Testing an IV method for reducing quality bias in demand systems estimations

Price elasticities of substitution estimated with demand systems and household survey data might suffer of an endogeneity bias due to measurement errors and unobserved quality substitution. The paper compares Deaton's and Cox and Wohlgenant's approaches commonly used to reduce this bias with a method based on instrumental variables. The test is conducted on selected food and beverage goods inVietnam. Price elasticities not corrected for quality substitution obtained with standard market prices are used as a benchmark. Results show that the instrumental variable method significantly reduces the endogeneity bias and performs better than the compared methods. Moreover, it is based on fewer assumptions and controls for measurement errors. Its main limitations are that it still implicitly assumes that price changes have no impact on quality and that it relies on identifying suitable instruments.