In a recently published article in the Harvard Business Review, economics professor Laura Boudreau and co-author Sylvain Chassang of Princeton University discussed “random rotation,” a method employers can use to shield employees from the stigma attached to testing positive for COVID-19 while helping to ensure a safe work environment.
In this Q&A, Boudreau elaborates on the benefits of the method, which could also be used to help employees with mental health issues and other workplace difficulties.
She also provides an update on her ongoing research in conjunction with Ada González-Torres of Ben-Gurion University of the Negev and Chassang on how reporting systems that employ plausible deniability could improve the reporting of sexual harassment in a factory in Bangladesh.
You suggest that organizations can learn from indirect survey methods to encourage employees to report their COVID-19 risk. Why do you think this?
Early detection of possible cases is critical to prevent outbreaks of COVID-19 in the workplace. But there is the concern that employees may not want to report potential exposure due to fear of being stigmatized by their coworkers. In this context, we can learn from indirect survey methods developed by social scientists, in particular randomized response methods. These methods provide survey respondents who submit sensitive answers with plausible deniability.
Random rotation is based on the idea that you can “garble” information to protect privacy. Let’s assume there's a survey where you ask the question: "Have you been exposed to COVID?" If you say yes, your response is always recorded as a yes. If you say no, then sometimes, your response is randomly flipped to a yes. What this approach does is make sure that whoever is conducting the survey does not know whether you’ve said yes because you were randomly flipped to say yes, or because you actually have been exposed. In this way, the survey provides plausible deniability about reporting exposure.
In random rotation, based on survey responses, employees temporarily rotate out of the workplace for, say, one week. Thus, workers concerned that they may be sick can temporarily isolate. The fact that some healthy employees are randomly rotated out of the workplace makes it safe to self-report their exposure. An important complement of random rotation is a paid sick-leave or work-from-home program that limits the potential financial losses from being rotated out of the workplace.
In which work environments is random rotation most effective?
This type of policy could be especially valuable when the risks of infection are very high or if employees are economically vulnerable. Take, for example, the case of the meatpacking industry or other industries where workers typically work physically close together; a small amount of Covid entering the workplace could rapidly cause an outbreak. Further, economically vulnerable workers may fear losing their job if they bring up potential illness.
The situation is different for organizations with office-based workforces, which may have open offices but also allow employees to work from home. Many are already operating at below capacity. For these types of workplaces, the opportunity cost of sending healthy employees home is low, so random rotation is relatively easy to implement.
We generally think, though, that random rotation is most beneficial for frontline industries that don't have the option of having a fully home-based workforce and that are more exposed to the virus.
How could random rotation help increase the awareness of mental health issues in the workplace?
In the case of mental health, or burnout, you can think about policies that target work relief or other support based on reporting. Employers can incorporate this garbling method into the reporting process to encourage truthful reporting.
For example, a hospital may say, “I want physicians who can't function highly because they're so burned out to have an extra day off a week.” But if the hospital directly asks its physicians whether they need work relief, some physicians may not want to report that they do because they are concerned that it may burden their colleagues or would reveal that they are struggling with burnout. The hospital could use a random rotation schedule in which physicians are regularly asked whether they need work relief in a survey that garbles their responses, and the employer uses the recorded responses to assign some physicians to an extra day off. If the hospital is concerned that physicans may abuse the system, it could limit the number of work relief days per physician per year.
It’s worth noting that in this and other scenarios, the employer needs to be willing to accept some additional costs to implement this policy, in which some employees who don't need help are also going to get it. The policy itself will vary depending on the setting, but at the core is that the implementation provides plausible deniability and that privacy is key in terms of increasing willingness to report.
What are you learning about the effectiveness of reporting systems that provide plausible deniability from your continuing study of factory workers in Bangladesh?
Sylvain Chassang, Ada González-Torres, and I are working on a project with an anonymous worker helpline in Bangladesh that was set up by apparel buyers to facilitate reporting by workers of problems at their factories.
In the first part of the project, we aim to understand how the extent of plausible deniability provided by the reporting system affects truth-telling when workers face the threat of retaliation by managers. We are also interested in how trust affects workers’ willingness to report. If someone exerts more time and effort to build rapport with you, are you more likely to report or to share your experience? From a policymaker’s perspective, how do the costs of taking extra time to build trust compare to the benefits in terms of increased information transmission?
Based on workers’ reports under systems with more versus less plausible deniability and trust, we want to calculate what we call sensitive statistics. We know that our data is wrong here. We know that we're underestimating these problems, and in our context, especially sexual harassment. From a policy maker's perspective, we want to know how we can calculate more informative statistics, for example, not only about the true rate of workers who've experienced sexual harassment, but also about what share of managers in this sector are perpetrating it and who is aware of this type of behavior.
Finally, in the third part of the project, we aim to introduce enhancements to the helpline’s grievance resolution system to help workers overcome barriers to reporting. One of these builds on the garbling methods that we’ve been discussing and one of these focuses on alleviating first-mover disadvantages in reporting issues such as sexual harassment.