Experimental Design Details
Participants were 334 managers in 16 factories in and around Dhaka, Bangladesh. Managers with incomplete data were excluded, leaving a final sample of 277. There were 10 factories in the first group (October 2016) and six factories in the second group (March 2017). Prior to participation, managers in the first group were informed at the time of the experiment that a social dialogue program would be introduced in the factory in the coming months.
In each factory, participants were invited to a managerial meeting to hear data about their factory. All participants were instructed to call a particular number using their cell phone. After the informed consent process, they were asked to use their phone to complete an automated survey.
The survey software randomly assigned each manager to complete one of two 10-question surveys. In the treatment condition, managers were asked to imagine their workers’ preferences, daily routines and life aspirations (e.g., “What do you think workers in your factory prefer to have for breakfast?”). Managers in the control condition answered the same questions about themselves (e.g., “What do you prefer to have for breakfast?”).
Next, managers were shown two graphs of worker-reported survey data, summarized across several factories. A positive worker perception was paired with a negative worker perception. The positive finding is that 78 percent of workers report being satisfied with their supervisor. The negative finding is that 53 percent of workers report verbal abuse by their supervisor.
After seeing graphs depicting the two findings, managers called a given number to complete a second survey. The second survey captured their reactions to the data presented and their attitudes about verbal abuse (e.g., “How common is verbal abuse in this factory?” “How appropriate do you think it is to yell or use harsh language with workers in this factory?”). In each factory, all managers participated at the same time, though they were asked not to discuss the surveys or data with each other until everyone had completed the second survey.