• Selected
• Entire List

Abstract:
This study experimentally tests the effect of information transparency on the probability of coordination failure in global games with finite signals. Prior theory has shown that in global games with unique equilibrium, the effect of information transparency is ambiguous. We find that in global games where the signal space is finite, increased transparency has two effects. First, increasing the level of transparency usually destroys uniqueness and precipitates multiple equilibria, so that the effect of transparency on coordination depends crucially upon which equilibrium is actually attained. Second, the level of transparency determines which of these equilibria is risk dominant. We find that increased transparency facilitates coordination only if it switches the risk-dominant equilibrium from the secure equilibrium to the efficient equilibrium. When the converse is true, improved transparency can be dysfunctional because it increases the probability of coordination failure.

Abstract:
A continuing goal of experiments is to understand risky decisions when the decisions are important. Often a decision’s importance relates to the magnitude of the associated monetary stake. Khaneman and Tversky (1979) argue that risky decisions in high stakes environments can be informed using questionnaires with hypothetical choices (since subjects have no incentive to answer questions falsely.) However, results reported by Holt and Laury (2002, henceforth HL), as well as replications by Harrison et al. (2005) suggest that questionnaire responses and decisions in "low" monetary payoff environments do not effectively predict decisions in "high" monetary payoff environments. Thus, a potential implication of the HL results is that studying decisions in high stakes environments requires using high stakes. Here we describe and implement a procedure for studying high-stakes behavior in a low-stakes environment. We use the binary-lottery reward technique (introduced by Berg, et al (1986)) to induce preferences in a way that is consistent with the decisions reported by HL under a variety of stake sizes. The resulting decisions, all made in low-stakes environments, reflect surprisingly well the noisy choice behavior reported by HL’s subjects even in their high-stakes environment. This finding is important because inducing preferences evidently requires substantially less cost than paying people to participate in extremely high-stakes games.

Inducing preferences, high-stakes, experiments

Abstract:
Noncompliance is a quantitatively important phenomenon that significantly affects revenue sources for governments. This phenomenon raises challenging questions about the determinants of tax reporting and about the appropriate design of a tax system. This paper provides specific empirical insights using an experimental approach to evaluate the effects of systematic sales tax monitoring and the determinants of sales tax compliance. The results indicate that if perfect monitoring is instituted without other complementary policies, an increase in tax revenues is not the likely outcome. Once people have chosen their level of tax compliance, a stepped up policy of increased monitoring aimed at reducing fiscal fraud may not necessarily increase tax revenues. The reference-dependent effect observed in the data suggests that individuals will try to recover their losses following any policy changes even if it means taking more risks.

sales tax, perfect monitoring, experimental economics, reference-dependent effect