the sovereign yield curve and the macroeconomy in...

28
THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA YIFENG YAN* Xi’an Jiaotong University JUE GUO Xi’an Jiaotong University Abstract. A dynamic Nelson–Siegel model is adopted to estimate three time-varying factors of yield curves, the level, the slope and the curvature, and a vector autoregressive model is built to study interactions between macro variables and the yield curve. Results show that, first, money supply growth is a more effective instrument to curb inflation than the monetary policy interest rate; however, the central bank also adjusts the interest rate to stabilize money supply. Second, invest- ment is an important measure to stimulate the Chinese economy, but it also pushes up money supply growth, which results in higher inflation. Third, the yield curve reacts significantly to innovations to investment growth and money supply growth. The segmentation of China’s bond market hinders the efficient implementation of monetary policy, and the monetary policy transmission mechanism is still weak in China. Finally, interactions between the yield curve and the macroeconomy in China are nearly unidirectional. Macroeconomic variables reshape the yield curve, but direct adjustments of the yield curve do not significantly change macroeconomic variables. Due to the incomplete liber- alization of financial markets, there exists a wide disjunction between the real economy and financial markets in China. JEL: E43, E44, E51, E52, E62, G12 1. INTRODUCTION In 1981, China resumed Treasury bond issuance, and, according to the Bank for International Settlements, by July 2012 China’s bond market had become the world’s third largest. On June 1997, the People’s Bank of China (PBC) instructed commercial banks to leave the exchange bond market to prevent credit funds from speculating in the stock market. Since then, China’s bond market has been segmented into the exchange bond market and the interbank bond market. The interbank bond market is open to institutional investors, including, for example, commercial banks, security companies and insurance companies. The exchange bond market is open to individuals and non-bank financial institutions. In addition, listed commercial banks have been allowed to trade in the exchange bond market since December 2010. The bond market in China remains segmented because regulation, the bond custodian system and clearing in these two markets are separate from each other. The interbank bond market is supervised by the PBC, and the exchange bond market is supervised by *Address for Correspondence: School of Management, Xi’an Jiaotong University, No. 28, Xianning West Road, Xi’an 710049, China. E-mail: [email protected]. We are very grateful to the editor and two anonymous referees for helpful comments and suggestions. We also thank the National Natural Science Foundation of China (No. 71173169) for financial support. Pacific Economic Review, 20: 3 (2015) pp. 415–441 doi: 10.1111/1468-0106.12063 © 2015 Wiley Publishing Asia Pty Ltd

Upload: others

Post on 22-Oct-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

  • THE SOVEREIGN YIELD CURVE AND THEMACROECONOMY IN CHINA

    YIFENG YAN* Xi’an Jiaotong UniversityJU’E GUO Xi’an Jiaotong University

    Abstract. A dynamic Nelson–Siegel model is adopted to estimate three time-varying factors of yieldcurves, the level, the slope and the curvature, and a vector autoregressive model is built to studyinteractions between macro variables and the yield curve. Results show that, first, money supplygrowth is a more effective instrument to curb inflation than the monetary policy interest rate;however, the central bank also adjusts the interest rate to stabilize money supply. Second, invest-ment is an important measure to stimulate the Chinese economy, but it also pushes up money supplygrowth, which results in higher inflation. Third, the yield curve reacts significantly to innovations toinvestment growth and money supply growth. The segmentation of China’s bond market hinders theefficient implementation of monetary policy, and the monetary policy transmission mechanism isstill weak in China. Finally, interactions between the yield curve and the macroeconomy in China arenearly unidirectional. Macroeconomic variables reshape the yield curve, but direct adjustments ofthe yield curve do not significantly change macroeconomic variables. Due to the incomplete liber-alization of financial markets, there exists a wide disjunction between the real economy and financialmarkets in China.

    JEL: E43, E44, E51, E52, E62, G12

    1. INTRODUCTION

    In 1981, China resumed Treasury bond issuance, and, according to the Bank forInternational Settlements, by July 2012 China’s bond market had become theworld’s third largest. On June 1997, the People’s Bank of China (PBC)instructed commercial banks to leave the exchange bond market to preventcredit funds from speculating in the stock market. Since then, China’s bondmarket has been segmented into the exchange bond market and the interbankbond market. The interbank bond market is open to institutional investors,including, for example, commercial banks, security companies and insurancecompanies. The exchange bond market is open to individuals and non-bankfinancial institutions. In addition, listed commercial banks have been allowed totrade in the exchange bond market since December 2010. The bond market inChina remains segmented because regulation, the bond custodian system andclearing in these two markets are separate from each other. The interbank bondmarket is supervised by the PBC, and the exchange bond market is supervised by

    *Address for Correspondence: School of Management, Xi’an Jiaotong University, No. 28, XianningWest Road, Xi’an 710049, China. E-mail: [email protected]. We are very grateful to theeditor and two anonymous referees for helpful comments and suggestions. We also thank theNational Natural Science Foundation of China (No. 71173169) for financial support.

    bs_bs_banner

    Pacific Economic Review, 20: 3 (2015) pp. 415–441doi: 10.1111/1468-0106.12063

    © 2015 Wiley Publishing Asia Pty Ltd

  • the China Securities Regulatory Commission. The China Central Depositoryand Clearing Company undertakes the function of centralized depository andhandles settlement for the interbank bond market; the China Securities Deposi-tory and Clearing Corporation does the same for the exchange bond market.

    The major types of bonds available in China’s bond market include Treasurybonds, central bank bills, financial bonds, enterprise bonds and corporatebonds. Central bank bills are short-term securities issued in the interbank bondmarket by the PBC as a monetary policy instrument. Financial bonds are issuedby the policy banks, commercial banks and other financial institutions. Centralbank bills and financial bonds are the most actively traded bonds in China.Enterprise bonds are issued by large-sized state-owned companies, while corpo-rate bonds can be issued by any company. Enterprise bonds can be traded onboth bond markets, but corporate bonds are only traded on the exchangemarket. Enterprise bonds have a much larger outstanding amount and arecurrently more actively traded than corporate bonds. The interbank market ismuch larger than the exchange market, and commercial banks dominate tradingactivity in the interbank bond market.

    China’s central bank initially conducted its monetary policy by directly con-trolling the credit quotas of each commercial bank. In 1998, the PBC terminatedthe system of national bank credit quotas. Since then, the central bank hasmoved towards applying indirect ways to control money supply. The majormonetary policy instruments comprise open market operations, reserve require-ments, deposit and lending interest rates, refinancing and rediscount rates. Openmarket operations are implemented principally through repo transactions ofTreasury bonds as well as issuing and trading of central bank bills in theinterbank market. Central bank bills were introduced into China’s interbankmarket in 2003. They are the most liquid bonds because of the large amount ofthese bills outstanding in the market and the regular weekly issuances. Thedeposit and lending interest rates were under the control of the PBC before 2013.On July 2013, the PBC announced the removal of lending interest rate control,but the deposit interest rates have not been fully liberalized yet. China’s centralbank also conducts monetary policy through window guidance and a standinglending facility, which was established on November 2013.

    Generally, China’s financial markets are strictly regulated, and Chinese mon-etary policy is mainly conducted through administrative and quantitative meas-ures (Koivu, 2009). He and Wang (2012) find that Chinese market interest ratesare sensitive to changes in the benchmark deposit interest rates and changes inthe reserve requirements, but not particularly reactive to open market opera-tions. Market-based monetary policy instruments of the PBC, such as the reporate and the benchmark lending rate, have little impact on the Chinese economy,and non-market-based measures, such as growth rates of total loan and moneysupply, are effective in adjusting the real economy and the price level (see e.g.Qin et al., 2005; Siklos and Zhang, 2010; He et al., 2013).

    Despite the inefficiency of market-based monetary policy, Porter and Cassola(2011) find evidence of an emerging transmission channel, and that severalnecessary elements to move towards the application of indirect monetary policy

    Y. YAN AND J. GUO416

    © 2015 Wiley Publishing Asia Pty Ltd

  • are already in place with further liberalization of Chinese financial markets.Koivu (2009) also points out that Chinese loan demand has become moredependent on interest rates since 2001. Although bond yields are not fullyefficient because of regulation, liquidity and segmentation, Chinese bond yieldscontain considerable information about the state of the economy and act as anintermediary in monetary policy transmission: changes in PBC bill rates (aproxy monetary policy variable) influence the structure of Treasury, financialand corporate bond yield curves, which are then associated with changes inoutput and inflation (Porter and Cassola, 2011).

    The existing papers studying the transmission mechanism of Chinese mon-etary policy mainly focus on the efficiency and improvement of Chinese mon-etary policy implementation as well as the comparison of monetary policyinstruments. The functioning and effectiveness of bond markets, especially theirinteractions with monetary policy and the real economy, have not been inves-tigated extensively in the existing studies, with the exception of Porter andCassola (2011) and Fan and Johansson (2010). Fan and Johansson (2010) modelChinese bond yields using the 1-year deposit interest rate as a state variable in anaffine framework. Their results suggest that a macro-finance approach for ana-lysing bond yields also works in a Chinese institutional setting that differssignificantly from that of the USA and Europe. Their main effort is to incorpo-rate monetary policy into a model of China’s yield curve, and they suggest futureresearch should include alternative economic variables such as inflation andother measures of monetary policy as explanatory variables when analysingChina’s bond market. Porter and Cassola (2011) proceed in this direction. Theyincorporate macroeconomic variables (industrial value added growth and con-sumer price index inflation) as well as the 3-month PBC bill rate in their analysis,and study the interaction between the policy bank bonds market and the realeconomy.

    The present paper aims to extend and deepen our understanding of howmacro variables as well as market-based and non-market-based measures ofmonetary policy interact with Chinese bond yields. The paper provides newevidence from the Chinese economy on the controversy over the relationshipsbetween macro variables and yield curves. This study differentiates from existingstudies in the following aspects: first, considering that the sovereign yield curveprovides a fundamental benchmark in the economy, we study the interactionbetween the real economy and the Treasury bonds market instead of the policybank bonds market; second, we expand the scope of the real economy byincluding investment and consumption in addition to output and inflationbecause they are major drivers of the economy (see e.g. Mehrotra et al., 2013);third, because of the importance of quantitative monetary policy measures inChina, both money supply and monetary policy interest rates are studied in thepresent paper; finally, we incorporate fiscal policy and foreign trade variablesinto our study because they play important roles in China’s economy (see e.g.Chow, 2006; He et al., 2009).

    Methodologically, we follow the framework of Afonso and Martins (2012),who use a dynamic Nelson–Siegel model to estimate yield curves and a vector

    THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA 417

    © 2015 Wiley Publishing Asia Pty Ltd

  • autoregressive (VAR) system to investigate linkages between financial variablesand the real economy. The Nelson–Siegel method captures the level, slope andcurvature of the yield curve (Nelson and Siegel, 1987), and plays a very impor-tant role in studies of the term structure of interest rates (see e.g. Gurkaynakaet al., 2007). Diebold and Li (2006) and Diebold et al. (2006) further develop adynamic Nelson–Siegel model with time-varying level, slope and curvaturefactors. The parsimonious Nelson–Siegel model provides good forecastability ofthe yield curve (see e.g. Diebold and Li, 2006; Vicente and Tabak, 2008; Yu andSalyards, 2009), and follows Zellner’s (1992) ‘KISS’ (i.e. keep it sophisticatedlysimple) principle of forecasting. However, very few articles have used thedynamic Nelson–Siegel model to study Chinese bond markets. Luo et al. (2012)find that the dynamic Nelson–Siegel model fits and forecasts the term structureof Chinese Treasury yields very well, and demonstrate that time-varying factorsof the model may be interpreted as the level, slope and curvature of the yieldcurve. Porter and Cassola (2011) also adopt the dynamic Nelson–Siegel modelto study Chinese bond markets.

    A large subset of the literature attempts to identify and distinguish deter-minants or influencing factors of the level, slope and curvature of the yieldcurve. The level captures the long-maturity yield, the slope captures behav-iours of short-maturity yields and the curvature captures behaviours of mid-maturity yields (Porter and Cassola, 2011). Accordingly, the level is typicallyassociated with the medium-term or long-run nominal anchor, namely, thetarget or expectation for long-term inflation; the slope is associated withchanges in the short rate and the reaction of monetary policy to the cyclicalstate of an economy; the curvature reflects the difference between the spreadof intermediate and short maturities, and the spread of long and intermediatematurities (see e.g. Rudebusch and Wu, 2008; Bekaert et al., 2010; Afonso andMartins, 2012; Aguiar-Conraria et al., 2012). However, the empirical evidenceremains somewhat mixed. Besides inflation-related variables, the level is alsoinfluenced by, for instance, the aggregate supply shocks from the privatesector (Wu, 2003), shocks to the monetary policy interest rate (Sultan, 2005),technology shocks (Wu, 2006) and shocks to the marginal rate of substitutionbetween consumption and leisure (Evans and Marshall, 2007). In addition tomonetary policy shocks (see e.g. Wu, 2006; Farka and DaSilva, 2011), theslope can also be affected by variables such as output, employment (Lu andWu, 2009) and other business cycle indicators. The curvature captures themonetary policy stance of the central bank (see e.g. Dewachter and Lyrio,2006; Bekaert et al., 2010; Lengwiler and Lenz, 2010) and is impacted byunemployment (Huse, 2011). Moreover, fiscal behaviour, such as governmentdebt burden, plays a role in determining the level and slope of the yield curve(see e.g. Benigno and Missale, 2004; Afonso and Martins, 2012). In an inter-national context, Diebold et al. (2008) find that global yield factors exist andare economically important for Germany, Japan, the UK and the USA;the temporal decline in the global level factor reflects the reduction ofinflation and movements in the global slope factor reflect the global businesscycle.

    Y. YAN AND J. GUO418

    © 2015 Wiley Publishing Asia Pty Ltd

  • The yield curve also contains future information on, for instance, real activityand inflation, and could provide an even better prediction than professionalmacroeconomic forecasters (Rudebusch and Williams, 2009). Some studies findthat the yield spread and the slope can be used to predict recession and inflation(see e.g. Ivanova et al., 2000; Mehl, 2009). Favero et al. (2012) and Moench(2012) point out that the level, slope and curvature can all contribute in fore-casting output and inflation. The stability of the predictive power of the yieldcurve for output and inflation is supported by many studies (see e.g. Schich,2002; Estrella, 2005), although there is some conflicting evidence (see e.g. Anget al., 2006; Nobili, 2007).

    The samples in this study start in January 2002 and end in December 2012.Using the dynamic Nelson–Siegel model, we estimate three time-varying latentfactors (level, slope and curvature) of the yield curve under a state-space frame-work. Furthermore, we establish a VAR system of macro variables and the threelatent factors, and explore the relationships among variables through impulseresponse analysis.

    The following are the main findings and conclusions in this study. First,Chinese authorities conduct monetary and fiscal policies to boost economicgrowth and stabilize prices. To curb inflation, money supply growth is a moreeffective instrument than interest rates; however, the central bank also adjuststhe interest rate to stabilize money supply. Second, investment is an importantmeans of stimulating the Chinese economy, but it also pushes up money supplygrowth, which results in higher inflation. Third, the yield curve reacts signifi-cantly to innovations to investment growth and money supply growth, but theopen market operations through the monetary policy interest rate conducted inthe interbank market do not reshape the yield curve in the exchange market. Thesegmentation of China’s bond market hinders implementation of monetarypolicy, and the monetary policy transmission mechanism is still weak in China.Fourth, interactions between the yield curve and the macroeconomy in Chinaare nearly unidirectional rather than bidirectional. Macroeconomic variablesreshape the yield curve, but direct adjustments of the yield curve do not signifi-cantly affect macroeconomic variables. Due to the incomplete liberalization offinancial markets, there exists a wide disjunction between the real economy andfinancial markets in China.

    The remainder of this paper is organized as follows. Section 2 introducesthe methodology, consisting of the dynamic Nelson–Siegel model and VARspecifications. An empirical analysis is conducted in Section 3 and Section 4concludes.

    2. METHODOLOGY

    This paper follows the methodology of Afonso and Martins (2012), whichcomprises two steps. First, three latent factors (level, slope and curvature) ofyield curves are estimated under the framework of the dynamic Nelson–Siegelmodel proposed by Diebold et al. (2006). Second, a VAR model is used toexplore linkages between the sovereign yield curve and the macroeconomy.

    THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA 419

    © 2015 Wiley Publishing Asia Pty Ltd

  • Our choice of a parsimonious Nelson–Siegel model to estimate three latentfactors rather than an arbitrage-free model is inspired by Diebold and Li (2006,pp. 361–2) and Diebold et al. (2006, p. 333), who state that:

    It is not obvious that use of arbitrage-free models is necessary or desirable for pro-ducing good forecasts; an a priori restriction may be violated in the data due toilliquidity in thinly traded regions of the yield curve; if the no-arbitrage restriction doesindeed hold for the data, then it will at least approximately be captured by the dynamicNelson–Siegel model which is a flexible approximation to the data.

    Coroneo et al. (2011) provide empirical evidence that the Nelson–Siegelmodel performs as well as its no-arbitrage counterpart in forecasting.

    2.1. The yield curve latent factors

    To gather information from the cross-section of yields with various maturities atany point in time, principal component analysis or the Nelson–Siegel factormodel are widely used. In the spirit of Diebold et al. (2006), the Nelson–Siegelformulation imposes some economically-motivated restrictions, including posi-tive forward rates at all horizons and a discount factor that approaches zero asmaturity increases, and these restrictions may be necessary in the analysis of yieldcurve dynamics. The Nelson–Siegel representation of the yield curve parsimoni-ously estimates the yield curve at any point in time using the following equation:

    ye e

    eτ β βλτ

    βλτ

    λτ λτλτ( ) = + −⎛

    ⎝⎜⎞⎠⎟

    + − −⎛⎝⎜

    ⎞⎠⎟

    − −−

    1 2 31 1

    , (1)

    where y(τ) denotes a set of yields, τ denotes maturity, and β1, β2, β3 and λ areparameters.

    Diebold and Li (2006) extends the former Nelson–Siegel model into adynamic latent factor model, where β1, β2 and β3 are time-varying, and can beinterpreted as the level (L), slope (S) and curvature (C) of the yield curve,respectively, at any point in time. Thus, we obtain:

    y L Se

    Ce

    et t t tτ λτ λτ

    λτ λτλτ( ) = + −⎛

    ⎝⎜⎞⎠⎟

    + − −⎛⎝⎜

    ⎞⎠⎟

    − −−1 1 . (2)

    Loadings of Lt are equal to one for all maturities, and other two-factorloadings approach zero as maturities increase, so Lt is a long-term factor, and isclosely related to the long end of the yield curve; meanwhile, it may be inter-preted as the level of the yield curve as an increase in Lt increases all yieldsequally. Loadings of St start at one and monotonically decrease to zero asmaturities increase; therefore, St is a short-term factor, and it may be interpretedas the slope of the yield curve, which can be defined as the long-term yield minusthe short-term yield. An increase in St increases short yields more than longyields, thereby changing the slope of the yield curve. Loadings of Ct start at zero

    Y. YAN AND J. GUO420

    © 2015 Wiley Publishing Asia Pty Ltd

  • and increase to a maximum at the middle of the yield curve, then return to zeroas maturities increase, so Ct is a medium-term factor, and it may be interpretedas the curvature of the yield curve. In other words, an increase in Ct increasesmedium-term yields, but barely moves very short and very long yields, therebychanging the curvature of the yield curve.

    As in Diebold et al. (2006), if the stochastic process of Lt, St and Ct is assumedto be a VAR process of order one, then dynamics of a large set of yields withvarious maturities can be presented as a state-space model. The state transitionequation is:

    L

    S

    C

    a a a

    a a a

    a a a

    t L

    t S

    t C

    −−−

    ⎢⎢⎢

    ⎥⎥⎥

    =⎡

    ⎢⎢⎢

    μμμ

    11 12 13

    21 22 23

    31 32 33

    ⎤⎤

    ⎥⎥⎥

    −−−

    ⎢⎢⎢

    ⎥⎥⎥

    +( )( )( )

    ⎢⎢⎢

    L

    S

    C

    L

    S

    C

    t L

    t S

    t C

    t

    t

    t

    1

    1

    1

    μμμ

    ηηη

    ⎤⎤

    ⎥⎥⎥, (3)

    where t = 1, 2, . . . , T, μL, μS and μC denote means of the three factors, and ηt(L),ηt(S) and ηt(C) are innovations to the system.

    The measurement equation, which expresses a set of N yields as differentcombinations of the three latent factors, is given as:

    y

    y

    y

    e ee

    t

    t

    t N

    ττ

    τ

    λτ λτ

    λτ λτλτ

    11

    2

    1

    11 11 1

    ( )( )

    ( )

    ⎢⎢⎢⎢

    ⎥⎥⎥⎥

    =

    − − −− −

    11

    22

    2

    11 1

    11 1

    2 2

    − − −

    − − −

    − −−

    − −−

    e ee

    e ee

    N N

    N N

    λτ λτλτ

    λτ λτλτ

    λτ λτ

    λτ λτ

    � � �

    NN

    L

    S

    C

    t

    t

    t

    t

    t

    t N

    ⎢⎢⎢⎢⎢⎢⎢⎢⎢

    ⎥⎥⎥⎥⎥⎥⎥⎥⎥

    ⎢⎢⎢

    ⎥⎥⎥

    +

    ( )( )

    ε τε τ

    ε τ

    1

    2

    �(( )

    ⎢⎢⎢⎢

    ⎥⎥⎥⎥

    (4)

    where t = 1, 2, . . . , T and εt(τ1), εt(τ2), . . . , εt(τN) are measurement errors.In matrix notation, the former transition equation and measurement equation

    comprise the following state-space system:

    f A ft t t− −( ) = ( ) +−m m h1 (5)

    y ft t= +L et , (6)

    where ft = (Lt, St, Ct)′ is a 3-dimensional state vector, yt = (yt(τ1), yt(τ2), . . . ,yt(τN))′ is an N-dimensional observation vector, A and Λ are 3 × 3 and N × 3coefficient matrices, and {ηt} and {εt} are 3-dimensional and N-dimensionalwhite noise series.

    For the Kalman filter to be linearly least-squares optimal, it is assumed thatthe white noise transition and measurement disturbances are orthogonal to eachother and uncorrelated with the state vector:

    he

    t

    t

    QH

    ⎡⎣⎢

    ⎤⎦⎥

    ∼ ⎡⎣⎢

    ⎤⎦⎥

    ⎡⎣⎢

    ⎤⎦⎥

    ⎛⎝⎜

    ⎞⎠⎟

    WN0

    0

    0

    0, (7)

    THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA 421

    © 2015 Wiley Publishing Asia Pty Ltd

  • E ft t′( ) =h 0 (8)

    E ft t′( ) =e 0. (9)

    It is further assumed that the variance–covariance matrix H of innovations tothe measurement equation is diagonal and the variance–covariance matrix Q ofinnovations to the transition equation is non-diagonal. The assumption ofdiagonal H implies that deviations of observed yields from the fitted yield curveare uncorrelated with each other, and makes estimation more tractable given alarge dimension of an observation vector. Meanwhile, the assumption of non-diagonal Q allows the innovations to the three latent factors to be correlated.

    The former state-space system can be estimated using the Kalman filter.Given initial values of the parameters, including μ, A, λ, Q and H, the Kalmanfilter calculates one-step-ahead forecast errors and variances of the forecasterrors from t = 2 through to t = T. A log-likelihood function is computed usingthese forecast errors and variances. Then, a standard numerical method, such asthe Berndt–Hall–Hall–Hausman (BHHH), algorithm is applied to maximize thelog-likelihood function by a convergence criterion (e.g. the change in the log-likelihood function is not larger than 10−1 from one iteration to the next.), andthe maximum likelihood estimates of parameters can be obtained after conver-gence. Finally, the Kalman smoother is used to compute the three latent factorsbased on complete data set information from t = T through to t = 2. (Harvey,1990 and Durbin and Koopman, 2012 provide comprehensive introductions tothe state-space model and the Kalman filter.)

    2.2. Vector autoregressive setting

    A VAR model is used to investigate the relationship between the yield curve andthe macroeconomy. As in previous studies (see e.g. Chen, 2009; Huse, 2011;Favero et al., 2012), besides the three latent factors, that is, level (L), slope (S)and curvature (C), we also incorporate the following variables in the VARsystem: inflation (pi), output growth (p), consumption growth (c), investmentgrowth (i), foreign trade growth (e), government expenditure growth (f), mon-etary policy interest rate (r) and monetary aggregate growth (m).

    A VAR(p) model is in the following form:

    X c V Xt i t t= + +−=∑ ii

    p

    1

    e , (10)

    where Xt = (pit, pt, ct, it, et, ft, rt, mt, Lt, St, Ct)′ is an 11-dimensional vector, c is an11-dimensional vector of intercepts, Vi is an 11 × 11 coefficient matrix and εt isan 11-dimensional vector of disturbances.

    The Cholesky decomposition of the variance–covariance matrix of innova-tions in an impulse response analysis implies that the innovations to variables inthe front rows of Xt have contemporaneous impacts on variables in the back

    Y. YAN AND J. GUO422

    © 2015 Wiley Publishing Asia Pty Ltd

  • rows of Xt, but the innovations to variables in the back rows do not havecontemporaneous impacts on variables in the front rows, so variables should beordered from the most exogenous to the least exogenous. As in Afonso andMartins (2012), it is assumed that financial variables may be instantaneouslyaffected by shocks to macroeconomic variables but the latter are not affectedcontemporaneously by the former; therefore, the monetary policy interest rate,monetary aggregate growth and the three latent factors are placed in the last fivepositions. Considering that the the monetary policy interest rate as an instru-ment instantaneously affects monetary aggregate growth and that the monetarypolicy interest rate does not instantaneously impact macroeconomic variablesbecause of monetary policy lags, we place the monetary policy interest ratebefore monetary aggregate growth and after macroeconomic variables. More-over, we assume that shocks to macroeconomic variables may instantaneouslyimpact fiscal variables because of the automatic stabilizing function of fiscalpolicy, but that a shock to fiscal variables does not have any immediate macro-economic effect because of fiscal policy lags, so we place government expendi-ture growth in the position immediately before the monetary policy interest rate.Finally, we assume that shocks to inflation and output may immediately affectsectors of consumption, investment and foreign trade, but that shocks to sectorsof consumption, investment and foreign trade do not have immediate impactson inflation and output because of price rigidity and capacity adjustment lags;thus, we put growth of consumption, investment and foreign trade behindinflation and output growth.

    3. EMPIRICAL ANALYSIS

    3.1. Data

    Because there are no existing zero coupon yield curves published by the People’sBank of China, we have to use the original Treasury bonds trading data1 toestimate zero coupon rates for 17 maturities of 3, 6, 9, 12, 15, 18, 21, 24, 30, 36,48, 60, 72, 84, 96, 108 and 120 months, as in Diebold et al. (2006), who adopt thecommonly-used Nelson–Siegel–Svensson method. Then the calculated zerocoupon rates are used to compute the three latent factors: level, slope andcurvature.

    As in previous studies (see e.g. Dewachter and Lyrio, 2006; Diebold and Li,2006; Diebold et al., 2008), we use start-of-month Treasury bonds trading datafor the period 2002:1–2012:12 for China’s exchange bond market to calculatezero coupon rates. There are two reasons for choosing 2002 as the start of thesample period: first, the rapid development of China’s bond market started after2002, and there were not enough Treasury bonds traded per day in China’s bondmarket before 2002 for the model estimation; second, some of China’s monthly

    1 To calculate zero coupon rates, this research collects the net closing prices of all Treasury bondstraded on the first trading day of each month from 2002 to 2012, as well as basic informationregarding each bond, including par value, coupon rate, coupon payment frequency and time tomaturity.

    THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA 423

    © 2015 Wiley Publishing Asia Pty Ltd

  • macroeconomic statistics, such as the M2 growth rate and fixed investmentgrowth rate, are not available for years prior to 2002.

    Macroeconomic variables are comprised of the year-on-year percentage rateof the change in the consumer price index (pi),2 the growth rate of industrialproduction (p), the growth rate of total retail sales of social consumer goods (c),the growth rate of fixed asset investment (i), the growth rate of the total volumeof imports and exports (e), the growth rate of government expenditure (f), theweighted average interest rate of 7-day bond-pledged repos (r, the monetarypolicy interest rate) and the M2 growth rate (m). All of the growth rates men-tioned above are year-on-year percentage growth rates. The GDP growth rate isa better indicator of output growth than the industrial production growth rate,but it is only calculated quarterly or annually, not monthly. Industrial produc-tion growth is a reasonable proxy for output growth (see e.g. Chen, 2009; Huse,2011; Favero et al., 2012). Data sources are outlined in the Appendix.

    3.2. Evaluating zero coupon rates

    Before computing the three latent factors using the dynamic Nelson–Siegelmodel, we need to fit a yield curve for each day in the sample period using theNelson–Siegel–Svensson formula:

    ye e

    eeτ β β

    λ τβ

    λ τβ

    λ τ λ τλ τ( ) = + −⎛

    ⎝⎜⎞⎠⎟

    + − −⎛⎝⎜

    ⎞⎠⎟

    + −− −

    −−

    1 21

    31

    41 1 11 1

    1

    λλ τλ τ

    λ τ

    22

    2

    −⎛⎝⎜

    ⎞⎠⎟

    −e , (11)

    where y(τ) denotes a set of yields, τ denotes maturity, and β1, β2, β3, β4, λ1 and λ2are parameters. The estimated zero coupon rates for each day in the sampleperiod are illustrated in Figure 1.

    To further validate the efficiency of the Nelson–Siegel model, we conductprincipal component analysis, which shows that the first three principal com-ponents account for more than 99% of variations in zero-coupon yields; they arethe level, slope and curvature factor, respectively, according to their factorloadings.

    3.3. Estimating the dynamic Nelson–Siegel model

    Because we consider 17 maturities, yt is a 17-dimensional vector, and there are 36parameters to be estimated using numerical optimization (see Equation 5, 6 and7): 3 elements of the 3-dimensional mean state vector μ, 9 elements of the 3 × 3transition matrix A, 1 element (λ) of the 17 × 3 measurement matrix Λ, 6elements of the 3 × 3 symmetric variance–covariance matrix of the transitionsystem innovations, Q, and 17 elements of the 17 × 17 diagonal variance–covariance matrix of the measurement innovations, H.

    2 The year-on-year percentage rate of change of the consumer price index is frequently quoted as theindicator of the inflation rate.

    Y. YAN AND J. GUO424

    © 2015 Wiley Publishing Asia Pty Ltd

  • Given the initial values of the state-space model, we use the Kalman filter tocalculate the Gaussian log-likelihood function, and iterate the BHHH algorithmto maximize the log-likelihood function according to a convergence criterion of10−1 for the change in the log-likelihood function from one iteration to the next.We impose non-negativity on all estimated variances through estimating logvariances, and initialize all variances at 1. We estimate the three latent factorsand λ for each day using the Nelson–Siegel model, then initialize the mean statevector μ using the means of the estimated three latent factors series, initialize λusing the mean of the estimated λ series, and initialize the transition equationmatrix A using the VAR(1) coefficient matrix of the estimated three latentfactors series.

    The estimate of λ is 0.207339, which implies that the curvature Ct reaches itsmaximum at maturity of 9 months and that the slope St decays quickly. Figure 2illustrates loadings of the level, slope and curvature at each maturity. Comparedwith estimates of λ in Diebold and Li (2006), Diebold et al. (2006) and Afonsoand Martins (2012) implying maximums of Ct at 23, 29 or 48 months for theUSA and at 43 months for Germany, the maturity of 9 months at which Ct

    6

    5

    4

    3

    2

    1

    0

    Zero

    rate

    s

    January2012January2010

    January2008January2006

    January2004January2002

    Time0

    24

    68

    10

    Maturities (i

    n years)

    Figure 1. Zero coupon rates at each month, 2002:1–2012:12Notes: The figure shows the estimated zero coupon rates with maximalmaturity of 10 years at each month from January 2002 to December 2012;time, the month when zero coupon rates are calculated; maturities, maturitiesof 3, 6, 9, 12, 15, 18, 21, 24, 30, 36, 48, 60, 72, 84, 96, 108 and 120 months,here expressed in years; zerorates, corresponding zero coupon rates, expressedin percentages.

    THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA 425

    © 2015 Wiley Publishing Asia Pty Ltd

  • reaches its maximum is much smaller in China. This pattern is consistent withtrading by much less patient individuals in the exchange market (Porter andCassola, 2011).

    Figure 3 shows the estimates of the three latent factors which are calculatedusing the Kalman smoother after convergence of maximum-likelihood estima-tion. The level peaks during high inflation periods from the end of 2003 to

    1.0

    0.8

    0.6

    0.4

    0.2

    0.0

    0 20 40 60 80 100 120Maturities (in months)

    Level loading

    Slope loading

    Curvature loading

    1–e–λτλτ

    1–e–λτ–e–λτλτ

    Figure 2. Loadings of the level, slope and curvature factors, 2002:1–2012:12Notes: The figure shows the loading of each latent factor at each maturity,expressed in months; LEVEL, the level of the yield curve; SLOPE, the slopeof the yield curve; CURVATURE, the curvature of the yield curve.

    64

    20

    –2–4

    –6

    2002 2004 2006 2008 2010 2012

    LEVELSLOPECURVATURE

    Figure 3. Estimates of the level, slope and curvature factors of yield curves,2002:1–2012:12Notes: The figure shows the values of the three latent factors at each monthfrom January 2002 to December 2012; LEVEL, the level of the yield curve;SLOPE, the slope of the yield curve; CURVATURE, the curvature of theyield curve.

    Y. YAN AND J. GUO426

    © 2015 Wiley Publishing Asia Pty Ltd

  • mid-2005 and from the second half of 2007 to the end of 2008, decreases sharplywith the deflation at the beginning of 2009, and during other periods the level israther stable. The slope is consistently negative, which implies ascending yieldcurves, and it has an apparent negative correlation with the level from 2003 to2008. The curvature displays much higher variation than the level and the slope,with an apparent positive correlation with the level since 2005.

    To further check the efficiency of our estimates of the three latent factors, wecompare the three latent factors with their empirical proxies:

    Level yt= ( )120 (12)

    Slope y yt t= −( ) ( )3 120 (13)

    Curvature y y yt t t= − −( ) ( ) ( )2 9 3 120 , (14)

    where yt(τ) denotes a zero coupon rate with maturity of τ months, and the 9months maturity of yt(9) is chosen according to the estimate of λ. Figure 4 showsthat the three latent factors move in line with their corresponding empiricalproxies. Furthermore, the correlations between the level, slope, curvature andtheir empirical counterparts are 96, 80 and 69%, respectively, and they are at asimilar size to previous studies (see e.g. Afonso and Martins, 2012).

    3.4. Impulse response analysis

    After obtaining the three latent factors of yield curves, we continue to estimatea VAR system consisting of eight macro variables and the three latent factors.Based on the Schwarz and Hannan–Quinn information criteria together with aresidual autocorrelation analysis, we estimate a VAR(1) model for the data.Then we report impulse response functions to Cholesky one standard deviationinnovations along with 95% confidence intervals. Three categories of impulseresponses will be considered in turn: macroeconomic responses to macroeco-nomic surprises, yield curve responses to macroeconomic surprises and macro-economic responses to yield curve surprises.

    3.4.1. Interactions among macroeconomic variablesFigure 5 displays significant macroeconomic responses to macroeconomicshocks. First, a positive shock to inflation leads to rising interest rates becausemonetary policy is tightened to control inflation. Consumption growth immedi-ately increases following a positive inflation surprise because rising pricesdirectly increase payment for consumer goods. Second, a positive shock toindustrial production growth pushes up foreign trade because foreign countriesare important resource providers and consumption markets for China rightnow. Enhancing industrial production growth also increases inflation anddecreases money supply growth because stimulative monetary policies areexpected to quit under a flourishing economy. Third, a positive shock to invest-

    THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA 427

    © 2015 Wiley Publishing Asia Pty Ltd

  • ment growth raises industrial production growth gradually and enhances moneysupply growth because of rising loans for financing the investment. Risinginvestment growth also leads to more rapid growth of consumption afterapproximately 6 months. Fourth, a positive shock to foreign trade growthdecreases money supply growth. When foreign trade performance is good,monetary policy will be tightened. Fifth, a positive shock to governmentexpenditure growth reduces the growth of foreign trade because domesticdemand is expanded by government. Higher government expenditure growthalso leads to an immediate increase in money supply, and active fiscal policy isoften accompanied by easing monetary policy. Sixth, a positive shock to mon-etary policy interest rate decreases money supply growth, as expected. Surpris-ingly, a rising monetary policy interest rate pushes up inflation slightly ratherthan reducing inflation; this result is consistent with Diebold et al. (2006), whoattribute this finding to the market’s future inflation expectations boosted by thecentral bank’s concern about overheating and inflationary pressures. Finally, a

    0.02

    50.

    040

    y (1

    20)

    2002 2004 2006 2008 2010 2012

    2.5

    3.5

    4.5

    LEVE

    LLEVELy (120)

    –0.0

    3–0

    .01

    y (3

    ) – y

    (120

    )

    2002 2004 2006 2008 2010 2012

    –5–3

    –10

    SLO

    PE

    SLOPEy (3) – y (120)

    –0.0

    20–0

    .005

    2002 2004 2006 2008 2010 2012

    –20

    12

    3C

    URV

    ATU

    RECURVATURE2y (9) – y (3) – y (120)

    2y (9

    ) – y

    (3) –

    y (1

    20)

    Figure 4. The level, slope, curvature factors and their empirical proxiesNotes: y(3), a zero coupon rate with maturity of 3 months; y(9), a zerocoupon rate with maturity of 9 months; y(120), a zero coupon rate withmaturity of 120 months; LEVEL, the level of the yield curve; SLOPE, theslope of the yield curve; CURVATURE, the curvature of the yield curve.

    Y. YAN AND J. GUO428

    © 2015 Wiley Publishing Asia Pty Ltd

  • positive shock to money supply growth increases inflation gradually andenhances consumption growth after approximately 6 months.

    The previous interactions among macroeconomic variables indicate the fol-lowing general results. First, Chinese monetary policy is mainly conductedthrough quantitative measures instead of market-based measures, and moneysupply growth is a more effective instrument to curb inflation than the monetarypolicy interest rate. Second, investment is still an important measure to stimulatethe Chinese economy, but it also pushes up money supply growth, which willresult in higher inflation. Adjusting the interest rate is an option for the centralbank to stabilize money supply. Finally, Chinese authorities conduct monetaryand fiscal policies to boost economic growth and stabilize prices.

    3.4.2. Yield curve responses to macroeconomic surprisesThe responses of yield curves to macroeconomic surprises are shown inFigure 6. First, a surprise increase in inflation leads to a very brief rise in thecurvature (a more concave yield curve) during the 7th and 10th months. Theresponse of the level to a positive inflation shock is not significant, although thelevel tends to rise with inflation. The long-term inflation is influenced little bytransitory changes in short-term inflation. This reaction is consistent with Porter

    0.0

    0.4

    0.8

    5 10 15 20 25

    Response of c to pi

    –0.0

    50.

    050.

    155 10 15 20 25

    Response of r to pi

    –0.1

    0.1

    0.3

    5 10 15 20 25

    Response of pi to p

    –0.6

    –0.2

    5 10 15 20 25

    Response of m to p

    02

    46

    5 10 15 20 25

    Response of e to p–0

    .20.

    20.

    61.

    0

    5 10 15 20 25

    Response of p to i–0

    .50.

    00.

    5

    5 10 15 20 25

    Response of c to i

    –0.4

    0.0

    0.4

    5 10 15 20 25

    Response of m to i

    –6–4

    –20

    1

    5 10 15 20 25

    Response of e to f

    –0.6

    –0.2

    0.2

    5 10 15 20 25

    –0.2

    0.0

    0.2

    0.4

    5 10 15 20 25

    –0.1

    0.1

    0.2

    5 10 15 20 25

    –0.4

    –0.2

    0.0

    0.2

    5 10 15 20 25

    –0.1

    0.1

    0.3

    5 10 15 20 25–0.

    10.

    10.

    35 10 15 20 25

    Response of m to e

    Response of m to f Response of pi to r Response of m to r Response of c to mResponse of pi to m

    Figure 5. Macro surprises and responsesNotes: pi, inflation; p, growth rate of industrial production; c, growth rate oftotal retail sales of social consumer goods; i, growth rate of fixed assetinvestment; e, growth rate of total volume of imports and exports; f, growthrate of government expenditure; r, the weighted average interest rate of 7-daybond-pledged repos; m, M2 growth rate.

    THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA 429

    © 2015 Wiley Publishing Asia Pty Ltd

  • and Cassola (2011), who point out that long-term inflation is more stronglyanchored in China than in the USA. Second, there are no statistically significantresponses of the yield curve to innovations to industrial production growth andconsumption growth. Third, a surprise increase in investment growth pushes upthe level and reduces the slope (a steeper yield curve). Higher investment growthenhances the market’s long-term perception of inflation. Short-term yields

    –0.0

    40.0

    00.

    040.

    08

    5 10 15 20 25

    Response of L to pi

    –0.0

    20.

    020.

    065 10 15 20 25

    Response of L to p

    –0.0

    6–0

    .02

    0.02

    5 10 15 20 25

    Response of L to c

    0.00

    0.05

    0.10

    5 10 15 20 25

    Response of L to i

    –0.0

    40.

    000.

    04

    5 10 15 20 25

    Response of L to e–0

    .04

    0.00

    0.04

    5 10 15 20 25

    Response of L to f–0

    .04

    0.00

    0.04

    5 10 15 20 25

    Response of L to r

    –0.0

    20.

    020.

    065 10 15 20 25

    Response of L to m

    –0.1

    5–0

    .05

    0.05

    5 10 15 20 25

    Response of S to pi

    –0.1

    00.

    00

    5 10 15 20 25

    Response of S to p

    –0.1

    00.

    000.

    10

    5 10 15 20 25

    Response of S to c

    –0.2

    5–0

    .15

    –0.0

    50.

    05

    5 10 15 20 25

    Response of S to i

    –0.0

    50.

    05

    5 10 15 20 25

    Response of S to e

    –0.1

    00.

    005 10 15 20 25

    Response of S to f

    –0.0

    50.0

    00.

    050.

    10

    5 10 15 20 25

    Response of S to r

    –0.1

    00.

    000.

    05

    5 10 15 20 25

    Response of S to m

    –0.0

    50.

    05

    5 10 15 20 25

    Response of C to pi

    –0.1

    50.

    100.

    00

    5 10 15 20 25

    Response of C to p

    –0.1

    50.

    000.

    10

    5 10 15 20 25

    Response of C to c

    –0.1

    00.

    000.

    10

    5 10 15 20 25

    Response of C to e

    –0.2

    0–0

    .05

    0.10

    5 10 15 20 25

    Response of C to f

    –0.0

    50.

    150.

    05

    5 10 15 20 25

    Response of C to r

    –0.1

    00.

    000.

    10

    5 10 15 20 25

    Response of C to m

    –0.1

    00.0

    00.

    10

    5 10 15 20 25

    Response of C to i

    Figure 6. Macro surprises and the level, slope, curvature responsesNotes: pi, inflation; p, growth rate of industrial production; c, growth rate oftotal retail sales of social consumer goods; i, growth rate of fixed assetinvestment; e, growth rate of total volume of imports and exports; f, growthrate of government expenditure; r, the weighted average interest rate of 7-daybond-pledged repos; m, M2 growth rate; L, the level of the yield curve; S, theslope of the yield curve; C, the curvature of the yield curve.

    Y. YAN AND J. GUO430

    © 2015 Wiley Publishing Asia Pty Ltd

  • decrease because projects with lower returns are carried out due to expandedinvestment; as a result, the slope is lowered and the yield curve becomes steeper.Fourth, the yield curve is not affected significantly by innovations to foreigntrade and government expenditure growth. This result reflects inefficiency ofimport–export and fiscal policies in shaping the yield curve. Fifth, there is nosignificant response of the yield curve to an innovation to monetary policyinterest rate. In practice, the open market operation instrument of the monetarypolicy interest rate is determined in China’s interbank market, and the yieldcurve in this study is based on China’s exchange market. The result that thesurprise increase in monetary policy interest rate does not affect the yield curvereflects not only the segmentation between the interbank bond market and theexchange bond market but also the weakness of the transmission channel ofmonetary policy in China. Finally, a positive innovation to money supplygrowth increases the level and decreases the slope, but the responses are barelysignificant. Higher money supply growth pushes up the market’s perception oflong-term inflation; as a result, the level of the yield curve rises. The cost ofshort-term financing is lowered because of monetary expansion, so the slope ofthe yield curve decreases.

    The previous yield curve responses to macroeconomic surprises reflect thefollowing important conclusions. First, China’s long-term inflation is stronglyanchored and is affected little by transitory changes of short-term inflation.Second, macroeconomic variables, including industrial production growth,import–export growth and government expenditure growth, do not have signifi-cant influences on the yield curve in China. Third, there are significant responsesof the yield curve to investment growth and money supply growth. Higherinvestment growth increases long-term inflation and decreases short-term ratesof return. Higher money supply growth also pushes up long-term inflation andreduces short-term financing costs. Finally, the open market operation con-ducted in China’s interbank market through monetary policy interest rates doesnot reshape the yield curve in China’s exchange market. The result reflects thatthe segmentation of China’s bond market hinders the efficient implementationof monetary policy, and that the transmission channel of monetary policy is stillweak in China.

    3.4.3. Macroeconomic responses to yield curve surprisesNow consider responses of macroeconomic variables to yield curve shocks (seeFigure 7). A positive shock to the level leads to an increase in consumptiongrowth, and such a reaction is consistent with Zhang and Wan (2002), who statethat higher expected inflation will discourage asset accumulation and increasecurrent spending. Surprisingly, there are no statistically significant responses ofmacroeconomic variables to a positive shock to the slope. A positive shock tothe curvature pushes up consumption growth and reduces the monetary policyinterest rate, and the responses are barely significant.

    The previous macroeconomic responses to yield curve surprises presentimportant observations. First, an increase in the level, which is interpreted as themarket’s perception of long-term inflation, does not enhance short-term infla-

    THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA 431

    © 2015 Wiley Publishing Asia Pty Ltd

  • tion or hamper industrial production growth as expected, and higher requiredreturn on investment does not reduce investment growth either. In addition, thehigher level induces no significant responses of the monetary policy interest rateand money supply growth. It appears that the monetary authority is not con-cerned about the level of the yield curve in the exchange market. Perhaps themonetary authority focuses on short-term inflation rather than long-term infla-

    –0.4

    –0.2

    0.0

    0.2

    Response of pi to L

    5 10 15 20 25 –0.

    5–0

    .3–0

    .10.

    1

    Response of p to L

    5 10 15 20 25

    –0.2

    0.2

    0.6

    Response of c to L

    5 10 15 20 25

    –0.5

    0.5

    1.5

    Response of i to L

    5 10 15 20 25

    –2.0

    –1.0

    0.0

    1.0

    Response of e to L

    5 10 15 20 25

    –1.5

    –0.5

    0.5

    Response of f to L

    5 10 15 20 25

    –0.1

    00.

    000.

    10Response of r to L

    5 10 15 20 25 –0.

    40.

    00.

    20.

    4 Response of m to L

    5 10 15 20 25 –0.

    100.

    000.

    10

    Response of pi to S

    5 10 15 20 25

    –0.2

    0.0

    0.2

    0.4

    Response of p to S

    5 10 15 20 25

    –0.4

    –0.2

    0.0

    0.2

    Response of c to s

    5 10 15 20 25

    –0.5

    0.0

    0.5

    1.0

    Response of i to S

    5 10 15 20 25

    –1.5

    –0.5

    0.5

    Response of e to S

    5 10 15 20 25–1

    .00.

    01.

    0

    Response of f to S

    5 10 15 20 25 –0.

    100.

    000.

    10

    Response of r to S

    5 10 15 20 25

    –0.2

    0.0

    0.2

    Response of m to s

    5 10 15 20 25 –0.

    25–0

    .10

    0.00

    Response of pi to S

    5 10 15 20 25

    –0.4

    –0.2

    0.0

    Response of p to C

    5 10 15 20 25

    –0.2

    0.2

    0.4

    Response of c to C

    5 10 15 20 25

    –1.5

    –0.5

    0.5

    Response of e to C

    5 10 15 20 25

    –2.5

    –1.5

    –0.5

    0.5

    Response of f to C

    5 10 15 20 25

    –0.1

    5–0

    .05

    Response of r to C

    5 10 15 20 25

    –0.1

    0.1

    0.3

    Response of m to C

    5 10 15 20 25

    –1.0

    0.0

    1.0

    Response of i to C

    5 10 15 20 25

    Figure 7. The level, slope, curvature surprises and the macro responsesNotes: pi, inflation; p, growth rate of industrial production; c, growth rate oftotal retail sales of social consumer goods; i, growth rate of fixed assetinvestment; e, growth rate of total volume of imports and exports; f, growthrate of government expenditure; r, the weighted average interest rate of 7-daybond-pledged repos; m, M2 growth rate; L, the level of the yield curve; S, theslope of the yield curve; C, the curvature of the yield curve.

    Y. YAN AND J. GUO432

    © 2015 Wiley Publishing Asia Pty Ltd

  • tion, or the yield curve in the exchange market is not an efficient benchmarkyield curve for conducting monetary policy. Besides, the increasing level doesnot affect investment growth significantly, which means the yield curve does notaccurately reflect the financing cost. As China’s interest rates have not been fullyliberalized yet, in practice, the regulated official deposit and loan interest ratesare better indicators of the financing cost. Second, there are no significantresponses of the monetary policy interest rate and money supply growth to theslope surprise, and this result reaffirms that the interrelation between openmarket operations conducted in the interbank market and the yield curve in theexchange market is very weak. Third, it is shown in Subsection 3.4.2 that thecurvature reacts significantly to innovations to inflation and industrial produc-tion growth, but in this subsection we do not find significant responses ofinflation and industrial production growth to an innovation to the curvature.Finally, it is evident that interactions between the yield curve and themacroeconomy in China are nearly unidirectional rather than bidirectional.Macroeconomic variables reshape the yield curve, but direct adjustments of theyield curve do not lead to significant changes in macroeconomic variables. Dueto the incomplete liberalization of interest rates and market segmentation, theyield curve does not accurately reflect the cost of capital, and it is not an effectivebenchmark yield curve. As a result, there exists a wide disjunction between thereal economy and financial markets in China.

    3.5. Forecast error variance decompositions

    Panel 1 in Table 1 shows that inflation surprises explain more than 70% of thevariance of the errors in forecasting inflation at a 4-month horizon. Meanwhile,innovations to industrial production growth stay around 20% for forecast hori-zons of 4 months and beyond. The money supply innovations become more andmore important in explaining that forecast error variance from 2.93% at a4-month horizon to above 12% at a 24-month horizon. The importance ofinnovations to the level of the yield curve keeps rising and reaches above 5% ata 24-month horizon.

    From panel 2 in Table 1, more than 72% of the forecast error variance ofindustrial production growth is explained by innovations to industrial produc-tion growth itself during a 24-month horizon. Besides, the importance of inno-vations to investment growth stabilizes at around 12% and that of innovationsto the monetary policy interest rate remains at around 4.5%.

    Panel 3 in Table 1 shows that approximately 73% of the variance of the errorsin forecasting consumption growth at a 4-month horizon is explained by inno-vations to consumption growth itself. From the 4-month horizon onwards, thepart explained by consumption growth surprises reduces gradually to approxi-mately 40% at a 24-month horizon. This reduction is mainly attributed toinnovations to investment growth, money supply growth and the level factor.The importance of innovations to investment growth stays above 16% from the16-month horizon onwards. Innovations to the level factor and inflation alsoexplain some of the forecast error variance; both contributions are above 10%.

    THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA 433

    © 2015 Wiley Publishing Asia Pty Ltd

  • Tab

    le1.

    For

    ecas

    ter

    ror

    vari

    ance

    deco

    mpo

    siti

    ons

    Per

    iod

    pip

    ci

    ef

    rm

    LS

    C

    1.F

    orec

    asti

    ngin

    flati

    on(p

    i)4

    70.8

    418

    .56

    0.68

    0.74

    1.37

    0.03

    3.53

    2.93

    0.61

    0.47

    0.24

    857

    .20

    22.8

    01.

    144.

    251.

    400.

    132.

    507.

    550.

    430.

    601.

    9912

    53.0

    021

    .99

    1.35

    5.65

    1.18

    0.37

    1.78

    10.4

    11.

    360.

    482.

    4316

    51.6

    320

    .40

    1.32

    5.52

    1.01

    0.65

    1.50

    12.0

    43.

    120.

    482.

    3420

    50.9

    119

    .18

    1.26

    5.17

    1.00

    0.89

    1.39

    12.7

    14.

    760.

    502.

    2324

    50.2

    918

    .56

    1.22

    5.08

    1.11

    1.04

    1.35

    12.8

    05.

    880.

    512.

    162.

    For

    ecas

    ting

    grow

    thra

    teof

    indu

    stri

    alpr

    oduc

    tion

    (p)

    40.

    4580

    .13

    1.47

    11.3

    00.

    910.

    223.

    270.

    380.

    540.

    500.

    838

    0.57

    74.9

    51.

    3212

    .68

    2.14

    0.34

    4.24

    1.07

    1.43

    0.46

    0.79

    120.

    5973

    .66

    1.30

    12.8

    22.

    220.

    334.

    461.

    491.

    870.

    480.

    7816

    0.60

    73.3

    21.

    2912

    .77

    2.22

    0.35

    4.53

    1.57

    2.07

    0.48

    0.78

    200.

    6573

    .15

    1.30

    12.7

    72.

    260.

    364.

    541.

    572.

    130.

    480.

    8024

    0.74

    72.9

    81.

    3012

    .79

    2.29

    0.36

    4.52

    1.59

    2.13

    0.48

    0.81

    3.F

    orec

    asti

    nggr

    owth

    rate

    ofto

    talr

    etai

    lsal

    esof

    soci

    alco

    nsum

    ergo

    ods

    (c)

    47.

    363.

    8472

    .78

    3.38

    0.79

    1.43

    0.92

    0.73

    4.94

    0.65

    3.17

    87.

    325.

    8654

    .94

    9.12

    0.63

    1.23

    1.14

    3.39

    12.6

    70.

    922.

    7912

    8.15

    6.38

    45.3

    914

    .66

    0.57

    1.51

    1.21

    5.42

    13.1

    60.

    802.

    7516

    9.38

    6.46

    41.6

    116

    .44

    0.60

    1.82

    1.24

    6.76

    12.2

    50.

    742.

    7220

    10.2

    66.

    3640

    .03

    16.6

    60.

    682.

    051.

    247.

    5211

    .79

    0.73

    2.68

    2410

    .71

    6.26

    39.3

    616

    .52

    0.79

    2.20

    1.24

    7.84

    11.6

    90.

    732.

    654.

    For

    ecas

    ting

    grow

    thra

    teof

    fixed

    asse

    tin

    vest

    men

    t(i

    )4

    2.47

    0.39

    2.57

    88.6

    01.

    890.

    031.

    210.

    661.

    980.

    100.

    108

    2.29

    0.88

    2.78

    85.6

    81.

    940.

    051.

    561.

    452.

    900.

    140.

    3412

    2.26

    0.93

    2.79

    85.1

    11.

    900.

    111.

    721.

    792.

    870.

    130.

    3916

    2.25

    0.94

    2.78

    84.8

    81.

    920.

    161.

    781.

    902.

    860.

    140.

    3920

    2.25

    0.98

    2.78

    84.7

    51.

    960.

    191.

    791.

    922.

    860.

    140.

    3924

    2.26

    1.03

    2.77

    84.6

    71.

    990.

    191.

    791.

    922.

    860.

    140.

    39

    Y. YAN AND J. GUO434

    © 2015 Wiley Publishing Asia Pty Ltd

  • 5.F

    orec

    asti

    nggr

    owth

    rate

    ofto

    talv

    olum

    eof

    impo

    rts

    and

    expo

    rts

    (e)

    40.

    8431

    .16

    0.46

    2.13

    51.0

    712

    .66

    0.23

    0.27

    0.41

    0.04

    0.72

    80.

    7832

    .29

    0.43

    4.51

    46.4

    311

    .38

    0.52

    1.09

    1.75

    0.05

    0.78

    120.

    9231

    .82

    0.42

    5.01

    44.7

    210

    .83

    0.76

    1.78

    2.95

    0.05

    0.75

    160.

    9531

    .43

    0.41

    4.99

    44.1

    210

    .68

    0.87

    2.02

    3.72

    0.06

    0.74

    200.

    9631

    .32

    0.42

    5.02

    43.8

    510

    .62

    0.91

    2.03

    4.06

    0.07

    0.75

    241.

    0331

    .31

    0.43

    5.13

    43.6

    210

    .56

    0.91

    2.04

    4.13

    0.07

    0.77

    6.F

    orec

    asti

    nggr

    owth

    rate

    ofgo

    vern

    men

    tex

    pend

    itur

    e(f

    )4

    5.24

    2.90

    4.87

    1.78

    5.94

    77.0

    01.

    360.

    010.

    220.

    200.

    498

    6.31

    2.84

    4.75

    2.34

    5.93

    75.1

    71.

    500.

    010.

    460.

    210.

    4912

    6.79

    2.88

    4.71

    2.39

    5.88

    74.5

    11.

    560.

    050.

    520.

    220.

    5016

    7.06

    2.93

    4.69

    2.38

    5.85

    74.1

    11.

    570.

    100.

    580.

    220.

    5120

    7.21

    2.96

    4.68

    2.38

    5.83

    73.8

    71.

    560.

    140.

    650.

    220.

    5124

    7.27

    2.96

    4.67

    2.37

    5.82

    73.7

    41.

    560.

    160.

    700.

    220.

    517.

    For

    ecas

    ting

    the

    wei

    ghte

    dav

    erag

    ein

    tere

    stra

    teof

    7-da

    ybo

    nd-p

    ledg

    edre

    pos

    (r)

    47.

    301.

    533.

    071.

    950.

    010.

    0581

    .32

    0.23

    0.23

    0.63

    3.68

    816

    .00

    3.37

    3.96

    2.04

    0.17

    0.07

    65.5

    90.

    582.

    070.

    555.

    6112

    20.2

    95.

    083.

    621.

    960.

    440.

    0656

    .76

    1.23

    4.98

    0.48

    5.11

    1622

    .08

    5.68

    3.38

    1.89

    0.51

    0.06

    52.3

    51.

    876.

    920.

    484.

    7920

    22.7

    05.

    713.

    251.

    900.

    490.

    0850

    .30

    2.23

    8.23

    0.48

    4.62

    2422

    .80

    5.63

    3.19

    2.01

    0.49

    0.10

    49.3

    92.

    349.

    010.

    494.

    548.

    For

    ecas

    ting

    M2

    grow

    thra

    te(m

    )4

    1.51

    5.06

    1.80

    8.31

    7.83

    8.57

    5.16

    61.0

    10.

    510.

    220.

    038

    1.14

    15.0

    01.

    375.

    8016

    .42

    11.5

    36.

    5140

    .67

    0.86

    0.23

    0.46

    122.

    6321

    .95

    1.38

    5.49

    18.8

    210

    .72

    5.60

    31.5

    10.

    710.

    190.

    9816

    4.89

    24.4

    61.

    426.

    1918

    .32

    9.48

    4.84

    28.0

    50.

    950.

    161.

    2420

    6.69

    24.6

    61.

    416.

    4617

    .36

    8.75

    4.47

    27.0

    71.

    650.

    171.

    3124

    7.69

    24.2

    21.

    396.

    3716

    .74

    8.46

    4.32

    26.8

    42.

    470.

    181.

    31

    THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA 435

    © 2015 Wiley Publishing Asia Pty Ltd

  • Tab

    le1.

    Con

    tinu

    ed

    Per

    iod

    pip

    ci

    ef

    rm

    LS

    C

    9.F

    orec

    asti

    ngth

    ele

    velo

    fth

    eyi

    eld

    curv

    e(L

    )4

    0.48

    2.91

    0.10

    14.5

    70.

    070.

    470.

    080.

    9579

    .50

    0.02

    0.87

    82.

    855.

    790.

    6425

    .18

    0.05

    0.69

    0.11

    2.93

    60.1

    10.

    021.

    6312

    6.00

    7.48

    1.00

    28.2

    10.

    050.

    940.

    175.

    2248

    .98

    0.02

    1.93

    168.

    628.

    011.

    1428

    .32

    0.05

    1.22

    0.21

    7.02

    43.3

    30.

    042.

    0420

    10.2

    77.

    961.

    1727

    .65

    0.10

    1.46

    0.24

    8.09

    40.9

    70.

    072.

    0424

    11.0

    47.

    791.

    1627

    .07

    0.20

    1.63

    0.25

    8.55

    40.2

    10.

    082.

    0110

    .F

    orec

    asti

    ngth

    esl

    ope

    ofth

    eyi

    eld

    curv

    e(S

    )4

    0.18

    1.65

    3.49

    8.36

    0.34

    2.09

    1.36

    1.45

    17.2

    549

    .20

    14.6

    48

    0.64

    1.37

    2.17

    19.7

    50.

    882.

    111.

    614.

    5429

    .86

    28.0

    39.

    0416

    3.20

    1.86

    2.00

    26.2

    51.

    273.

    021.

    606.

    9726

    .85

    19.7

    97.

    2120

    4.18

    1.90

    1.99

    26.2

    81.

    373.

    241.

    587.

    6325

    .82

    19.0

    27.

    0024

    4.72

    1.88

    1.98

    26.0

    61.

    473.

    381.

    577.

    9525

    .40

    18.7

    06.

    9011

    .F

    orec

    asti

    ngth

    ecu

    rvat

    ure

    ofth

    eyi

    eld

    curv

    e(C

    )4

    0.28

    1.29

    1.02

    0.97

    0.45

    1.32

    1.27

    0.33

    13.9

    355

    .78

    23.3

    68

    2.07

    2.92

    2.55

    1.98

    0.53

    1.15

    1.58

    0.31

    13.2

    547

    .88

    25.7

    712

    3.91

    4.22

    2.58

    1.95

    0.91

    1.10

    1.63

    0.53

    13.3

    245

    .38

    24.4

    716

    5.07

    4.79

    2.53

    1.92

    1.01

    1.07

    1.60

    0.93

    13.4

    343

    .94

    23.7

    220

    5.69

    4.91

    2.49

    1.90

    1.00

    1.06

    1.57

    1.20

    13.6

    943

    .16

    23.3

    324

    5.92

    4.90

    2.48

    1.89

    0.99

    1.07

    1.55

    1.32

    13.9

    442

    .80

    23.1

    3

    Not

    es:

    pi,i

    nflat

    ion;

    p,gr

    owth

    rate

    ofin

    dust

    rial

    prod

    ucti

    on;c

    ,gro

    wth

    rate

    ofto

    talr

    etai

    lsal

    esof

    soci

    alco

    nsum

    ergo

    ods;

    i,gr

    owth

    rate

    offix

    edas

    set

    inve

    stm

    ent;

    e,gr

    owth

    rate

    ofto

    talv

    olum

    eof

    impo

    rts

    and

    expo

    rts;

    f,gr

    owth

    rate

    ofgo

    vern

    men

    texp

    endi

    ture

    ;r,t

    hew

    eigh

    ted

    aver

    age

    inte

    rest

    rate

    of7-

    day

    bond

    -ple

    dged

    repo

    s;m

    ,M2

    grow

    thra

    te;

    L,t

    hele

    velo

    fth

    eyi

    eld

    curv

    e;S

    ,the

    slop

    eof

    the

    yiel

    dcu

    rve;

    C,t

    hecu

    rvat

    ure

    ofth

    eyi

    eld

    curv

    e.E

    ach

    row

    show

    sth

    epe

    rcen

    tage

    ofth

    eva

    rian

    ceof

    the

    erro

    rin

    fore

    cast

    ing

    the

    vari

    able

    men

    tion

    edin

    the

    titl

    eof

    the

    tabl

    e,at

    each

    fore

    cast

    ing

    hori

    zon

    (in

    mon

    ths)

    give

    nin

    the

    first

    colu

    mn.

    Y. YAN AND J. GUO436

    © 2015 Wiley Publishing Asia Pty Ltd

  • Panel 4 in Table 1 presents the forecast error variance decomposition ofinvestment growth. Almost all of the forecast error variance is explained byinnovations to investment growth itself. Innovations to other variables accountfor a negligible part of the forecast error variance. Among them, the contribu-tions of surprises to the level factor and consumption growth stay above 2.5%.

    Panel 5 in Table 1 shows that the variance of the errors in forecasting import–export growth is explained by surprises to import–export growth itself, and, toa lesser extent, by surprises to industrial production growth. The importance ofimport–export growth surprises falls gradually, from approximately 51% at a4-month horizon to approximately 44% at a 24-month horizon, while that ofinnovations to industrial production growth stabilizes at around 31% for fore-cast horizons of 4 months and beyond. Innovations to government expendituregrowth account for approximately 11% from the 8-month horizon onwards, andthe importance of surprises to investment growth rises from approximately 2%at a 4-month horizon to above 5% at a 24-month horizon.

    From panel 6 in Table 1, more than 73% of the variance of the errors inforecasting government expenditure growth is explained by innovations to gov-ernment expenditure growth itself. The importance of surprises to industrialproduction growth and import–export growth remains above 5% from the4-month horizon onwards.

    Panel 7 in Table 1 shows that approximately 81% of the forecast error vari-ance of the monetary policy interest rate is explained by innovations to themonetary policy interest rate itself at a 4-month horizon; meanwhile, approxi-mately 7% is attributed to inflation surprises. The importance of surprises to themonetary policy interest rate diminishes below 50% at a 24-month horizon;meanwhile, the contribution of the inflation surprises increases to above 22%. Inaddition, the part explained by innovations to the level factor becomes increas-ingly large, and approaches 10% at a 24-month horizon.

    From panel 8 in Table 1, we can see that approximately 61% of the forecasterror variance of money supply growth is explained by surprises to moneysupply growth at a 4-month horizon; meanwhile, approximately 5% of theforecast error variance is attributed to innovations to industrial productiongrowth. At a 24-month horizon, the importance of surprises to money supplygrowth falls gradually below 27%; meanwhile, that of innovations to industrialproduction growth exceeds 24%. The part explained by surprises to import–export growth rises from below 8% at a 4-month horizon to nearly 17% at a24-month horizon. Innovations to government expenditure growth account forapproximately 8% of the variance of the errors in forecasting money supplygrowth at a 24-month horizon, while the importance of innovations to themonetary policy interest rate stabilizes at around 5%.

    Panel 9 of Table 1 presents the forecast error variance decomposition of thelevel of the yield curve. At a 4-month horizon, innovations to the level explainnearly 80% of the variance of the errors in forecasting the level; meanwhile,approximately 15% of the forecast error variance is attributed to investmentsurprises. From the 4-month horizon onwards, the importance of innovations tothe level reduces to approximately 40%, while that of investment surprises

    THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA 437

    © 2015 Wiley Publishing Asia Pty Ltd

  • stabilizes at around 27%. The part explained by inflation surprises rises fromapproximately 0.5% at a 4-month horizon to above 11% at a 24-month horizon,and innovations to industrial production growth and money supply growth alsoaccount for approximately 8 and 8.6%, respectively, at a 24-month horizon.

    Panel 10 in Table 1 shows that at a 4-month horizon, almost half of thevariance of the errors in forecasting the slope is attributed to the slope surprises;meanwhile, approximately 17% is explained by the level surprises and approxi-mately 8% is driven by investment shocks. The importance of the slope surprisesfalls below 19% at a 24-month horizon, and that of the level surprises is rela-tively stable at around 25%. Investment surprises become another driver of thevariance of the errors in forecasting the slope at a 24-month horizon, and theimportance stabilizes at around 26%.

    From panel 11 in Table 1, the curvature surprises explain approximately 23%of the variance of the errors in forecasting the curvature at a 4-month horizon,and the importance of the curvature surprises stabilizes at around 23% at a24-month horizon after increasing slightly at an 8-month horizon. The slopesurprises are the main driver of the variance of the errors in forecasting thecurvature. The slope surprises account for approximately 56% of the forecasterror variance at a 4-month horizon, and their importance reduces gradually toapproximately 43% at a 24-month horizon. Moreover, the importance of thelevel surprises stabilizes at around 13% for forecast horizons of 4 months andbeyond.

    4. CONCLUSION

    This paper studies the sovereign yield curve and its interactions with themacroeconomy in China for the period January 2002–December 2012. Almostall variations in China’s zero-coupon yields can be explained by the first threeprincipal components, which can be interpreted as the level, slope and curvatureof the yield curve. We estimate these three time-varying latent factors in thedynamic Nelson–Siegel model. We further establish a VAR system consisting ofmacroeconomic variables and the three factors to explore interactions amongvariables by impulse response analysis. Our findings are summarized as follows.

    Chinese authorities conduct monetary and fiscal policies to boost economicgrowth and stabilize prices. In China, money supply growth is a more effectiveinstrument to curb inflation than the monetary policy interest rate, and mon-etary policy is mainly conducted through quantitative measures instead ofmarket-based measures. Investment is still an important measure to stimulatethe Chinese economy, but it also pushes up money supply growth, which resultsin higher inflation. Adjusting the monetary policy interest rate is an option forthe central bank to stabilize money supply. However, raising the monetarypolicy interest rate pushes inflation up slightly rather than reducing inflation,because the market’s future inflation expectations are boosted by the centralbank’s concern about overheating and inflationary pressures.

    China’s long-term inflation is strongly anchored and is affected little bytransitory changes in short-term inflation. Macroeconomic variables, including

    Y. YAN AND J. GUO438

    © 2015 Wiley Publishing Asia Pty Ltd

  • industrial production growth, import–export growth and government expendi-ture growth, do not have significant influence on the yield curve in China, butthe yield curve reacts significantly to innovations to investment growth andmoney supply growth. Both higher investment growth and higher money supplygrowth push up long-term inflation and decrease short-term yields. The openmarket operations through the monetary policy interest rate conducted in theinterbank market do not reshape the yield curve in the exchange market. Thesegmentation of China’s bond market hinders the efficient implementation ofmonetary policy, and the monetary policy transmission mechanism is still weakin China.

    Interactions between the yield curve and the macroeconomy in China arenearly unidirectional rather than bidirectional. Macroeconomic variablesreshape the yield curve, but direct adjustments of the yield curve do not result insignificant changes in macroeconomic variables. Due to the incomplete liberali-zation of interest rates and market segmentation, the yield curve does notaccurately reflect the cost of capital , and it is not an efficient benchmark yieldcurve. As a result, there exists a wide disjunction between the real economy andfinancial markets in China.

    China’s bond market is segmented into the exchange bond market and theinterbank bond market. The present study has focused on interactions betweenthe yield curve and the macroeconomy, and has revealed that the market seg-mentation affects the transmission of monetary policy. However, it is still anopen question whether the segmentation in bond markets influences the inter-actions between macro variables and yield curves. We leave this topic for futureresearch.

    REFERENCES

    Afonso, A. and M. M. F. Martins (2012) ‘Level, Slope, Curvature of the Sovereign Yield Curve, andFiscal Behaviour’, Journal of Banking & Finance 36, 1789–807.

    Aguiar-Conraria, L., M. M. F. Martins and M. J. Soares (2012) ‘The Yield Curve and the Macro-economy across Time and Frequencies’, Journal of Economic Dynamics & Control 36, 1950–70.

    Ang, A., M. Plazzesi and M. Wei (2006) ‘What Does the Yield Curve Tell Us about GDP Growth?’,Journal of Econometrics 131, 359–403.

    Bekaert, G., S. Cho and A. Moreno (2010) ‘New Keynesian Macroeconomics and the TermStructure’, Journal of Money Credit and Banking 42, 33–62.

    Benigno, P. and A. Missale (2004) ‘High Public Debt in Currency Crises: Fundamentals versusSignaling Effects’, Journal of International Money and Finance 23, 165–88.

    Chen, S.-S. (2009) ‘Predicting the Bear Stock Market: Macroeconomic Variables as Leading Indi-cators’, Journal of Banking & Finance 33, 211–23.

    Chow, G. C. (2006) ‘Globalization and China’s Economic Development’, Pacific Economic Review11, 271–85.

    Coroneo, L., K. Nyholm and R. Vidova-Koleva (2011) ‘How Arbitrage-Free Is the Nelson-SiegelModel?’, Journal of Empirical Finance 18, 393–407.

    Dewachter, H. and M. Lyrio (2006) ‘Macro Factors and the Term Structure of Interest Rates’,Journal of Money Credit and Banking 38, 119–40.

    Diebold, F., G. Rudebusch and S. Boragan Aruoba (2006) ‘The Macroeconomy and the YieldCurve: A Dynamic Latent Factor Approach’, Journal of Econometrics 131, 309–38.

    Diebold, F. X. and C. L. Li (2006) ‘Forecasting the Term Structure of Government Bond Yields’,Journal of Econometrics 130, 337–64.

    THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA 439

    © 2015 Wiley Publishing Asia Pty Ltd

  • Diebold, F. X., C. Li and V. Z. Yue (2008) ‘Global Yield Curve Dynamics and Interactions:A Dynamic Nelson-Siegel Approach’, Journal of Econometrics 146, 351–363. Fed ReserveBank Atlanta.

    Durbin, J. and S. J. Koopman (2012) Time Series Analysis by State Space Methods, Vol. 38. Oxford,UK: Oxford University Press.

    Estrella, A. (2005) ‘Why Does the Yield Curve Predict Output and Inflation?’, Economic Journal115, 722–44.

    Evans, C. L. and D. A. Marshall (2007) ‘Economic Determinants of the Nominal Treasury YieldCurve’, Journal of Monetary Economics 54, 1986–2003.

    Fan, L. and A. C. Johansson (2010) ‘China’s Official Rates and Bond Yields’, Journal of Banking &Finance 34, 996–1007.

    Farka, M. and A. DaSilva (2011) ‘The Fed and the Term Structure: Addressing Simultaneity withinA Structural VAR Model’, Journal of Empirical Finance 18, 935–52.

    Favero, C. A., L. Niu and L. Sala (2012) ‘Term Structure Forecasting: No-Arbitrage Restrictionsversus Large Information Set’, Journal of Forecasting 31, 124–56.

    Gurkaynaka, R. S., B. Sack and J. H. Wright (2007) ‘The US Treasury Yield Curve: 1961 to thePresent’, Journal of Monetary Economics 54, 2291–304.

    Harvey, A. C. (1990) Forecasting, Structural Time Series Models and the Kalman Filter. Cambridge,UK: Cambridge University Press.

    He, D. and H. L. Wang (2012) ‘Dual-Track Interest Rates and the Conduct of Monetary Policy inChina’, China Economic Review 23, 928–47.

    He, D., Z. Zhang and W. Zhang (2009) ‘How Large Will Be the Effect of China’s Fiscal StimulusPackage on Output and Employment?’, Pacific Economic Review 14, 730–44.

    He, Q., P. H. Leung and T. T. L. Chong (2013) ‘Factor-Augmented VAR Analysis of the MonetaryPolicy in China’, China Economic Review 25, 88–104.

    Huse, C. (2011) ‘Term Structure Modelling with Observable State Variables’, Journal of Banking &Finance 35, 3240–52.

    Ivanova, D., K. Lahiri and F. Seitz (2000) ‘Interest Rate Spreads As Predictors of German Inflationand Business Cycles’, International Journal of Forecasting 16, 39–58.

    Koivu, T. (2009) ‘Has the Chinese Economy Become More Sensitive to Interest Rates? StudyingCredit Demand in China’, China Economic Review 20, 455–70.

    Lengwiler, Y. and C. Lenz (2010) ‘Intelligible Factors for the Yield Curve’, Journal of Econometrics157, 481–91.

    Lu, B. and L. Wu (2009) ‘Macroeconomic Releases and the Interest Rate Term Structure’, Journalof Monetary Economics 56, 872–84.

    Luo, X., H. Han and J. E. Zhang (2012) ‘Forecasting the Term Structure of Chinese TreasuryYields’, Pacific-Basin Finance Journal 20, 639–59.

    Mehl, A. (2009) ‘The Yield Curve as a Predictor and Emerging Economies’, Open Economies Review20, 683–716.

    Mehrotra, A., R. Nuutilainen and J. Paakkonen (2013) ‘Changing Economic Structures andImpacts of Shocks: Evidence from A Dynamic Stochastic General Equilibrium Model forChina’, Pacific Economic Review 18, 92–107.

    Moench, E. (2012) ‘Term Structure Surprises: The Predictive Content of Curvature, Level, andSlope’, Journal of Applied Econometrics 27, 574–602.

    Nelson, C. R. and A. F. Siegel (1987) ‘Parsimonious Modeling of Yield Curves’, Journal of Business60, 473–89.

    Nobili, A. (2007) ‘Assessing the Predictive Power of Financial Spreads in the Euro Area: DoesParameters Instability Matter?’, Empirical Economics 33, 177–95.

    Porter, N. and N. Cassola (2011) Understanding Chinese Bond Yields and Their Role in MonetaryPolicy, Working Paper, International Monetary Fund.

    Qin, D., P. Quising, X. H. He and S. G. Liu (2005) ‘Modeling Monetary Transmission and Policyin China’, Journal of Policy Modeling 27, 157–75.

    Rudebusch, G. D. and J. C. Williams (2009) ‘Forecasting Recessions: The Puzzle of theEnduring Power of the Yield Curve’, Journal of Business & Economic Statistics 27, 492–503.

    Rudebusch, G. D. and T. Wu (2008) ‘A Macro-Finance Model of the Term Structure, MonetaryPolicy and the Economy’, Economic Journal 118, 906–26.

    Schich, S. (2002) ‘Assessing the Stability of the Relationship between Yield Curves and Inflation’,Applied Economics Letters 9, 485–91.

    Siklos, P. L. and Y. Zhang (2010) ‘Identifying the Shocks Driving Inflation in China’, PacificEconomic Review 15, 204–23.

    Y. YAN AND J. GUO440

    © 2015 Wiley Publishing Asia Pty Ltd

  • Sultan, J. (2005) ‘Information Content of the Fed Funds Rates’, Journal of Futures Markets 25,753–74.

    Vicente, J. and B. M. Tabak (2008) ‘Forecasting Bond Yields in the Brazilian Fixed Income Market’,International Journal of Forecasting 24, 490–7.

    Wu, T. (2003) ‘Stylized Facts on Nominal Term Structure and Business Cycles: An Empirical VARStudy’, Applied Economics 35, 901–6.

    Wu, T. (2006) ‘Macro Factors and the Affine Term Structure of Interest Rates’, Journal of MoneyCredit and Banking 38, 1847–75.

    Yu, W.-C. and D. M. Salyards (2009) ‘Parsimonious Modeling and Forecasting of Corporate YieldCurve’, Journal of Forecasting 28, 73–88.

    Zellner, A. (1992) ‘Statistics, Science and Public-Policy’, Journal of the American Statistical Asso-ciation 87, 1–6.

    Zhang, Y. and G. H. Wan (2002) ‘Household Consumption and Monetary Policy in China’, ChinaEconomic Review 13, 27–52.

    APPENDIX

    Data sources

    Treasury bonds trading data on the first trading day of each month fromJanuary 2002 to December 2012: The Shanghai Stock Exchange (http://www.sse.com.cn).

    Macroeconomic variables including the consumer price index, the growth rateof industrial production, the growth rate of total retail sales of social consumergoods, the growth rate of fixed asset investment, the growth rate of total volumeof imports and exports, the growth rate of government expenditure, theweighted average interest rate of 7-day bond-pledged repos and the M2 growthrate are collected mainly from the China Stock Market & Accounting Researchdatabase (http://www.gtarsc.com/) and partly from the National Bureau ofStatistics of China (http://www.stats.gov.cn/tjsj/ndsj/).

    THE SOVEREIGN YIELD CURVE AND THE MACROECONOMY IN CHINA 441

    © 2015 Wiley Publishing Asia Pty Ltd

  • 本文献由“学霸图书馆-文献云下载”收集自网络,仅供学习交流使用。

    学霸图书馆(www.xuebalib.com)是一个“整合众多图书馆数据库资源,

    提供一站式文献检索和下载服务”的24 小时在线不限IP

    图书馆。

    图书馆致力于便利、促进学习与科研,提供最强文献下载服务。

    图书馆导航:

    图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具

    http://www.xuebalib.com/cloud/http://www.xuebalib.com/http://www.xuebalib.com/cloud/http://www.xuebalib.com/http://www.xuebalib.com/vip.htmlhttp://www.xuebalib.com/db.phphttp://www.xuebalib.com/zixun/2014-08-15/44.htmlhttp://www.xuebalib.com/