Abstract

This paper contributes to the growing literature on gen)etic epidemiology and health economics by employing an innovative approach to estimate local average treatment effects (LATE) using genetic information. Using data from the Understanding Society dataset, we apply a flexible instrumental variables (IV) framework that leverages individual genetic variants as instruments to estimate heterogeneous effects of arthritis on equivalized income. Our findings reveal significant negative effects of arthritis on income within specific genetic subgroups and positive and significant effects on other subgroups, possibly driven by the low-income status of its members who may rely heavily on welfare programs. This heterogeneity in the effects underscore the importance of subgroup analysis, which would otherwise remain hidden under standard 2SLS estimation. To increase the credibility of our analysis as well as the precision of our estimates, we build on the testing approach proposed by Apfel et al. (2023) to distinguish valid instruments from invalid ones under specific conditions (namely that a relative majority of instruments is valid conditional on compliance) and implement it in the context of genetic data. Additionally, we benchmark our results using Mendelian Randomization (MR), an alternative IV approach that confirms the robustness of our findings. This study provides new insights into the economic consequences of chronic conditions like arthritis, with important implications for public health policy and economic interventions.