Using double-debiased machine learning to estimate the impact of Covid-19 vaccination on mortality and staff absences in elderly care homes

Girma, Sourafel; Paton, David

Abstract

Machine learning approaches provide an alternative to traditional fixed effects estimators in causal inference. In particular, double-debiased machine learning (DDML) can control for confounders without making subjective judgements about appropriate functional forms. In this paper, we use DDML to examine the impact of differential Covid-19 vaccination rates on care home mortality and other outcomes. Our approach accommodates fixed effects to account for unobserved heterogeneity. In contrast to standard fixed effects estimates, the DDML results provide some evidence that higher vaccination take-up amongst residents, but not staff, reduced Covid mortality in elderly care homes. However, this effect was relatively small, is not robust to alternative measures of mortality and was restricted to the initial vaccination roll-out period.

Citation

Girma, S., & Paton, D. (2024). Using double-debiased machine learning to estimate the impact of Covid-19 vaccination on mortality and staff absences in elderly care homes. European Economic Review, 170, Article 104882. https://doi.org/10.1016/j.euroecorev.2024.104882

• Journal Article Type: Article
• Acceptance Date: Oct 1, 2024
• Online Publication Date: Oct 6, 2024
• Publication Date: 2024-11
• Deposit Date: Oct 25, 2024
• Publicly Available Date: Oct 1, 2024
• Journal: European Economic Review
• Print ISSN: 0014-2921
• Electronic ISSN: 1873-572X
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Volume: 170
• Article Number: 104882
• DOI: https://doi.org/10.1016/j.euroecorev.2024.104882
• Keywords: machine learning; vaccines; care homes; Covid-19.
• Public URL: https://nottingham-repository.worktribe.com/output/40868366
• Publisher URL: https://www.sciencedirect.com/science/article/pii/S0014292124002113