Summary of a Speech Given by Kiyohiko G. Nishimura, Member of the Policy Board, at a Dinner Meeting at the The Brookings Institution on July 2, 2007

July 3, 2007
Bank of Japan

Figures [PDF 363KB]

1. Increased Diversity and Reduced Volatility
2. Increased Uncertainty in the Policy Transmission Mechanism

It is my great pleasure to speak before distinguished guests tonight at The Brookings Institution, where I was a Research Fellow more than a quarter-century ago. I feel quite fortunate to come "home" to where my career in economics and economic policy started, and to talk about vital issues of mutual interest, that is, policy challenges that the central bank of the second-largest national economy faces in an increasingly diverse world. It should be noted, though, that the views expressed here are my own as a Policy Board member and should in no way be interpreted as the agreed-upon view of the Policy Board.

One buzzword to describe the world economy today might be "great diversity." In real economic activity, the world is now substantially diversified: there are multiple engines of growth in various regions around the globe. This is in stark contrast with the 1990s, when the world depended on one powerful locomotive of growth, the U.S. economy. In the financial world, the increase in diversity is phenomenal. Many novel financial products are being introduced and sold to a continuing flow of newcomers to the financial world, located in various regions of the globe. Risk appears to be diversified worldwide and ultimately shared by many people.

In the first half of this speech, I will discuss the current state of the Japanese economy from this viewpoint of diversity. In fact, the Japanese economy has also experienced increased diversity in both real economic and financial activity. This diversity currently contributes to reducing volatility in real GDP growth and foreign exchange markets. However, it also deepens uncertainty about the future development of financial markets.

In the second half of the speech, I will discuss policy challenges that the Bank of Japan faces. Increasing diversity in Japan has a distinctive local trait, stemming from the continuing restructuring of regional economies, which are still struggling with the negative legacy from the long stagnation of the 1990s. I will show that the Bank faces increased uncertainty about the transmission mechanism of monetary policy. I will then argue that the flexible gradualism which the Bank has now adopted is a prudent way to cope with such deepened uncertainty.

1. Increased Diversity and Reduced Volatility

(1) Reduced Volatility in Real GDP Growth and Increased Diversity in Domestic and Foreign Demand

Let me begin by outlining the current state of the Japanese economy. The current recovery is now the longest expansion of the past 50 years. However, the current recovery differs distinctively from past business cycles. It is characterized by a substantial decline in the volatility of key macroeconomic variables.

Chart 1 shows real GDP growth. The solid line represents year-on-year real GDP growth. The chart shows increasingly stable growth since the current recovery started some five years ago. There are some ups and downs, but on average, the Japanese economy has grown a little more than 2 percent annually in the past three years. This chart also depicts the growth rate of labor input in the manufacturing sector, which tracks real GDP growth relatively well since 1998. The most recent data on manufacturing labor input are those for April and they show decent growth following some weakness in the last quarter. These data suggest that the Japanese recovery remains on the right track.

This chart also reveals that large fluctuations in real GDP growth, found for years in Japan, seem to have subsided recently. Small changes in the year-on-year rate of real GDP growth in the last twelve quarters are in stark contrast with large swings in the twelve quarters of the 1999-2001 period.

A remarkable fact is that this reduced volatility in real GDP growth coincides with a substantial increase in the diversity of demand for final goods and services.

To show this, I calculated pair-wise correlations of year-on-year output growth for all 19 final goods categories produced in Japan, in three periods: 1999/Q1-2001/Q4, 2002/Q1-2004/Q4, and 2005/Q1-2007/Q1. Table 1-A shows the result. Out of the 171 pairs, the number of statistically significant positive correlations drastically dropped from 62 in the 1999-2001 period to only 17 in the 2005-2007 period. In contrast, the number of statistically significant negative correlations substantially increased from 6 in the 1999-2001 period to 16 in the 2005-2007 period. Thus, stable growth in recent years is a result of small ups and downs in the demand for various products which have cancelled out one another or averaged out through the law of large numbers.

This reduced correlation in the demand for various products is not confined to goods markets: it is also found in service markets. A similar calculation of pair-wise correlations of service activities in the same periods reveals almost the same picture, as shown in Table 1-B. Out of 78 pair-wise correlations among all 13 service categories in Japan, the number of statistically significant positive correlations is 13 for the 1999-2001 period but only 8 for the 2005-2007 period.¹

As we all know, the hallmark of business cycles is the output co-movement of various products and services. This classical characteristic has been found in most Japanese business cycles until the period of the IT bubble and its bursting in 1999-2001. However, this co-movement is conspicuously absent after 2005/Q1.

The diversity of demand comes from both at home and abroad. Firstly, the increased diversity in foreign demand for Japanese goods and services is likely to be the product of a worldwide trend toward diversity. Let me examine this issue a bit more closely. I again calculate pair-wise correlations, this time of the export growth of all 6 major goods categories produced in Japan, as shown in Table 1-E. Out of the 15 pairs, all but one show a statistically significant positive correlation in the 1999-2001 period. In contrast, the number of statistically significant positive correlations in the 2005-2007 period is currently 5, down from 14. This shows a growing diversity in world product demand, which coincides with the increasing diversity of growth sources I mentioned earlier. There is greater heterogeneity in demand for products than before. A similar diversity is found in exports categorized by destination, as shown in Table 1-F. Thus, while Japanese auto exports to the U.S. are reduced somewhat these days, their reduction is offset by an increase in exports to all other regions.

Secondly, although the so-called restructuring of Japanese industries at least at large corporations is largely over, the restructuring of small to medium-sized regional firms remains in progress to differing degrees, depending on the industry concerned. Moreover, under the pressure of large deficits, regional governments have started their own fiscal consolidation, which has a dragging effect on regional economies and some industries like construction that depend heavily on public expenditure. Although it is hard to quantify the magnitude of this home-made divergence effect,² we can sense its growing importance through the quarterly reports submitted by our regional branches and summarized in the Bank's Regional Economic Report, which is the counterpart of the U.S. Beige Book.

1Tables 1-C and 1-D also reveal demand diversity for import goods, both by goods category and by country of origin.
2Regional data are too noisy and too infrequent so that even this simple pair-wise correlation analysis cannot be applied.

(2) Reduced Volatility in Foreign Exchange Markets and Growing Diversity in Japanese Households' Financial Asset Positions

So far, I have explained reduced volatility in real GDP growth and its possible sources. The reduced volatility is also found in foreign exchange markets. Chart 2 shows the implied volatility calculated from forex option prices. The black line is the implied volatility of the current period (July 2006-June 2007), while the grey line shows the volatility one year before (July 2005-June 2006). The chart clearly shows that implied volatility has been markedly reduced this year. Also, the chart reveals that this year's "market turbulence" (March-April) had a smaller effect on foreign exchange markets than last year's "global risk reduction" (May-June).

The decline of volatility is not only a Japanese phenomenon; rather, it is observed worldwide. Still, it has a uniquely Japanese flavor.

Japanese households were considered very conservative investors for a long period. Their preference for cash and bank deposits over risky assets was noted for years, and their home bias in investing in risky assets was legendary even at the height of the banking crisis in the late 1990s. However, a substantial shift in their investment behavior has been observed in recent years. Their holdings of foreign currency denominated assets have risen from around 0.5 percent in the 1994-96 period to 2.8 percent in March 2007. The slow but steady departure from the former home bias is having a sizable impact in international financial markets, since Japanese households hold 1,500 trillion yen in financial assets. In particular, their holdings of investment trusts denominated in foreign currency show a phenomenal increase from 9 trillion yen in March 2004 to 31 trillion yen in March 2007.

The arrival of Japanese households as major investors seems to have affected foreign exchange markets. The black line in Chart 3 depicts the movement of Japanese yen short positions of Japanese retail investors (households) in foreign exchange margin trades. In the same chart, the grey line shows the short positions of professional investors in the International Monetary Market (IMM) released by the Commodity Futures Trade Commission. There are two remarkable features in this chart. First, Japanese retail traders' positions are sizable when compared with professional investors' positions in the IMM. Second, Japanese retail investors are distinctively contrarian. A clear negative correlation is found between the positions of professional investors in the IMM and Japanese retail investors.

Years ago, it was the gnomes of Zurich who shook the foreign exchange markets. They have now been replaced by the housewives of Tokyo, who speculate in various currencies. However, whereas the gnomes of Zurich were accused in their day of destabilizing markets, the housewives of Tokyo are apparently acting to stabilize them. Their presence seems to lie behind the marked decline in (perceived) volatility in yen-dollar exchange rates in Chart 2. The housewives are betting against professional investors in the IMM, and seem to be profiting from their trading so far.

(3) Near-Term Prospects: "Diversity and Stability" -- But How Far Can They Go?

Let me wrap up the first section of my speech with a note on the near-term prospects for the real side of the Japanese economy and the world.

Firstly, we now seem to have many sources of demand in world product markets. This means that as of late, economic conditions have been rather resilient, capable of absorbing economic downturns and upturns here and there through the law of large numbers. However, there are various geopolitical risks that may have large effects if they materialize, and we should watch them carefully.

Secondly, the financial diversity we now observe is surely far more beneficial than harmful to economic welfare. Risks are now shared by many instead of the few. Various arbitrage opportunities, enabled by the ceaseless introduction of new financial products, have so far become absorbers of various shocks in different regions of the world.

We should note, however, that we are in a transition period from one financial structure of little diversification to another of great diversification. In the end, we are sure to be better off in a well-diversified economy. But we have not yet reached the stable state of a diversified economy. Being in a transition period, the economy may be resilient to small scattered shocks but remains vulnerable to unforeseen snowballing shocks.

In sum, we now face an increase in a kind of uncertainty different from that in the past, because of the initial success in diversification. In a transition period, information about the entire system is scarce and rapidly becomes obsolete. The distribution of final risk sharing is one piece of such scarce information. In addition, we do not know how far the contrarian strategies of Japanese retail investors can go. A sudden change in their behavior is likely to shift the direction and the magnitude of trading in many foreign exchange markets. This is the kind of risk that cannot be diversified away.

2. Increased Uncertainty in the Policy Transmission Mechanism

(1) Increased Uncertainty About Both the IS and Phillips Curves

Let me turn from the real side of the economy to the monetary front, especially the inflation front. Here, we again face increased uncertainty: uncertainty has increased markedly regarding how monetary policy works in the Japanese economy.

As a frame of reference, I take up a simplistic mainstream-textbook characterization of an economy: a reduced-form two-equation model consisting of an IS curve and a Phillips curve. The fundamental problem that the Bank of Japan now faces is that both the IS and Phillips curves have become increasingly uncertain or "blurred" with respect to their most important coefficients: the GDP-gap sensitivity of inflation in the Phillips curve and the real interest rate sensitivity of the GDP gap in the IS curve.

To illustrate the problem, I conducted a simple 10-year rolling quarterly estimation of the IS and Phillips curves from 1990/Q1 to the present. (It should be noted that this estimation is only for illustrative purposes in this speech and of course in no way represents how the Bank of Japan forecasts the Japanese economy.) In the Phillips curve equation, the inflation rate³ is regressed on its own lag, the GDP gap estimated by the Bank of Japan staff, and import prices (rate of change from the previous quarter). In the IS curve equation, the GDP gap is regressed on its own one-quarter lag, the real interest rate⁴ (one-quarter-lagged), and the rate of potential GDP growth estimated again by the Bank's staff.⁵

The result is astonishing. Take the Phillips curve first. The most important coefficient of the Phillips curve, that is, the sensitivity of inflation to the GDP gap, averaged 0.067 for the 1990s until 2000/Q2 and was significantly different from zero. The sensitivity drops to an average of 0.059 after 2000/Q3.⁶ Although it is not a big decline, the standard error of the estimated coefficient also increased dramatically, and the GDP-gap sensitivity of inflation is no longer significantly different from zero. The two-sigma band of the coefficient is currently between -0.05 and +0.16. This means that we can no longer say as confidently as in the 1990s that the Phillips curve relationship of a positive inflation-output correlation exists right now, at least in this simple framework. In addition, the estimated coefficient had declined further in the past several quarters.

The result is no less surprising for the IS curve. From the viewpoint of monetary policy, the crucial coefficient in the IS curve is that of the real interest rate, that is, the interest sensitivity of the GDP gap. The coefficient of the real interest rate increased in absolute terms from -0.2 in 1990 (the starting year) to -0.4 around 2005/Q1 and was significantly different from zero. Then, the standard error increased dramatically after 2005/Q2. For example, the standard error of the interest sensitivity coefficient in the 1997-2006 estimation is more than treble that in the 1985-96 estimation. The two-sigma band of the coefficient is between -0.87 and +0.06 right now, implying that the null hypothesis of no interest rate effect on the GDP gap cannot be rejected. To sum up, although there were significant real interest rate effects on the GDP gap until 2005, they seem to have disappeared recently in this framework.

It should be noted that estimating the GDP gap and potential GDP is a very difficult business and there may be substantial measurement errors. Thus, the result is far short of conclusive. I myself expect that, as time goes by, future improvement in GDP-gap and potential GDP estimation will eventually put the Phillips and IS curves back on their feet, so to speak. However, one thing is clear: at this moment, we face greater uncertainty than ever about the transmission mechanism of monetary policy.

One may wonder what happens to the more "traditional" description of the monetary policy transmission mechanism through changes in monetary aggregates. The relationship between monetary aggregates and economic activity becomes even looser than the IS and Phillips curves. The money stock, or M₂+CDs in the case of Japan, has had a negative correlation -- let me repeat to make it clear, a negative correlation -- with nominal GDP in the past decade. The growth of M₂+CDs increased when nominal GDP growth declined in 1997-98 and 2001, and M₂+CDs growth decreased when nominal GDP growth showed a recovery in 1999-2000 and from 2002 onward. In addition, the relationship between the monetary base that the Bank of Japan can control and the money stock such as M₂+CDs is looser than ever, as the experience after the lifting of the quantitative easing policy suggests. I still believe that inflation is a monetary phenomenon, when the money stock is appropriately defined. Great uncertainty is looming here in appropriately defining the money stock.

3The inflation rate is the year-on-year rate of change in the CPI (excluding fresh food), which is widely considered to be representative of Japanese consumer price inflation and on which the Japanese version of Treasury inflation protected securities (TIPS) is based. The basic characteristics of this rolling estimation exercise do not change if we use the trimmed-mean CPI inflation rate referred to later in this speech instead of CPI (excluding fresh food) inflation.
4The real interest rate here is defined as the annualized overnight call rate minus the year-on-year rate of CPI (excluding fresh food) inflation.
5These two equations were estimated by the Full Information Maximum Likelihood (FIML) method.
6The decline in the GDP gap-sensitivity of inflation is more pronounced in the Ordinary Least Squares estimation of the same equation. The coefficient averaged 0.71 for the 1990s and dropped to an average of 0.36 after 2000/Q3. As in the FIML case, the coefficient is no longer significantly different from zero after 2000/Q3.

(2) Increased Diversity and Uncertainty with regard to the IS and Phillips Curves

Does the increased diversity and reduced volatility discussed earlier have any bearing on this increased uncertainty about the transmission mechanism of monetary policy? There recently seem to have been some effects on the Phillips curve, that is, there has been a further decline (though small) in the GDP-gap sensitivity of inflation in the past several quarters.

I showed earlier that domestic demand and foreign demand for Japanese final products and services are increasingly uncorrelated with one another and in some cases recently have become negatively correlated. In a very competitive economy like Japan's, firms are always looking at their competitors. If demand conditions change rather uniformly across firms, they adjust their prices to the new conditions, since they know their competitors have the same incentive to adjust. In contrast, if demand conditions are diverse across firms, they are likely to shun price adjustments, fearing that their competitors, facing different demand conditions, will not follow them, and thus they will lose customers to their competitors. The more competitive the market is and the more diverse demand conditions are across firms, the less frequently prices are adjusted to changes in demand conditions.⁷

Moreover, we have witnessed a steady shift of bargaining power in intermediate product/service markets from upstream firms to downstream firms. These downstream firms are fearful of losing customers in a very competitive final goods market if they adjust their prices to a substantial increase in commodity prices, including crude oil. Thus, they jawbone upstream firms into absorbing increased costs by improving productivity. This competitive behavior is likely to be behind the current flattening of the Phillips curve and at the same time behind the apparent lack of pass-through of high import costs, in spite of the hike in commodity prices and the substantial depreciation of the yen.

The diversity may also explain increased uncertainty with regard to the sensitivity of the GDP gap to a change in the real interest rate. What lies behind the diversity in domestic demand is the "polarization" of the Japanese economy. At one pole of the Japanese economy, we have profit-making exporting firms that have benefited from global demand diversity and stability. However, at the other pole, especially in regional economies, the belated restructuring of regional firms is now gaining momentum, partly pushed by the fiscal consolidation of regional governments. Cash-rich exporting firms can finance their hefty investment bills by themselves, and thus their investment is not as sensitive to the borrowing rate as before. Firms in the midst of regional restructuring feel the pinch, although the GDP gap has firmly been on the positive side in recent quarters. Property markets are also polarized. Tokyo households may find their property more highly valued today than a couple of years ago, while households in Hokkaido, the rural northernmost of Japan's four major islands, may continue to see the value of their property moving in a downward direction. The diversity, when it is translated into polarization, is likely to make the effect of the real interest rate on macroeconomic variables like the GDP gap uncertain.

7This argument is an old one developed in K. G. Nishimura, "Rational Expectations and Price Rigidity in a Monopolistically Competitive Market," Review of Economic Studies, 53 (1986), pp. 283-292. Unfortunately, although the interplay of imperfect information and competitive pressure is an important factor in the real economy, it is largely ignored in mainstream macroeconomics. It should be added that the argument here is immune to the confusion between absolute and relative prices that plagued recent discussion of the effects of globalization on the flattening of the Phillips curve. An explicit macroeconomic model based on the argument developed here is found in K. G. Nishimura, Imperfect Competition, Differential Information, and Microfoundations of Macroeconomics, Oxford University Press, 1992.

(3) Flexible Gradualism: Keeping Up with the Trend in General Economic Conditions

What then is a sound monetary policy for a central banker facing this great uncertainty about the transmission mechanism of monetary policy? I will conclude my speech with some thoughts on this most difficult problem.

Firstly, prescriptions based on the now-popular mainstream-textbook model consisting of the IS and Phillips curves augmented by some form of Taylor rules are likely to be less useful than before in providing good guidance, since both the IS and Phillips curves have become uncertain at this moment, at least in the framework described before. We may still have non-flat IS and Phillips curves in more sophisticated models, but the increased uncertainty about the slope of the IS and Phillips curves is not likely to disappear soon.

Secondly, however, I should make it clear that Japanese prices are in fact responding to the trend in general economic conditions, albeit gradually. Let me present another chart, Chart 4, which juxtaposes the CPI inflation rate and a proxy of the trend in general economic conditions.

Here I use the trimmed-mean CPI inflation rate to avoid the arbitrariness of excluding particular goods and services to obtain "core" inflation. The methodology of purging the extremes on both sides of the product-wise inflation distribution is the same as the trimmed-mean PCE inflation rate calculated and published monthly by the Federal Reserve Bank of Dallas, except that the methodology is applied to the CPI here rather than PCE deflators.⁸ For your information, the April year-on-year trimmed-mean PCE inflation rate in the United States was 2.3 percent, while the core PCE inflation rate excluding food and energy was 2.0 percent.

A remarkable feature of the Japanese CPI is that the trimmed-mean inflation rate of service prices has been very stable at zero since 1999.⁹ Thus, the price movement of goods produced by the manufacturing sector dominates overall trimmed-mean CPI inflation. Therefore, I use employment growth in the manufacturing sector as a proxy for the trend in general economic conditions. In Japan, where employment adjusts very gradually, employment growth is a good proxy for the trend in general economic conditions.

Chart 4 clearly reveals that although Japanese price inflation fails to comply with the mainstream-textbook Phillips curve relationship, it responds to or at least co-moves with the gradual improvement of general economic conditions that is represented by manufacturing-sector employment growth. The trimmed-mean CPI inflation rate entered positive territory when the Bank of Japan ended its five-year-long unorthodox quantitative easing policy. There have recently been small two-month dips into negative territory, but such ups and downs are inevitable with this type of inflation measure. Reflecting improved economic conditions, it quickly recovered and moved into positive territory again in April. This gradual inflation movement may be another manifestation of a popular saying that everything in Japan moves slowly, but steadily. Thus, it is of utmost importance to discern the trend in general economic conditions in predicting the future course of CPI inflation.

On one hand, the current diversity in foreign demand and increased stability of GDP growth suggest that this positive trend will continue for some time. GDP growth is rather modest compared with that in other parts of the world, but general economic conditions are now on a favorable trend. A slow but steady upward pressure on prices is clearly felt in many markets for goods and services, though it does not yet show up in CPI inflation.

On the other hand, however, domestic demand may change and increase steadily and broadly as regional restructuring approaches its end and retired baby boomers begin to purchase big-ticket items and services. Then, firms may find it easier to adjust their prices upward, reflecting excess demand and increasing material costs. Service prices, which have been remarkably stable for years, may show upward movement as service labor costs (especially part-timers' hourly wages) start increasing.

These considerations lead to a pragmatic approach: flexible gradualism, in which the policy rate should be adjusted gradually to the trend of general economic conditions and the timing of policy change should depend on the pace of economic improvement. To stand pat for a long period of time is not a prudent strategy, since the acceleration of economic activity may in the future come to require a large adjustment in the policy rate, causing unnecessary swings in economic activity and prices. However, policy adjustment cannot be pre-scheduled or occur at some fixed interval, since it should be in line with the general economic improvement, which never follows a fixed schedule. In fact, as Chart 4 shows, there was a slight pause in the trend of general economic improvement in the last quarter of last year. A prudent policymaker should take this pause seriously and wait until the clouds over the trend have lifted somewhat. That is what the majority of the Policy Board did last December and this January.

This pragmatic approach may be regarded as more classical in its spirit, in keeping with the trend of general economic conditions to minimize future fluctuations in economic activity, than mainstream-textbook prescriptions based on a perceived ability to influence the economy with the IS-Phillips curve transmission mechanism.

The economy is changing, and I do not expect that the present uncertainty regarding the IS and Phillips curves will continue forever. Eventually, we will return to a regime where the mainstream model holds up well. However, a great deal of uncertainty remains about the time needed to get there. In the meantime, much prudence is required.

Thank you for your kind attention.

8I use the trimmed-mean inflation rate here only for expository purposes, with no intention to state that it is better than the current Japanese core CPI inflation rate excluding fresh food. In fact, S. Shiratsuka showed that both the trimmed-mean CPI inflation rate and the CPI (excluding fresh food) inflation rate are among the best predictors of the future trend of the CPI (all inclusive) inflation rate, while a U.S.-type CPI (excluding food and energy) inflation rate is the worst. See S. Shiratsuka, "Core Indicators of Japan's Consumer Price Index," Bank of Japan Review, 2006-E-7, November 2006, pp. 1-9.
9The average of year-on-year changes of the trimmed-mean inflation rate of service prices was 0.08 percent with a standard deviation of 0.11 percent from January 1999 to April 2007. In contrast, the average for the trimmed-mean inflation rate of goods was -0.61 percent and the standard deviation was 0.56 percent.

The Bank of Japan

Increased Diversity and Deepened Uncertainty: Policy Challenges in a Zero-Inflation Economy

Contents