Testing residential construction and real estate forecast accuracy within a regional structural equation model
This study examines the historical accuracy of 11 variables in the housing block of the Borderplex Econometric Forecasting Model for the El Paso metropolitan area. Forecasts in three year increments are combined to form a total of 15 observations for each variable between the years 1998 through 2003. The structural forecasts are true ex-ante, implying that they are generated using data from before the forecast period. ^ In order to compare two forecast models, they should be made “ex-ante and at the same time” (West 1996). As regional forecast accuracy is a relatively new area of research there are no structural forecast models suitable for direct comparison to the Borderplex Econometric Forecasting Model. As a result, a random walk is used as a benchmark. In variables which exhibit positive growth rates over the study period, a random walk with drift is added to increase the competitiveness of the benchmark. Furthermore, as there is no absolute standard for forecast accuracy of the housing sector, the analysis herein is limited to reporting the predictive accuracy of the structural equations, as opposed to evaluating accuracy. (Abstract shortened by UMI.) ^
Kelley, Brian Wayne, "Testing residential construction and real estate forecast accuracy within a regional structural equation model" (2005). ETD Collection for University of Texas, El Paso. AAI1427693.