Evaluación de modelos de predicción para la venta de viviendas[Evaluation of forecasting models for house sales]

The aim of this working paper is assessing the predictive ability of different econometric models with forecasting windows of 3, 6 and 12 months, in order to improve housing sales forecasts published by the Chilean Chamber of Construction. To do so, five different families of models are estimated, among which Bayesian Vector Autorregresive (BVAR) stands due to a wide acceptance in the last decade. The main result of this paper shows that, in most cases, BVAR models provide more accurate predictions than classical models. This is consistent with the evidence found in several macroeconomic and sectoral applications of this type of models.