This thesis develops a variety of econometric approaches in order to examine model specification and weighting choices in an era when computational power and efficiency is becoming less of a binding constraint and 'big data' becomes widely available. We apply our methodological contributions to housing markets, given their recently verified importance with regard to international financial stability. In particular, we develop two original forecasting routines based on high dimensional datasets, take simulated and empirical approaches to specifying large sets of spatial weighting matrices and we contrast structural (panel and single country based) vector autoregressions which are identified and analyzed in a number of ways. The appendices are reserved for introducing published econometric software and replication codes developed as a by-product of the main body of work.