Do past decisions influence future decisions?

Abstract

The random lottery incentive (RLI) mechanism (which involves subjects taking many decisions, only a randomly chosen one of which determines the payment to the subject) is widely used in many experiments, and hence its validity is of crucial importance to the inferences that can be drawn from the experiment. Its validity has been investigated in two main ways: first, in simple experiments where subjects were presented with just two decision problems, and second, in experiments where subjects were presented with many problems. In the first set, the hypothesis under test was whether the decisions made on the two problems were connected; in the second set, whether the set of all (previous) decisions affected the current decision. This article combines and extends these two ways by investigating, in experiments with many questions, whether the immediately preceding decision affects the current decision. This is a cognitively less-demanding hypothesis than that all previous decisions are taken into account, but allows for an indirect effect of previous decisions on current ones. Reassuringly, we find little effect and hence our results complement the previous evidence indicating that the RLI mechanism is robust and can safely be used.

Keywords:

• contamination
• experiments
• nonexpected utility preferences
• random lottery incentive mechanism
• separation

JEL Classification:

• C91
• D81

Notes

¹ Lee (2008) is also relevant.

² http://people.few.eur.nl/wakker/miscella/debates/randomlinc.htm

³ For otherwise there could be no contamination.

⁴ See Wakker (2010), though we should note that he prefers to call it rank-dependent utility (or even Prospect Theory).