Dynamic incentives and productivity - Evidence from field and lab experiments

A firm pays a monthly bonus to its workers and adjusts the target every month according to how fast the workers worked in the previous month: if they worked faster, the target becomes harder. How do workers react to such an incentive system? Forward-thinking workers should work less hard than they potentially could to keep an easy-to-reach target. This unintended side effect results in lower worker effort and thus lower overall productivity. Our project aims to understand such "dynamic incentives", i.e., positive or negative effects of incentives on behaviour in later periods.

It is important to understand both the direct and the indirect effects of incentives since they affect individual decisions on a daily basis. Incentive systems are widely used in the workplace, usually with the aim of achieving higher productivity. Incentive systems are also used outside the workplace, e.g., by governments that use tax and benefit systems to induce people to take up a job or to save. Understanding and making incentive systems more effective is thus of central importance for many aspects of society, in particular with regard to increasing productivity. Productivity growth in the UK has stalled for the last 10 years. Well-designed incentive systems could play a part in increasing overall productivity. Moreover, incentives that are more effective could help mitigating health-cost increases or improving educational outcomes.

While we already have well-developed theoretical models of behaviour under dynamic incentives, we have very little empirical knowledge about them. This project aims to fill this gap. We aim to find out how important dynamic effects of incentive on productivity really are and which characteristics of workers and of the workplace weaken or strengthen them. We we will also design and test ways how firms can avoid negative dynamic effects or harness positive dynamic effects of incentives.

The research is undertaken in partnership with a firm that runs several warehouses with thousands of workers. The firm has allowed us to randomly allocate workers to different incentive schemes. This gives us the unique opportunity to test the effect of dynamic incentives in a very clean and direct way. In the first part of the project, some workers receive incentives that have direct (static) positive effects but negative dynamic effects. Others receive the same direct incentives but without any dynamic incentives. Comparing the output of the two groups will allow us to measure the importance of dynamic incentives. By collecting other data on workers and the workplace, we will be able to identify which factors weaken or strengthen dynamic incentives.

In the second part of the project, we use the fact that dynamic incentives are inherently harder to understand, as they need thinking about indirect effects. We will study whether the firm could reduce the negative dynamic effects by making the incentive system more complex. It might well be possible to do so while retaining the direct (static) incentive effect. Firms would benefit from this through increased productivity and workers would benefit through higher bonus payments. We use laboratory experiment with the same workers to study this question. This will also allow us to link the behaviour in the lab with behaviour on the shop floor.


In the last part of the project, we will study how we can use within-day changes of incentives to harness positive dynamic effects. The growing use of technology by firms has made such "real-time" incentive changes possible in many workplaces. For example, varying incentives during the day could influence agents to work hardest when a high productivity is particularly valuable or could enhance task engagement by making the task less boring. We will first try out different incentive schemes on an online casual-work platform like Amazon Mechanical Turk before rolling out the most successful schemes in the warehouse of our partner firm.

This project aims to understand how well-designed incentives can increase productivity. We focus in particular on the impact of dynamic aspects of incentives. Given that the UK's productivity growth has been lacking recently, research on how to increase productivity is urgently needed. Our project focuses on a microeconomic, within-firm level of the analysis of productivity and thus complements the more prevalent, macroeconomic research on productivity.

From the very beginning, we have developed this research project jointly with a partner firm. We have thus ensured that the research question and the insights are important for non-academic users and that the project has direct impact. Potential users are all employers, private or public, that use some kind of incentives for their workers, i.e., almost all employers. Our research will help them design incentives such that they can avoid negative dynamic effects and harness positive dynamic effects, to increase overall productivity.

We study dynamic incentives, i.e., direct or indirect effects of incentives on behaviour in later periods. Almost all real-world incentive systems feature such dynamic incentives. As a case in point, it is a pressing problem for our partner firm how to increase worker productivity, in particular through improved incentives that avoid negative dynamic effects. The insights from our research will be directly implemented across the firm affecting thousands of workers. But also all other firms using incentive schemes will benefit from a better understanding of dynamic incentives and insights into how to reduce negative dynamic effects and harnessing positive dynamic effects.

All of our projects will generate findings that are generalizable from this one firm to other firms or other users of incentive systems. In Project 1, we focus on how different worker or workplace characteristics strengthen or weaken dynamic incentives. While the overall effect of dynamic incentives will differ between settings, the effect of different moderating factors will be more constant. This will make our insights useful outside the partner firm. In Project 2, we study how to reduce negative dynamic effects by making the incentive scheme more complex. This method can be easily applied in other firms. Project 3 studies the use of real-time incentives to harness positive dynamic effects of incentives. For example, we will test whether varying incentives over time is able to push agents to work hardest when effort is particularly valuable. We will also test whether one can use incentives to enhance task engagement by making the task less boring. The growing use of technology has made real-time incentive changes possible in many workplaces. Similar incentive systems are already used in the gig-economy, e.g., the surge pricing of Uber. Moreover, Project 3 will develop and optimize the use of Amazon Mechanical Turk as a test bed for incentive systems before rolling them out in the field. This will be highly useful for any user of incentive systems who does not want to roll out an untested new incentive system.

Dynamic incentives do not just arise when firms try to elicit effort from workers. Improved incentive design is a potentially crucial tool for addressing pressing societal challenges like mitigating increased health costs, improving educational outcomes, and avoiding financial crises. Incentives for doctors are often not aligned with the interests of patients or the NHS. Similarly, the incentives for individual bankers or banks were often not aligned with the interests of the regulator. The implications also extend beyond the workplace, to include, e.g., insights into the design of incentives for social insurance programs.