Experimental Design
Using four hypothetical workplace scenarios, the survey experiment allows identifying the causal effects of health-related productivity loss on the appropriateness of working longer hours across different occupations. For this purpose, survey participants assess the appropriateness of an employee X finishing the work week on time depending on the productivity that is impaired by ill-health or not. The comparison of treatment conditions reveals changes in the social pressure on X to work longer hours due to variation in productivity (high vs. low without reason vs. low due to sickness vs. low due to injury) and gender (male employee vs. female employee), which implies a 4x2 design. The use of different real-world contexts informs about the generalizability of the findings across different occupational settings.
To identify social pressure to work long hours, respondents are asked first about their views regarding what the broader population perceives as socially appropriate. In a second step, they report on their own choices that could be different from their beliefs about social norms.
As a well-established tool to foster successful identification of social norm effects, survey responses are incentivized by a monetary reward that participants could get if the response matches the most common answer in all the other survey participants (in a given scenario and treatment condition). This idea follows a concept by Krupka and Weber (2013) to measure social norms regarding appropriate behavior via an incentivized experimental approach. In this application of their approach, participants can take part in different lotteries and win one 50-Euro prize per scenario, i.e. up to 200 Euro.
To test the main hypothesis of the experiment, the social appropriateness of finishing the work week on time is compared across the high-productivity baseline condition and the low-productivity main treatment condition. This comparison provides causal evidence on the idea that less healthy individuals are pressured to compensate the potential impairments on labor productivity by spending some additional time at the workplace. Importantly, it is not clear in the main treatment condition why the employee has not been very productive during the work week. This aligns with many real-world contexts when it is not observable if somebody has minor health problems or not.
To test whether social pressure to work long hours is still higher, compared to the baseline, if low health as the reason is known, two additional low-productivity treatment conditions provide information on the health impairment of the employee. For example, it could be seen as more appropriate to finish the work week early for the purpose of recovery. Furthermore, it could be that survey respondents consider it as less appropriate if employees with sickness stay longer at the workplace, given that many people are concerned about contagious diseases. To find out more about this, two different health-related reasons for low productivity are given in two additional treatment conditions: sickness (with a cold) and hand injury (due to an accident).
In addition to the random manipulation of health-related productivity loss, the gender of employee X varies randomly to allow identifying the role of gender differences in this context. There are several reasons why gender could be relevant for behavioral choices in the context of health problems at work. For example, pressure on women to work long hours could be higher due to gender-specific expectations and possible stigmatization of female employees as being less productive compared to male employees. However, due to societal trends regarding gender equality and significant changes in people’s beliefs concerning gender discrimination, it is not clear whether such gender-specific expectations are prevalent in today’s labor market.
To learn more about the generalizability of the findings across occupational settings, four different work scenarios are used and vary in several respects: First, the number of co-workers is manipulated, which could reflect differences in group pressure to work long hours. Since concerns about contagion at work could also be relevant in this context, another work setting describes a job that takes place outdoors. Furthermore, working-from-home is of particular interest, given that this promises insights about possible differences in workplace behavior due to recent trends.
Finally, the dataset is complemented by a set of questions about the survey respondents themselves, including their attitudes towards work, health and gender, all of which could play a role when assessing behavioral choices in the labor market. Of particular interest are attitudes towards ill-health in employees and concerns about contagious diseases.