Forecasting Washington State's Housing Market: A Comparative Analysis of Time Series Models and Economic Indicators

Abstract

This study examines Washington State's housing market by analyzing the relationship between new private housing units authorized by building permits and various economic indicators. Initially, an Ordinary Least Squares regression model was employed, which indicated significant autocorrelation in the residuals, as identified by the Durbin-Watson statistic. To address the autocorrelation and non-stationarity detected through the Augmented Dickey-Fuller and Ljung-Box tests, we employed differencing followed by ARIMA and Holt-Winters Exponential Smoothing models. The Holt-Winters model, in particular, proved more effective, showing lower prediction errors and providing a more accurate forecast, thus emerging as the preferred method for forecasting in this market context.