本文修改當期景氣衰退指標（current depth of recession, CDR），重建 CDR 在景氣擴張期的資料，再由門檻模型（threshold model）的估計，調整正、負值範圍，消除景氣區分由外生決定 的缺點。本文以 7 大工業國、香港、馬來西亞、菲律賓、南韓以 及台灣等國家，自 1959 年至 2005 年的季資料為樣本，分別建構 以 CDR、修正 CDR（modified CDR, MCDR）為門檻變數的門檻 模型，進行樣本外預測效能的比較，評估 2 種門檻模型樣本外預 測的效能。實證結果發現，MCDR 的預測效能比 CDR 為佳，適 合納入門檻模型作為門檻變數。

This study aims to modifies the Current Depth of Recession (CDR) for proving that the modified CDR (MCDR) is more suitable to be threshold variable than the CDR. We first rebuild the CDR data in the expansion period, and then adjust the positive and negative range of the CDR based on the estimation of threshold model. The quarterly data of 12 countries including G-7 countries, Hong Kong, Malaysia, Phillip, South Korea and Taiwan over 1959 - 2005 are used to construct two types of TAR models which adopt the CDR and the MCDR as threshold variables. To examine the efficiency of out-ofsample forecast, we find that MCDR is more proper for being the threshold variable than CDR.

當期景氣衰退指標;景氣循環;樣本外預測;門檻模型

Current Depth of Recession; Business Cycle; Out-ofSample Forecasting; Threshold Model