PhD defence - Carl-Emil Pless

The empirical literature in economics has produced robust evidence revealing fundamental tensions between theoretical predictions and observed behavior, between laboratory and field findings, and among methodological approaches to understanding choice. Standard economic theory predicts rational utility maximization, but people systematically exhibit multiple behavioral biases simultaneously. Controlled experiments find frequent choice inconsistencies, while real-world scanner data suggest that behavior appears close to rational. Finally, behavioral interventions show large effects in academic settings but modest impacts when implemented by policymakers. This thesis addresses these contradictions through four independent chapters that develop theoretical, methodological, and empirical approaches in an attempt to reconcile these conflicting findings about consumer behavior.

Chapter 1 develops a unified theoretical framework for analyzing multiple behavioral biases simultaneously. Building on substantial empirical evidence that people typically exhibit several biases at once, the chapter presents a dynamic optimization model that integrates reference dependence, loss aversion, and present bias. The model shows that analyzing biases in isolation can mislead both theory and policy, as multiple biases may amplify or mask individual effects. Comparative statics reveal that increasing loss aversion or present bias may decrease current consumption through their impact on future reference points.

Chapter 2 provides empirical evidence on tax policy effectiveness using household scanner data from the YouGov Shopper Panel. Exploiting the COVID-19 border closures as a quasi-natural experiment, the chapter delivers the first direct household-level estimates of cross-border shopping’s impact on sin tax policy. Using a difference-in-differences design, the analysis finds that Danish households living near the German border increased their sin-good expenditure by 17.6% when cross-border shopping became unavailable. These findings suggest that while cross-border shopping dilutes sin taxes’ public policy goals, its impact on public health outcomes may be overstated when demand adjustments are driven by availability rather than standard tax effects.

Chapter 3 addresses the methodological challenges of applying revealed preference analysis to high-dimensional scanner data. Despite the attractiveness of scanner data, the fact that prices are only observed when consumers transact means that more than 98% of the price information needed to construct choice sets remains unobserved. The chapter develops a generalization of Afriat’s Theorem that accommodates partial price observability through support prices. This methodological contribution enables nonparametric revealed preference testing even when most prices are unobserved, allowing rigorous demand analysis without imposing restrictive parametric assumptions. The framework bridges the gap between the granularity of scanner data and revealed preference approaches to consumer demand analysis.

Chapter 4 uses the revealed preference framework to investigate heterogeneity in decision-making quality. Recognizing that computational constraints limit large-scale implementation of the chapter 3 approach, this chapter develops a tractable resampled efficiency index. Combining this measure with comprehensive psychometric data reveals that personality traits and emotional states significantly relate to real-world decision-making ability. Consumers reporting stress, anxiety, or financial concerns tend to make more efficient economic choices than their less worried counterparts, even after controlling for socioeconomic characteristics. On average, consumers exhibit inefficiencies equivalent to approximately half their budget.