We estimate 13 fruit demand functions using a Korean micro consumption panel data set taking into account consumers’ quality choice behavior, and investigate whether individual household’s unit values are more appropriate than aggregate price indices as the price variables of the demand functions. Three types of demand models are estimated: 1) the one that employs unit values as price variables, dealing with the quality shading issue of Deaton (1988), 2) the one that employs the quality-adjusted prices and requires using both unit values and price indices, and 3) the one that employs price indices without considering the quality choice issues. We filled many missing unit values using a multiple imputation method. Models 1 and 2, taking into account the quality choice behavior of consumers, showed quite similar own price elasticity estimates. Also, Model 1 identified small but statistically significant quality shading phenomena. On the other hand, Model 3, which applies the usual price indices only, showed relatively unstable and possibly biased elasticity estimates.