Modelos estocásticos para el estudio y predicción de índices de precios de viviendas en el Ecuador

Abstract

The construction industry has a dynamic influence on the economy. In this field, it is known that construction projects carry a risk associated with fluctuations in the prices of construction materials. Estimating future values of construction price indices is important because contractors and builders use these values for budgeting. The aim of this research was to design stochastic models to explain and predict the construction price indices in single-family homes (IPCOU) and multi-family homes (IPCOM) in Ecuador. The design of univariate and multivariate models that consider exogenous variables was contemplated. Nine and eight predictors were found to be statistically significant in predicting the IPCOU and IPCOM, respectively. A set of fourteen potential models were subjected to training and validation, applying the cross-validation technique for time series. In the testing stage, it resulted that the Autoregressive Integrated Moving Average model known as ARIMA(1,1,2) and the multiple regression model to predict IPCOU values, and the ARIMA(2,1,2) model to predict the IPCOM, have good predictive capacity for the forecast horizon contemplated in this study. IPCOU and IPCOM prediction intervals were generated up to the year 2023, with a confidence level of 95%.