This paper describes determinants of Japan’s productivity growth, reform outputs, and speculation on future potential growth. It investigates the macroeconomic implications of alternative fiscal strategies; examines the issues surrounding the Bank of Japan’s understanding of price stability and monetary policy; examines the impact of Japan’s inflation and price developments; and describes the reengagement of Japanese banks with the rest of Asia for maintaining financial and macroeconomic stability in both Japan and in the other Asian countries.

Abstract

This paper describes determinants of Japan’s productivity growth, reform outputs, and speculation on future potential growth. It investigates the macroeconomic implications of alternative fiscal strategies; examines the issues surrounding the Bank of Japan’s understanding of price stability and monetary policy; examines the impact of Japan’s inflation and price developments; and describes the reengagement of Japanese banks with the rest of Asia for maintaining financial and macroeconomic stability in both Japan and in the other Asian countries.

III. Practical Issues surrounding the New Understanding of Price Stability 25

1. This chapter examines two issues surrounding the Bank of Japan’s understanding of price stability in the context of its new framework for monetary policy. First, the range of inflation—zero to two percent—is lower than the inflation reference ranges used at other central banks. Second, while the range focuses on headline CPI inflation, other price indicators could also be useful in assessing medium-term inflation pressures. This chapter shows that inflation has indeed been lower in Japan than in most industrial countries, suggesting a lower range may be appropriate. It also examines a number of price indices and finds that core CPI performs well in describing the medium-term inflation outlook.26

A. Japanese Inflation in the Historical and International Context

2. Over much of the recent history, Japan’s inflation has been in the middle of the range of G-7 countries’ experiences (Table 1). Japan has been somewhat unusual in that it has experienced periods of both substantially higher-than average and lower-than average inflation in the G7.

Table 1.

International Inflation Trends, 1960-2005

Average annual percent change in CPI

article image
Source: Haver Analytics, WEO database, and staff estimates.

PPP weighted average inflation of the non-Japan G7. Weights prior to 1970 are based on 1970-1972 shares.

IFS definition. Not available prior to 1968.

3. During the 1960s—and for most of the 1970s, Japan’s inflation rate was significantly above the average due to a number of factors. Importantly, the economy was growing robustly, and prices were converging to levels in the other industrial countries. At the same time, labor unions were strong—representing 35 percent of workers—and negotiated indexed wage contracts, which fueled cost-push inflationary pressures.

4. Economic changes after the 1970s helped to spark a sharp disinflationary trend in Japan that was larger than in other industrial economies. Cautious monetary policy, a decline in unionization, greater trade openness, an appreciating currency, and lower oil prices reduce inflation from a high of 23¼ percent in 1974 to 0 percent in 1987. While most industrial countries saw a decline in inflation, Japan’s was the largest and fastest. Other countries, on average, saw a decline of only around 10 percentage points over same period.

5. In most other advanced economies, inflation declined gradually through a two-stage process that left inflation at a low, but positive, level (Figure 1). In the first round, medium-term inflation fell from above 10 percent in the late 1970s to an average near 4 percent in the late-1980s.27 A second round of disinflation started in the mid-1990s and brought inflation down to around 2 percent, where it has remained over the past several years.

Figure 1.
Figure 1.

Medium-Term Inflation Trends

3-year centered average of y/y change in CPI (using monthly data)

Citation: IMF Staff Country Reports 2006, 276; 10.5089/9781451820683.002.A003

Source: Haver, IFS and WEO. IFS defintion for industrial countries.1/ Austria, Belgium, Denmark, Germany, Luxembourg, Netherlands, and Switzerland.

6. Japanese inflation has been lower than in other advanced economies since the early 1980s. 28 A substantial difference (around 2¼ percentage points) emerged during 1983–1994, even prior to Japan’s decade of deflation. This difference was smaller in comparison to a group of relatively “low” inflation countries (Austria, Belgium, Denmark, Germany, Luxembourg, Netherlands, and Switzerland), but a gap of around ¾ percentage point existed over this period.

7. Since the mid-1990s, inflation has converged across countries (excluding Japan) to near the levels of traditional “low” inflation countries. Over the 1990s, most of the gap (of around 1½ percentage points) between relatively “high” and relatively “low” inflation countries disappeared. This convergence has occurred despite differing economic structures: across both small and large countries and with countries that have high and low fiscal deficits (Figure 2). Even the level of product regulation does not appear to have affected inflation.29

Figure 2.
Figure 2.

Fiscal Balance and Inflation, 1996-2004

General government balance in industrial countries (avg. over the period, in percent of GDP) vs. year-on-year change in CPI.

Citation: IMF Staff Country Reports 2006, 276; 10.5089/9781451820683.002.A003

Source: WEO

8. Reflecting these developments, central banks came to a broadly similar view regarding the desirable rate of inflation. Most central banks surveyed in Mahadeva and Sterne (2002) and Roger and Scott (2005) have an inflation range above 1 percent and mainly in the range of 1–4 percent.30 Central banks appear willing to keep inflation low, but not “too low.” This caution is partly explained by Japan’s deflation experience, but other central banks have faced operational difficulties, and recent research has highlighted the dangers of letting inflation drift too close to zero percent (WEO, 1999).

9. The historical experience suggests that price stability in Japan would be consistent with inflation in the low end of the range of other industrial countries. It is difficult to speculate when inflation in Japan would converge to the rates of “low” inflation countries. Notwithstanding the strong growth picture, the pace may be gradual, if Japanese inflation expectations are colored by the recent deflationary period (Fukui, 2006).

B. Methods for Identifying and Communicating Price Pressures

10. Medium-term trends in the CPI now play an important role as a part of the BoJ’s understanding of price stability. The CPI has a number of advantages over other indicators (such as the GDP deflator): it is available quickly and with high frequency, is not subject to frequent revision, is well-understood, and is immediately relevant to consumers.

11. However, the usefulness of the CPI may be limited by statistical issues that impart an upward bias to measured inflation. As a fixed weight index, the CPI uses expenditure shares that are updated infrequently (in Japan, every five years). This imparts an upward bias because it fails to account for the fact that consumers tend to substitute relatively less expensive items for expensive ones.31 For Japan, the BoJ has indicated that recent methodological changes have reduced the bias significantly, and some now estimate the bias at between 15–30 basis points (Nishimura, 2006). By comparison, the upward bias in the CPI in other countries remains more pronounced, at around ½–1 percentage points.

12. Moreover, it is useful to glean additional information regarding underlying price developments by examining various other indices. Because headline CPI is somewhat noisy and concentrated on just consumption goods, it is possible to miss unfolding price pressures in other sectors of the economy. Indeed, most central banks have found it useful to report on a variety of price indices, to reflect developments across the economy. Even inflation-targeting central banks that target the overall CPI inflation rate also look to a range of indicators to help guide and communicate policy (Whitesell, 2005), such as:

  • Core indicators: trend inflation is derived by eliminating selected, volatile components of the CPI (such as fresh food or food and energy).

  • Multiple-price indicators: price trends can also be discerned from an examination of a broad range of indices, such as the GDP price deflator and producer prices, which may reflect price pressures in various sectors.

  • Robust indicators: recently, attempts have been made to measure underlying inflation through indices that are robust to economic changes or to excessive volatility.

Core indicators

13. Different core inflation measures track medium-term trends well (Figure 3). Core inflation manages to smooth out the data over most (but not all) of the transient spikes in the contemporaneous data. Indeed, the Japanese definition of core inflation, which excludes only fresh food, is marginally smoother than the traditional definition (CPI less food and energy). Importantly, they both indicate that at present medium-term inflation (as defined in Figure 1) is trending up.

Figure 3.
Figure 3.

Core inflation trends

Citation: IMF Staff Country Reports 2006, 276; 10.5089/9781451820683.002.A003

Source: CEIC and OECD.

14. The ease-of-calculation has made core measures a standard part of inflation analysis, but they are not without disadvantages. Their chief advantage is their simplicity; they are easy to calculate, easy to understand, and widely used. This makes them useful for communication. However, this approach can prove problematic; first, core indices can exclude a large percentage of the overall index.32 Second, core CPI can still remain volatile. Finally, the composition of goods in the core CPI often does not have a theoretical justification.

Multiple-price indicators

15. Because inflation pressures may develop in a specific sector of the economy, it is also useful to examine indices other than the CPI. Since the CPI covers consumption goods, it may miss developments in the business or external sector. To help address this shortcoming, alternative indices such as the producer price index or GDP deflator can also shed light on current developments.

16. Business sector indicators do not appear to match overall price developments well, but even these confirm that the effects of deflation are waning. Japan has two price indices of the business sector (Corporate Goods Price Index, CGPI, and Corporate Service Price Index, CSPI) that measure prices paid and received by business for most goods and services.33 While inflation in the business sector moves somewhat in line with medium-term CPI changes, there are gaps that persist over time (Figure 4). Indeed, during the late 1990s and into the early 2000s, they seem to show greater deflation than the headline measure of CPI. More recently, however, both the CGPI and CSPI are also trending upward.

Figure 4.
Figure 4.

Inflation trends in the business sector

year-on-year percent change in corporate goods and services

Citation: IMF Staff Country Reports 2006, 276; 10.5089/9781451820683.002.A003

Source: CEIC.

17. Similarly, GDP-based price indicators also follow trend inflation, albeit with more noise and a lag (Figure 5). To understand developments in the broader economy, it can be useful to look at the deflators for private consumption expenditure and total domestic demand.34 These GDP-based series show mild, ongoing deflation through 2005. However, this may be due to the broader range of goods included in the GDP-indices (for example, investment goods in the total domestic demand deflator) and methodological differences. (The GDP data are chain-weighted; the CPI is fixed-base.)

Figure 5.
Figure 5.

Trends in GDP deflators

year-on-year percent changes in the deflators (n.s.a.)

Citation: IMF Staff Country Reports 2006, 276; 10.5089/9781451820683.002.A003

Source: CEIC and Haver.

18. However, price indices from the business sector or the national accounts are more complex than CPI-based measures, which detracts from their usefulness. GDP-based measures include a broader range of goods (for example, investment goods and government services), are subject to frequent revisions, and are available only with a long delay. All of these complications making their use difficult for policy makers, especially as real-time indicators. The business-sector price deflators can include the prices of intermediate goods, whose ultimate impact depends on how these goods are used in final production.

Robust price measurement

19. Besides the core and multiple price indicator approaches, other analytical indices can be informative about underlying trends.35 The first approach, (the “trimmed mean” estimator) excludes those components of CPI that have the largest effect on the headline rate. This “trimmed mean” index assumes that large price changes reflect temporary shocks to the trend rate (Bryan and Cecchetti, 1994). The second approach constructs a Fisher chain-weighted price index, which closely approximates an “ideal” measure from consumer demand theory (Fisher, 1922). Notably, both match trend inflation well (Figure 6).

Figure 6.
Figure 6.

Alternate measures of core inflation

year-on-year pct chg in trimmed mean and core Fisher indices

Citation: IMF Staff Country Reports 2006, 276; 10.5089/9781451820683.002.A003

Source: Statistics Office and staff estimates.

20. However, the complexity of these indices can present additional communication challenges. In particular, the novel approach to their construction and relatively new nature would pose communications difficulties. For example, some studies have shown that the best forecast of medium-term inflation (using the trimmed mean estimator) requires removing as much as 75 percent from the CPI. However, such trimming would likely lead to questions as to whether “too much” information is being discarded. Furthermore, these two estimators disagree about inflation developments since 2000, so even robust approaches can offer different views on trend inflation.

Evaluation of the Three Approaches

21. A number of criteria can help serve as a guide to selecting which indicator is the most useful in assessing and communicating about inflation. Roger (1998) and Wynne (1999) suggest that an ideal price index should be: timely, well-understood and trusted by the public, seldom revised, have a low or negligible bias, reflect the forward-looking component of inflation expectations, and be theoretically-based.

22. By these criteria, core CPI ranks well (Table 3). Core inflation lacks only a link to theory or an estimate of forward-looking price developments. However, the lack of such elements also means that the index is simple to understand. Furthermore, the estimated bias of core CPI is low, with a substitution bias of only around ¼ percent.36

Table 2.

Assessing Trend Inflation Measures

article image

Detailed information on the GDP deflators are unavailable.

Has been used extensively in public documents in Japan.

Pct. pnt. difference from headline Fisher chain-weighted index (1995-present).

For CGPI/CSPI: geometric average of business services and final demand goods.For GDP: difference between private consumption deflator and Fisher index.

23. Importantly, the core CPI approach also appears superior to every other approach, other than the trimmed mean. The CGPI and CSPI may be nearly as useful as core indices; however, there can be substantial deviations between consumer and producer price inflation. Meanwhile, the GDP-based measure (private consumption deflator) and Fisher chain-weighted indices are not likely to be readily understood by the general public. Furthermore, both indices are based on survey data and can be revised substantially over time. Only the trimmed mean estimate of inflation is available as readily as core CPI, but the estimated bias is marginally lower, and research has shown that it is not much better at forecasting future inflation than the core CPI. At the same time, it may be difficult for the central bank to explain and justify the trimming process. (For example, why certain items were excluded or why a trimming percentage was chosen.)

C. Summary

24. This chapter examines the impact of Japan’s inflation history on the new understanding of price stability. The definition of price stability (zero to two percent) is lower than that used in other central banks. However, Japan’s inflation has also been lower than in most industrial countries, occurring even prior to the onset of deflation.

25. Second, the chapter also reviews several approaches to examining price developments and concludes that core inflation works well in explaining trend inflation. The core and headline measures of CPI will agree over the medium-term, which is the reference period for the BoJ’s understanding of price stability. Nevertheless, it can be useful to have a set of indicators to help explain the current conjuncture. In that regard, the core CPI is smooth and matches some estimates of trend inflation. It is also well-understood by the public, which is a useful feature in a price index that is used as part of a central bank’s communication strategy.

Annex. Robust Measures of Trend Inflation

Fisher chain-weighted price index

1. A significant source of bias in the CPI arises from the use of expenditure weights that are updated only periodically. In Japan, the updating process occurs once every five years. (The next is scheduled for August 2006.) Over time, these weights will become less representative of current expenditure patterns, introducing errors.

2. The chain-weighting procedure attempts to fix these shortcomings by using weights that are updated each period. The resulting index is linked together with its level in the previous period, so the index is constantly freshened with weights reflecting the most recent data.

3. While there are many chain-weighting approaches, the Fisher index turns out to have optimal statistical properties. The Fisher chain-weighted index is a geometric average of a Laspeyres (previous period weights) and Paasche (current period weights) indices. Diewert (1976) shows that the Fisher version of chain-weighting does a better job at representing a true “cost-of-living” index than either the Laspeyres or Paasche indices.

4. The Fisher CPI is constructed using monthly data starting in 1980. First, detailed CPI data (on 598 items) are matched to 41 expenditure categories in the monthly Family Income and Expenditure Survey for worker’s households (from the Nomura Research Institute). Then, the expenditure and price series are seasonally-adjusted used the Census X12 seasonal adjustment process. A quantity series is calculated from the seasonally-adjusted price and expenditure data. Since the FIES expenditure data does not include a “fresh food” category, the “core” Fisher price index is calculated by excluding the two items: the one with the largest positive and the one with the most negative contribution to headline inflation, where the contribution to growth formula is taken from Whelan (2000). In general, the procedure typically results in excluding categories that account for around 5-10 percent of the overall index (similar to the core CPI).

Trimmed mean price index

5. Not only does the CPI suffer from statistical biases, it is typically noisy—making it difficult to determine in “real-time” the underlying trend inflation rate. Not only are the month-to-month changes in the CPI large, but the changes in individual price components can also be significant. Bryan and Cecchetti (1999) report that the standard deviation of the individual components of inflation are nearly six times higher than the average inflation rate. Shiratsuka (1997) also notes that the distribution of prices are non-normally distributed, with “fat-tails” (i.e., large deviations of the mean are common).

6. This noisiness suggests that a weighted average is not the best measure of the mean of the distribution. Noisy, or high variance, data suggests that the probability distribution cannot be represented with an exponential probability density function (e.g., normal). Instead, the mean of these fat-tailed distributions may not be well-defined. In these cases, it is better to use an alternate estimator, such as the sample median.

7. To address these statistical problems, Bryan and Cecchetti (1997) recommend taking a special weighted average of the index that excludes extreme values. They also show that this statistical solution has useful economic and econometric properties. The trimmed mean is less noisy, allowing policy makers to overlook the monthly blips. It also provides a better forecast for inflation, typically even better than core inflation.37 For Japan, Bryan and Cecchetti find that the optimal trimming excludes 76 percent of the sample (the 38 percent with the largest positive and negative inflation rates).

8. Table A.1 reports the results of the 15 products that are most likely to be trimmed.

Table A.1:

Trimming Results: 15 most likely products to be trimmed, 1986-2005 Ranked by the percentage of months trimmed from the sample

article image

Primarily recreational durables, but also including medical appliances and automobiles.

References

  • Bank of Japan, 2006, “Background Note to: The Bank’s Thinking on Price Stability,” March 10, available on the BoJ web site: www.boj.or.jp.

    • Search Google Scholar
    • Export Citation
  • Bryan, Michael F. Bryan and Stephen G. Cecchetti, 1994, “Measuring core inflation,” in Monetary Policy, ed. N. Gregory Mankiw, (University of Chicago Press).

    • Search Google Scholar
    • Export Citation
  • Bryan, Michael F. Bryan and Stephen G. Cecchetti, 1999, “The Monthly Measurement of Core Inflation in Japan,” Bank of Japan: Institute for Monetary and Economic Studies (IMES) Discussion Paper Series, No. 99-E-4 (February).

    • Search Google Scholar
    • Export Citation
  • Conway, Paul, Véronique Janod, and Giuseppe Nicoletti, 2005, “Product Market Regulation in OECD Countries: 1998 to 2003,” OECD Economics Department Working Paper, ECO/WKP (2005)6, April.

    • Search Google Scholar
    • Export Citation
  • Diewert, W. Erwin, 1976, “Exact and Superlative Index Numbers,” Journal of Econometrics Vol. 46 (May) 11545.

  • Fukui, Toshihiko, 2006, “New Framework for the Conduct of Monetary Policy: Toward Achieving Sustainable Economic Growth with Price Stability,” Speech to the Japan Chamber of Commerce and Industry, Tokyo (March 16).

    • Search Google Scholar
    • Export Citation
  • Mahadeva, Lavan and Gabrieal Sterne, 2002, “The role of short-run inflation targets and forecasts in disinflation,” Centre for Central Banking Studies, No. 167 (Bank of England).

    • Search Google Scholar
    • Export Citation
  • Moulton, Brent, 1997, “Bias in the Consumer Price Index: What is the Evidence,” Journal of Economic Perspectives, Vol. 10, No. 4 (August) 159177.

    • Search Google Scholar
    • Export Citation
  • Nishimura, Kiyohiko, 2006, “The New Policy Framework of the Bank of Japan: Central Banking in an Uncertain World,” Speech at Uppsala University, June 27.

    • Search Google Scholar
    • Export Citation
  • Roger, Scott, 1998, “Core Inflation: Concepts, Uses and Measurement,” Reserve Bank of New Zealand Discussion Paper, No. G98/9, July.

    • Search Google Scholar
    • Export Citation
  • Roger, Scott and Mark Stone, 2005, “On Target? The International Experience with Achieving Inflation Targets,” IMF Working Paper, No. 05/163, August.

    • Search Google Scholar
    • Export Citation
  • Shiratsuka, Shigenori, 1997, “Inflation Measures for Monetary Policy: Measuring Underlying Inflation Trend and Its Implication for Monetary Policy Implementation,” Institute for Monetary and Economic Studies Discussion Paper Series, No. 97-E-7, August.

    • Search Google Scholar
    • Export Citation
  • Shiratsuka, Shigenori, 1999, “Measurement Errors in the Japanese Consumer Price Index,” Monetary and Economic Studies, Vol., 17, No. 3 (Institute for Monetary and Economic Studies, Bank of Japan) December.

    • Search Google Scholar
    • Export Citation
  • WEO, 1999, “Safeguarding Macroeconomic Stability at Low Inflation,” World Economic Outlook, Chapter 4, October.

  • Whelan, Karl, 2000, “A Guide to the Use of Chain Aggregated NIPA Data,” Finance and Economics Discussion Series, No. 2000-35 (June, Federal Reserve Board of Governors).

    • Search Google Scholar
    • Export Citation
  • Wynne, Mark, 1999, “Core Inflation: A Review of Some Conceptual Issues,” Federal Reserve Bank of Dallas Working Paper Series, No. 9903 (June).

    • Search Google Scholar
    • Export Citation
25

Prepared by Christopher Faulkner-MacDonagh.

26

For more discussion of the understanding, see the accompanying staff report. Since the understanding is defined over the medium to long-term, both headline and core CPI refer to the same concept. However, there is a need to explain how the current conjuncture is affecting future inflation.

27

In this paper, medium-term inflation is measured by taking the three year centered moving average of the year-on-year changes. The latest data suggest that the average rate of medium-term inflation on this basis is 2 percent for all advanced economies—1½ percent for “low” inflation countries (Figure 1).

28

This point is also made in Bank of Japan (2006).

29

Results are from a simple regression of the average inflation rate for OECD economies from 1996-2004 on the level of product market regulations (both economy-wide and in the non-manufacturing sector) in 2003. Descriptions of the data are available in Conway, Janod, and Nicoletti (2005).

30

The Reserve Bank of New Zealand had a target range of 0-2 percent until 1997, when it was extended to 0-3 percent, and then further lifted to 1-3 percent in 2002. The Swiss National Bank equates price stability as annual inflation of less than 2 percent. The ECB aims to keep inflation below, but close to, 2 percent over the medium term.

31

See Moulton (1997), who also provides a useful overview of other biases.

32

In Japan, fresh food comprises only 5 percent of the CPI, but in other countries many more items are excluded. For example, in the United States, the core CPI excludes just under 25 percent of the overall index.

33

The CGPI is comprised of the prices of goods traded among companies, including both domestically-traded goods (910 items,74 percent of the index) and those that are exported (222 items, 14 percent of the index) and imported (293 items, 12 percent of the index). In a similar manner, the CSPI focuses on the prices of services traded among companies, although it excludes some services that are difficult to measure at a monthly frequency (such as imputed interest of financial services, wholesale, or retail trade) or services to individuals. It, too, includes domestic and imported services.

34

These indices more closely reflect domestic price developments than the GDP deflator, because the latter includes the effects of import prices, but with a negative sign. A large increase in the price of imported oil shows up as a drag on the GDP deflator, even though consumers and businesses are paying higher prices.

35

See Appendix A for details on the construction of both indices. The results shown here remove the top and bottom 10 percent of the index that is the most volatile.

36

Estimates in Shiratsuka (1999) suggest that the substitution bias (which is corrected by a chain-weighted formula) accounts for around one-half of the total bias (substitution plus others). This gives an upper bound to the overall bias of the CPI index of possibly around ½ percent.

37

Measured by the root mean squared forecast error of the regression of trimmed inflation on the 36-month moving average of CPI inflation.

Japan: Selected Issues
Author: International Monetary Fund
  • View in gallery

    Medium-Term Inflation Trends

    3-year centered average of y/y change in CPI (using monthly data)

  • View in gallery

    Fiscal Balance and Inflation, 1996-2004

    General government balance in industrial countries (avg. over the period, in percent of GDP) vs. year-on-year change in CPI.

  • View in gallery

    Core inflation trends

  • View in gallery

    Inflation trends in the business sector

    year-on-year percent change in corporate goods and services

  • View in gallery

    Trends in GDP deflators

    year-on-year percent changes in the deflators (n.s.a.)

  • View in gallery

    Alternate measures of core inflation

    year-on-year pct chg in trimmed mean and core Fisher indices