Experiments on Performance-related Pay and Stress: Sample Selection, 2022

Bender, Keith and Andelic, Nicole and Allan, Julia and Powell, Daniel and Theodossiou, Ioannis (2023). Experiments on Performance-related Pay and Stress: Sample Selection, 2022. [Data Collection]. Colchester, Essex: UK Data Service. 10.5255/UKDA-SN-856873

The 18th Century economist Adam Smith observed in his book Wealth of Nations, "Workmen...when liberally paid by the piece, are very apt to overwork themselves, and to ruin their health and constitution in a few years". Thus, even as early as 1776, economists observed a link between work and health - specifically here, how the method of pay affects health. Recently a number of research papers by economists and healthcare academics finds a strong relationship between performance-related pay (PRP) (i.e. being paid according to how well you perform) and health in large surveys of workers and in case studies. Those being paid by PRP typically show poorer health. However, it is not currently known whether PRP causes poor health in workers or whether people with poor health are simply more likely to work in PRP jobs. The present study will investigate this. Similarly, little is known about how PRP may affect health. The present study will investigate the possibility that workers paid by PRP report poorer health because PRP work is stressful in ways that are detrimental to health. Economic theory would predict several potential pathways linking PRP to health, but research by two of the investigators (Bender and Theodossiou, 2014) suggests that because PRP induces increased effort at work and generates an uncertain stream of pay, stress will be higher for PRP workers than for workers paid under a typical salary scheme. Given the large and growing medical literature on the relationship between (particularly low-grade and constant) stress and physical and mental health, it is possible that if PRP increases stress, it will have negative health outcomes, all things equal. The investigators designed a small pilot study (Allan et al. 2017) to take a look at this linkage in an experimental framework. Experiments have several advantages over survey data. First, they can explicitly control for the possibility that people select PRP jobs by randomising PRP across subjects so we know the direction of causality. Second, they allow us to measure stress both subjectively (by asking subjects how stressed they are) and objectively (by measuring salivary cortisol, a hormone that the body creates in larger quantities in response to acute stress). With only 40 individuals and a simple design, the investigators found significant differences in subjective stress and cortisol levels between the PRP and nonPRP groups. The overarching goal of this project is to investigate this link more comprehensively. It is broken into several distinct projects. First, it will expand the subject pool and employ a crossover design to test the robustness of the findings in the pilot study. Second, it will investigate how important self-selection is in moderating the stress response to PRP. Most workers choose their payment contract, and so allowing subjects to choose the payment type and then measuring stress will examine how important this issue is. Third, we will look at anticipation effects of PRP to see if knowing in advance that the day ahead involves PRP work increases a worker's stress. Fourth, we will look at how intensity (via increased rewards and penalties) and monitoring by supervisors and peers can impact the stress generated by PRP. Finally, we will analyse a large survey of British workers where there is both information on the method of pay and noncortisol biomarkers related to stress. This is a useful parallel analysis to see whether the controlled conditions of the lab are found in the 'real world' where we cannot guarantee exogeneity. In sum, this research will help establish the link between PRP and stress. Given the potential for stress to affect physical and mental health, the implementation of PRP may have significant impacts on workers through their individual health, on firms through lost productivity and on society through lost economic output and increased public health costs.

Data description (abstract)

As described in the overarching funded project description, the focus of the research is to examine the relationship between performance-related pay and stress using experimental methods. Importantly, stress is measured here both subjectively (i.e. asking the stress level of participants) and objectively (i.e. collection of salivary cortisol). The basic experiment is generally the same across all of the experiments in the project. Subjects are recruited by using a database at the university where the experiments took place. All experimental sessions happened at 1400 to control for the diurnal patterns of cortisol. Participants were also told to avoid exercising, smoking, drinking caffeine and eating two hours before the experiment since these also affect cortisol. During the experiment, participants are told generally about the experiment and enter a 10-minute relaxation period where they can colour in a mindfulness colouring page if they wish. Then they are asked questions about their subjective stress and a baseline cortisol measurement. Next, participants are given three example maths questions to complete in their own time and then allocated their employment contract. The performance-related pay (PRP) contract is piece rate based on the number of correct answers while the minimum performance contract (nonPRP) is a flat rate once ten questions are answered correctly. Then, participants are given 10 minutes to complete as many questions of maths problems (up to 50) without the use of a calculator. The computer screen gives the number of correct answers and for the nonPRP group, a banner is shown when they answer 10 questions correctly. After the task, a screen shows how much they earned from their performance, including the show up fee. Another measure of cortisol is taken and a survey administered to collect information on subjective stress and to collect demographic information. Participants had a 10-minute rest period of colouring after which a third cortisol sample was taken. Then participants had one final 10-minute rest period after which a fourth and final cortisol sample was taken. Finally, participants were called into the control room and given their payment. Cortisol samples were frozen and after all were collected were sent to a laboratory for analysis. In the second experiment, we want to allow for a more ‘real life’ example from the labour market – namely that workers generally select into PRP jobs. It might be that this selection stems from an ability to handle stress better and so this experiment tests whether selection negates the effects of PRP on stress as found in the first experiment when the contract was randomised. Thus there are a couple additional changes in this experiment compared to the general description above. 1) Pre-testing of the experiment showed that participants overestimated their performance. Thus, for this experiment, participants were invited to a morning session where they completed two 10-minute tasks, where they were told in one how much they would have been paid if in the PRP contract and the other they were told how much they would have earned in the nonPRP contract. 2) As in the previous experiment, participants were paid £5 if they choose and met the minimum performance contract and were paid £0.20 for each correct answer when they chose the PRP contract. However, the show-up fee was increased to £7.50 to increase the sample pool of participants. The data suggest that PRP is significantly associated with higher subjective stress, but the effect on cortisol is less robust with only weakly significant effects found.

Downloads

Altmetric