Macroeconomic Forecasting Using Pooled International Data
重新检验了Garcia-Ferrer等人用池化技术预测九国产出增长率的模型，发现状态空间方法下的简约固定参数模型（不含自回归部分）预测效果优于AR(3)LI模型。
In a recent article, Garcia-Ferrer, Highfield, Palm, and Zellner (1987) used pooling techniques to forecast annual output growth rates in nine countries. Examining various model specifications, they found that, among the fixed-parameter models investigated, a third-order autoregressive (AR) model with leading indicators, AR(3)LI, produces the best forecasts overall. A reexamination using state-space methods shows that a more parsimonious fixed-parameter model, which does not have an autoregressive part, leads generally to better forecasts than the fixed-parameter AR(3)LI model.