Essays on Belief Updating and Biases

Abstract

Through four independent research papers, this dissertation contributes to a better understanding of how individuals integrate new incoming information into their beliefs. <br /> Chapter 1, starts with the consideration that, in many relevant decision contexts, individuals are affected by a wide array of behavioral biases. This chapter investigates the impact of social learning on a broad range of behavioral biases, reflecting economically relevant settings. Through an online experiment, I document how social learning can amplify errors stemming from behavioral biases, leading to worse group outcomes. This is due to wrong beliefs leading individuals to engage in social learning sub-optimally. These results shed light on settings where cognitive biases affect decision-making in the presence of social learning, such as interpreting statistical information or making investment decisions. My results suggest that social learning often does not eliminate, and sometimes exacerbates, the impact of cognitive biases in such settings. <br /> In Chapter 2, we study how contingent thinking affects belief updating. This type of assessment is pervasive in real-world contexts, whether within economic contexts (e.g., contingent contracts or acquiring information through experimentation) or in other domains (e.g., a doctor considering what they would learn from running a test on a patient). Overall, our results show that contingent thinking increases deviations from Bayesian updating, and this effect can be attributed to hypothetical thinking. We also investigate how the features of the information structure affect this effect and find that reasoning fully offsets the negative impact of hypothetical thinking when the signal-generating process is symmetric but not when it is asymmetric. <br /> Chapter 3 also studies mechanisms of belief updating in an abstract setting. The chapter is motivated by the idea that agents undertaking economic decisions are exposed to an ever-increasing amount of information sources. Hence, this chapter investigates how the number of available information sources impacts agents’ ability to (i) select reliable sources and (ii) effectively use their content to update their beliefs. I set up an online experiment informed by a simple automata decision-making and belief-updating model. In line with theoretical prediction, participants’ source selection performances deteriorate as the number of available sources increases. Also, their belief-updating performance worsens, showing a trade-off between source selection and belief-updating performances. These results may help guide policy-making decisions by providing evidence on externalities of information production. <br /> Chapter 4 starting point is that overconfidence is one of the most ubiquitous cognitive biases. There is copious evidence of overconfidence being relevant in a diverse set of economic domains. In this chapter, I relate the recent concept of cognitive uncertainty with overconfidence. Cognitive uncertainty represents a decision maker’s uncertainty about their action's optimality. <br /> I present a simple model of overconfidence based on cognitive uncertainty. The model relates these concepts theoretically and generates testable predictions. I propose an experimental paradigm to cleanly identify such theoretical relationships. In particular, I focus on overplacement and find that, as predicted, cognitive uncertainty is inversely related to overplacement. By exogenously manipulating cognitive uncertainty through compound choices, I show a causal relationship with overplacement. <br /> Considered jointly, the four chapters of this thesis point towards two key insights. First, studying the interaction between systematic mistakes in beliefs about oneself and others, and other (belief) biases, is key to understanding how these biases evolve and persist over time. This sheds light on how some biases may persist in the aggregate, even when individuals are exposed to feedback. Second, theoretically irrelevant features can have a great impact on how effectively individuals integrate new information into their beliefs. Specifically, referring to Chapters 2 and 3, updating beliefs contingently as opposed to conditionally and having access to more or fewer information sources largely affect mistakes in belief updating.