We empirically compare the accuracy and precision of representative Least Squares, Maximum Likelihood and Bayesian methods of estimation. Using an approach similar to the jackknife, each method is repeatedly applied to subsamples of a data set on the property market in Bogotá, Colombia to generate multiple estimates of the underlying explanatory spatial hedonic model. The estimates are then used to predict prices at a fixed set of locations. A nonparametric comparison of the estimates and predictions suggests that the Bayesian method performs best overall, but that the Likelihood method is most suited to estimation of the independent variable coefficients. Significant heterogeneity exists in the specific test results.