This dissertation is composed of three papers on distinct topics, each studying a different aspect of decision making. In chapter 1, we study why, even without material incentives for coordination or learning, social interactions influence individual decision making. Identifying why conformity arises absent explicit incentives faces the challenge that any rationalizing theory must rely on unobservable preferences or beliefs. Therefore, empirical distinction requires theories that make predictions beyond the basic dynamic of conformity. To that end, we propose and test a theory of self signaling in peer effects. The model generates (partial) conformity as a response to how others' choices inform one's own self-image. The mechanism of self-signaling for peer effects delivers unique, falsifiable predictions that we test experimentally. The theory predicts that the anticipation of learning others' choices and the garbling of information on others' choices will both deeply influence behavior. In two real-effort lab experiments we not only find treatment effects closely in line with the model's unique predictions, but also document the importance of self image in social comparisons.

In chapter 2, we attempt to identify present-biased procrastination in tax filing behavior. Our exercise uses dynamic discrete choice techniques to develop a counterfactual benchmark for filing behavior under the assumption of exponential discounting. Deviations between this counterfactual benchmark and actual behavior provide potential `missing-mass' evidence of present bias. In a sample of around 22,000 low-income tax filers we demonstrate substantial deviations between exponentially-predicted and realized behavior, particularly as the tax deadline approaches. Present-biased preferences not only provide qualitatively better in-sample fit than exponential discounting, but also have improved out-of-sample predictive power for responsiveness of filing times to the 2008 Economic Stimulus Act recovery payments. Additional experimental data from around 1100 individuals demonstrates a link between experimentally measured present bias and deviations from exponential discounting in tax filing behavior.

In chapter 3, we provide a universal condition for rationalizability by risk-averse expected utility preference in a demand-based framework with multiple commodities. Our test can be viewed as a natural counterpart of a classical test of expected utility, due to Fishburn (1975), in a demand setting.

This dissertation explores the interplay between heterogeneity in gain-loss attitudes and a leading model of expectations-based reference dependence (Koszegi and Rabin, 2006, 2007) (KR). Reference-dependent preferences posit that individuals consider how outcomes differ from some reference point (e.g. the status quo or an expectation) rather than evaluating the outcomes in isolation; these models have helped rationalize a number of empirical regularities inconsistent with neoclassical theories in economics.

Since the development of this KR expectations-based mechanism, a number of studies have sought to test the comparative static predictions in the exchange (Ericson and Fuster, 2011; Goette, Harms, et al., 2016; Heffetz and List, 2014) and labor supply (Abeler et al., 2011; Gneezy et al., 2017) contexts. The mixed experimental results initially cast doubt on KR’s ability to predict behavior. Importantly, these tests were all conducted under an implicit assumption of universal loss aversion – that individuals weight losses below the reference point more than gains above. Recent work, however, documents a substantial fraction of gain-lovers (Chapman, Snowberg, et al., 2018), meaning these experiments were incidentally testing a joint hypothesis of the KR model and loss aversion.

Throughout these chapters, experiments are specifically designed to overcome this potential confound by measuring gain-loss attitudes in a first stage and relating these measures to theoretical predictions in a second stage. The results provide evidence in support of the KR predictions after accounting for this underlying heterogeneity. Moreover, a Bayesian re-analysis of the exchange experiments suggests that a highly heterogeneous distribution of gain-loss attitudes best rationalizes the mixed results, with an estimated posterior indicating 35% to 55% of participants to be gain-loving.

This dissertation studies a notion called access-based beliefs, the idea that access to an alternative has a direct impact on beliefs of its value. As an example, individual's may lower their beliefs of alternatives to which they have low access, a phenomenon colloquially known as sour grapes. Using multiple settings in the lab and field, I demonstrate existence of access-based beliefs consistent with the sour grapes hypothesis and provide evidence for how it may be prevalent in real-world settings with an emphasis on the poor. In chapter 1, in an environment where learning channels are controlled for, I provide a proof-of-concept of access-based beliefs, showing that lower access to a subjective lottery reduces revealed beliefs of its value. In chapter 2, I study access-based beliefs in a real-world setting. I show that lowering access to a real-world investment (an SAT prep package) lowers beliefs of its expected effectiveness and reduces investment in it amongst a subsample of the population that would be expected to have the highest demand for it. In chapter 3, I explore the notion of relative access, that is, the value of an alternative is a function of whether or not others possess it. I find that a small but significant portion of individuals display a preference for conformity or non-conformity in consumption based on the societal distribution.

This dissertation investigates the persistence of strategic sophistication across different scenarios. In daily life, people often make decisions based on their beliefs about other people’s actions, and sometimes engage in iterated steps of reasoning. Open questions exist for how contextual aspects influence strategic choices. When given additional information about their opponents, or the strategic environment, do people’s strategic choices change accordingly? This dissertation uses experimental methods to answer these questions. This work helps to inform foundational models of strategic decision-making. It points to the significant effects of contextual variables on strategic behaviors that must be incorporated into models of choices.

In this dissertation I axiomatize and extend current behavioral economic models for risk preferences. Based on my theoretical findings, I also design experiment to test these theories.

In Chapter 1, I provide axiomatic foundations for salience and regret theory, and then evaluate their empirical validity. In Chapter 2, which is co-authored with Professor Charles Sprenger, we extend the model of salience theory into a multi-dimensional setting and test the new model using both existed dataset and novel experiments. In Chapter 3, I propose a method to test cumulative prospect theory.