Misspecified heterogeneity in alternative-specific constants: Omitted variables in discrete choice models for environmental valuation

Discrete choice models are widely used in environmental valuation. In this field, many logit-based specifications interact individual-specific variables with attributes, yet omit the interaction of the former with alternative-specific constants (ASCs). This oversight constitutes a fundamental econometric error that, surprisingly, persists even in leading field journals. In this note, we study the multinomial logit (MNL) model and derive a first-order asymptotic approximation for the omitted variable bias. Within this MNL setting, we demonstrate that neglecting ASC interactions generally contaminates all parameter estimates, yielding inconsistent estimates of both attribute coefficients and willingness to pay.