دشواری نکته سنجی آنچه که خیلی کساد است: قوانین تیلور و سیاست های پولی ژاپنی

Observers have relied increasingly on simple reaction functions, such as the Taylor rule, to assess the conduct of monetary policy. Applying this approach to deflationary or near-zero inflation environments is problematic, however, and this paper examines two shortcomings of particular relevance to the Japanese case of the last decade. One is the unusually high degree of uncertainty associated with potential output in an environment of prolonged stagnation and deflation. Consequently, reaction function-based assessments of Japanese monetary policy are so sensitive to the chosen gauge of potential output as to be unreliable. The second shortcoming is the neglect of policy expectations, which become critically important as nominal interest rates approach zero. Using long-term bond yields, we identify five episodes since 1996 characterized by abrupt declines in Japanese inflation expectations. Policies undertaken by the Bank of Japan during this period did little to stabilize expectations, and the August 2000 interest-rate increase appears to have intensified deflationary concerns.

Monetary policy must be evaluated, both by those conducting the policy and by those in markets and in the political realm holding it accountable. In recent years, the academic literature as well as the financial press and pundits have relied increasingly on simple monetary policy reaction functions, such as the eponymous rule of Taylor (1993), to make that assessment. Using (seemingly) readily available data and requiring little or no estimation, an analyst using a Taylor (or similar) rule can give a simple yes or no answer to the question of whether policy at a given moment is too tight or too loose. For academics, Taylor-like reaction functions are appealing to the extent that they can be interpreted as reasonable approximations to the optimal instrument rules derived from simple, micro-founded macro models.2 As with most current forms of comfort in macroeconomics, however, Taylor rule analysis becomes problematic when confronting a low-inflation environment, and particularly when facing the experience of Japan’s Great Recession. Many commentators have focused on the obvious difficulty that when the central bank’s nominal instrument interest rate nears zero, it is impossible to evaluate actual policy measures by means of a Taylor rule that might suggest negative instrument interest rates.3 But this self-evident concern is misplaced in at least two ways. First, even when the instrument interest rate is zero, the same theoretical approaches that underlie Taylor rule analysis imply that a credible commitment to a future course of interest rates should have the same effect as a movement in the interest rate today. 4 Thus, the issue for monetary policy assessment when interest rates near zero becomes how to discern whether such a commitment exists and is believed by the market and households. Second, even if there is sufficient room left in nominal interest rates to make an assessment based on some version of a Taylor rule, uncertainty over the potential rate of growth in the economy is likely to rise as inflation approaches zero. This is both a statistical artifact, a result of the econometric methods employed to make most “top-down” estimates of potential output, and a substantive problem, reflecting the difficulty of distinguishing between structural change and deficient aggregate demand in times of financial distress (almost inevitably coincident with low inflation/deflationary periods). The Taylor rule assessment depends upon an estimate of potential output to such a degree that its results are extremely fragile to variation or uncertainty about this estimate, contrary to the usual blithe assumption that potential output is known with certainty. Both of these issues played critical roles in the debate over what happened in Japan over the last decade, and particularly in the assessment of monetary policy there. In the early part of the Great Recession, from 1992 to 1998, differing assessments of the downturn’s severity between various Japanese government agencies and the market, based on differing estimates of potential growth, determined the relative willingness to undertake countercyclical macroeconomic stimulus.5 The Bank of Japan (BoJ), though lacking legal independence until April 1998, participated in these debates, and had to make its monetary policy judgments on the basis of some estimate of the output gap (or a similar measure of economic slack). From the time that the zero nominal interest rate bound was approached in early 1998, however, the BoJ was confronted by the need to undertake a policy of quantitative measures, and to consider purchases of “unconventional” assets in order to pursue its policy goals. Thus, the difficulty in assessment of monetary policy became one of whether the BoJ’s actions constituted a credible commitment to a future path of interest rates, and then whether that commitment was having the desired effect. Admittedly, Japan’s descent into and then persistence of deflation is a particularly difficult case for discerning whether or not monetary policy is too tight. It is through such difficult cases, however, that the limitations of Taylor rule assessments of monetary policy are most clearly illustrated. Therefore, we have undertaken this study to demonstrate the difficulties of discerning whether monetary policy was too tight in Japan during the Great Recession, at least by utilizing the reaction–function approach.6 Proceeding chronologically if not logically, we first address the difficulties presented by uncertainty over potential output in Japan. We catalog the variety of approaches undertaken by previous investigators to estimate potential output for use in Taylor rule assessments of Japanese monetary policy, and show the dispersion of findings that have resulted in the Taylor rule assessments. We then discuss the real reasons for uncertainty about Japanese potential output over this period, and trace how these account for the major divergences in ex-post evaluations of BoJ monetary policy. Turning to an assessment of policy expectations, we propose gauging the credibility of anti-deflationary monetary policy by looking for occurrences of “deflation scares,” along the lines of Goodfriend’s (1993) analysis of “inflation scares” in the U.S. After jointly examining a chronology of the policy measures taken by the Bank of Japan and the behavior of long-term interest rates over the past several years, we conclude that the central bank’s actions were ineffective at preventing declines in long-term inflation expectations.

We have considered two issues regarding the applicability and informativeness of Taylor rules as a means of monetary policy assessment in the low-inflation environment of Japan’s Great Recesssion. First, we addressed the difficulties presented by uncertainty over potential output in Japan. A wide variety of approaches were undertaken by previous investigators to estimate potential output for use in Taylor-rule examinations of Japanese monetary policy, and these have resulted in a wide spread of assessments, as we document. Given the real (substantive and statistical) reasons for uncertainty about Japanese potential output over this period, we argue that these account for the major divergences in ex-post evaluations of BoJ monetary policy. Our conclusion is that while one can argue that there is a preferable means of estimating potential output for this purpose,27 that argument is hardly settled, and even after picking any one method there remains great uncertainty, sufficient to render the Taylor-rule assessments of Japanese monetary policy unreliable. We then considered the communication of future commitments about the course of interest rates through the BoJ’s quantitative and unconventional measures since the nominal interest rate neared zero. Our assessment of the impact of developments in the current stance and approach of Japanese monetary policy on expectations begins with a timeline of developments drawn up from official releases by the Bank of Japan. We find that deflation scares have occurred in Japan five times since 1996, notably prompted by an interest-rate increase in August 2000 that reversed a commitment to a future “zero-interest rate policy” until deflationary concerns would be dispelled, and one that persisted through much of 2002, seemingly associated with the BoJ’s public pronouncements regarding deflation and the inefficacy of monetary policy. Consistent with the evidence in Kuttner and Posen (2001b), and some a priori arguments made by the BoJ, we find no evidence that the various quantitative measures and expansions of eligible assets for open-market operations undertaken by the BoJ over this period had an impact on long-term bond yields. The reason, presumably, is that these measures are not expected to have a significant impact either on the future path of the policy rate, or on inflation. Taking ex-post Taylor-rule evaluation of monetary policy at face value, of course, ignores a vast range of issues in the monetary policy challenge in Japan at present. A partial list would include the asymmetries of deflation versus low inflation, the transmission mechanism breakdowns in the financial system, the economic and political interactions between monetary policy and financial reform, and the external effects and constraints of any exchange-rate movements for the yen. It also ignores broader normative approaches to evaluating the welfare of any given monetary policy. In a sense, this is precisely the point of this paper: that monetary policy evaluation cannot be done by rote rules any more than policymaking can.