This paper evaluates the performances of Japanese leading indicators in predicting business cycle turning points. We extract the business cycle component in leading indicators using the frequency selective filter proposed by Baxter and King (1999), and we try to clarify empirically whether or not the leading composite index and its component series truly lead the business cycle turning point dates officially determined by the Japanese government. We argue that if we utilize the evaluated properties of the component series, we may construct a composite leading indicator which has some desirable properties as requested. As an illustration we provide one such example.