Performance effects of managerial discretion

2016-07-01

2016-09-30

Key outcome variables: Turnover, sales
Other outcome variables: other indicators for store performance, e.g. absenteeism, shrinkage/waste

Understanding the costs and benefits of discretionary compared to rule-based decision-making is at the core of organizational economics. We set up a field experiment in a network of 241 supermarkets to generate causally interpretable evidence on the trade-offs involved.

Consider a team headed by a store manager M who reports to his principal P. M observes both the aggregate output of the team and individual contributions of its members, whereas P observes only the aggregate output. Should P distribute the reward equally between the team members, thus ignoring M's knowledge of the members' individual contributions, or should P give M the discretion to distribute the reward?

We will allow the managers in randomly selected stores to decide on the bonus distribution within their stores, while in the rest of the stores the bonus will be distributed equally. The size of the bonus to be awarded to each store is determined by a firm-wide rule, which is based on the store's results and performance and is known to all. Our treatment will start on July 1st 2016.

What is the effect of giving discretion over bonus distribution on the store performance outcomes, in particular, sales and employee turnover? On one hand, discretion will make use of the manager's knowledge, which should improve performance. On the other hand, there is a risk of favouritism and a possibility that managers may overlook important interactions between individual contributions.

We expect the treatment effect to be heterogeneous. Given our previous work with our study firm, we expect, among others, the following dimensions of heterogeneity:

1. Our previous treatment (https://www.socialscienceregistry.org/trials/826) in randomly selected store managers asked store managers "to do what they can" to reduce employee turnover. We expect that managers in the treatment stores have experienced a sense of empowerment and will hence be more likely to use the opportunities given proactively, leading to stronger treatment effects. They may also be more concerned with reducing personnel turnover. As a result, the effect of discretion on turnover may be stronger in this group.

2. Ability or willingness of managers to recognize and reward good performance, as measured by a survey. Managers with higher such ability will be using their discretion more effectively. We thus expect a larger effect of discretion on sales in such stores.

3. The amount of discretion previously given to store managers by their regional manager. We see from interviews with regional managers that there is significant variation in the existing discretion that store managers have to distribute the bonus pools in their stores. We expect that higher existing levels of discretion (pre-treatment) reduce the effect of our new managerial discretion treatment.

4. The exposure of store managers to their workers through meetings and other communications pre-treatment. As manager-employee communication should improve the manager's knowledge of their employees, we expect stronger effects of the discretion treatment in stores where managers and employees communicate more often.

5. Store size. Larger stores have more hierarchy and less direct contact between the manager and employees, thus limiting the manager's ability to gather individual performance information. Since the positive effect of the managerial discretion treatment relies on this information, we expect to see less of it in larger stores.

Stratified randomization; done in office by a computer

Stores of a retail chain

Yes

241 (stores)

241

66 stores: Manager has no discretion
175 stores: Manager has discretion