We evaluate the impact of and demand for mid-level management training in a group of garment factories in Bangladesh. Industry participants recognise that there exists a shortage of skills needed to fill in lower-level management positions. There are several reasons firms may fail to provide socially optimal levels of training, and the project is designed to distinguish among these possibilities. Firms training programs currently on offer may not be effective, or relatedly, firms may believe they are ineffective even if they are. Learning about effectiveness may be complicated by complementarities between various components of training programs. Alternatively, labour markets may be fluid enough that workers are able to capture all of the benefits of the training by leaving for other factories.
Entrepreneurs intending to build large and more effective organisations must establish hierarchies and delegate decision-making to lower levels of the chain of command. As a result, effective management requires clear but nuanced communication both within and across layers of these hierarchies (for a review see, e.g., Gibbons (2010)). Formal schooling may provide certain essential tools, giving workers the language, reading, and writing skills that underlie most effective communication. But workers in low-income countries typically receive only low levels of education, often of poor quality. This makes construction of management hierarchies a particularly challenging task in developing countries.
We offer a group of factories the opportunity to participate in a training programme modelled on the GIZ Female Supervisor Training (FST) program. The FST program provides six weeks of training designed to train machine operators to be line supervisors. For existing supervisors, the training is shortened to three weeks, divided into three modules of one week each: Production, Quality Control and HR / Social Compliance.
The factories were randomly sorted into four groups of eight. We asked each factory to select six production lines to participate in the study. We next randomized the selected lines into one of four groups. Two lines were selected to receive training in all of the modules in round 1 (“full treatment”), two were selected to receive training in all modules in round 2 (“control”), and one each were selected to receive the HR training (Production and Quality training) in round 1 and the Production and Quality (HR) training in round two.
We assess training effects using surveys and detailed production data. The surveys were conducted at three points of time: just before the first training round, two months after first training round, and just before the second training round. Production data was collected for a period starting two months before the first training and continuing for a period of 13 months. Importantly, we collect data on all lines in the units participating in the study, whether those lines were nominated for training or not. Thus, we have a very clean randomized control group against which to measure the short-run effects of training and can obtain credible estimates of effects over a somewhat longer period of time using a set of comparison lines that were not selected by the factories to receive training. We measure changes in three dimensions: efficiency, absenteeism, and the rate of quality defects.