The ratio of household expenditure on a particular food to the quantity consumed is often used as a proxy for market price in cross-sectional demand studies. These unit values are likely to give biased estimates of price elasticities, so Deaton (1990) developed procedures for correcting these biases. However, empirical evidence on the bias created by unit values in demand systems is lacking so in this paper we use data collected specially to carry out comparisons with the results of using market prices. Our findings suggest that unit values, whether used in naïve or improved estimation procedures, provide poor approximations to the elasticities calculated with actual market price data.