Management Accounting Section Midyear Meeting

Do MTurkers Collude in Interactive Online Experiments?

• In Event: Ethics

Abstract

One of the issues that can potentially affect the internal validity of interactive online experiments is collusion: participants could act upon information shared through channels that are external to the experimental design. Using an experiment that provides strong incentives for participants to collude, I measure how prevalent collusion is among MTurk workers and whether collusion depends on experimental design choices. In the treatments that resembled the design of most other interactive online experiments, I find no evidence that MTurk workers collude. This suggests collusion is not a concern for data quality in the vast majority of interactive online experiments. However, in the treatment in which participants were explicitly told that communication with others is allowed, I find evidence of collusion. This suggests that collusion is possible and that it should not be ignored as a potential threat to data quality in experiments in which participants have strong incentives to collude.

Author

• Razvan S. Ghita, University of Southern Denmark