers and firm performance.
Following rehumanization, managers accurately interpret data about labor management practices and firm outcomes.
Factories develop organizational norms related to sexual harassment, women as supervisors, supervisor accountability for decisions and clear communication about pay.
The factory introduces high performance systems such as incentive pay, accountability and promotional ladders without adversely affecting workers.

Rehumanization is indicated if managers and supervisors believe that workers can feel physical and emotional pain, are intelligent, think at a high level and understand complicated ideas and workers feel less angry, frustrated small or unimportant in interactions with supervisors and more willing to voice opinions at work.
Information processing is indicated by accurately recalling details about HR systems and worker reports and manager beliefs about the relationship between labor management practices such as paying workers as promised, treating workers respectfully and providing safe and comfortable working conditions with firm performance including product quality and productivity.
The factory has clear policies concerning the review of supervisor decisions related to workers and supervisors believe that they are accountable for decisions, has clear rules about promotion that value the traits typically possessed by women and women are more likely to be promoted, and has low tolerance for sexual harassment.
Firm performance is measured by productivity, product quality, order size, relationship with buyers, absenteeism, late-coming and workforce turnover. Positive outcomes for workers include reduced sexual harassment, verbal abuse, accidents and injuries, excess hours and burnout.

Factories sustain and take ownership of investments in humane workplace investments after the intervention ends.

Sustainability is indicated when firms undertake investments that are not donor supported, employ high performing management techniques, are socially compliant, have reduced workforce turnover and high job satisfaction.

The study will include factories in Dhaka and Chittagong, Bangladesh. In Dhaka, we will start with a baseline survey in all factories, then our training partner will implement the rehumanization training (known as Restart) in Batch 1 factories. Next we will do a midline survey in all factories, then Batch 2 factories will receive Restart training. Next we will do an endline survey in all factories. Finally, all factories will receive the management training (known as BBW). In Chittagong, all factories will receive the Restart training, followed by a baseline survey, then Batch 3 factories will receive BBW training. Next, all factories will have a midline survey, and then Batch 4 factories will receive BBW training. Finally, all factories will have an endline survey.

The 8 factories in Bangladesh will receive one of two training protocols.

Four factories in Chittagong will participate in a study of the worker portion of the training. Before any training occurs, participating factories will be randomly assigned to one of two batches: batch 1 and batch 2. A baseline data collection will be conducted in all participating Chittagong factories. Factories in batch 1 will then receive worker training. After a midline data collection, factories in batch 2 will be provided the worker training. Once worker training is complete, an endline data collection will be conducted. Chittagong factories will then be optionally offered the manager portion of the training.

Factories in Dhaka will participate in a study of the manager portion of the training. Prior to any manager training, all factories in Dhaka will complete the worker portion of the training. Participating factories will then be randomly assigned to one of two batches: batch 1 and batch 2. A baseline data collection will be conducted in all participating factories. Factories in batch 1 will then receive manager training. This will be followed by a midline data collection. After the midline, manager training will be conducted in factories in batch 2. The study ends with an endline data collection.

We will be recruiting participants from 8 Bangladeshi apparel factories for which the management has already decided to introduce bonus pay or piece rate pay but have not yet implemented the change in pay practice. Participants in the training will include the director/general manager, HR & compliance managers, production manager, welfare officer and a selection of supervisors. A selection of workers will be included in the training. Ages of participants will range from 18 to 80 (though we do not place an upper age limit). There is no specification of gender.

Languages in Bangladesh will be English and Bengali.

We will also recruit a random sample of up to 120 workers and 60 supervisors per participating factory to report on work and life experiences. Participants will range from 18 to 80 (though we will not place an upper age limit) and language will be Bengali. There is no specification on gender but the sample is likely to be disproportionately female given the gender composition of most apparel factories in Bangladesh.

Statistical analysis will begin with analysis of the baseline. The purpose of the baseline analysis is to determine whether there are systematic differences between the treatment and control facilities. We will further conduct SEM analysis of the data to test theories about the relationship between individual characteristics and work outcomes.

The regression equation for the reduced form in the baseline analysis is given by

Z_ij=α+γX_i+βT_i

where Z_ij is outcome j for individual i, X_i is a vector of individual characteristics and T_i is a binary variable defined at T_i=1 if participant i is in a Batch 1 facility and T_i=0 otherwise. Errors will be clustered by facility. The coefficient of interest is β. We will test whether β=0. A finding that β≠0 indicates that the Batch 1 and Batch 2 facilities are systematically different in some dimensions at the baseline. We will also conduct Simultaneous Equation Modeling (SEM) to test for causal linkages in the theoretical model.

Analysis of the fully integrated data set will be conducted similarly to the analysis of the baseline with the exception that we will include a variable in the regression analysis that identifies whether the data is from the baseline, midline or endline.

The regression equation for the reduced form at the endline is given by

Z_ijt=α+γX_it+θ〖Treat〗_it

where Z_ijt is indicator j as reported by participant i at time t, X_it is a vector of individual characteristics i at time t, 〖Treat〗_it=1 if individual i is employed in a factory that had received training before time t and zero otherwise. The above equation is estimated with a random effects panel estimator with time and factory fixed effects. Standard errors are clustered by factory.

We will also conduct SEM analysis relating variables that exhibit a reduced form treatment effect to identify the causal chain that produces a treatment effect on work outcomes. Standard errors will be clustered at the facility level.

Randomization done in office by a computer.

Randomized at the factory level.

Yes

8 factories

Participants in the training will be 160 factory managers, supervisors and workers. Participants in the survey will be 1600 workers, supervisors and managers.

Participants in the training will be 160 factory managers, supervisors and workers.
Participants in the survey will be 1600 workers, supervisors and managers.

1760 participants with three observations each.

For our variables of interest, we have in other studies measured a standard deviation ranging from a low of 1.06 (necessitating a sample of 179) to a high of 49.1 (necessitating a sample of 1665) to detect a treatment effect at the 5% level.