Psychological factors in welfare and policy design

This thesis investigates the role of psychological factors in the design of optimal policy, focusing on mental health and the social safety net. Around 1 billion people suffer from mental disorders [WHO, 2022], and those with poor mental health are disproportionately likely to live in poverty [Lund et al., 2010]. Mental disorders cause significant disturbances in cognition, emotion regulation, and everyday functioning [Hammar and Årdal, 2009], yet their role in economic policy design remains understudied. Chapters 1 to 3 build on my Job Market Paper, which focuses on whether social assistance effectively reaches people with poor mental health. • Chapter 1 develops a theoretical framework showing how take-up responses to policy separately identify the marginal value of benefits (need) and the cost of barriers. • Chapter 2 presents new empirical facts about mental health and the targeting of social assistance using Dutch administrative data. • Chapter 3 combines the theory and empirics to show that people with poor mental health have a 2× higher need for benefits yet face a 64% higher cost from barriers. I also show that reducing barriers would be twice as effective as increasing benefits. While Chapters 1 to 3 take a revealed preference approach, a question remains: should the planner normatively respect the observed choices of people with poor mental health? Chapter 4 (with Daniel Reck) generalises this idea, tackling the fundamental challenge of behavioural welfare economics: psychological factors can cause inconsistencies, forcing policymakers to take a stand on which choices reflect an individual’s true normative preferences. We show that incorporating normative uncertainty leads to a structured welfare criterion, and explore how the resulting notion of robustness shapes optimal policy in several examples. Throughout the thesis, I argue that understanding psychological mechanisms, and their normative consequences, is essential for designing effective policies.