The challenges of measuring substitution elasticities in food scanner data.

Abstract

Demand-based price indices for food inflation and consumer welfare depend primarily on credible estimates of the CES substitution elasticity (σ). In the Feenstra–Broda–Weinstein (FBW) system, σ is jointly identified with the inverse supply elasticity (ω), which helps interpret supply-side responses. In high-frequency scanner data with substantial product churn and unbalanced panels, standard implementations that effectively use all products as instruments are fragile and can severely inflate σ. Using Monte Carlo simulations calibrated to UK scanner panels and applications to three Kantar categories, we show large upward bias in σ under FBW-WLS. We propose a transparent common-product IV strategy that yields more plausible elasticities and clearer inflation and welfare implications.