LAUSR

Abstract Canonical demand studies and fiscal policy simulations rest on the assumption that consumers react symmetrically to price increases and decreases. Such assumption has been challenged from both the empirical and theoretical points of view. We propose theoretically consistent empirical specifications to estimate discrete choice models (random utility DCM) and continuous demand systems (EASI and AIDS demand systems) that allow for reference prices and asymmetric own- and cross-price demand response. Our application focuses on the demand for sugar-sweetened beverages in Great Britain, using transaction-level household purchase data and different product aggregation levels. We find substantial evidence of asymmetric consumer response and loss aversion, with a stronger response when prices rise above their reference level. Our results hold for both DCMs on highly differentiated products and demand systems on aggregate product categories, and are robust to alternative model and reference price specifications. Simulations of taxes and subsidies on soft drinks shows that ignoring asymmetry may lead to biases, especially when predicting price cuts.