Predicting future forestland area: a comparison of econometric approaches.

摘要: Predictions of future forestland area are an important component forest policy analyses. In this article, we test the ability econometric land use models to accurately forecast area. We construct a panel data set for Alabama consisting county and time-series observation period 1964 1992. estimate using restricted sets-namely, from early periods-and out-of-sample values dependent independent variables precise tests model's forecasting accuracy. Three model specifications examined: ordinary least squares, dummy (fixed effects), error components (random effects). find that produces more accurate forecasts at state level than other specifications. This result is related completely control cross-sectional variation in variables. suggests estimated parameters better capture temporal relationship between economic