Speech given by Kazuo Ueda, Member of the Policy Board, at a Conference Sponsored by the National Bureau of Economic Research, European Institute of Japanese Studies, Tokyo University Center for International Research on the Japanese Economy, and the Center for Economic Policy Research, held at Swedish Embassy in Tokyo on September 22, 2000

September 22, 2000
Bank of Japan

Contents

1. 1. The Nature of the 1997-98 Recession
2. 2. The Monetary Policy Response to the Credit Crunch
3. 3. The Adoption of the Zero Interest Rate Policy
4. 4. The Transmission Channels of the ZIRP
5. 5. The Ending of the ZIRP
   1. (1) Arguments against the Rate Hike
   2. (2) Arguments for the Rate Hike
6. 6. Some Final Comments
7. References:

The Bank of Japan (BOJ) has gone through a unique experience since a few years ago. When I joined the bank's newly formulated policy board in April 1998, the overnight call market rate, the key policy instrument of the BOJ, was already below 0.5%. The economy was in the midst of the most serious recession in the post war period, although it took us a little while to realize this. We guided the call rate down to virtually zero in the first quarter of 1999 and followed up by promising to keep it there until deflationary concerns had been dispelled. We finally increased the rate up to 25 basis points last month after having kept the zero rate for one and a half years.

Today, I would like to discuss some of the key aspects of the evolvement of our thinking on monetary policy during the last 2-3 years. In so doing, I would like to focus specifically on the characteristics of the 1997-98 recession, the transmission process of monetary policy near zero rate, and backgrounds for the rate hike in August.

I must confess that I am putting myself in an extremely difficult position, especially on the rate hike issue, because, as reported, I was a dissentient in the rate hike decision. The policy board, however, makes a collective decision. So, I presume that I am responsible for both the rate hike decision of the board and my vote against it. Explaining the two positions in a coherent manner is a tough task. If I succeeded, I could be proving that I am not capable of thinking logically. But I will try my best. To anticipate, the case against the rate hike was straightforward. It was, however, not impossible to come up with arguments to support it.

1. The Nature of the 1997-98 Recession

At the outset, it is appropriate to discuss briefly the nature of the recession that started in 1997IV, which is what we were trying to respond to. Clearly, the most important characteristic of the recession was its credit-crunch aspect caused by the slow and inappropriate handling of the bad loan problem. In addition, the Asian economic crisis, a premature tightening of fiscal policy in 1997 and the Russian crisis in 1998 added to the severity of the recession.

More specifically, the onset of the Asian economic crisis and the absence of fundamental measures to address the bad loan problem finally resulted in the failure of three medium- to large-sized financial institutions in the fall of 1997. In particular, there was a small default in the call market in the case of the failure of Sanyo Securities, the first of the three that went under. The default generated a panic in the money and financial markets, became a trigger for the failure of the other two, and led to sharp increases in risk premiums and the demand for liquidity across the financial system.

Japanese banks, already suffering from bad loans, now were facing difficulty in raising funds. Naturally, they started calling in their loans to nonfinancial companies, especially small firms. Business fixed investment started to decline immediately with the onset of the recession. As can be seen in Figures [PDF 404KB]-Table 1, the decline in fixed investment is the major story about the recession on the aggregate demand side.

Strains in the financial system eased somewhat in the first half of 1998, but resurfaced with the increasing awareness of the problem of the Long-Term Credit Bank of Japan and later with the Russian economic crisis. Even large companies were feeling the pressure of the credit crunch. Many said in our interviews that "all the banks except the main are saying that they will not be able to roll over the loans at the year end." Of course, they had to cut back on their investment.

Such was the background for our policy making during the period.

2. The Monetary Policy Response to the Credit Crunch

Central banks around the world have responded to credit crunches in one way or another. The justification for the response has not been unique. They may have been worried about systemic risks or about the risk of generating a serious deflation in the future.

Anyway, a variety of instruments have been used by central banks to mitigate credit crunches. The typical instrument has been the injection of liquidity to the system. This has usually been supported by lowering of short-term interest rates.¹ Depending on the circumstances, more unorthodox interventions have been applied. During the collapse of the CP market in the early 1970s the Fed announced to lend automatically to the banks that extended loans to companies having difficulty rolling over their CPs. The Fed even announced to lend directly to nonfinancial corporations if necessary, though apparently they did not do this.

In the fall of 1998 the Fed injected large amounts of reserves to the financial system accompanied by three consecutive lowering of the federal funds rate. The Fed, reportedly, also participated in a meeting to discuss a bailout package for the Long-Term Capital Management.

During 1997-98, the BOJ also injected huge reserves (Figures [PDF 404KB]-Figure 1) and cut the call rate in September 1998, from around 0.43 % to 0.25 %.² Several other attempts were made to ease the credit crunch. The BOJ supplied liquidity to the CP and corporate bond markets by essentially lending short-term funds to banks against these instruments.

In addition, the BOJ carried out a version of twist operations to flatten the yield curve in the money market. The operations were a response to the steepening of the curve during the credit crunch at horizons of a few months. They also were a way to lower the Japan premium on dollar interest rates by letting Japanese banks acquire funds in the yen money market and then swap them into dollars.

All these measures were attempts to address the credit crunch nature of the recession or to do something about the impaired ability of the private financial system to carry out financial intermediation.

1. Goodfriend (2000) points out the signaling role of a rate cut in a liquidity crunch. Because the accommodation of increased demand for liquidity usually takes place automatically, he argues that it is better to have an accompanying rate cut to show the central bank's commitment to maintain financial stability.
2. Incidentally, this was a period of extreme volatility in the demand for money. It is almost possible to detect a negative correlation between monetary aggregates and the pace of real economic activity. Hence, use of a money supply rule, for example, the McCallum rule for the monetary base, would very likely have aggravated the economic instability during the period.

3. The Adoption of the Zero Interest Rate Policy

The BOJ's monetary policy, together with the re-capitalization of the credit guarantee associations and private banks, eased somewhat the strains in the financial system. But the financial instability did not disappear. Toward the end of 1998 and early 1999 the long-term interest rate increased and the yen appreciated sharply, and to levels probably unjustified by the strength of the economy at the time. To be sure, the real side of the economy was stabilizing temporarily thanks to the return of the fiscal stimulus; but there were uncertainties about the second half of the year and after. The rate of inflation was around zero. We worried that a dip in real economic activity from that point on might lead to a vicious cycle between declines in the rate of inflation and real economic activity.

Such a consideration led us to ease monetary policy in an unprecedented way, i.e., to lower the overnight call rate down to zero. The decision was made in February 1999. Eventually, the overnight rate settled at 0.01 % for lenders and 0.02 % or above for borrowers. We followed up in April by declaring that the "zero rate will be maintained until deflationary concerns are dispelled." The combination, the zero rate and the commitment to maintain it until a certain set of conditions are met, has been sometimes called the zero interest rate policy (ZIRP) outside the BOJ. I will use the terminology in this talk.

The ZIRP produced significant impacts on money and financial markets and then on the rest of the economy. The yield curve flattened considerably all the way out to 10 years. Rates on instruments with maturities of less than a year were virtually zero during most of 1999. The 10 year JGB rate had been between 1.6 and 2.0 % for the last one and a half years. Stock prices rose sharply in 1999. Differences in yield between corporate and government bonds narrowed substantially. Thus, the ZIRP forced investors to take risks that led to declines in risk premiums in many places. One of the few exceptions has been the bank loan market, where the growth rate of loans is still negative. This evidences the seriousness of the bad loan problem we have been facing.

Over time, the ZIRP, rising exports, fiscal stimulus and investment in IT related areas have produced a slow recovery from perhaps the most serious recession in the post war period. Incidentally, I think that the time pattern of the current recovery can be more clearly seen in movements in the index of production than in many problems ridden GDP statistics.

4. The Transmission Channels of the ZIRP

In retrospect, the ZIRP had exerted much larger effects on the economy than many had thought at the time of its inception. What were the reasons?

Needless to say, the ZIRP involved a lowering of the overnight rate by about 20 basis points, which produced the usual chain of impacts. But the effects of the policy seem to have been larger than justified by the small change in the short rate. The commitment component must have played a large role. I think that there have been essentially three mechanisms through which the commitment exerted strong effects on the economy. They all concern the effects on the term structure of interest rates and, through them, on the rest of the economy.

First, the commitment minimized policy uncertainties especially at the early stage of the ZIRP. In March 1999 some traders in the money market feared that the zero rate and accompanying liquidity injections by the BOJ would be only a temporary policy measure designed to let financial institutions go over the fiscal year end. Such a perception limited the effect of the zero rate on longer-term interest rates. Hence, in April the commitment to maintain the zero rate until deflationary concerns had been dispelled was announced. As a result, interest rates across the yield curve declined.³

Second, the commitment mitigated liquidity concerns of financial institutions. With the ZIRP, differences in funding costs among banks in the money market virtually disappeared. To see why, let us note that the zero rate on average meant zero for all financial institutions given that there were no negative rates.⁴ And, the zero was expected to stay for at least a few months by market participants because of the commitment part of the ZIRP. Any serious liquidity problem would drive up the interest rate, which would be counteracted by the BOJ. Hence, the ZIRP became a powerful tool to contain the liquidity concerns prevailing in the economy at the time. Some of the pickup in consumption and investment in 1999, I think, was due to the easing of liquidity constraints faced by households and firms. This feature of the ZIRP goes a long way toward explaining why a mere 20 basis point cut in the overnight rate produced such a large impact on the economy --a perspective consistent with the credit crunch type interpretation of the recession.

Third, there is a controversial academic part to the ZIRP. In a sense the ZIRP was an attempt to produce a larger stimulative effect on the economy than the zero rate. Even with no explicit commitment from a central bank, however, rational market participants would expect that the zero rate will be maintained as long as it is appropriate to do so.⁵ Hence, in order to go beyond that, the central bank has to commit to maintaining the zero rate into the future even after it becomes appropriate to raise the interest rate under a certain criterion.

This point is made briefly by Woodford (1999). Independently, Reifschneider & Williams (1999) carry out simulations using the FRB/US model in which the performance of two policy rules is compared. One rule simply dictates that the federal funds rate is equal to the Taylor rule rate when the latter is positive and is zero when the Taylor rule rate is negative. The other policy rule is similar except that it contains a commitment under which once the federal funds rate hits zero, the zero rate is maintained until the Taylor rule rate exceeds a certain positive number. The simulation shows that the second rule delivers a better performance. Intuitively, the second rule borrows easing from future monetary policy and by so doing compensates for the central bank's inability to lower the rate below zero when desirable.

The ZIRP in this sense comes close to other methods to go beyond the zero rate proposed by academic economists. For example, the proposal for the BOJ to buy large amounts of JGBs is an attempt to affect the long-term interest rate. Unless the BOJ can move the risk premium on long-term bonds systematically, the BOJ should be able to achieve the same result by making commitments about future short rates. Or, the Krugman proposal to reflate the economy can be implemented by simply announcing to maintain the zero rate long enough.⁶

What then did the statement "keep the zero rate until deflationary concerns have been dispelled" exactly mean? I must confess that there was some ambiguity in our commitment. We certainly had the first and second channels above in mind from early on. We knew of the third argument. But we rejected the adoption of the extreme Krugman type proposal if it meant a pursuit of a high inflation target like 4 or 5 %. Benchmarking our policy against the Taylor rule did not seem easy given wide ranges of the interest rate coming out of the rule. I will come back to this point again later.

Our stance became a more informal one. Many of the board members made remarks to the effect that " the zero rate will be kept until the risk of the rate of inflation falling down to a large negative number becomes small enough." This was further paraphrased to "inflation would fall a lot when aggregate demand goes down sharply. Hence, we would wait until we are confident that domestic private demand is on a sustained recovery path." But such a stance left room for a fairly wide range of interpretation, as we shall see below.

1. 3Tinsley (1999) discusses what central banks can do to generate stimulative effects when short rates are near zero. He suggests that the central bank writes put options on short term securities in order to minimize the expectation that short rates will be raised in the near future. In a sense we achieved the same end using a less formal approach.
2. 4Actually, there were some foreign banks that enjoyed negative funding rates. They entered into yen-dollar swap arrangements with Japanese banks at favorable terms and thus took advantage of the existence of the large Japan premium.
3. 5This argument probably assumes too much rationality on the part of market participants or the central bank. Therefore, the commitment that ensures the first channel does not seem meaningless.
4. 6Similarly, the ZIRP in this sense seems to overlap at least partially with proposals to generate a weaker yen.

5. The Ending of the ZIRP

By the second quarter of this year it became apparent that, with high probability, the economy was going to grow in fiscal 2000 at a much higher rate than the government's forecast of 1%. Naturally, there was much discussion among the board members, as shown in the published minutes, about the appropriate timing to end the ZIRP. There remained, however, the question of how much growth would be enough. Below I will highlight some of the key arguments for and against a rate hike and hope to illustrate the difficulties we faced.

(1) Arguments against the Rate Hike

In a sense it was straightforward to argue against the rate hike. Although the economy has started to recover, it had been in a long and serious slump. Prices are, with the exception of WPI, still falling. Hence, there is no rush to increase interest rates.

More formally, we may calculate the Taylor rule rate.⁷ The rate turns out to be negative under many sets of assumptions. For example, Hayakawa & Maeda (2000) present estimates of output gap under several assumptions. The most standard one, i.e., the one based on a production function with a linear trend for the productivity term, results in a number like 8 to 9 % gap for 2000. One needs to know the neutral level of the gap in terms of its effect on the rate of inflation. The rate of CPI inflation was around zero toward the end of 1996 when the gap was between 4 to 5 %. This implies that the gap is larger, on the deflationary side, by about 4 % than the neutral level. With a coefficient of 50% on the gap in the Taylor rule formula, the gap term already contributes -2 % to the interest rate. The inflation term also contributes negatively but with the size of the contribution depending on the target rate of inflation and the price index used. The growth rate of potential output consistent with the gap estimate is below 2%. Thus, there is no chance for the Taylor rule rate to become positive under such assumptions.

To be sure, the above calculation is a rough one at best. But it does indicate the possibility that the optimal level of the policy rate is still negative. Moreover, under the Reifschneider-Williams scenario discussed above we needed to wait until the rate exceeded a certain positive number rather than zero.

1. 7I am not suggesting that the Taylor rule is the benchmark we use. I am using the rule just as an illustration of an analytical approach to the determination of the optimal rate.

(2) Arguments for the Rate Hike

One can point out a number of problems with the above type of calculation. Perhaps the most serious one is the following. Yes, standard inferences result in estimates of a very large ouput gap, which would be putting large negative pressure on prices. In fact, this was exactly the line of logic that guided us to the adoption of ZIRP. To our surprise, prices did not decrease much in 1999. Standard Phillips curve type calculations would have suggested a decline of CPI or WPI by about 2% in 1999. Actually, as of December 1999 core CPI declined only by 0.1%, and WPI, by 0.5% over a year earlier. Surely, there were a rise in oil prices and the lagged effect of the depreciation of the yen in 1998. These factors alone, however, did not seem to explain the forecast error. Either there was a serious overestimate of output gap or something was missing from the price equation.

It is not difficult to find reasons for possible upward biases in output gap estimates. A significant portion of investment in structures during the late 1980s and the early 1990s may have been just wastes --of course, the counterpart of the bad loan problem. Instead of using the actual capital stock through a production function, we could regress actual output on a nonlinear time trend and, say, various measures of demand pressure in the goods market, for example, the "supply & demand conditions for products" diffusion index in our Tankan (Short-Term Economic Survey of Enterprises) to estimate potential output and the gap. This method results in a much smaller output gap estimate. Depending on what is assumed about other parameters in the Taylor rule formula, it is not impossible to come up with a Taylor rule rate that is already positive.

We could go a little further here. Uncertainties about the size of output gap have important implications for the Taylor rule. Orphanides, Porter, Reifschneider, Tetlow & Finan (1999) report interesting simulation results. They suggest policy attenuation, i.e., to decrease the response of the interest rate to the gap in the presence of large measurement errors with respect to output gap. Moreover, when measurement errors are very large, it is better to respond to the difference between the growth rate of actual and that of potential output rather than to output gap. In other words, the interest rate now responds to the change in the output gap rather than its level. They call this the growth rate rule.⁸

Most of the board members now believe that, assuming that no major negative shocks hit the economy, the economy will grow in fiscal 2000 at a higher rate than some of the largest, but reasonable estimates of the growth rate of potential output. Then, the growth rate rule, assuming that its usefulness carries over in the neighborhood of the zero bound on the interest rate, seems to give us a positive interest rate under wide range of assumptions.⁹

I hasten to add that these are not exactly the ways proponents of the rate increase presented their arguments. I also realize that these are rather weak cases for the rate hike decision, but they do illustrate the difficulty of using the Taylor rule type benchmark in the face of large measurement errors in output gap. Even with measurement errors the rule seems to get the change in the optimal level of the interest rate roughly right. Assuming that one's previous decision was correct, the central bank can use the rule to decide on policy changes. But in order to end the zero rate, the BOJ needed to know the optimal level of the interest rate accurately, which turned out to be very hard.¹⁰

1. 8In a later paper Orphanides (2000) develops this into what he calls the natural growth targeting rule.
2. 9See also footnote 10.
3. 10Even if the zero rate had been a correct decision a quarter ago, it may have been because the Taylor rule rate was negative. An upward movement in the Taylor rule rate from there is no assurance to end the zero rate because the Taylor rule rate may still be negative. This line of thinking suggests that one needs a special check to see whether the result of Orphanides and others (1999) holds true at the zero bound.

6. Some Final Comments

Let me summarize the major messages of my talk. I think that the ZIRP was a unique experiment in the history of the BOJ not just because the level of the overnight rate was zero but because it involved some commitment about the future course of monetary policy. The effects of the policy on the economy were significant as a result of this commitment to affect expectations of market participants. The commitment was a useful way to address the liquidity crunch aspect of the recession. It also helped to contain expectations of premature rate hikes especially at the early stage of the ZIRP. Given the way the ZIRP was lifted in August 2000, it is probably fair to say that we did not complete the Reifschneider-Williams type experiment.¹¹

It was easy to make a case against the rate hike proposal in August. As I have hopefully demonstrated, however, the analytical foundation of such an argument is less tight than one might think. And, it is not impossible to analytically support the rate hike. The indeterminacy here results from our inadequate understanding of the supply side of the economy. I would add that other central banks including the Fed have been facing similar difficulties. We also have not developed effective tools to deal with mis-measurements and uncertainties in our policy making. I can promise that the BOJ will continue to carry out more researches on these issues, but would like to invite people here, especially academic participants, to join the effort.

1. 11There is room to continue the experiment with the 0.25 % rate rather than the zero rate.

References:

• Goodfriend, M. (2000) "Financial Stability, Deflation, and Monetary Policy," paper Presented at the Ninth International Conference at the Institute of Monetary and Economic Studies, The Bank of Japan.
• Hayakawa, H. & E. Maeda (2000) "Understanding Japan's Financial and Economic Developments Since Autumun 1997," Working Paper 00-1, Research & Statistics Department, The Bank of Japan.
• Orphanides, A. (2000) "The Quest for Prosperity Without Inflation," ECB Working Paper, No.15.
• Orphanides, A., R. Porter, D. Reifschneider, R. Tetlow & F. Finan (1999) "Errors in the Measurement of the Output Gap and the Design of Monetary Policy," Federal Reserve Board.
• Reifschneider, D. & J. Williams (1999) "Three Lessons for Monetary Policy in a Low Inflation Era," Federal Reserve Board.
• Tinsley, P. (1999) "Short Rate Expectations, Term Premiums, and Central Bank Use of Derivatives to Reduce Policy Uncertainty," Finance and Economics Discussion Series, 1999-14, Federal Reserve Board.
• Woodford, M. (1999) "Commentary: How Should Monetary Policy Be Conducted in an Era of Price Stability?" paper presented at the Kansas City Fed conference.