The exact measurement of the welfare costs of tax and price reform requires a detailed knowledge of individual preferences. Typically, first-order approximations of welfare costs are calculated avoiding detailed knowledge of substitution effects. The authors derive second-order approximations which, unlike first-order approximations, require knowledge of the distribution of substitution elasticities. This paper asks to what extent simple approximations can be used to measure the welfare costs of tax reform and evaluates the magnitude of the biases for a plausible size tax reform. In the authors' empirical examples, first-order approximations display systematic biases; second-order approximations always work well.