Quasi-experiments rely on design choices: a reflection on Renson and Westreich’s ‘What makes a study design quasi-experimental?’
International Journal of Epidemiology · 2026-10-14 · 55 (6)
DOI: 10.1093/ije/dyag207 ↗Abstract
Abstract Renson and Westreich’s recent article questions whether difference-in-differences designs should be considered quasi-experimental. While we agree that difference-in-differences requires careful interpretation and should not be uncritically privileged, we argue that this is not sufficient grounds for excluding it from the quasi-experimental category. We argue that quasi-experiments are defined not by the strength and transparency of their justification for causality, but by the presence of explicit design choices. Design choices vary in strength. At one end of the spectrum are designs that guarantee exogeneity, such as focusing only randomized allocation to an exposure; at the other are approaches that rely primarily on statistical adjustment. Difference-in-differences occupies an intermediate position because it combines an explicit design element: the selection of a suitable comparator, with the strong functional-form assumption of parallel trends. We further argue that no epidemiological design, including difference-in-differences, Mendelian randomization, or randomized trials, merits special status solely because of its label. Instead, the credibility of causal inferences depends on the quality of the specific design and analytical choices used to justify identifying assumptions.
EUVIMED