Pay for Performance – Evidence from a large discount retailer v2

This field experiment consists of two projects.

Project 1: The first project is a follow up on the field experiment registered under the ID AEARCTR-0000961. We hold the circumstances constant and incentivize the same normalized key performance variable (average receipt) for a group of randomly assigned employees of the lower (shop) management instead of the middle (district) management in the previous experiment. In the previous experiment we found that on average the bonus payment did not affect the average receipt. We hypothesize that lower level store managers have a more direct operational control to influence the average receipt and therefore we investigate the hypothesis that the bonus has a positive impact on the average receipt when paid to the store managers.

Project 2: In project 2 we investigate the role of relative performance information and complexity in incentive design by adding one additional treatment to project 1. We randomly assign the store managers to either being paid based on a normalized key performance variable as in project 1 (change of the average receipt to the same month in the previous year relative to the change of the average receipt of all stores in Germany normalized to the change relative to the months before the experiment) or to a substantially simpler key performance variable (average receipt compared to the average receipt in the same month of the previous year). In both cases, the bonus is paid out only when the performance measure exceeds a threshold value.

We hypothesize that the change in the normalization can affect the store manager’s performance in different ways. First, the normalization shifts the threshold value after which a bonus is paid out. While in treatment 1 this threshold varies with the nationwide performance of the firm, this is not the case in treatment 2. This will allow us to study whether and to what extent store managers strategically adapt their effort reaction based on the distance to the threshold.

But we also hypothesize that there can be at least two countervailing behavioral effects: on the one hand, using the development of the nation-wide average receipt as a normalization provides a better insurance to employees (Holmström (1979)’s informativeness principle) and this, in turn, could behaviorally trigger a more positively reciprocal reaction of the store managers. On the other hand, the normalization increases the complexity of the incentive scheme which may decrease the perceived influence store managers have on the key performance variable.