Articleidicec.1999.1607,availableonlineathtp://www.idealibrary.comonIdeAl Monetary Growth and Inflation in China A Reexamination T Mohammad s Hasan School of Financial Studies and Law, Sheffield Hallam University, City Campus Pond Street, Sheffield, S 1WB, United Kingdom Received September 24, 1996, revised June 8, 1999 Hasan, Mohammad S -Monetary Growth and Inflation in China: A Reexamination Using the notion of cointegration theory and its implied vector error correction mod- eling strategy, this paper reexamines the relationship between monetary forces and inflation in mainland China. Contrary to most recent research in this area, these results based on unit root and cointegration tests indicate a reliable long-run relationship between the general price level and the money stock, as well as between inflation and mone wth, Our findings also suggest a bi-directional or feedback relationship between inflation and monetary growth. J. Comp. Econ., December 1999, 27 (4), pp. 669-685 School of Financial Studies and Law, Sheffield Hallam University, City Campus, Pond Street, Sheffield, SI IWB, United Kingdom. 0 1999 Academic Press Journal of Economic Literature Classification Numbers: P24, P52, Oll L INTRODUCTION The nature of the relationship between monetary aggregates and inflation in mainland China has been the subject of ongoing debate among researchers. Chow (1987) contends that the quantity theory of money appears to be a plausible explanation of the inflationary process in China over the period 1952-1983. In contrast, Peebles(1992), while recognizing the institutional differences of the Chinese economy from other highly developed market economies, argues that the quantity theory is no help in understanding the historical association between money and prices in China. The situation became more clouded when Huang (1995)reported that monetary forces explain price movements in China in the The author thanks John Bonin and two anonymous referees for providing useful criticisms and Press All rights of reproduction in any form reserved
Monetary Growth and Inflation in China: A Reexamination1 Mohammad S. Hasan School of Financial Studies and Law, Sheffield Hallam University, City Campus, Pond Street, Sheffield, S1 1WB, United Kingdom E-mail: m.s.hasan@shu.ac.uk Received September 24, 1996; revised June 8, 1999 Hasan, Mohammad S.—Monetary Growth and Inflation in China: A Reexamination Using the notion of cointegration theory and its implied vector error correction modeling strategy, this paper reexamines the relationship between monetary forces and inflation in mainland China. Contrary to most recent research in this area, these results based on unit root and cointegration tests indicate a reliable long-run relationship between the general price level and the money stock, as well as between inflation and monetary growth. Our findings also suggest a bi-directional or feedback relationship between inflation and monetary growth. J. Comp. Econ., December 1999, 27(4), pp. 669–685. School of Financial Studies and Law, Sheffield Hallam University, City Campus, Pond Street, Sheffield, S1 1WB, United Kingdom. © 1999 Academic Press Journal of Economic Literature Classification Numbers: P24, P52, O11. 1. INTRODUCTION The nature of the relationship between monetary aggregates and inflation in mainland China has been the subject of ongoing debate among researchers. Chow (1987) contends that the quantity theory of money appears to be a plausible explanation of the inflationary process in China over the period 1952–1983. In contrast, Peebles (1992), while recognizing the institutional differences of the Chinese economy from other highly developed market economies, argues that the quantity theory is no help in understanding the historical association between money and prices in China. The situation became more clouded when Huang (1995) reported that monetary forces explain price movements in China in the 1 The author thanks John Bonin and two anonymous referees for providing useful criticisms and suggestions. Journal of Comparative Economics 27, 669–685 (1999) Article ID jcec.1999.1607, available online at http://www.idealibrary.com on 669 0147-5967/99 $30.00 Copyright © 1999 by Academic Press All rights of reproduction in any form reserved
MOHAMMAD S HASAN prereform perio reform economy. In contrast, the empirical work of Blejer et al. (1991)on the demand function in the m documents the existence of a relatively stable relationship between a range of regates, inflation, and real lr v After its inception, the Chinese economy experienced relative price stabilit more than 30 years. Following the economic reform in 1979, price increases remained a substantial and common phenomena almost throughout the 1980s and mid-1990s. Inflationary pressure rose to its first postreform peak of 6%per annum in 1980: in the second half of the 1980s. inflation continued to accelerate and reached double digits with a peak of 18.5% in 1989. Inflationary pressure subsided concurrent with faltering growth for the next three years. But ar inflationary upsurge again hit the economy in 1992 and reached its historical apex of 21.7% in 1995 as fast real growth resumed, e.g., 13.5% in 1994 Contemporaneously, the stock of broad money increased by 268% during the period from 1979 to 1985. Broad money rose by 30% in 1992 and at a similar rate in 1993, narrow money and currency in circulation were growing at a rate of% and 46%, respectively, by the end of the first quarter of 1993( Harrold and Lall 1993). Although reserve requirement, credit ceiling, variable interest rates, and other administrative levers have been introduced to control monetary and credit aggregates in the postreform period, their role and effectiveness were problem atic. The postreform economy has experienced insufficient control over monetary and credit aggregates. The activities of nonbank financial institutions(NBFls) and disintermediation have loosened the link between the credit plan and mon- etary aggregates. Unbridled monetary growth in the postreform period was blamed as an important contributor to this inflationary upsurge The aim of this paper is to delineate both the short-run and long-run monetary dynamics of inflation in the Chinese economy. This approach has significant differences from others both theoretically and methodologically, First, this study adopts the theoretical framework of a conventional monetarist model( Chen 1989, Chow, 1987; Huang, 1995)but extends the analysis by incorporating a more general model of inflation to capture the institutional features of the Chinese economy. Second, the study uses the cointegration ogy with However, the exception applies during the period of the Great Leap Forward(1958-1961)when infiation achieved a peak of 16. 2% Harrold and Lall (1993)have noted that the postreform credit explosion outside the credit plan as induced by diversion of funds from the specialized banks to a variety of fast growing NFBls and into the direct hands of enterprises through the interbank market See, for example, Country Profile: China Mongolia, The Economic Intelligence Unit, 1994-1995, p. 17. The country report also suggests that the government often had to resort to printing noney to cover the budget deficit as selling treasury bonds to state employees through paro deductions became an unpopular practice
prereform period reasonably while they have no predictive content in the postreform economy. In contrast, the empirical work of Blejer et al. (1991) on the money demand function in the postreform period from 1983QI to 1988QIII documents the existence of a relatively stable relationship between a range of monetary aggregates, inflation, and real income. After its inception, the Chinese economy experienced relative price stability for more than 30 years.2 Following the economic reform in 1979, price increases remained a substantial and common phenomena almost throughout the 1980s and mid-1990s. Inflationary pressure rose to its first postreform peak of 6% per annum in 1980; in the second half of the 1980s, inflation continued to accelerate and reached double digits with a peak of 18.5% in 1989. Inflationary pressure subsided concurrent with faltering growth for the next three years. But an inflationary upsurge again hit the economy in 1992 and reached its historical apex of 21.7% in 1995 as fast real growth resumed, e.g., 13.5% in 1994. Contemporaneously, the stock of broad money increased by 268% during the period from 1979 to 1985. Broad money rose by 30% in 1992 and at a similar rate in 1993; narrow money and currency in circulation were growing at a rate of 41% and 46%, respectively, by the end of the first quarter of 1993 (Harrold and Lall, 1993). Although reserve requirement, credit ceiling, variable interest rates, and other administrative levers have been introduced to control monetary and credit aggregates in the postreform period, their role and effectiveness were problematic. The postreform economy has experienced insufficient control over monetary and credit aggregates. The activities of nonbank financial institutions (NBFIs) and disintermediation have loosened the link between the credit plan and monetary aggregates.3 Unbridled monetary growth in the postreform period was blamed as an important contributor to this inflationary upsurge.4 The aim of this paper is to delineate both the short-run and long-run monetary dynamics of inflation in the Chinese economy. This approach has significant differences from others both theoretically and methodologically. First, this study adopts the theoretical framework of a conventional monetarist model (Chen, 1989; Chow, 1987; Huang, 1995) but extends the analysis by incorporating a more general model of inflation to capture the institutional features of the Chinese economy. Second, the study uses the cointegration methodology with an 2 However, the exception applies during the period of the Great Leap Forward (1958–1961) when inflation achieved a peak of 16.2%. 3 Harrold and Lall (1993) have noted that the postreform credit explosion outside the credit plan was induced by diversion of funds from the specialized banks to a variety of fast growing NFBIs and into the direct hands of enterprises through the interbank market. 4 See, for example, Country Profile: China Mongolia, The Economic Intelligence Unit, 1994–1995, p. 17. The country report also suggests that the government often had to resort to printing money to cover the budget deficit as selling treasury bonds to state employees through payroll deductions became an unpopular practice. 670 MOHAMMAD S. HASAN
MONEY AND INFLATION N CHINA 671 error correction model(ECm) to identify short-run and long-run interdependen cies and the causal linkage between prices and money stock. This approach avoids the spurious regression problem and offers a parsimonious time series approach based on a more general inflation model with a rich dynamic structure Third, it is a well known fact that the official price indices in China do not provide a proper yardstick for measuring the overall extent and character of inflation due to inflationary repression over a long period of time. Previous studies concerning the money-price relationship have focused on the official price indices(Chow, 1987; Huang, 1995). In contrast, we use a measure of the true price index to explore an otherwise hidden relationship between money and prices. Despite the institutional differences between the Chinese economy and other Western market economies, the statistical techniques, when supplemented with the true price index, unravel the monetary dynamics of the inflationary process in China The paper is organized as follows. Section 2 presents a more general model designed to estimate the monetary dynamics of inflation and discuss the institu- tional issues in China that affect the application of the monetarists'modeling strategy. Sections 3 and 4 present the estimates, while the final section offers a 2. THE MODEL We start with a simple and transparent quantity theory model to explain the monetary dynamics of inflation in China. Irving Fishers(1911)celebrated quantity equation of exchange was responsible for assigning monetary forces the principal role in the determination of the price level where M,v, P, and y are the quantity of money, the income velocity, the price level, and real income, respectively. Classical, monetarist, and new classical economists invoke several assumptions to convert this simple identity to rticulated theory. The classical assumptions of full employment equilibrium fully flexible price and wages, Friedman's statement of the natural rate of unemployment, and the rational expectation hypothesis of the New Classical economists characterize a time path where long-run monetary growth only determines the rate of inflation. If we interpret the quantity theory as a long-run equilibrium and supplement it with a short-run error correction mechanism obtain a dynamic path of the inflation rate that cannot deviate too far from the path of long-run solution values. More specifically, combining the notion of tegration and the error correction mechanism, we specify the following del to capture the dynamic essence of the inflation rate
error correction model (ECM) to identify short-run and long-run interdependencies and the causal linkage between prices and money stock. This approach avoids the spurious regression problem and offers a parsimonious time series approach based on a more general inflation model with a rich dynamic structure. Third, it is a well known fact that the official price indices in China do not provide a proper yardstick for measuring the overall extent and character of inflation due to inflationary repression over a long period of time. Previous studies concerning the money–price relationship have focused on the official price indices (Chow, 1987; Huang, 1995). In contrast, we use a measure of the true price index to explore an otherwise hidden relationship between money and prices. Despite the institutional differences between the Chinese economy and other Western market economies, the statistical techniques, when supplemented with the true price index, unravel the monetary dynamics of the inflationary process in China. The paper is organized as follows. Section 2 presents a more general model designed to estimate the monetary dynamics of inflation and discuss the institutional issues in China that affect the application of the monetarists’ modeling strategy. Sections 3 and 4 present the estimates, while the final section offers a summary and conclusion. 2. THE MODEL We start with a simple and transparent quantity theory model to explain the monetary dynamics of inflation in China. Irving Fisher’s (1911) celebrated quantity equation of exchange was responsible for assigning monetary forces the principal role in the determination of the price level, MV 5 PY, (1) where M, V, P, and Y are the quantity of money, the income velocity, the price level, and real income, respectively. Classical, monetarist, and new classical economists invoke several assumptions to convert this simple identity to an articulated theory. The classical assumptions of full employment equilibrium, fully flexible price and wages, Friedman’s statement of the natural rate of unemployment, and the rational expectation hypothesis of the New Classical economists characterize a time path where long-run monetary growth only determines the rate of inflation. If we interpret the quantity theory as a long-run equilibrium and supplement it with a short-run error correction mechanism, we obtain a dynamic path of the inflation rate that cannot deviate too far from the path of long-run solution values. More specifically, combining the notion of cointegration and the error correction mechanism, we specify the following model to capture the dynamic essence of the inflation rate, MONEY AND INFLATION IN CHINA 671
MOHAMMAD S HASAN P=0o+a,M,+ a2V,,+ Er △P,= ∑n△-,+v where the growth rates of real output, money, and income velocity are assumed to be exogenously determined at time period t and E, is an identically and independently distributed error term. Equation (2) is the cointegration regression and depicts the equilibrium relationship; Eq. ( 3)is the error correction equation and describes the process of adjustment to equilibrium, with Er-I being the equilibrium error of the previous period The application of quantity theory models of inflation to a developing semi monetized economy is a somewhat controversial issue. However, Chow (1987) and Duck(1993) provide good justifications for using such models in a devel oping country, while Chow(1987) has estimated a variant of a quantity theory model for the Chinese economy. In order to alleviate the problem of omitted variable bias, we specify a general model of inflation that subsumes the aggregate demand and supply factors, as well as monetary forces(Darrat, 1994; Gordon 1982; Huang,1995).Let P1=δ+81(L)M1+82(Lg1+63(LW1+δ4(LAP1+85(LIP4+Ep,(4) where the right-hand variables are the money stock(M), economywide excess demand pressure proxied by the output gap(g), wages(), agricultural pro- ductivity(AP), and a measure of industrial productivity (IP). The 8()s are lag polynomials in the lag operator and st is a serially independent error with zero mean. Equation (4)is s-curve inflation equation augmented by the inclusion of supply-side factors(Gordon, 1982) If inflation and the right-hand variables, such as growth of monetary aggregates are cointegrated, P is proportional to M from a statistical point of view. If the variables are cointegrated, the next test seeks to ascertain whether the coefficient of E-I in Eq. (3)or the coeficient of s-l in a similar error correction equation is negative and statistically significant so as to specify the short-run dynamics of the system. Granger(1988) has shown that finding cointegration also implies the pres- ence of Granger causality between cointegrated variables, at least in one direction Several institutional issues affect the mechanical application of the quantity eory as well as the general modeling strategy and deserve discussion. First over a long period of time, retail price growth in China was slowed or even halted by administrative price controls, usually accompanied by administrative wage controls. In the postreform period, price liberalization progressed in a piecemeal fashion and with considerable delays. Previous research has attempted to mea- sure the extent and character of true inflationary or deflationary pressure in the economy( Chen and Hou, 1986, Feltenstein and Farhadian, 1987, Feltenstein and
Pt 5 a0 1 a1Mt 1 a2Vt 2 a3Yt 1 «t (2) DPt 5 2g«t21 1 O i51 s di DMt2s 1 O i51 s bi DVt2s 2 O i51 s hi DYt2s 1 vt, (3) where the growth rates of real output, money, and income velocity are assumed to be exogenously determined at time period t and «t is an identically and independently distributed error term. Equation (2) is the cointegration regression and depicts the equilibrium relationship; Eq. (3) is the error correction equation and describes the process of adjustment to equilibrium, with «t21 being the equilibrium error of the previous period. The application of quantity theory models of inflation to a developing semimonetized economy is a somewhat controversial issue. However, Chow (1987) and Duck (1993) provide good justifications for using such models in a developing country, while Chow (1987) has estimated a variant of a quantity theory model for the Chinese economy. In order to alleviate the problem of omitted variable bias, we specify a general model of inflation that subsumes the aggregate demand and supply factors, as well as monetary forces (Darrat, 1994; Gordon, 1982; Huang, 1995). Let Pt 5 d0 1 d1~L! Mt 1 d2~L! gt 1 d3~L!Wt 1 d4~L!APt 1 d5~L!IPt 1 jt, (4) where the right-hand variables are the money stock (M), economywide excess demand pressure proxied by the output gap ( g), wages (W), agricultural productivity (AP), and a measure of industrial productivity (IP). The d(L)s are lag polynomials in the lag operator and jt is a serially independent error with zero mean. Equation (4) is a Phillips-curve inflation equation augmented by the inclusion of supply-side factors (Gordon, 1982). If inflation and the right-hand variables, such as growth of monetary aggregates are cointegrated, P is proportional to M from a statistical point of view. If the variables are cointegrated, the next test seeks to ascertain whether the coefficient of «t21 in Eq. (3) or the coefficient of jt21 in a similar error correction equation is negative and statistically significant so as to specify the short-run dynamics of the system. Granger (1988) has shown that finding cointegration also implies the presence of Granger causality between cointegrated variables, at least in one direction. Several institutional issues affect the mechanical application of the quantity theory as well as the general modeling strategy and deserve discussion. First, over a long period of time, retail price growth in China was slowed or even halted by administrative price controls, usually accompanied by administrative wage controls. In the postreform period, price liberalization progressed in a piecemeal fashion and with considerable delays. Previous research has attempted to measure the extent and character of true inflationary or deflationary pressure in the economy (Chen and Hou, 1986; Feltenstein and Farhadian, 1987; Feltenstein and 672 MOHAMMAD S. HASAN
MONEY AND INFLATION N CHINA Ha, 1991). However, Feltenstein and Ha(1991)develop a measure of the true price index in China from a money demand function and this performed better in a simulation exercise. Li and Leung(1994) have updated the annual data based on this approach. We use their measure of the price index to avoid the measurement problem of price and inflation encountered in the empirical work in China. Second, the assumption of exogeneity of monetary aggregates, Mo or M3,in a reforming centrally planned economy is important in the quantity theory analysis Considering the importance of the state enterprise borrowing requirement, the nonstate enterprise borrowing requirement, and the public sector borrowing requirement in the money supply process, most studies recognize explicitly the endogeneity of money with respect to income(Chen, 1989; Gong, 1986; Li and Leung, 1994; Portes and Santorum, 1987). In contrast, Blejer et al.(1991) recommended the use of broader monetary aggregates as more suitable interme- diate targets of monetary policy. In a recent study, Hasan and Taghavi(1996) found a feedback relationship between the narrow money stock(Mo) and real income while the broad money stock(M,) is statistically exogenous in the information set that contains the money stock, price, real income, and the real interest rate. However, if we consider the fact that the cash and credit plans are assembled principally by the Peoples Bank of China(PBC)and the State Planning Commission(SPC), and the original versions of both or any modifi cations during the year have to be approved by the State Council, control of the money supply by the central authority enhances its exogenous characteristics Indeed, the central authorities in China set monetary growth to maintain price money stock, M,, as a proxy variable of m prefer to use a measure c S Feltenstein and Ha( 1991) proposed constructing a true price index on the basis of the equations logP=logP+alog(M/PR)0≤a≤1 m5-m5-1=B(mr-m-1)0≤B≤1 where Pr is the true price, P is the official retail price, R is the real volume of consumer retail sales, and M2 is the stock of currency plus household bank deposits. The true rate of infation is given by T,- log Pr-- where Ti denotes the expected rate of infiation in the true price index. a and B are parameters of repressed inflation and expectation adjustment, respectively. The first equation relates the true price index to the official price level and an index of monetary overhang Feltenstein and Ha(1991) hypothesized that if a= 0, the true and official rates of inflation would be equal. On the other hand, a= I implies that a 10% increase in the ratio of money to retail sales will cause the true rate of inflation to be 10%. The second equation hypothesized that inflationary expectations follow an adaptive pattern in which the change in the expected value of the true rate of inflation is proportional to the deviation between actual and expected inflation in the last period Feltenstein and Ha(1991)suggest a simultaneous search process to find the optimal values of a and B using a log-likelihood criterion that maximizes the log-likelihood function. Li and Leung(1994) found the values of a and B to be 0.49 and 0.61, respectively. Using these parameter values and yearly data over the period from 1952 to 1989, they generated a true price index b Under the traditional central planning model, the amount of currency in circulation, in the economy, Mo, was viewed as the principal determinant of inflation(Chow, 1987; Portes and
Ha, 1991). However, Feltenstein and Ha (1991) develop a measure of the true price index in China from a money demand function and this performed better in a simulation exercise. Li and Leung (1994) have updated the annual data based on this approach.5 We use their measure of the price index to avoid the measurement problem of price and inflation encountered in the empirical work in China. Second, the assumption of exogeneity of monetary aggregates, M0 or M3, in a reforming centrally planned economy is important in the quantity theory analysis. Considering the importance of the state enterprise borrowing requirement, the nonstate enterprise borrowing requirement, and the public sector borrowing requirement in the money supply process, most studies recognize explicitly the endogeneity of money with respect to income (Chen, 1989; Gong, 1986; Li and Leung, 1994; Portes and Santorum, 1987). In contrast, Blejer et al. (1991) recommended the use of broader monetary aggregates as more suitable intermediate targets of monetary policy. In a recent study, Hasan and Taghavi (1996) found a feedback relationship between the narrow money stock (M0) and real income while the broad money stock (M3) is statistically exogenous in the information set that contains the money stock, price, real income, and the real interest rate. However, if we consider the fact that the cash and credit plans are assembled principally by the People’s Bank of China (PBC) and the State Planning Commission (SPC), and the original versions of both or any modifi- cations during the year have to be approved by the State Council, control of the money supply by the central authority enhances its exogenous characteristics. Indeed, the central authorities in China set monetary growth to maintain price stability (Chen, 1989). Given these facts, we prefer to use a measure of the broad money stock, M3, as a proxy variable of monetary aggregates.6 5 Feltenstein and Ha (1991) proposed constructing a true price index on the basis of the equations logPT 5 logP 1 alog~M2/PR! 0 # a # 1 p T e 2 p T21 e 5 b~pT 2 p T21 e ! 0 # b # 1, where PT is the true price, P is the official retail price, R is the real volume of consumer retail sales, and M2 is the stock of currency plus household bank deposits. The true rate of inflation is given by pt 5 log PT 2 logPT21 where pT e denotes the expected rate of inflation in the true price index. a and b are parameters of repressed inflation and expectation adjustment, respectively. The first equation relates the true price index to the official price level and an index of monetary overhang. Feltenstein and Ha (1991) hypothesized that if a 5 0, the true and official rates of inflation would be equal. On the other hand, a 5 1 implies that a 10% increase in the ratio of money to retail sales will cause the true rate of inflation to be 10%. The second equation hypothesized that inflationary expectations follow an adaptive pattern in which the change in the expected value of the true rate of inflation is proportional to the deviation between actual and expected inflation in the last period. Feltenstein and Ha (1991) suggest a simultaneous search process to find the optimal values of a and b using a log-likelihood criterion that maximizes the log-likelihood function. Li and Leung (1994) found the values of a and b to be 0.49 and 0.61, respectively. Using these parameter values and yearly data over the period from 1952 to 1989, they generated a true price index. 6 Under the traditional central planning model, the amount of currency in circulation, in the economy, M0, was viewed as the principal determinant of inflation (Chow, 1987; Portes and MONEY AND INFLATION IN CHINA 673
MOHAMMAD S HASAN The third issue is the assumption of constant and stable income velocity. Chen and Hou(1986) pointed out that income velocity is generally quite stable in China compared with income velocity in market economies, because of the role of the state banking system in handling the receipts and expenditures of industrial and commercial enterprises. However, the Chinese banking authorities indicated that income velocity, V, declined during the postreform period from 1980 to 1982 Chen and Hou, 1986). Chow(1987) found that the elasticity of the price level with respect to the ratio of money stock to real output, Mo/Y, is smaller than unity, which appears to suggest that velocity is not constant. Although the declining trend in income velocity, PY/M, in the 1980s certainly reduces the significance of the traditional quantity theory prediction that monetary growth is the sole cause of inflation, this trend does not necessarily eliminate the validity of the quantity theory in the long run. As long as a decline in income velocity is will be less than proportional to a change in money stoo oa not completely offset by an increase in money stock, change in the price level Finally, we do not claim that Chinese inflation can be explained in a classical paradigm, but rather that the empirical techniques we describe here will help to identify and estimate how much of the movement in price can be explained by monetary forces. Our general model is more acceptable and subject to less criticism than a pure quantity theory model 3. EMPIRICAL RESULTS To estimate the general inflation model, we identify six variables. They broad money (M,), the true price level (P), an output gap(g), wages(h agricultural productivity(AP), and a measure of industrial productivity (IP) Descriptions and sources for the variables and data used for estimation are given in the data appendix. The sample spanned the period from 1952 to 1993. Figure I plots the logarithms of M and P. A causal glance at the graphs clearly indicate that both series are trending together in the long run Santorum, 1987). Blejer et al. (1991), however, noted that currency demand has an unusual patte nges in expected inflation. The uncertainty surrounding the precise timing of the phases of the response pattern and difficulties in observing changes in expected inflation restrict the usefulness of currency as an intermediate target variable. Moreover, using M3 has the added advantage that the overall credit plan of the monetary authority is incorporated This study uses 42 annual observations for the following reasons. First, the impact and adjustment lags of various macroeconomic relations, such as money income and the money-price relationship, are too long for monthly quarterly observations to reflect the actual correlation between these macroeconomic variables. Although annual observations yield smaller degrees of freedom, the noisy effects associated with monthly or even quarterly observations tend to average out with annual data which better approximate the money-income or money-price relationship(Masih and Masih, 1995; Spencer, 1989). Second, Hakkio and Rush(1991)and Berg and Jayaneti(1993) contend that cointegration is a long run concept and, hence, requires a long span of data to give it much explanatory power. Finally, Chinese national income data are available only annually
The third issue is the assumption of constant and stable income velocity. Chen and Hou (1986) pointed out that income velocity is generally quite stable in China compared with income velocity in market economies, because of the role of the state banking system in handling the receipts and expenditures of industrial and commercial enterprises. However, the Chinese banking authorities indicated that income velocity, V, declined during the postreform period from 1980 to 1982 (Chen and Hou, 1986). Chow (1987) found that the elasticity of the price level with respect to the ratio of money stock to real output, M0/Y, is smaller than unity, which appears to suggest that velocity is not constant. Although the declining trend in income velocity, PY/M, in the 1980’s certainly reduces the significance of the traditional quantity theory prediction that monetary growth is the sole cause of inflation, this trend does not necessarily eliminate the validity of the quantity theory in the long run. As long as a decline in income velocity is not completely offset by an increase in money stock, change in the price level will be less than proportional to a change in money stock. Finally, we do not claim that Chinese inflation can be explained in a classical paradigm, but rather that the empirical techniques we describe here will help to identify and estimate how much of the movement in price can be explained by monetary forces. Our general model is more acceptable and subject to less criticism than a pure quantity theory model. 3. EMPIRICAL RESULTS To estimate the general inflation model, we identify six variables. They are broad money (M3), the true price level (P), an output gap ( g), wages (W), agricultural productivity (AP), and a measure of industrial productivity (IP). Descriptions and sources for the variables and data used for estimation are given in the data appendix. The sample spanned the period from 1952 to 1993.7 Figure 1 plots the logarithms of M and P. A causal glance at the graphs clearly indicates that both series are trending together in the long run. Santorum, 1987). Blejer et al. (1991), however, noted that currency demand has an unusual pattern of responses to changes in expected inflation. The uncertainty surrounding the precise timing of the phases of the response pattern and difficulties in observing changes in expected inflation restrict the usefulness of currency as an intermediate target variable. Moreover, using M3 has the added advantage that the overall credit plan of the monetary authority is incorporated. 7 This study uses 42 annual observations for the following reasons. First, the impact and adjustment lags of various macroeconomic relations, such as money income and the money–price relationship, are too long for monthly or even quarterly observations to reflect the actual correlation between these macroeconomic variables. Although annual observations yield smaller degrees of freedom, the noisy effects associated with monthly or even quarterly observations tend to average out with annual data which better approximate the money–income or money–price relationship (Masih and Masih, 1995; Spencer, 1989). Second, Hakkio and Rush (1991) and Berg and Jayaneti (1993) contend that cointegration is a long run concept and, hence, requires a long span of data to give it much explanatory power. Finally, Chinese national income data are available only annually. 674 MOHAMMAD S. HASAN
MONEY AND INFLATION N CHINA 7.0 12 65 6.0 5.0 6 4.5 556065707580859 FIG. 1 Log level of money stock(-)and prices(---). As a first step, the data were checked for stationarity using the Augmented Dickey-Fuller(ADF) test in each of the six variables. The number of augmentation terms in the ADF regression was determined by examining the significance of the final lag, up to three, and the serial correlation of residuals. The results of the ADF tests, as reported in Table 1, indicate that each of the six variables is nonstationary in level but not in first difference form. In the next step, the data series are further checked for the presence of cointegration using Hansen's(1990) procedure to see whether stochastic trends of these variables move together in the long run. Given the relatively small sample size, we apply the single equation estimation method. Under the single equation method, although statistical inference on the parameter of the integrating vector is possible, readers are urged to exercise some caution Hansen(1990)has noted that the power of the Engle-Granger(1987)tests falls substantially as the size of the system increases. Hansen(1990)has suggested a simple test of cointegration by applying a Cochrane-Orcutt procedure to correct for first-order serial correlation in the residuals of t cointegration equation. The procedure is modified to take into account possible structural break in the postreform period; e.g., Huang(1995)iden- tifies 1979 as the turning point of economic liberalization The results of the cointegration regression as well as the cointegration tests are reported in Table 2. The Dickey-Fuller t-statistics for both price and money
As a first step, the data were checked for stationarity using the Augmented Dickey–Fuller (ADF) test in each of the six variables. The number of augmentation terms in the ADF regression was determined by examining the significance of the final lag, up to three, and the serial correlation of residuals. The results of the ADF tests, as reported in Table 1, indicate that each of the six variables is nonstationary in level but not in first difference form. In the next step, the data series are further checked for the presence of cointegration using Hansen’s (1990) procedure to see whether stochastic trends of these variables move together in the long run. Given the relatively small sample size, we apply the single equation estimation method. Under the single equation method, although statistical inference on the parameter of the cointegrating vector is possible, readers are urged to exercise some caution. Hansen (1990) has noted that the power of the Engle–Granger (1987) tests falls substantially as the size of the system increases. Hansen (1990) has suggested a simple test of cointegration by applying a Cochrane–Orcutt procedure to correct for first-order serial correlation in the residuals of the cointegration equation. The procedure is modified to take into account a possible structural break in the postreform period; e.g., Huang (1995) identifies 1979 as the turning point of economic liberalization. The results of the cointegration regression as well as the cointegration tests are reported in Table 2. The Dickey–Fuller t-statistics for both price and money FIG. 1. Log level of money stock (—) and prices (- - -). MONEY AND INFLATION IN CHINA 675
MOHAMMAD S HASAN TABLE 1 Unit Root Test Results A. Stationarity Variable First difference PM 0.162(-3.52) 4.16(-2.93) -473(-3.52) 998(-2.9 0.395(-3.52 5.13(-2.93) 527(-3.52) 2.88(-2.93) 0.940(-3.52) 442(-2.93) -4.32(-3.53) -2.53(-2.93) 247(-3.52 -5.15(-2.93) 524(-3.53) 925(-2.93) 2.82(-3.52) 3.83(-2.93) 461(-3.53) 0.458(-2.93) -6.46(-2.93) 6.41(-3.52) I, and I- are the t-statistics based on augmented Dickey-Fuller(ADF)regression with allowance for a constant and trend, respectively. Figures in parentheses are McKinnons(1991) critical value equations reject the null hypothesis of noncointegration. The estimated coeffi cients are plausible in static multivariate regressions. The results suggest that, in most cases, variables are statistically significant and have the signs predicted by the theory Given the fact that a residual-based testing procedure, such as Engle-Granger ( 1987), is alleged to have a low power to detect an otherwise dormant long-run relationship, we proceed to test for the presence of cointegrating vectors using the Johansen and Juselius (JJ)maximum likelihood procedure. For further details of the JJ method, see Johansen(1988), Johansen and Juselius(1990). Since annual data loyed, a maximum lag length of two is used in the Johansen vector autoregressive(VAR) model. Results of Johansen's maximum eigenvalue and trace tests are presented in Table 3. Cheung and Lai(1993) have argued that Johansen's likelihood ratio(Lr)tests are derived from asymptotic results and standard inferences in finite samples may not be appropriate Johansens LR tests re biased toward finding cointegration too often in finite samples when asymp- totic critical values are used. The finite sample bias of Johansens test is a positive function of T/(T-nk), where T, n, and k signify the sample size, the number of variables in the estimated system and the lag length, respectively Reimers(1991), and Reinsel and Ahn(1992) have suggested adjusting Jo- hansens test statistics by a scaling factor of (T-nk)/T and comparing them with their asymptotic critical values. Following Reinsel and Ahn(1992), th computed test statistics were adjusted using the scaling factor. We have used the critical values furnished by Charemza and Deadman (1992). Following Cheung and Lai(1993), the critical values were also adjusted by multiplying by the degree-of-freedom correction factor, T/(T-nk), and then comparing with the asymptotic test statistics The results turn out to be similar
equations reject the null hypothesis of noncointegration.8 The estimated coeffi- cients are plausible in static multivariate regressions. The results suggest that, in most cases, variables are statistically significant and have the signs predicted by the theory. Given the fact that a residual-based testing procedure, such as Engle–Granger (1987), is alleged to have a low power to detect an otherwise dormant long-run relationship, we proceed to test for the presence of cointegrating vectors using the Johansen and Juselius (JJ) maximum likelihood procedure. For further details of the JJ method, see Johansen (1988), Johansen and Juselius (1990). Since annual data are employed, a maximum lag length of two is used in the Johansen vector autoregressive (VAR) model. Results of Johansen’s maximum eigenvalue and trace tests are presented in Table 3. Cheung and Lai (1993) have argued that Johansen’s likelihood ratio (LR) tests are derived from asymptotic results and standard inferences in finite samples may not be appropriate. Johansen’s LR tests are biased toward finding cointegration too often in finite samples when asymptotic critical values are used. The finite sample bias of Johansen’s test is a positive function of T/(T 2 nk), where T, n, and k signify the sample size, the number of variables in the estimated system and the lag length, respectively. Reimers (1991), and Reinsel and Ahn (1992) have suggested adjusting Johansen’s test statistics by a scaling factor of (T 2 nk)/T and comparing them with their asymptotic critical values. Following Reinsel and Ahn (1992), the computed test statistics were adjusted using the scaling factor.9 8 We have used the critical values furnished by Charemza and Deadman (1992). 9 Following Cheung and Lai (1993), the critical values were also adjusted by multiplying by the degree-of-freedom correction factor, T/(T-nk), and then comparing with the asymptotic test statistics. The results turn out to be similar. TABLE 1 Unit Root Test Results A. Stationarity testa Variable Level First difference tm tp tm tp P 1.79 (22.93) 0.162 (23.52) 24.16 (22.93) 24.73 (23.52) M 0.998 (22.93) 20.395 (23.52) 25.13 (22.93) 25.27 (23.52) W 2.88 (22.93) 0.940 (23.52) 24.42 (22.93) 24.32 (23.53) g 22.53 (22.93) 22.47 (23.52) 25.15 (22.93) 25.24 (23.53) AP 0.925 (22.93) 22.82 (23.52) 23.83 (22.93) 24.61 (23.53) IP 0.458 (22.93) 21.15 (23.52) 26.46 (22.93) 26.41 (23.52) a tm and t p are the t-statistics based on augmented Dickey–Fuller (ADF) regression with allowance for a constant and trend, respectively. Figures in parentheses are McKinnon’s (1991) critical value. 676 MOHAMMAD S. HASAN
MONEY AND INFLATION IN CHINA TABLE 2 Multivariate Cointegration Regression Rest Coefficient on lag of P M Constant 0433(0.611) 0.363(7.53) 0.043(0.159) 0.345°(-3.84 0011(0.0752) 0.997 -4.91 ADF(I) 5.51 ADF(2) Notes. Figures in the parentheses undemeath the coefficients are the i-ratios, Dw and Se are the Durbin-Watson statistic and the standard error of the regression, respectively. The regressions als include a time trend and an intercept dummy for 1979. Significant at the 1% level b Significant at the 5% level Results of Johansens maximum eigenvalue test indicate that there exist at least one cointegrating relationship in the six-variate VAR system, since r=0 clearly rejected in favor of r= I by the 95% critical value. The trace test, on the other hand, suggests that there are two cointegrating vectors in the model since r s I is clearly rejected in favor of r=2; but r s 2 cannot be rejected by the 95% critical values. Since there is growing evidence in favor of the robustness of the trace statistic compared to the maximal eigenvalue statistic Cheung and Lai, 1993; Kasa, 1992, Luintel and Paudyal, 1998), we accept the race test results that tend to suggest there are at least two stationary relationships between price, money stock, wage, output gap, agricultural and industrial pro- ductivity variables. Given that there are (n-r) common trends within the system, we can conclude that there exist at most four common trends within the vector such that current price is moving into alignment with the trend values of explanatory variables in Eq (4) The finding of cointegration among these macroeconomic variables eral implications. First, consistent with economic theory, it indicates that money, wages, and prices have a long-run equilibrium relationship that may be exploited by the monetary authorities in the formulation of monetary policy. Second, the
Results of Johansen’s maximum eigenvalue test indicate that there exist at least one cointegrating relationship in the six-variate VAR system, since r 5 0 is clearly rejected in favor of r 5 1 by the 95% critical value. The trace test, on the other hand, suggests that there are two cointegrating vectors in the model since r # 1 is clearly rejected in favor of r 5 2; but r # 2 cannot be rejected by the 95% critical values. Since there is growing evidence in favor of the robustness of the trace statistic compared to the maximal eigenvalue statistic (Cheung and Lai, 1993; Kasa, 1992; Luintel and Paudyal, 1998), we accept the trace test results that tend to suggest there are at least two stationary relationships between price, money stock, wage, output gap, agricultural and industrial productivity variables. Given that there are (n 2 r) common trends within the system, we can conclude that there exist at most four common trends within the vector such that current price is moving into alignment with the trend values of explanatory variables in Eq. (4). The finding of cointegration among these macroeconomic variables has several implications. First, consistent with economic theory, it indicates that money, wages, and prices have a long-run equilibrium relationship that may be exploited by the monetary authorities in the formulation of monetary policy. Second, the TABLE 2 Multivariate Cointegration Regression Results Coefficient on lag of Dependent variable P M Constant 0.433 (0.611) 3.78b (2.38) P 1.38a (6.43) M 0.363a (7.53) W 0.527a (5.85) 0.043 (0.159) g 20.345a (23.84) 1.23a (6.63) AP 0.011 (0.0752) 20.653b (22.16) IP 20.133b (22.00) 20.254 (21.61) R2 0.995 0.997 DW 1.59 1.76 SE 0.038 0.076 Cointegration test DF 24.91b 26.20a ADF(1) 25.51a 25.06b ADF(2) 25.28b 23.96 Notes. Figures in the parentheses underneath the coefficients are the t-ratios; DW and SE are the Durbin–Watson statistic and the standard error of the regression, respectively. The regressions also include a time trend and an intercept dummy for 1979. a Significant at the 1% level. b Significant at the 5% level. MONEY AND INFLATION IN CHINA 677
MOHAMMAD S HASAN TABLE 3 Johansen's Maximum Likelihood Procedure for Cointegration Tests Vector LR statistics 95%critical values H H r=1 45.31 40.53 102.56 30.28 82.30 75.98 22.90 52.02 53.48 r三3 12.79 29.11 r≤4 11.65 16.32 20.18 r≤5 4.66 Note. Columns I and 2 are null and alternative hypotheses to test the value of r, of cointegration relations. Amas and Auase are two alternative test statistics based on m value and trace of the stochastic matrix, respectively. *denotes rejection of the null hy evidence of cointegration also rules out the possibility of spurious correlations and granger noncausality between price and the explanatory variables in Eq (4 Third, the dynamic modeling of the inflationary process has a valid error correction representation with a cointegrating constraint embedded in it. In sum, the above results suggest that, contrary to Huang(1995), a stable long-run linear relationship between price and the money stock does exist during both the prereform and postreform period in the mainland China. Following economic reform, as the economy is increasingly exposed to market forces, the nature and strength of relationships between the price level and money stock, and between the price level and money income ratio have been increased substan- 4. TEST RESULTS FOR GRANGER CAUSALITY Following the Granger representation theorem, the above unit root and cointe gration test results also imply that inflation and monetary growth can be given a dynamic specification using the error-correction models +∑83△logW+∑,△g-+∑8AP-
evidence of cointegration also rules out the possibility of spurious correlations and Granger noncausality between price and the explanatory variables in Eq. (4). Third, the dynamic modeling of the inflationary process has a valid errorcorrection representation with a cointegrating constraint embedded in it. In sum, the above results suggest that, contrary to Huang (1995), a stable long-run linear relationship between price and the money stock does exist during both the prereform and postreform period in the mainland China. Following economic reform, as the economy is increasingly exposed to market forces, the nature and strength of relationships between the price level and money stock, and between the price level and money income ratio have been increased substantially. 4. TEST RESULTS FOR GRANGER CAUSALITY Following the Granger representation theorem, the above unit root and cointegration test results also imply that inflation and monetary growth can be given a dynamic specification using the error-correction models: Dlog Pt 5 d0 1 w1jt21 1 O s51 n1 d1sDlog Pt2s 1 O s51 n2 d2sDlog Mt2s 1 O s51 n3 d3sDlog Wt2s 1 O s51 n4 d4sDgt2s 1 O s51 n5 d5sAPt2s 1 O s51 n6 d6sDIPt2s 1 n1t (5) TABLE 3 Johansen’s Maximum Likelihood Procedure for Cointegration Tests Vector LR statistics 95% critical values H0: H1: lmax ltrace lmax ltrace r 5 0 r 5 1 45.31* 127.62* 40.53 102.56 r # 1 r 5 2 30.28 82.30* 34.40 75.98 r # 2 r 5 3 22.90 52.02 28.27 53.48 r # 3 r 5 4 12.79 29.11 22.04 34.87 r # 4 r 5 5 11.65 16.32 15.87 20.18 r # 5 r 5 6 4.66 4.66 9.16 9.16 Note. Columns 1 and 2 are null and alternative hypotheses to test the value of r, i.e., the number of cointegration relations. lmax and ltrace are two alternative test statistics based on maximal eigenvalue and trace of the stochastic matrix, respectively. * denotes rejection of the null hypothesis at the 5% significance levels. 678 MOHAMMAD S. HASAN
