Bhar and Hamori, International Journal of Applied Economics, 6(1), March 2009, 77-89 77 Growth Rate of Domestic Credit and Output: Evidence of the Asymmetric Relationship between Japan and the United States Ramaprasad Bhar a and Shigeyuki Hamori b a The University of New South Wales, Australia and b Kobe University, Japan Abstract This paper uses a Markov-switching vector autoregression (MS-VAR) model to capture the nonlinearity of business cycles in the United States and Japan. We found a clear difference between the empirical results of the two countries, which might reflect the difference in their respective financial systems. We also found the existence of a comovement between the two countries over the last several decades. Keywords: Markov-switching vector autoregression (MS-VAR), domestic credit, asymmetric effect, concordance measure JEL Classification: C22, E44 1. Introduction Monetary economists have frequently expressed the view that the financial system is not only an important source of business cycles but also the propagation mechanism for cyclical fluctuations. This implies that the financial system itself causes the endogenously arising economic volatility. This view has certain significant empirical implications. Several recent recessions have been associated with credit crunches (or disintermediation). In the event of a decline in the volume of bank-extended credit, credit crunches are often accompanied by an increased incidence of non-price rationing of credit (Azariadis and Smith, 1998). Because of the importance of the credit variable, many researchers are in the process of exploring its effects on the business cycle. Bernanke and Gertler (1989), Kiyotaki and Moore (1997), and Azariadis and Smith (1998) are some examples of related significant theoretical research. Bernanke and Gertler (1989) is an important theoretical work that analyzes the credit view of monetary policy transmission. A characteristic of their model is that the balance sheet positions of the borrower play an important role. Taking into consideration the asymmetric information between borrowers and lenders, Bernanke and Gertler show that the agency costs of undertaking physical investments have certain accelerator effects on investment. When the economy is stable, strong balance sheets of borrowers may expand investment demand and amplify the upturn. On the other hand, when the economy is unstable, weak balance sheets may decrease investment demand and
Bhar and Hamori, International Journal of Applied Economics, 6(1), March 2009, 77-89 78 amplify the downturn. They also suggest that the agency problem may only bind on the down side, and thus, the aggregate effects of productivity shocks may be asymmetric. Kiyotaki and Moore (1997) theoretically analyze the manner in which credit constraints interact with the aggregate economic activity over the business cycle. A characteristic of their model is that the prices of collateralized assets (e.g., land, buildings, and machinery) affect the borrower s credit limits and the size of credit limits affects these prices. The dynamic interaction between credit limits and asset prices cause the effects of shocks to persist, amplify, and spill over to other sectors. Kiyotaki and Moore also show that even a small, temporary shock to technology can generate large, persistent movements of output and asset prices. Azariadis and Smith (1998) analyzed the relationship between credit and production using a simple dynamic general equilibrium model. A characteristic of their model is that capital investments are credit financed, and credit markets are characterized by the presence of an adverse selection problem. They show that the existence of two equilibrium regimes is possible. One is the Walrasian regime in which economic activity is recovering and the rising interest rates are accompanied by a credit market. The other is the credit rationing regime in which economic activity and interest rates are declining, and credit constraints are binding. Parallel to the theoretical developments, there has been considerable empirical research on the role of credit in the process of business cycles. Japan, Latin America, and Scandinavia have each experienced major problems in their banking sectors, which coincided with recent severe recessions. The role of banks in both such crises and subsequent recoveries is likely to be an important research subject. Empirical evidence at the aggregate level for this transmission channel is found in Bernanke and Blinder (1992), Bernanke and Gertler (1995), and Christiano et al. (1996). These researches tend to show that the effects of monetary policy or other shocks have an asymmetric effect on the economy. These effects arise from the fact that lending is pro-cyclical, and therefore, credit constraints become more binding during a downturn, whereas they do not have an equally symmetric positive effect during an upturn. This paper empirically analyzes the relationship between domestic credit and output and examines if any asymmetric effects can be found in the economy. The characteristics of this paper are threefold. First, we analyze the relationship between domestic credit and output using the Markov-switching vector autoregression (MS-VAR) model, which offers the following advantage: it allows parameters to vary across regimes. Hence, this model is being widely applied in financial economics (See Guidolin and Timmermann, 2005a, 2005b; Guidolin and Ono, 2005). The basic feature of MS-VAR can serve as a powerful tool for empirically analyzing asymmetry. This paper empirically analyzes the data of Japan and the United States. These two countries are compared as they have very different financial systems. Historically, Japan has adopted the indirect financial system and the United States, the direct financial system. Thus, we analyze whether this difference of financial structure between the two countries has some influence on the relationship between domestic credit and output. Finally, using the concordance
Bhar and Hamori, International Journal of Applied Economics, 6(1), March 2009, 77-89 79 measure, we analyze the comovement of the business cycles between Japan and the United States. The concordance measure is a nonparametric statistic, first proposed by Harding and Pagan (1999), and later extended by McDermott and Scott (1999) owing to its distributional properties. 2. Model 2.1 MS-VAR Let y 1(t) denote the growth rate of bank lending and y 2(t), the growth rate of industrial production. Next, in order to characterize the asymmetric movement of bank lending and output growth rates, we use the Markov switching (MSW) model as follows: 1 y1,t α0,s α t 1,S α t 2,S t ε1,s t y1,t 1, (1) y = 2,t β0,s β t 1,S β + t 2,S ε t 2,S y2,t 1 t 2 ε1,s 0 σ t 1,S σ t 1,2,St ~N, 2. (2) 2,S 0 ε t 1,2,S σ σ t 2,St Here, S { 0,1} t indicates the MS variable defining the regime occurring at a given time. The transition between the two states is governed by the time-invariant transition probability matrix: p 1 p 1 p22 p22 11 11. (3) It is clear that we have to estimate 20 parameters in this model. 2.2 Concordance Measure The recursive estimation process generates the probability that a particular month is in a state i ( i = 1, 2 ). Using these probability state estimates, we form the concordance statistics. The following is a brief description of this statistic. The concordance measure is a nonparametric statistic, first proposed by Harding and Pagan (1999), and later extended by McDermott and Scott (1999), owing to its distributional properties. This statistic has been successfully applied in studies on the comovements of the prices of seemingly unrelated commodities (e.g., Cashin, McDermott, and Scott (1999)). The concordance measure statistic between the two series x and x j is defined by i
Bhar and Hamori, International Journal of Applied Economics, 6(1), March 2009, 77-89 80 T { ( ) ( 1 )( 1 S ) t 1, } 1 C = S S + S ij, = it, jt, it, jt T, (4) where T is the number of observations in each series and S it, is a binary variable taking on the value 0 when the corresponding value of x i is below a certain reference level; otherwise it is 1. S j, t is similarly defined. In this paper, we are dealing with the probability series, and choose 0.25 as a reference value. This implies that when the estimated state probability is less than 25%, we consider it low and assign 0 to the corresponding S variable. In order to conduct a statistical significance test of the computed concordance statistic between the two series, Cashin, McDermott, and Scott (1999) propose and conduct a simulation experiment to establish the validity of their approach. We follow this guideline and compute the critical value of the concordance statistic for 10%, 5%, and 1% levels of significance under the assumption that a Brownian motion without a drift generates the probability state realization. 3. Data This paper uses the quarterly data of real GDP and domestic credit for the United States and Japan over the period between the first quarter of 1957 and the first quarter of 2008. The data source is the International Financial Statistics (IFS) CD-ROM of the International Monetary Fund (IMF). 1 The quarterly growth rates are calculated as the first difference of the logarithmic value. Since the first difference of each data set is used for empirical analysis, we use the data starting from the second quarter of 1957. 4. Empirical Results Tables 1 and 2 show the empirical results of the MSW model for Japan and the United States, respectively. In both the countries, state 1 corresponds to the low variance state and state 2 corresponds to the high variance state. The transition probabilities are significant in Tables 1 and 2, as expected in an MSW model. In Table 1, for Japan, the average duration of regime 1 (low variance state) is 66.17 quarters. The covariance in this state is estimated to be 0.20657 and is statistically significant. This tends to indicate that for Japan, the output growth was supported with positive interaction from the credit market during the relatively calm period, which, on average, lasted longer than the other state. We consider this phenomenon to be intuitively correct. In addition, in this state, the lagged coefficient of the credit growth in the output growth equation is estimated to be 0.16213, and is statistically significant. In other words, the credit growth significantly influences the real growth rate during the low variance state in Japan. In Table 1, the covariance term in regime 2 is estimated to be 12.60342, but is statistically significant. The lagged credit growth coefficient is estimated to be 0.44038
Bhar and Hamori, International Journal of Applied Economics, 6(1), March 2009, 77-89 81 and is significant in the output growth equation. This indicates that credit growth influences the output growth rate in the high variance state as well. An interesting fact is that in Japan, domestic credit is a significant explanatory variable for output growth in both the states. In Table 2, the classification of regimes 1 and 2 in the case of the United States is similar to that in the case of Japan, i.e., low and high variance states. Note that the differences in the variances for the United States are not as pronounced as those for Japan. The covariance terms are estimated to be 0.05121 in state 1 and 0.02578 in state 2; however, they are not statistically significant in either state. Another interesting fact is that the average duration of state 2 is longer than that of state 1 in the United States, contrary to the case of Japan. In Table 2, for the United States, the lagged credit growth coefficient in the output growth equation is estimated to be 0.06330, but it is not significant in state 1. On the other hand, the lagged credit growth coefficient in the output growth equation is estimated to be 0.13828, and is statistically significant in state 2. As is clear from the empirical results, there is a clear difference between the empirical results for Japan and the United States. In the case of Japan, domestic credit growth is a significant explanatory variable for output growth in both states. For the United States, however, domestic credit growth is a significant explanatory variable only for the high variance state. In other words, the relationship between domestic credit and business cycle is symmetric for Japan but asymmetric for the United States. These empirical results might reflect the difference between the financial systems of Japan and the United States. There are two methods for the flow of funds between the household and firm: one is the indirect financial system and the other, the direct financial system. Since the financial structure of Japan is centered on indirect financing, the financial dependence of firms on banks is high; thus, banks have mainly taken over corporate governance. On the other hand, the financial structure of the United States is centered on direct financing, and thus, the stock market has taken charge of corporate governance. Thus, the movement of domestic credit may be more accurately reflected in the business sector of Japan rather than that of the United States. Figures 1 and 2 show the estimated filtered probability that the economy of each country is in a relatively high variance state. The figures clearly indicate that the probabilities that the economies are in the same state tend to move together internationally. This might indicate that the economies of Japan and the United States are likely to have been integrated over the last several decades. Table 3 shows the concordance statistics. 2 The concordance statistic for the full sample period is 0.7537, and is statistically significant at the 1% level. This implies that the business cycles between the two countries are in the same phase. Next, we split the sample into two subsample periods in order to ascertain if there is any change of concordance between the two countries. The concordance statistic for the first subsample is 0.8571, and is statistically significant at the 1% level, whereas that for the
Bhar and Hamori, International Journal of Applied Economics, 6(1), March 2009, 77-89 82 second subsample is 0.6571, and is statistically significant at the 5% level. These results support the co-movement between the two countries over the last several decades, which is consistent with the observations from Figures 1 and 2. 5. Concluding Remarks There exist many theoretical models in which credit markets propagate shocks to the economy because of the existence of asymmetric information. The procyclicality of bank lending results in the amplification of the business cycle. This has a more pronounced effect during recessions and thus leads to asymmetric effects of monetary policy over time. To capture this type of nonlinearity, we use an MS-VAR model. In this model, parameters switch according to an unobservable state variable that is assumed to capture the changing credit or economic regimes and is estimated together with the model parameters. We found a clear difference between the empirical results of Japan and the United States. For Japan, domestic credit growth is a significant explanatory variable for output growth not only in the low variance state but also in the high variance state. For the United States, however, domestic credit growth is a significant explanatory variable only for the high variance state. These empirical results might reflect the difference between the financial systems of Japan and the United States. Since the financial structure of Japan is centered on indirect financing, the financial dependence of firms on banks is high, and thus, banks have mainly take charge of corporate governance. On the other hand, the financial structure of the United States is centered on direct financing, and thus, the stock market has taken over corporate governance. Thus, the movements of domestic credit are more accurately reflected in the business sector of Japan rather than that of the United States. We have also observed co-movement between the two countries over the last several decades. Endnotes * Ramaprasad Bhar, School of Banking and Finance, The University of New South Wales Sydney 2052, AUSTRALIA, Email: R.Bhar@unsw.edu.au. Shigeyuki Hamori, Faculty of Economics, Kobe University, 2-1, Rokkodai, Nada-Ku, Kobe, 657-8501, JAPAN, Email: hamori@econ.kobe-u.ac.jp. We are grateful to an anonymous referee and the editor for many helpful comments and suggestions. 1. The IFS code of each variable is as follows: real GDP for Japan: 15899BVRZF...; real GDP for the United States: 11199BVRZF...; domestic credit for Japan: 15832...ZF; and domestic credit for the United States: 11132...ZF. 2. Bhar and Hamori (2007) analysed the co-movement of the price of risk in G-7 equity markets using the concordance measure.
Bhar and Hamori, International Journal of Applied Economics, 6(1), March 2009, 77-89 83 References Azariadis, C. and B. Smith. 1998. "Financial Intermediation and Regime Switching in Business Cycles," American Economic Review, 88, 516-536. Bernanke, B. and M. Gertler. 1989. "Agency Costs, Net Worth and Business Fluctuations, " American Economic Review, 79, 14-31. Bernanke, B. and A. S. Blinder. 1992. "The Federal Funds Rate and the Channels of Monetary Transmission," American Economic Review, 82, 901-921. Bernanke, B. and M. Gertler. 1995. "Inside the Black Box: The Credit Channel of Monetary Policy Transmission," Journal of Economic Perspectives, 9, 27-48. Bhar, R. and S. Hamori. 2007. "Co-movement in the Price of Risk of Aggregate Equity Markets," Economic Systems, 31, 256-271. Brock, W., D. Dechert, J. Scheinkman, and B. LeBaron. 1996. "A Test for Independence Based on Correlation Dimension," Econometric Reviews, 15 (3), 197-235. Cashin, P., C. J. McDermott, and A. Scott. 1999. "The Myth of Co-moving Commodity Prices," Reserve Bank of New Zealand Discussion Paper G99/9, Wellington. Christiano, L. J., M. Eichenbaum, and C. Evans. 1996. "The Effects of Monetary Policy Shocks: Evidence from the Flow of Funds," Review of Economics and Statistics, 78, 16-34. Guidolin, M. and A. Timmermann. 2005a. "An Econometric Model of Nonlinear Dynamics in the Joint Distribution of Stocks and Bond Returns," Working Paper, #2005-003A, Federal Reserve Bank of St. Louis. Guidolin, M. and A. Timmermann. 2005b. "International Asset Allocation under Regime Switching, Skewness and Kurtosis Preferences," Working Paper, #2005-034A, Federal Reserve Bank of St. Louis. Guidolin, M. and S. Ono. 2005. "Are the Dynamic Linkages Between the Macroeconomy and Asset Prices Time-varying?" Working Paper, #2005-056A, Federal Reserve Bank of St. Louis. Harding, D. and A. Pagan. 1999. "Dissecting the Cycle," Melbourne Institute Working Paper, No.13/99, Mel borne: University of Melbourne Australia. Kaufmann, S. and M. T. Valderrama. 2004. "The Role of Bank Lending in Marketbased and Bank-based Financial Systems," Quarterly Review of Economic Policy, 88-97.
Bhar and Hamori, International Journal of Applied Economics, 6(1), March 2009, 77-89 84 Kiyotaki, N. and J. Moore. 1997. "Credit Cycles," Journal of Political Economy, 105, 211-248. McDermott, C. J. and A. Scott. 1999. "Concordance in Business Cycles," Reserve Bank of New Zealand Discussion Paper G99/7, Wellington.
Bhar and Hamori, International Journal of Applied Economics, 6(1), March 2009, 77-89 85 Table 1. Parameter Estimates of the Model: Growth Rate of Domestic Credit and Output Two State Regime Dependent Relationship (Japan) Regime 1: Credit Growth Output Growth Regime 2: Intercept 0.00480 0.00433 (3.63) (4.72) Credit Growth ( 1) 0.58434 0.16213 (7.84) (3.37) Output Growth ( 1) 0.05827 0.03636 (-0.82) (-0.70) Variance 1.23842 0.63559 (7.59) (8.23) Co-variance 0.20657 (2.69) Intercept 0.04249 0.04658 (6.68) (3.17) Credit Growth ( 1) 0.04351 0.44038 (0.46) (-1.83) Output Growth ( 1) 0.02666 0.34893 (0.77) (-4.11) Variance 26.96949 174.94860 (7.30) (7.15) Co-Variance 12.60342 (1.92) Transition probability matrix p i,j : j i j= 1 j= 2 i= 1 0.98489 0.01780 (90.53) i= 2 0.01511 0.98220 (74.53) Average duration in a particular state (quarters) 66.17 56.17 Numbers in parentheses are student t-statistics. The data covers the period from 1957 Q1 to 2008 Q1. The RCM measure for Japan is 6.82.
Bhar and Hamori, International Journal of Applied Economics, 6(1), March 2009, 77-89 86 Table 2. Parameter Estimates of the Model: Growth Rate of Domestic Credit and Output Two State Regime Dependent Relationship (the United States) Regime 1: Credit Growth Output Growth Regime 2: Intercept 0.00739 0.00454 (3.51) (3.47) Credit Growth ( 1) 0.40472 0.06330 (4.11) (1.14) Output Growth ( 1) 0.27480 0.27816 (1.28) (2.17) Variance 0.68730 0.20088 (5.87) (6.43) Co-variance 0.05121 (1.20) Intercept 0.02399 0.00286 (10.75) (1.59) Credit Growth ( 1) 0.17230 0.13828 ( 2.07) (2.02) Output Growth ( 1) 0.27024 0.27808 (2.73) (3.74) Variance 1.47859 1.07032 (7.48) (7.44) Co-Variance 0.02578 (0.24) Transition probability matrix p i,j : j i j= 1 j= 2 i= 1 0.98039 0.01449 (70.50) i= 2 0.01961 0.98551 (91.40) Average duration in a particular state (quarters) 50.99 68.99 Numbers in parentheses are student t-statistics. The data covers the period from 1957 Q1 to 2008 Q1. The RCM measure for Japan is 17.06.
Bhar and Hamori, International Journal of Applied Economics, 6(1), March 2009, 77-89 87 Table 3. Concordance Statistics Sample Period Concordance statistics 1% Critical Value 5% Critical Value 10% Critical Value 1957Q3-2008Q1 0.7537 *** 0.6234 0.5831 0.5597 1957Q3-1981Q4 0.8571 *** 0.6738 0.6184 0.5855 1982Q1-2008Q1 0.6571 ** 0.6683 0.6146 0.5827 *** shows the significance at the 1% level. ** shows the significance at the 5% level. * shows the significance at the 10% level.
Bhar and Hamori, International Journal of Applied Economics, 6(1), March 2009, 77-89 88 Figure 1 Estimated Filtered Probability for Japan: Pr (S t = 2) High Volatility State Japan 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1957Q3 1962Q1 1966Q3 1971Q1 1975Q3 1980Q1 1984Q3 1989Q1 1993Q3 1998Q1 2002Q3 2007Q1
Bhar and Hamori, International Journal of Applied Economics, 6(1), March 2009, 77-89 89 Figure 2 Estimated Filtered Probability for the United States: Pr (S t = 2) High Volatility State for USA 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1957Q3 1960Q3 1963Q3 1966Q3 1969Q3 1972Q3 1975Q3 1978Q3 1981Q3 1984Q3 1987Q3 1990Q3 1993Q3 1996Q3 1999Q3 2002Q3 2005Q3