Empirical Tests of Non-standard Preferences for Beliefs and Information

Abstract

This thesis consists of three chapters that conduct empirical tests of behavioural microeconomic theory. The first is a study of observational data, using reduced form and structural models, and the other two are based on laboratory experiments. The connecting thread between all three chapters is their focus on preferences for beliefs and information. Chapter 1, Biased Recall and The Dynamics of Beliefs: Evidence From Schools, uses reduced form and structural methods to test theories of motivated beliefs in the field. These theories predict biased recall of ego-relevant information and thus provide a compelling explanation for persistent overconfidence in a variety of contexts. However, they have remained largely untested in the field until very recently. This chapter does so in a critical context for the formation of beliefs about ability: schooling. Using within-individual variation, the analysis finds evidence of typical patterns of recall consistent with motivated beliefs theory: recall of grades is biased and becomes more biased as time passes, and biased recall is associated with overconfidence about future grades. A structural model is then estimated. The model allows for unobserved heterogeneity in tastes for biased recall, revealing large differences across types of students. The model is used to simulate the effects of an informational intervention correcting biases in recall, finding that it would reduce student welfare and widen existing attainment gaps. Chapter 2, Recall Distortion and Past Choices, inherits Chapter 1’s focus on recall and memory and marks the beginning of the part of the thesis that focuses on experiments. It is joint work with Rebecca Heath. The basic question addressed by this chapter is whether beliefs are distorted to enable optimism about the consequences of past choices. The experiment finds support for this hypothesis and obtains suggestive evidence that belief distortions increase after a few days due to memory loss. This has implications for the measurement of beliefs in many empirical settings, since it shows that beliefs can be endogenous to prior choices. The analysis also shows that individuals distort beliefs in precisely the same way even when they know they will find out the truth at the end of the experiment, implying that the ability to sustain belief distortions indefinitely does not influence the present motive to distort beliefs. Chapter 3, Temptation to Consume Information, was my job market paper. It makes several contributions to the growing literature on preferences and markets for information. In contrast to existing work, which focuses on when and why individuals avoid information, it instead explores whether individuals are tempted to consume information they would like to avoid. In doing so, it entails the first attempt to empirically examine whether preferences for information can be dynamically inconsistent. The analysis relies on an experiment, guided heavily by theory, to establish the prevalence of temptation for information. Almost half of all individuals, and 75% of those who want to avoid information, are classified as finding information tempting: undesirable but more difficult to resist at the time of access. This chapter introduces a methodological innovation to the empirical study of temptation by showing theoretically that, when facing a tempting but resistible choice, response time increases in self-control costs. Using this method, the analysis shows that preferences for commitment are highly predictive of experienced self-control costs. The findings in this chapter suggest that the provision of commitment devices to avoid information could be welfare-improving. They also raise new questions on the feasibility of strategic information avoidance when such devices are unavailable. The chapter closes with a discussion of implications for important contemporary digital markets, including social media and news.