Using spatial econometric specification in hedonic pricing model: Is there a problem?

Since the formal paper of Can (1992), several attention have been paid to correction of spatial pattern among residuals in hedonic pricing model. Recent development of the spatial econometric have facilitate the correction of such pattern. However such applications are usually based on strictly spatial context, ignoring that real estate data are in fact a collection of spatial data pooled over time. Such mechanical applications can introduce bias on estimated coefficients since temporal dimension is unidirectional effect as opposed to the multidirectional spatial effect. The paper addresses the possible bias generated on estimated coefficient when spatial weights matrix, as opposed to the spatio-temporal weights matrix, is used in hedonic applications. The results suggest that ignoring the temporal dimension and applying strictly spatial econometric tools can generate bias in estimated autoregressive coefficients while potentially generating other important problems.