Interaction Economics: Instruments that Measure Social Computational Systems
How to cite
Portner, Claus. 2018. "Interaction Economics: Instruments that Measure Social Computational Systems." AEA RCT Registry. AEARCTR-0003541. https://doi.org/10.1257/rct.3541-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
The purpose of this research is to test a set of fundamental economic theories that previously has not been tested experimentally. The main question is the extent to which workers require additional compensation for performing work that is less appealing and, if so, how large this effect is. The research is conducted using the online labor market Amazon's Mechanical Turk. For each experiment, we offered a job on Mechanical Turk and randomly vary the pay offered and the job characteristics. We observe workers once they click on our offered job. Once workers click on the offered job, they are presented with a description of the job and the randomly allocated job characteristics. We then observe whether the worker decides to perform the job and, if so, how much work they do. Once the experiment is over we surveyed the workers to collect basic demographic information (no identifiable information was collected).
Intervention
We ran experiments to test one of the oldest theories in economics—Adam Smith's theory of the Compensating Wage Differential. This theory has been influential since it was first proposed in the "Wealth of Nations." The theory has, however, never been validated by experimental data because the experiments required have been impractical. We demonstrate that we can measure four of the five dimensions of a job that Adam Smith defined in his theory:
1) Pleasantness of the work.
2) Amount of learning required to perform the job.
3) Probability of success in the job.
4) Constancy of employment available.
For each of these conditions, we designed experimental equivalents. We implemented this study with two different real-world tasks that are commonly found in Internet labor markets, and for each, we have designed manipulations to all four of these dimensions.
Experimental design
Once a worker clicks on our offered task, the worker will randomly be assigned to an experimental condition and a pay offered. Each experiment offers two or more conditions or treatments and a variety of levels of pay per task completed, and each worker will see one of these conditions and pay levels. We then observe whether the worker accepts the task with that condition and pay level, and, if so, how many tasks she performs. The number of tasks performed is the main outcome of interest. All information is gathered electronically by the system we have set up.
Design details
Cited by
No citation data available for this registration yet.