Essays in Dishonesty

Abstract

This thesis describes three different experiments investigating dishonesty. Chapter one investigates the use of default values and prompts in a tax filing system. Pre-populated fields simplify the process of filing taxes, thereby reducing the scope for errors. Such defaults may increase the scope for non-compliance if set incorrectly. The chapter describes an experiment investigating the effect of correct and incorrect defaults. The results show that setting defaults that underestimate taxpayers’ true liability produces a fall in compliance. Nudges designed to mitigate the adverse effect of pre-population are also described. Nudges using descriptive norms in a dynamic manner that react to taxpayer decisions raise compliance. The chapter concludes that the use of defaults is worthwhile only if the data is of sufficient quality. Chapter two describes a model for lying aversion containing cost elements in terms of the size of the lie told and in the positive deviation above a reference point reflecting the point at which someone becomes concerned about the credibility of the value being reported or about appearing boastful. An experiment based on a numeracy test where subjects have the ability to cheat by paying themselves for their performance is used to test the model. Two treatments are detailed using modal values from initial control sessions to set different reference points. The results show a greater propensity among subjects to report false values under the higher reference point consistent with the model. Chapter three details an experimental investigation into lying behaviour between two samples, one a sample of undergraduate student subjects the other of workers recruited through Amazon Mechanical Turk. Results from a senderreceiver game based on a lottery draw show a higher propensity to report partially false values among student subjects, consistent with a higher reputational concern on behalf of the workers compared to students.