آیا معافیت های ساختاری در اثر سیاست های پولی ژاپنی وجود دارد؟ ارزیابی مجدد اثرات سیاست دهه از دست رفته

This paper employs block recursive structural VAR models with Markov switching for modeling monetary policy and private sector behavior of the Japanese economy. By estimating the endogenous structural breaks, we investigate the existence, number, and nature of breaks possibly implied by the monetary policy adopted between 1975 and 2002. Results indicate that the Japanese economic system is best described by a non-absorbing two-state model, with major break happened around 1996. We also confirm that the interest rate monetary policy was effective before 1996, while monetary base shocks are identified as monetary policy shocks only after 1996. J. Japanese Int. Economies22 (3) (2008) 320–342.

After the bursting of the asset price bubble in the early 1990s, the Japanese economy experienced a long-lasting economic stagnation. The industrial production index once reached a high of 101.04 in May 1991, and sixty-eight months were needed to break this high in January 1997.1 The growth rate of real GDP from the first quarter of 1991 to the first quarter of 2002 was only 0.92% per annum, which is also very low.2 Lack of aggregate demand necessarily reduced job opportunities. As a result, the unemployment rate rose from 2.2 percent (February 1991) to 5.4 percent (February 2002).3 For these reasons, this period of turmoil is referred as the lost decade ( Hayashi and Prescott, 2002) or the great recession ( Kuttner and Posen, 2001). In order to explain this economic stagnation, a number of competing hypotheses emerged from differing points of view, and have been tested empirically. These views are roughly classified into either supply-shortage hypotheses or demand-shortage hypotheses.4 As proponents of the supply-shortage hypothesis, Hayashi and Prescott (2002) emphasize the importance of the decline in the growth rate of total factor productivity (TFP). They argue that by treating TFP as an exogenous factor, the neoclassical growth model accounts well for the Japanese lost decade of growth. Concerning the effect of low TFP growth on demand-shortage, Miyao (2006) confirms the positive and persistent relationship between them by estimating a four-variable VAR model with stock prices, call rates (interbank rates), TFP, and GDP gaps. By applying historical decomposition, Miyao also finds that GDP gaps in the post-1993 period were associated with a series of negative productivity shocks. Regarding one source of low TFP growth, Caballero et al. (2006) point to the misallocation of credit by banks to unprofitable borrowers, appropriately referred to as zombie firms. In order to prevent their own bankruptcy, banks hesitated to stop extending loans to zombies. Caballero et al. argue that this reduced the potential profits from new and more productive entrants. On the other hand, the hypotheses of demand-shortage proponents consist of three views. The first view is called a credit crunch hypothesis. This hypothesis, unlike Caballero et al. (2006), interprets a decline of the banking sector's financial intermediary function as a cause of demand-shortage. The banks could not supply the necessary amount of funds to companies when needed for business investment since banks were suffering from non-performing loan problems. The second view emphasizes the misconduct of fiscal policies. Kuttner and Posen, 2001 and Kuttner and Posen, 2002, for example, estimate the effects of fiscal spending and tax cuts with a structural VAR, and find that a tax cut was an effective measure for stimulating the economy in the 1990s. However, the tax burden increase in 1997 was more than any of the fiscal stimulus packages that were implemented in the 1990s (Kuttner and Posen, 2001, p. 128). Therefore, they conclude that an inadequate choice of fiscal policy aggravated stagnation in the lost decade. The third view focuses on change in the effectiveness of monetary policy, a hypothesis that we investigate in this paper. Particularly in the late 1990s, when the nominal interest rate approached the zero boundary, monetary policy did not seem to have any apparent positive effect on the economy (Krugman, 1998). Thus, proponents of this hypothesis argue that monetary policy lost its effect during this liquidity trap period. In applied works, such an event is called a structural break. Although there exist several empirical works on this issue, both definition and treatment of a structural break differs among researchers. Therefore, we investigate this problem in a comprehensive manner. The goal of this paper is to address the following four questions. (i) Was there a structural break in the Japanese economic system in the post-1975 period? (ii) If a structural break is suspected, did it occur once, twice, or several times? (iii) Should we model a structural break as a permanent or temporary event? (iv) Do those structural changes affect the effectiveness of Japanese monetary policy? To answer these questions, this paper estimates structural VAR models with time-varying coefficients and error variances, having the time variation governed by an exogenously evolving unobserved discrete state.5 In other words, we combine traditional VAR modeling with Hamilton's (1989) Markov switching specification. By comparing the results of several Markov switching VAR models, we examine the nature of structural breaks on the effects of monetary policy in detail. Our main findings are, first, confirmation that the Japanese economy has experienced the structural break in the post-1975 period. Second, the timing of major break is most likely in 1996, but the structural break is better described by a reversible state changes. Third, the interest rate monetary policy affects the economy in a conventional way in State 1 (which is dominant in the pre-1996 era), but its effects have significantly diminished in State 2 (which is the main state in the post-1996 period). Lastly, monetary base shocks are identified as monetary policy shocks only in State 2. Also it seems that the shocks do stimulate the economy, although they are not strong enough to generate an inflationary pressure. The paper is organized as follows. Section 2 discusses the related literature. Section 3 presents the time-varying structural VAR model used in this paper. Section 4 explains data and discusses the empirical results, then Section 5 concludes.

The Japanese economy experienced a long-lasting economic stagnation in the 1990s. This paper focused on one of the hypotheses that explains this extraordinary phenomenon. Our goal was to verify if there exist any structural breaks in the effects of monetary policy in the lost decade by using a sophisticated econometric model. In this paper, several block recursive structural VAR models were estimated to evaluate the policy shocks of the interest rate and the monetary base, under the emergence of severe deflation in the 1990s. Our VAR included the following five variables: output, consumer prices, interest rate, monetary base, and exchange rate. In order to consider the possibility that VAR parameters shift in the sample period, the law of state evolution was modeled by applying the Markov regime switching model. Through comparison of competing specifications, we selected the non-absorbing two-state MSVAR model as the best model to describe the economic situation in the post-1975 period. The estimated results are summarized as the following. First, a possibility of structural break in the post-1975 period is confirmed. Second, the timing of a major break is most likely in 1996, but the structural break is a reversible phenomenon. For convenience, we call a state which is dominant before December 1995 as State 1 and another state which prevails after January 1996 as State 2. Third, the interest rate monetary policy affects the economy in a conventional way in State 1, but its effect has significantly diminished in State 2. Lastly, monetary base works as a monetary policy indicator only in State 2. In addition, it seems that its shocks stimulate the economy although they are not strong enough to generate an inflationary pressure. Based on the results, we re-evaluated the BoJ's monetary policy in the 1990s as the following. During the first half of the decade, which starts with the collapse of the asset price bubble, the BoJ has kept propping out the Japanese economy through a series of interest rate cuts. Since these cuts were conducted prior to January 1996 (which is classified as State 1), our estimation result implies that the BoJ's policy was effective in propping up the Japanese economy. Even after the interest rate approaches the zero boundary and that the interest rate manipulation has become an ineffective policy instrument, the BoJ has kept stimulating the economy through expanding the monetary base. This corresponds to the period in State 2, during which the monetary base innovations are identified as monetary policy shocks. Although State 2 becomes the dominant state in the post- 1996 period, it is important to recall the reversible nature of our state classification. As our estimation has identified, there are several occasions in the post-1996 period which are classified as State 1. Thus, the BoJ's interest rate manipulation, such as an implementation or suspension of the ZIRP, surely had some effects on output even in the post-1996 period. In sum, the BoJ's policies in the second half of the decade were still effective in increasing output, although they are not strong enough to generate an inflationary pressure to escape from the trap of deflationary phase. As a final contribution of this paper, we would like to mention one future research topic. In particular, studying the evaluation of both monetary and fiscal policy effects along with the possibility of structural breaks. For example, Bayoumi (2001) uses a standard VAR model to explain the slowdown in Japanese growth in the 1990s. Specifically, Bayoumi has investigated four possible explanations of the extended slump in the 1990s, two of which are the inadequacy and misconduct of monetary and fiscal policies. In reality, however, several authors have reported a decline in the estimated spending multiplier in the 1990s (see Kamoi and Tachibanaki, 2001 and Kawade et al., 2004). Hence we believe that the type of examination we employed in this paper should be performed in order to properly evaluate these policy effects.