Sample Selection Bias and Repeat-Sales Index Estimates

Analysis of variations in house values among localitiesrequires reliable house value indices. Gatzlaff and Haurin(1994) indicate that traditional hedonic house value indexestimates, using only information from a sample of soldhomes to estimate value movements for the entire housingstock, may be subject to substantial bias. This paperextends previous work by adapting the censored sampleprocedure to the repeat-sales index estimation model. Usingdata from Dade County Florida, a house value indexconstructed from a sample of homes selling more than once,rather than all houses in a locality, is found to bebiased. the bias is shown to be highly correlated withchanges in economic conditions