Accurate estimation of price indices for residential property is an essential feature of real estate research, especially in view of recent efforts to forecast price trends for the 1990s. In this article, price trends are estimated by using the sales price, assessed value and date of sale for every residential property transaction between independent parties. This assessed value (AV) methodology is compared to the repeat sales (RS) method. This article develops a simple method for correcting the effect of the measurement errors associated with assessed value. We demonstrate that the large samples available with the AV method allow the measurement error problem to be reduced to negligible proportions. Using data on the Hartford, Connecticut metropolitan area, we find that price trends estimated from the AV and RS methods are substantially similar over a seven-year period. But the RS method is inefficient because it uses a relatively small subset of the data. Our results indicate that it remains inefficient when the researcher has a dataset much richer in repeat sales than ours.