Research

This study investigates the impact of impending termination on employee behavior, using comprehensive data from 3,340 employees of a retail chain. Employees who were dismissed exhibit a sharp increase in absenteeism around the time they were given notice. In contrast, employees who resigned show only a moderate increase in absenteeism toward the end of their employment relationship, which is, however, not particularly pronounced around the time they gave notice. The conclusion of a mutual termination agreement between the employer and the employee even tends to be followed by a decrease in absenteeism.

We conducted a field experiment to compare the effects of financial performance pay and monetary gifts on employee performance. We randomly assigned store managers at a German discount supermarket chain to receive either performance pay based on profit increases or an unconditional gift in the form of a lump-sum payment. Our findings indicate that, on average, performance pay significantly outperformed gifts, yielding profit increases of approximately 8%. However, managers' base wages appear to moderate the effectiveness of performance pay versus gifts. Performance pay appears to have a stronger performance effect for managers with higher wages. In contrast, for gifts, we find tentative evidence of a stronger effect at lower wages. We argue that the effectiveness of both incentives depends on a reference point defined by the desired bonus-to-wage ratio (i.e. performance-pay-to-wage or gift-to-wage ratio). For performance pay, which can vary based on effort, higher-wage managers must exert more effort to meet this ratio. For gifts, which are fixed amounts, lower-wage managers may perceive the same bonus as more generous, which can strengthen reciprocity.

The assignment of individuals with different observable characteristics to different treatments is a central question in designing optimal policies. We study this question in the context of increasing workers' performance via targeted incentives using machine learning algorithms with worker demographics, personality traits, and preferences as input. Running two large-scale experiments, we show that (i) performance can be predicted by accurately measured worker characteristics, (ii) a machine learning algorithm can detect heterogeneity in responses to different schemes, (iii) a targeted assignment of schemes to individuals increases performance significantly above the level of the single best scheme, and (iv) algorithmic assignment is more effective for workers who have a high likelihood to repeatedly interact with the employer or who provide more consistent survey answers.

This field experiment investigates how different levels of aggregation in relative performance information (RPI) impact employee performance in environments with multiple tasks. We randomly assign store employees of a retail chain to three groups: RPI on overall performance (control group), RPI on separate tasks, and RPI on both overall performance and separate tasks. We do not find evidence that providing separate task RPI instead of overall RPI affects performance or effort allocation. However, providing RPI on both overall performance and separate tasks seems to reduce performance, especially in the low-return task. This suggests that detailed RPI directs employees' attention to the smaller benefits of low-return tasks. We further find that only 30.5% of the employees accessed their performance reports, highlighting a distinction between providing RPI in the field and the laboratory. This study is based on a registered report accepted by the Journal of Accounting Research.

Monetary incentives are widely used to align employees' actions with the objectives of employers. We conduct a field experiment in a retail chain to evaluate whether an attendance bonus reduces employee absenteeism. The RCT assigned 346 apprentices for one year to either a monetary attendance bonus, a time-off bonus or a control group. We find that neither form of the bonus reduced absenteeism, but the monetary bonus increased absence by around 45%. This backfiring effect is persistent and driven by the most recently hired apprentices. Survey results reveal that the bonus shifted the perception of absenteeism as acceptable behavior.

Employees spend a considerable amount of their working time enjoying on-the-job leisure. While this demonstrates a management control problem, it can also function as a novel domain for bonuses. In this study, I investigate the effect of an unconditional bonus (gift) in the form of more off-the-job leisure time. In particular, I examine how the gift of an intentional reduction of working time affects employee work behavior compared to a cash gift. A real-effort laboratory experiment shows that a cash gift neither alters employees' on-the-job leisure time nor performance. A gift of more off-the-job leisure time, however, does reduce the on-the-job leisure time of employees and increases their performance. A follow-up vignette study among human resource (HR) professionals further provides external validity for these results. Moreover, it also displays the other positive influence of leisure time offered as gifts on several different employee outcomes such as satisfaction, commitment, and health.

We study the profit effects and interplay of two core accounting practices in a field experiment in a large retail chain. In a 2 × 2 factorial design, we vary (1) whether store managers obtain decision-facilitating information on a profit metric and (2) whether they receive performance pay based on the same metric. We find that both practices increase profits significantly. In contrast to reasoning based on standard economic theory, we do not find complementarity between both interventions. Rather, we detect evidence in line with an attention-directing role of both practices: the introduction of each raises attention to the underlying objective, which induces a countervailing substitution effect.

I study the performance effect of removing performance pay across various incentive domains. Subjects in the laboratory experiment receive performance pay in either the monetary or time domain during the first working period, after which the incentive is removed in the second working period. Removing performance pay decreases performance in all treatments compared to a baseline of no bonus. However, this effect is less pronounced in the time domain, where participants receive a time-off bonus that allows them to leave the laboratory earlier depending on their performance. A post-experimental questionnaire suggests reduced enjoyment of the task as a possible driver.

We investigate the causal effect of performance pay and conversations about performance in 224 stores of a retail chain implementing a field experiment with a 2x2 factorial design. In the performance pay treatments, managers receive a bonus, which is a simple linear function of the profits achieved above a threshold value. In the performance review treatments, managers have to report their activities undertaken to increase profits in regular meetings. We find that whereas performance pay did not yield significant profit increases, performance review conversations increased profits by about 7%. However, when additionally receiving performance pay, the positive effect of performance reviews vanished. We provide evidence from surveys and meeting protocols that performance pay changes the nature of conversations, leading to a stronger self-reliance of store managers, which undermines the value of the performance reviews.

We run two field experiments within a large retail chain showing that the effectiveness of performance pay crucially hinges on prior job experience. Introducing sales-based performance pay for district- and later for store-managers, we find negligible average treatment effects. Based on surveys and interviews, we develop a formal model demonstrating that the effect of performance pay decreases with experience and may even vanish in the limit. We provide empirical evidence in line with this hypothesis, for instance, finding positive treatment effects (only) in stores with low job experience.

This study investigates whether individuals engage in prosocial behavior when it requires an investment of their time, but not money. In a laboratory experiment with rigorous anonymity arrangements, senders receive their payoff at the beginning. They may then engage in a tedious task to increase the earnings of exogenously disadvantaged recipients who otherwise receive no earnings. We find that senders are willing to sacrifice time to benefit recipients. Whether or not the recipient is present in the laboratory during the working time does not alter this decision. However, in a treatment variation some senders also display antisocial behavior.

We investigate the impact of providing employees with information about the descriptive norm of absenteeism through a field experiment within a retail chain. The study involved 817 employees, from whom we selected those in the top quartile of absenteeism. Half of these 194 employees were randomly assigned to the treatment group, where employees received information about their own absence relative to the median absence rate in the firm. The intervention led to a significant decrease in one-day absences. In line with rational updating of norm beliefs, the effects are particularly pronounced in departments with high prior absence rates. The effects also tend to be stronger for employees with weaker career concerns, consistent with the view that intrinsic preferences for adhering to norms rather than fear of extrinsic norm enforcement play a pivotal role.

We conduct a field experiment in a retail chain to examine how two common designs of relative performance information (RPI) affect employee performance. Store employees were randomly assigned to one of three groups: a control group receiving information about their own performance only, a treatment group additionally receiving RPI based on an aggregate reference point (peer median), and a treatment group additionally receiving RPI based on the full performance distribution (peer deciles). The information was provided privately and without financial incentives. The average treatment effects of both RPI designs are statistically indistinguishable from zero. Responses are heterogeneous, however: Distribution RPI shows the clearest positive effects among employees with high prior performance, while we find no robust evidence that either design demotivates low performers. Thus, rather than polarizing performance, private RPI based on the performance distribution strengthened performance at the top.

We examine how a workplace health-oriented initiative built around a health-tracking smartwatch affects worker productivity in a field experiment at a large port. Smartwatches that continuously record physiological indicators during working hours were randomly assigned to gantry crane operators. We find that the initiative significantly reduces crane operators' productivity, lowering weekly throughput by approximately 10%. The decline occurs despite output-based pay and without a detectable treatment effect on perceived organizational monitoring, care, trust, or engagement, providing little evidence that the decline is explained by the distrust-and-crowding-out channel emphasized in the control literature. Instead, exploiting the staged activation of real-time notifications and variation in notification exposure, we find that the decline is most apparent in the period when the notification system is activated and among operators with greater notification exposure, a pattern most consistent with attention diversion. Our results show that a supportive, non-performance-related control system can impose immediate productivity costs through its information and interruption architecture alone, even when it is not perceived as surveillance. The paper offers implications for the design and governance of workplace health and wellness technologies.