Incentive Complexity, Bounded Rationality, and Effort Provision

How to cite

Huffman, David; Abeler, Johannes; Raymond, Collin. 2023. "Incentive Complexity, Bounded Rationality, and Effort Provision." AEA RCT Registry. AEARCTR-0011427. https://doi.org/10.1257/rct.11427-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

David Huffman* , Johannes Abeler , Collin Raymond

* Primary investigator

Registry ID
AEARCTR-0011427
DOI
https://doi.org/10.1257/rct.11427-1.0
First registered
Last updated
Status
Completed
Trial dates
2014-06-01 – 2017-06-30
Countries
United Kingdom of Great Britain and Northern Ireland ()
JEL codes
D8, D9, J2, J3
Registry record
http://www.socialscienceregistry.org/trials/11427

Abstract

Using field and laboratory experiments, we demonstrate that the complexity of incentive schemes and worker bounded rationality can interact to affect effort provision, by shrouding particular attributes of the incentives. In our setting, complexity leads workers to over-provide effort relative to a fully rational benchmark, and improves efficiency. We identify contract features, and facets of worker cognitive ability, that matter for shrouding. We find that even relatively small degrees of shrouding can cause large shifts in behavior. Our results illustrate important implications of complexity for designing and regulating workplace incentive contracts.

Intervention

There was a staggered roll out of a static incentive system paying piece rates for manual work.

In a field experiment denoted INDIVIDUAL trial, the intervention gave treatment workers a dynamic financial incentive to work slower, without any motives related to social or group pressure.

In a field experiment denoted GROUP trial, the intervention gave treatment workers a dynamic financial incentive to work slower, but also potential non-financial motives related to social or group pressure.

Experimental design

The introduction of static incentives happened on a staggered basis across areas of the firm's warehouse. The timing of introduction was not randomized. We compare before and after, and compare to a control warehouse that did not have a change in incentive schemes, to estimate the impact on worker effort of introducing static incentives.

The INDIVIDUAL trial randomly assigned new hires to treatment and control on a weekly basis. For each cohort of new hires the trial lasted 12 weeks. Weeks 1 to 3 were a baseline period in which incentives were the same for treatment and control. In weeks 4 to 6, treatment and control had the same contemporaneous incentives, but treatment had dynamic incentives to work slower. This is because, in weeks 7 to 9, the incentives of treatment workers were a function of their speeds in weeks 4 to 6, whereas incentives for control workers were independent of previous speed. Specifically, working faster in weeks 4 to 6 would make it harder to earn money in weeks 7 to 9 for treatment workers. In weeks 7 to 9, control workers had similar dynamic incentives, because in weeks 10 to 12 their incentives were a function of speed in weeks 7 to 9.

The GROUP trial randomly assigned a portion of all workers in the warehouse to treatment and control. Each month, treatment and control workers had the same contemporaneous incentives, but treatment workers had dynamic incentives. The average speed of all treatment workers doing a given task in a given month determined the incentives of both treatment and control workers in the next month for that task. Specifically, if treatment workers were faster, it would make it harder for everyone to earn incentive payments next month. Control workers did not influence anyone's incentives. The trial lasted 10 months, after which incentives were determined on the average speeds of all workers.

Design details

Randomization method
Randomization was done by the researchers using a computer.
Randomization unit
The individual worker was the unit.
Clusters
NA
Observations
Introduction of static incentives affected the entire workforce. Over the course of our analysis period we expected to have roughly 4,000 workers and ended up with 4,580 (including the control warehouse). We planned to have roughly 650 workers for treatment and control, respectively, in INDIVIDUAL (we ended up with 1,294 workers), and to assign 40 percent and 60 percent of the workforce to treatment and control, respectively, in GROUP (this yielded 1,447 workers).
Arms
We planned to have roughly 650 workers for treatment and control, respectively, in INDIVIDUAL (we ended up with 631 and 663, respectively), and to assign 40 percent and 60 percent of the workforce to treatment and control, respectively, in GROUP (this yielded 573 and 874, respectively).
Primary outcomes
Worker effort (output) as measured by units per hour.

Cited by

No citation data available for this registration yet.

Keywords