Comparing Univariate Forecasting Techniques in Property Markets

Abstract

Executive Summary. This article presents a visual comparison of out-of-sample turning point performance as well as a brief comparison of forecast accuracy statistics of spectral analysis against other univariate techniques such as ARIMA modeling and exponential smoothing. Conventional forecast accuracy statistics show that exponential smoothing models are highly comparable with, and generally outperform, the other more complex model structures but only in those property markets when a stable trend is present. However, the article also demonstrates that such models perform poorly in turning point prediction. By way of contrast, the research shows that both the ARIMA and spectral regression modeling processes are capable of predicting turning points in property markets.