Abstract
This paper focuses on problems of predicting business cycle turning points in a regional economy. In the 1980s, the work of Neftci[18]clearly suggested that turning points are naturally defined in nonlinear models of regime switching. However, Neftci's Sequential Probability Recursion (SPR), which failed to signal the turning points, yields fewer false signals and earlier detections of turning points. This paper shows that a posterior probability of turning points change with fluctuations in the economy, and Neftci's model can be extended to improve the performance of turning point forecasts in Japan.
In addition, the duration and the level of peaks and troughs in business cycles fluctuate with industrial compositions, industrial localization and the capital composition of corporations in each prefecture. We then illustrate the application of this "extended Neftci model" using CI (composite index) on the duration of business cycles in Aichi and Gifu Prefectures and compare the business cycles of both prefectures in regards to duration based on CI.
The results of the analysis are summarized as follows. First, the extended Neftci's approach in this paper correctly forecasts all of the past turning points before they occurred in Japan. Second, it also shows better performance relative to the probability of regime switching that signals the predicted turning points in Aichi and Gifu Prefectures.
JEL classification: E32, L11, R11, R15
