FORECASTING OFFICE RENTALS USING BOX-JENKINS METHODOLOGY

Abstract

Office rentals fluctuate in response to changes in numerous factors. As rentals are good indicators of office market condition, there are attempts made to forecast rentals. The most common approach used is the econometric simultaneous equation approach. Though time-series approaches are not as popular as the simultaneous equation approach, they give forecasts as accurate as this approach. In addition, the amount of data required is less than the amount of data required by the econometric approach. Thus, where data is lacking, time-series approaches are more suitable. Among the various time-series approaches, Box-Jenkins methodology provides the broadest class of models and flexibility. Box-Jenkins methodology is applied in this study to predict future office rentals. Generally there are two main types of models in Box-Jenkins methodology. They are 1) univariate models which can be in autoregressive form, moving average form or mixed autoregressive-moving average form and 2) multivariate models, often referred as transfer function models, are models using the cause-and-effect approach. Both types of models are used in this study. The output series that are of interest are the prime grade A, prime grade B and secondary quarterly office rentals. The input series comprise the gross domestic product of the "Financial & Business Services" sector and the vacancy rates. This study finds that the future rents in the twelve quarters ahead increase for the three different grades of office space.