景気動向のモデル分析―そのフロンティア―, Analysis of the Business Cycle Using Probability Models -the DFM and Possible Future Developments-

本論文では Dynamic Factor Model (DFM) を中心に景気動向のモデル分析のフロンティアを紹介するとともに，その発展方向について考察する．まず２節では DMF の基本型とその推定法について述べる．３節では景気変動が景気の拡大期と縮小期で異なることを想定し，レジーム間のスイッチ（景気の転換）を組み入れたモデルとその推定法について紹介する．さらに本論文の基本的な考え方としてレジームが観察可能であるときのモデル化について考える．４節では景況感をモデルに組み込み景気指標の作成に利用することを考えるとともに，その適用例を紹介する．５節においては４節のモデルをわが国のデータに適用し，2000年の景気の山を推定する．６節ではレジームスイッチの定式化を吟味する．特にスイッチの推移が１階のマルコフ課程に従うか，推移確率は他の外生変数に依存してないのか，そして推移確率を記述するパラメータが時間に関して不変かを統計的に調べる．７節は以上をまとめた結語に当てられる．, In this article, recent research developments in the field of business cycle analysis are summarized. The main focus is on the Dynamic Factor Model with Markov switching and a couple of possible future developments are proposed. In section 2, the prototype of DFM and its estimation procedures are introduced. In section 3, assuming that patterns of business cycle movements are differ between the expansion and the contraction periods, the switch mechanism between the regimes is described by the Markov Process. The fundamental idea in this paper is that the regimes are observable which greatly simplifies the estimation of the model. In section 4, data on business sentiment is incorporated into the DFM with a view to stabilizing the resulting business cycle. In section 5, the turning point of the business cycle in 2000 is estimated using the DFM proposed in section 4. In section 6, the formulation of the switch mechanism in Japanese business cycle is statistically explored paying particular attention to : whether the first order Markov process is appropriate, whether the transition probabilities depend on other exogenous variables and whether the parameters describing the probabilities are constant regardless of time. Section 7 summarizes the above and concludes this article.