Business cycle fluctuations in Japanese macroeconomic time series: 1980–2000

Abstract

This paper provides the robust stylized facts for recent Japanese business cycle fluctuations, by examining the cyclical component of macroeconomic time series based on the frequency domain analysis. The results confirm many of the conventional views on business cycles in Japan. Among the most interesting findings are that non-scheduled hours worked plays a key role as a buffer for labor inputs. Distinctively, because of the behavior of non-scheduled cash earnings and bonuses, wages in Japan are very sensitive to changes in the level of economic activity. Also significant, the relationship between money and output has changed dramatically after the collapse of the bubble economy in 1991.

Keywords:

• business cycle fluctuations
• Japan
• band-pass filter

JEL Classifications:

• E320

Acknowledgements

The author is grateful to Professor Gary Saxonhouse for his exceptional guidance and encouragement, and the Cabinet Office, the Japanese Government for giving the author an opportunity to study at University of Michigan. The comments of Professor Nakamura and seminar participants at ESRI are gratefully acknowledged. The author is also indebted to the anonymous referees and the editors for valuable comments. Any views expressed herein are those of the author and not of the Cabinet office in which he works. Needless to say, any remaining errors are his own.

Notes

1. See, for instance, Motonishi and Yoshikawa (1999) and Hayashi and Prescott (2002).

2. Saito (1997) discusses the Japanese business cycle after 1991 in terms of its duration and speed. Yoshizoe et al. (2003) is a remarkable study, empirically examining the business cycle fluctuations from a macroeconomic perspective. Gerlach and Yiu (2004) apply the band-pass filter as one of the techniques for estimating output gaps for Asian countries including Japan.

3. In the following study, we consider the cyclical component of real output as a proxy of business cycle based on Burnsand Mitchell (1946).

4. The approximate band-pass filter applied in this study is the BP ₁₂(6,32), corresponding to the definition of business cycle in Burnsand Mitchell (1946); that is, the cyclical component of real output in the range of six quarters to 32 quarters. See Baxter and King (1999) and Christiano and Fitzgerald (2003) for detailed discussion on the approximate band-pass filter. Harvey and Trimbur (2003) present an alternative procedure for extracting cyclical components of macroeconomic time series relying on model-based filters.

5. In the appendix, we also examine the longer time series after the Second World War, applying quarterly data on major aggregate series. Our findings confirm the cyclical regularities of series after the Second World War are mostly consistent with those for the past two decades.

6. Each series basically covers the period from 1980:q1 to 2004:q3. Data Sources are: Annual Report on National Accounts, Economic and Social Research Institute, Cabinet Office, Labor Force Survey, Statistics Bureau, Ministry of Internal Affairs and Communications, Monthly Labor Survey, Ministry of Welfare and Labor, and Financial and Economic Statistics Monthly, Bank of Japan.

7. See Rotemberg and Woodford (1996), Stock and Watson (1999), and King and Rebelo (2000) for the stylized facts for the US economy.

8. The lagging property of private non-residential investment is also reported in Saito (1997).

9. Saito (1997) points out the lagging mechanism of employment adjustments.

10. Abe (2002) empirically examines the relationship between the corporate governance structure, well-known for the main-bank system or cross-shareholding system, and employment adjustment for individual industry, and verifies how the presence of a peculiar management system in Japanese firms influences the slowness of employment adjustment in Japan. Wakita (1997) discusses the presence of causality from labor hoarding to working hours in Japan.

11. Stock and Watson (1999) conclude that ‘real wages have essentially no contemporaneous co-movement with the business cycle’, by referring the cross-correlation with output. Since the definition of series in each study is different, the statistics should be interpreted with care.

12. As for BOJ's behavior, existing studies including Kasa and Popper (1997) show that BOJ has conducted the counter shocks in response to the economy, through the interest rate policy and quantitative policy. As seen above, our results also show that both the call rate and monetary base seem to be offsetting the economic fluctuations.

13. The Taylor-rule type of policy response function we estimate in this study simply consists of the natural rate of interest, inflation gap and output gap:

where i* _t is the target nominal interest rate, r ⁿ _t is the natural rate of interest, π ^e _t is the expected inflation rate based on Consumer Price Index, π* is the implicit inflation target, and GAP _t is output gap estimated using the approximate band-pass filter.

14. Miyao (2000) also discusses the possible structural change in money-output relation after the collapse of the bubble economy based on a vector autoregression (VAR) analysis.

15. Other major series besides money have also been tested. The results show the possible causality in most cases.

16. There are some other techniques designed for de-trending and smoothing: removing liner trends, differencing, and the Hodrick-Prescott (1980) filter. See Baxter and King (1999) for detailed comparison among these filters.

17. Since we define the business cycle as fluctuations between six and 32 quarters in time domain, we apply λ = |2π/32| and = |2π/6| in frequency domain.