FORECASTING HOUSE PRICE INDEX IN TURKEY USING ARIMA TRANSFER FUNCTIONS AND ARTIFICIAL NEURAL NETWORKS (ANN)

Abstract

It is widely acknowledged that the Turkish economy was launched by the housing sector. As a result, both government policymakers and decision-makers in the building business are keenly interested in how housing prices change. Understanding the connections between changes in macroeconomic variables and changes in housing prices is crucial for properly tracking price changes in homes. With the use of autoregressive integrated moving average (ARIMA) and artificial neural network (ANN) models, the primary goal of this research is to predict Turkey's housing price index (HPI) with macroeconomic variables. Eleven macroeconomic parameters are chosen as the independent variables when predicting the housing price index (HPI). The quarterly time series data of these parameters are used for the time period between January 2010 to December 2022 to train the models. An in-depth analysis is done on the connection between these macroeconomic factors and changes in the housing prices index. 11 hypotheses are formed to test the model, from that GDP (β=0.001, p= 0.97), Gold prices (β=-0.00, p= 0.69), BORSA index (β=2.08, p= 0.52), monetary rate (β=-0.105, p= 0.27) and foreign trade (β=0.171, p= 0.26) are not affected on Turkey's HPI. Inflation rate (β=1.20, p= 0.000), an Exchange rate (β=0.09, p= 0.00), USD/TL (β=1.43, p= 0.000), Unemployment (β=-0.94, p= 0.005), Deposit interest rate (β=-0.42, p= 0.00) and money supply (β=-0.11, p= 0.04) are significantly affected in Turkey's housing price index. This study will help policy makers and investors who are taking investment decisions in Turkish real estate markets.