Sarah Clifford + Benson Tsz Kin Leung (External Seminar)

Talk 1: Sarah Clifford, “Tax enforcement using a hybrid between self- and third-party reporting”

Time: 14:30 – 15:15 BST

Abstract: We study behavioural responses to a widely-used tax enforcement policy that combines elements of self- and third-party reporting. Taxpayers self-report to the tax authority but must file documentation issued by a third-party to corroborate their claims. Exploiting salary-dependent cutoffs governing documentation requirements when claiming deductions for charitable contributions in Cyprus, we estimate that deductions increase by £0.7 when taxpayers can claim £1 more without documentation. Second, using a reform that retroactively shifted a threshold activating documentation requirements, we estimate that at least 64% of the response is purely a reporting adjustment. Finally, the reporting environment is the key driver of behaviour for a substantial part of the population who displays little responsiveness to financial incentives for giving.

Talk 2: Benson Tsz Kin Leung, “Learning in a Small/Big world”

Time: 15:15 – 16:00 BST

Abstract: This paper analyses how the complexity of the state space and the cognitive ability of individuals explain a wide range of abnormalities in learning behaviours. An individual tries to guess which of N possible states is true, but has limited cognitive ability that his belief is characterised by a finite automaton of size M. The learning problem is a small world if the complexity of the world N is much smaller than the cognitive ability of the individual M, and is a big world otherwise. I characterise the optimal learning mechanisms and show that first, the learning behaviour, i.e., the guesses of the individual, is close to a Bayesian individual in small worlds but is significantly different from Bayesian in big worlds. Second, it could be optimal for the individual to ignore some states in big worlds but not in small worlds. It explains ignorance learning behaviours, such as the uses of heuristic, correlation neglect, persistent over-confidence, inattentive learning, and other behaviours of model simplification or misspecification. Lastly, individuals are bound to agree, i.e., make the same guesses, in small worlds. In big worlds, individuals could disagree even when they receive a large sequence of public information.

Join via Teams