ブル相場・ベア相場のベイズ分析

概要

論文の詳細を見る
This paper investigates Japanese "Bull and Bear" market using Markov switching models of Hamilton (1989) from the Bayesian view. Gordon and St-Amour (2000) suggests a theoretical model of bull-bear markets and empirical outcomes using monthly security data. Maheu and McCurdy (2000) and Pagen and Sossounov (2002) also use monthly data for their analysis. From the perspective of trading, we need a study of bull-bear markets using higher frequency data. In this paper, we use Markov switching model with 2, 3, and 4 states; the two states model has bull and bear states, the three states one has additional jump state, and four states one has upper and lower jumps. From the Bayesian model selection, the tree states model with is preferred. Also we find the GARCH effects are not negligible and the cycles of bull-bear terms are determined by hyper-parameters of the prior distributions of transition matrices.欧文抄録: p.295