Answers are grounded in this paper and cite their sources.
The authors study the estimation of causal treatment effects on demand when treatment is randomly assigned but prices adjust in response to treatment. They show that regressions of demand on treatment or on treatment and price lead to biased estimates of the direct treatment effect. The sign of the bias in both cases depends on the sign of the correlation of price with treatment, and the two biases point in the same direction. In most cases including an endogenous price control reduces bias but does not remove it. The authors also show how to test whether bias from an endogenous price response arises and how to recover an unbiased direct treatment effect using a price instrument. They apply the approach to the estimation of feature advertising effects across several product categories using supermarket scanner data and show that the bias when not instrumenting price can be substantial.
No takes yet. Share an insight, caveat, or question.
Jeziorski et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: