ESRI Discussion Paper Series No.251
Forecasting of Coincident Composite Index from an Affine Term Structure Model: A Macro-Finance Approach

Tatsuo Ichikawa
Morgan Stanley MUFG Securities Co., Ltd. Executive Director
Graduate School of Social Science, Tokyo Metropolitan University
Hirokuni Iiboshi
Graduate School of Social Science, Tokyo Metropolitan University

The full text is written in Japanese.


In this paper, we propose and evaluate the forecasting model of Coincident Composite Index (CCI) using Affine Term Structure Model of interest rates that consider the tendency of term spread leading business cycle by approximately 24 months. Our forecasting model modified the macro finance model introduced by Ang, Piazzesi and Wei (2006) by incorporating multivariate Bandfilter with phase shift, introduced by Valle e Azevedo, Koopman and Rua (2006). The forecasting method of Ang e al. (2006) is VAR(1) and is suited of short term forecasts but not for long term forecasts. On the other hand, our model is suited for 1-2 year forward forecasts as we consider the phase shift between the cyclicality in the macro factors and variables. The empirical results using monthly CCI data shows our model dramatically improves accuracy in 12-24 month ahead forecasts. We also note that the CCI forecasting using the OLS estimation in 12-24 months ahead shows the coefficient of term spread is negative, but the macro finance model gives positive coefficient. It induces to the big difference of performance of forecasting between macro-finance models and OLS.

  • 1-6-1 Nagata-cho, Chiyoda-ku, Tokyo 100-8914, Japan.
    Tel: +81-3-5253-2111