ارزیابی سیاست های پولی که با نرخ بهره اسمی تقریبا صفر

The non-negativity constraint on nominal interest rates may have been a major factor behind a putative structural break in the effectiveness of monetary policy. To check for the existence of such a break without making prior assumptions about timing, and to enable comparison between pre- and post-break monetary policy, we employ an identified Markov switching VAR framework. Estimation results support the existence of a structural break around the time when the de facto zero nominal interest rate policy was resumed and the effectiveness of monetary policy is seen to weaken since then although slightly positive effects from monetary easing still exist. J. Japanese Int. Economies20 (3) (2006) 434–453.

It is often argued that the effectiveness of monetary policy has significantly weakened during the 1990s. When we look closely at the Japanese macroeconomy during the “lost decade” since 1990,1 monetary policy does not indeed seem to have had any obviously stimulatory effects on the economy. Miyao (2000) points out three reasons for this: “(i) the yen's appreciation to just over 80 yen per dollar, (ii) the Bank of Japan's actions to lower the official discount rate or the call rate to a record low below 1%, and (iii) the series of bank failures that disclosed the serious bad loan problem in Japan's financial sector.” Moreover, as concluded in Kimura et al. (2002), the introduction of the zero nominal interest rate may be expected to have lessened the effectiveness of monetary policy still further, by reducing the scope for easing interest rates or, to put it another way, by denying the economy the traditional interest rate channel for monetary policy transmission. The principal aim of this paper is to check whether, due to the introduction of the zero nominal interest rate, the effects of monetary policy on the real economy are characterized by a structural break. Given such a break, the further aim is then to improve our understanding of the currently prevailing relationship between monetary instruments and other economic variables, especially when nominal interest rates are almost zero. In order to check for the existence of such a break without making prior assumptions about timing, and to enable comparison between pre- and post-break monetary policy, a recently pioneered econometric technique known as identified Markov Switching Vector Autoregression (MSVAR) is employed in this paper. With this method, structural breaks are expressed in terms of Markovian regime shifts, where the latter are themselves one of the outputs of the estimation process. As long as the regimes identified by the Markov switching estimation are long-lived and distinct, it is appropriate to analyze the characteristics of a putative structural break by comparing the impulse responses of different regimes. This paper is the first attempt to employ the MSVAR for the analysis on the effectiveness of monetary policy in Japan.2 The paper is laid out as follows. Section 2 reviews the related literature: looking first at recent work on Markov switching regressions, and then turning to previous research that makes use of VARs to investigate monetary policy transmission mechanism. Section 3 provides the framework for analyses in this paper. The estimation process for the Markov switching model, the derivation of the impulse responses, and the use of bootstrapping to establish confidence intervals are explained. In Section 4, results from an MSVAR model with an explicit interest rate channel are demonstrated. Section 5 further constructs two MSVAR models with an implicit interest rate channel, which suit the current monetary policy scheme employed by the Bank of Japan, and results from these models are also summarized. The aim here is to enable us to attain insight into the effectiveness of the Bank of Japan's current “quantitative easing” policy. Section 6 concludes the paper

Several intriguing results are obtained from the three identified MSVAR models estimated in this paper. The findings may be summarized as follows. (1) There seems to be a structural break in the macroeconomic dynamics describing the monetary transmission mechanism around the time when the Bank of Japan resumed the de facto zero nominal interest rate policy in the mid 1990s. (2) Even though there is no room for monetary easing by means of lowering interest rates, monetary expansion seems to have some slightly positive but statistically insignificant effects. (3) The impact of monetary policy on the macroeconomy using monetary expansion becomes significantly weaker after the structural break, suggesting that within the regime currently prevailing, monetary policy is not fulfilling its desired role. In short, it is now less effective than it was before the mid-1990s at influencing output or the price level. These findings support similar conclusions of Miyao (2000)28 and Kimura et al. (2002) with regard to the reduced effectiveness of monetary policy. Recent research by Boivin and Giannoni (2003), on the other hand, presents an alternative point of view: “Recent studies using vector autoregressions (VARs) find that the impact of monetary policy ‘shocks’—defined as unexpected exogenous changes in the Federal funds rate—have had a much smaller impact on output and inflation since the beginning of the 1980s … An alternative interpretation could thus be that monetary policy itself has come to systematically respond more decisively to economic conditions, thereby moderating the real effects of demand fluctuations. In this case, the change in the responses to monetary shocks would reflect an improvement in the effectiveness of monetary policy.” Observing, however, that the putative structural break occurs along with the introduction of the de facto zero nominal interest rate policy, it is scarcely plausible to assume that smaller responses to monetary shocks are due to an improvement in the efficacy of monetary policy in Japan. It may therefore be safely concluded that the effectiveness of monetary policy has indeed deteriorated in Japan during the 1990s. However, there exist other possible reasons for the break during the 1990s. This break may be caused by the burst of the bubble expectation of the 1980s, huge negative demand shock, yen's appreciation, banking sector problems including the onset of the BIS regulation, credit contraction, and Zombi lending, lower technology growth, etc. Since the break in this paper is purely based on the stochastic process following the hidden Markov chain, we cannot identify the cause of the break concretely. Hence, in our future research, we would like to estimate VAR models with a larger number of endogenous variables so as to analyze various factors for the break in the 1990s. However, since the number of parameters is already large compared to the number of observations in regime switching models, we believe that the Bayesian estimation technique as in Geweke (1999) must be useful. At the same time, employing other non-linear econometric technique, such as the TAR (threshold autoregressive) model seems also beneficial because it can relate developments in some variables to the structural break directly. In any case, the results from the three variables VAR still indicate that the effects of monetary expansion are significantly smaller after the break. They suggest that the current problem confronting the Bank of Japan, namely the deflation under zero nominal interest rates, is not tractable. Monetary expansion without the transmission channel of nominal interest rates lacks impact on macroeconomic variables. The existence of this break in the macroeconomic dynamics has some very important implications for macroeconomic modeling. This is true for both VAR and DSGE models. Needless to say, the VAR analysis of macroeconomic dynamics, which does not take into consideration a possible structural break, may lead the researcher to misjudge the current state of the economy and to make the wrong policy prescription.29 Even the DSGE analysis, regardless of whether it is subject to a weak or strong economic interpretation (following the terminology of Geweke, 1999),30 a model whose parameters are determined based on the whole sample may fail to explain the current state of the economy correctly. The recent tendency for the DSGE analysis to become more data-oriented, i.e. to offer a “stronger” econometric interpretation, as in Ireland (1999) and Smets and Wouters (2002), should be encouraged as it makes DSGE more realistic and hence more relevant to policy analysis. However, caution always needs to be applied when using the model to analyze the current economic state because there may have been some structural breaks and the economic dynamics at that particular time may be quite different from those that prevailed in earlier periods. When conducting monetary policy, it is more important to recognize the current economic dynamics than the average economic dynamics which were prevalent in the past.