Chapter

Chapter 3 Sovereign Bonds: What Does the Yield Curve Tell Us?

Editor(s):
Alfred Schipke, Markus Rodlauer, and Longmei Zhang
Published Date:
March 2019
Share
  • ShareShare
Show Summary Details
Author(s)
CHEN Sally, ZHANG Longmei, Kevin Chow, Thomas Harjes and Nathan Porter 

China’s sovereign bond market has ballooned in recent years into the world’s third largest after the United States and Japan. But concerns about limited liquidity, investor participation, and market efficiency have consistently marred investor perceptions of the country’s government bonds as a suitable benchmark for the overall bond market.

Low liquidity in the medium and long end of the yield curve, as well as thin overall secondary market activity beyond the 10-year maturity, limit the reference function of government bonds for other instruments. And banks still hold roughly 70 percent of all government Treasuries outstanding—usually holding them until maturity—limiting the supply of recently issued on-the-run securities for secondary trading and “information processing” in the bond market.

These factors have caused many to doubt whether the government yield curve can accurately reflect market conditions and thus help develop the bond market and, more broadly, China’s capital markets—an important task in the country’s economic reform efforts.

Nonetheless, this chapter finds that various measures implemented in recent years have boosted the linkage between China government bond (CGB) curve factors, macro variables, and global financial market conditions. And despite liquidity concerns, the government bond yield curve seems to anticipate changes in macroeconomic conditions. Similarly, there is evidence that linkage between liquidity management by the People’s Bank of China (PBC) and CGB pricing is emerging, suggesting growing policy transmission via the bond market.

This chapter analyzes China’s government bond market—yield movements, changes in its slope and curvature, and responses and interactions with the aggregate economy—to better understand its liquidity and informational efficiency. It finds that with greater market liberalization, more market-based open market operations from the PBC, integration with the global financial system, and a more robust CGB market,1 the Chinese central government bond market can become a more efficient tool in the country’s crucial economic transition.

Characteristics of a Liquid, Robust Bond Market

An efficient and well-functioning government bond market underpins operations in the entire financial market of a country. Generally, the “risk-free” status of government bonds and their relatively higher liquidity can benchmark other fixed income securities in the same currency, facilitate hedging positions held in other markets, and foster efficient resource allocation.2

Moreover, because governments are considered the most creditworthy borrowers in an economy, government bond yields are by extension seen as proxies for nominal risk-free rates. Yield spreads of other debt—including local government and private corporate debt relative to the levels at which governments borrow— otherwise known as “risk spreads,” are often used to gauge the market’s assessment of the creditworthiness of other borrowers. Capital market participants also rely on the government yield curve to assess the cost of funds on different time horizons. For this, the bond market can be a useful long-term asset and risk management tool.

The latter is important for China as it further liberalizes its economy. A developed capital market would unburden a banking system that has thus far provided the bulk of financing to borrowers. A deep capital market would also expand the government’s ability to fund large infrastructure projects and urbanization programs, as well as meet greater social spending needs, such as for an aging population.

A well-developed government securities market can also boost the rest of the bond market by enhancing overall bond market liquidity. Government securities generally offer a range of maturities and are widely traded, facilitating construction of a risk-free yield curve. The most recently issued bonds—so-called on-the-run securities—are often more liquid than other, older bonds. In addition, deep, liquid repurchase (repo) and derivatives markets allow investors to take speculative positions on sovereign debt and thus reveal expectations on future interest rate movements—including growth and inflation outlooks, as well as government funding needs—enhancing bond market liquidity.

A liquid bond market, in turn, can provide useful information about macroeconomic developments. And the changing slope of the government yield curve and interest rate movements can help gauge investor expectations of macroeconomic fundamentals. A rich literature also shows that, empirically, the slope of the yield curve offers useful insights into investor expectations for growth and inflation prospects. For example, an inverted yield curve—where shorter rates are higher than longer rates—tends to be associated with greater likelihood of recession (or default); and a steeper nominal yield curve suggests expectations of faster growth and, possibly, rising inflation.

The efficiency of the bond market is also a critical part of indirect monetary policy transmission and implementation. Even in a bank-dominated financial system, the government bond market is important for transmitting the cost of banking system reserves—as influenced by policy decisions about interest rates and banking system reserves—and of the cost of commercially borrowed funds (given the dominant role of banks in the government bond market). These prices should then transmit directly to bond yields for subnational governments and corporations, subject to movement in their specific credit risk spreads. In turn, these changes influence banks’ decisions over lending, loan rates, and more broadly, other interest rates in the economy.

But the usefulness of a government yield curve is not determined by fiat; it depends critically on the efficiency with which bond prices adjust to new information. This process, including the movement of rates toward their equilibrium values, relies on market participants’ willingness to use government securities as reference rates, including their ability to act on their views and the usefulness of these sovereign bonds as hedging instruments. A liquid government bond market therefore is crucial to efficient capital allocation and information gathering, as much of the analysis and pricing activity in the bond markets revolve around the government yield curve.

Cassola and Porter (2011) have already found that China’s bond yields (sovereign and corporate) do contain considerable information about the state of the economy and seem to play a role in an emerging monetary policy transmission channel, although they are not fully efficient, given regulation, liquidity, and segmentation issues. Since then, much has happened: China’s financial market has deepened and is further integrating gradually with the global financial system.

Yet, questions of efficiency and concern for lack of liquidity in China’s bond market remain. Although, as noted, the central government bond market has expanded to become one of the largest globally, liquidity remains an issue amid limited investor participation. A propensity for the largest investors to hold to maturity and limited market liquidity beyond the 10-year sector have led many to doubt whether the yield curve can accurately reflect market conditions.

Meanwhile, despite formal interest rate liberalization, competition among banks in deposit taking and lending remains somewhat constrained. Beyond the central government bond market, well-connected state-owned enterprises still borrow at favorable rates (see Chapter 4 on credit bonds), and local governments and many large, private corporations issue debt with limited risk spreads relative to central government bonds due to perceptions of implicit guarantees and backstops by governments. Still, the Chinese government bond market is worthy of study. Its absolute size and growing linkage with the macroeconomy and gradually with the global financial system suggest that its developments are likely to have greater spillover and spillback between the domestic economy and the rest of the world (see Chapter 2 on China’s bond and global financial markets).

This chapter addresses the following questions: Are Chinese government bond yields informative about macroeconomic developments? What are the macrofinancial interactions between real variables and bond yields? Have recent market developments and structural reforms strengthened transmission channels? Have financial prices become more informative signals of business cycle dynamics to play a bigger role in allocating credit?

The chapter finds that since the 2008–09 global financial crisis, both domestic monetary policy and global financial conditions have had an increasing impact on the level, slope, and curvature of the yield curve, likely boosted by China’s growing nonbank and shadow banking sectors as the government launched large stimulus programs to support growth following the global financial crisis. In addition, the yield curve provides predictive power for real activities proxied by industrial production. Meanwhile, the monetary policy stance, proxied by the 7-day repo rate and the benchmark lending rate, has a significant and persistent impact on short-dated central government bonds, suggesting growing linkage between monetary policy and bond market pricing. Lastly, the transmission to inflation is limited, likely reflecting China’s flattening Philips curve.3

Overview of China’s Government Bond Market

China’s bond market has expanded sizably, from 17 percent of GDP in 2000 to 90 percent in 2017. Bonds in 2017 made up about one-third of total financing, from less than 10 percent in 1999 (Figure 3.1). Meanwhile, the importance of bank lending has been declining, with its share in total financing falling from 70 percent in 1999 to less than 50 percent in 2017.

Figure 3.1.China’s Financial Structure, 1999 and 2017

Sources: People’s Bank of China; and authors’ calculations.

This expansion in bond financing has been largely driven by relaxed policy on bond issuance and growing financing needs. For example, simplification of the approval procedure has boosted corporate bond issuance in the onshore market. In the public sector, gross issuance of local government bonds has increased markedly since the implementation of the debt-for-bond swap program, while bond issuance by the Ministry of Finance and policy banks (notably, the China Development Bank, the Export-Import Bank of China, and the Agricultural Development Bank of China) has continued to increase to support government finance and infrastructure investment (Figure 3.2).

Figure 3.2.Public Sector Bond Issuance, 2000–17

(Trillions of renminbi)

Sources: WIND Economic Database (www.wind.com.cn); and authors’ calculations.

As the size of the bond market has grown, the share of bonds from government entities has fallen relative to the total outstanding. Bonds and fixed income securities issued by central and local governments, policy banks, and the central bank fell from 78 percent in 2010 to 56 percent in 2017 (Figure 3.3). The decline is largely due to a suspension of central bank bill issuance as the PBC changed its tools for managing liquidity in the banking sector. Meanwhile, the share of private sector bonds increased to 45 percent of the total outstanding, thanks to active issuance of corporate bonds, bank negotiable certificates of deposit, and other types of fixed income securities such as convertible bonds and asset-backed securities.

Figure 3.3.Outstanding Bonds by Issuer, 2010 and 2017

(Percent)

Sources: WIND Economic Database (www.wind.com.cn); and authors’ calculations.

Currently, the liquidity of CGBs remains limited and unevenly distributed. In the more mature markets, such as the United States, most transactions—and by extension, private sector bond pricing references—take place in the intermediate-dated sector of the yield curve, such as 5- and 10-year bonds. By comparison, most transactions in CGBs are concentrated in the short end of the yield curve (Figure 3.4). For example, more than half of turnover is located in the 3-year or shorter segment. For longer-dated CGBs, liquidity is concentrated at the 4- and 5-year as well as the 9- and 10-year segments of the yield curve. Relatively thin liquidity in longer-dated bonds suggests that the slope of the yield curve could be largely driven by movements in short-dated instruments and the information content of longer-dated bonds could be limited.4

Figure 3.4.Central Government Bond Daily Turnover, by Maturity, 2011, 2014, and 2017

(Billions of renminbi)

Sources: WIND Economic Database (www.wind.com.cn); and authors’ calculations.

Note: Y = year.

Liquidity in the market for policy bank bonds, which are also considered risk free, is deeper by comparison; this deeper pool of trading liquidity has offered greater information efficiency in policy bank bonds.5 Specifically, liquidity here has been boosted by more frequent reopenings of on-the-run papers, helping to enhance the benchmark status of these securities and boosting trading activity. Based on evidence from the policy bank bond market, analysis by International Monetary Fund (IMF) staff suggests that improving trading liquidity is key to enhancing the CGB market’s information efficiency and its policy transmission mechanism.

Constructing and Dissecting the Different Yield Curve Components

There is a large amount of literature about the power of the slope (and by extension the curvature) to predict economic activity and inflation. The transmission mechanism largely involves monetary policy and inflation expectations (Afonso and Martins 2010). Bond yields tend to rise in response to stronger growth and higher inflation. Meanwhile, a flat or humped yield curve is generally associated with an uncertain economic outlook. Specifically, an inverted yield curve has historically been associated with recessions (Bauer and Mertens 2018).

To better understand the structure of China’s government yield curve and, in turn, its interaction with macroeconomic developments, this section decomposes the central government yield curve into three factors: level, slope, and curvature. The identification of these factors helps assess the linkage between yield curve and macroeconomic developments.

The identification exercise relies on the three-factor Nelson-Siegel model (Diebold and LI 2006).6 This class of factor model is one of the most popular used by academics and market participants. It is capable of replicating a variety of stylized facts from empirical yield curves. Specifically, the loading parameter λ can be varied over time; this variable, which determines the relative factor loadings, has been found to increase dramatically, becoming volatile ahead of recessions and declining afterward in both its level and volatility (JIAO and MA 2017). The factor loadings for level, slope, and curvature—each set to capture movements in these three factors—also afford users flexibility in reproducing a range of yield curve shapes, including changes in overall yield levels, as well as movements in long and short rates that determine the slope and curvature of the yield curve.

Data and Methodology

Data on the government yield curve are from the WIND database for both interbank and exchange traded bonds;7 these are monthly observations from 2002. Using the Nelson-Siegel model, the CGB yield curve is approximated by the following three factors:

The parameters of the defined yield curve are β1t (level), β2t (slope), and β3t (curvature). Their respective loadings are given as 1, (1eλ,τλ,τ),and(1eλ,τλ,τeλ,τ).

  • The first factor, with a loading of one, represents the long-term level of interest rates as an increase in β1t increases all yields equally, changing the level of the yield curve (Figure 3.5).

  • Loading for the second factor, as it declines from 1 to 0, reflects a short-term factor and is interpreted as negative of the “slope”: an increase in β2t raises short yields more than long yields, thus changing the slope of the yield curve. Here, a negative slope factor indicates a steep curve (that is, short rates are lower than long rates).

  • Loading for the third factor, starting at 0, increases before converging back to 0 at long maturities, represents a medium-term factor, and is interpreted as curvature. Since increases in β3t have minimal effects on long and short rates, an increase in β3t will increase the curvature of the yield curve.8

Figure 3.5.Estimated Factor Loadings, by Tenor

(Coefficient)

Source: Authors’ calculations.

Note: M = month; Y = year.

Results

Using estimation results, the implied fitted curve and the actual yield curve are close, indicating overall good fit provided by the model (Annex Figure 3.1.1). A robustness test using the most-liquid tenors—bills out to the 10-year note— offers qualitatively similar results (Annex Figure 3.1.2).

The time series of the three latent yield-curve factors—level, slope, and curvature—offer the following insights on China’s central government bond yield curve (Figure 3.6):

Figure 3.6.Estimated Factors for the Central Government Bond Yield Curve, 2002–17

(Percent)

Source: Authors’ calculations.

Level: There is some variability in the level of the yield curve over time. Higher rates—particularly at longer tenors—were seen earlier in the sample (around 2003), while, more recently, rates were lower by comparison (Annex Figure 3.1.3). That said, overall, yield levels were mostly stable since the global financial crisis, with a few cyclical swings seen since 2013.

Slope: The slope factor shows the typical pattern of a steep yield curve (that is, negative value), though it does occasionally flatten (approaching zero), including periods before the global financial crisis as well as over brief periods in more recent years. Notably, the dramatic flattening of the slope around the global financial crisis and its subsequent steepening are developments in line with other bond markets.

Slope changes matter, as they offer greater insight into yield movements. The short end of the curve exhibits greater variability relative to the longer end (Annex Table 3.1.1), likely reflecting reduced liquidity for longer-dated CGBs and the segment’s limited responses to changes in macro data, financial conditions, and policy.

Curvature: Findings here offer insight into yield movements across the three segments of the yield curve—the short end, proxied by the 1-month bill; the mid section, proxied by the 10-year note; and the long end, proxied by the 30-year note. The curvature factor is often negative, suggesting that the curve is usually upward sloping (that is, the yield on short-dated paper is lower relative to the middle of the curve, which is lower compared to the longer-dated paper; Annex Tables 3.1.2 and 3.1.4). Not surprisingly, it is also positively correlated with slope, given a steep yield curve observed generally.

Still, the curvature factor exhibits slightly different variability compared to the slope, suggesting that the different segments of the yield curve react differently to the same information (Annex Tables 3.1.2 and 3.1.4); again, this is not a surprise. Inflation and growth shocks, as documented by the literature, tend to have varying impacts on different segments of the yield curve. A negative growth shock tends to lower longer-dated yields more than short rates, while positive inflation shocks tend to lift short rates more than long rates, likely in anticipation of tighter monetary policy in the immediate future. Apart from this, limited trading liquidity in the intermediate-dated and longer-dated securities may have played a role in the CGB’s yield curvature. Indeed, longer-dated securities have narrower standard deviations than shorter-dated paper; this suggests that the information content in the longer-dated securities may be limited (Annex Table 3.1.1). Robustness tests using the most liquid tenors offer qualitatively similar results (Annex Figure 3.1.4).

Macrofinancial Linkage

To assess the interaction between the CGB yield curve and the macro economy, this section first looks at simple correlations between the estimated factors and macro variables as a starting point. It then uses vector autoregression to analyze the impact of shocks to monetary policy, economic activity, and global financial conditions on the yield curve. Finally, it estimates the extent to which shocks to the yield curve could be transmitted to the real economy.

Simple Correlations

Simple correlation suggests that the yield curve factors are correlated with monetary policy, macro indicators, and global financial conditions, though the degree of correlation varies (Table 3.1). Notably, the 7-day repo rate and the 10-year Treasury yield are both highly correlated with the slope factor. A closer look at these variables offers the following insights:

The level factor was correlated with movements in industrial production growth —a proxy for GDP growth. As China’s growth slowed after the boom years in the early 2000s, yield levels fell generally, though there were cyclical swings. Since late 2016, amid an economic upswing and recovery in industrial production— prompted in part by a rebound in producer price inflation—yields have increased (Figure 3.7).

Figure 3.7.Central Government Bond Level Factor and Industrial Production Growth, 2002–17

(Percent, left scale; year-over-year percent, right scale)

Sources: CEIC; and authors’ calculations.

Table 3.1Correlation Matrix of Central Government Bond Yield Curve Factors and Macro and Financial Variables
12345678
1. Level factor-
2. Slope factor0.72
3. Curvature factor-0.57-0.56
4. Industrial production growth0.450.50-0.22
5. Consumer price index0.26-0.180.170.30
6. 7-day repurchase rate-0.15-0.670.28-0.280.44
7. 10-year Treasury yield0.380.50-0.310.760.01-0.43
8. VIX0.030.23-0.13-0.01-0.05-0.160.03-
Sources: Bloomberg L.P.; CEIC; and authors’ estimates.Note: VIX = Chicago Board Options Exchange Volatility Index.
Sources: Bloomberg L.P.; CEIC; and authors’ estimates.Note: VIX = Chicago Board Options Exchange Volatility Index.

The high negative correlation between the repo rate and the slope suggests that short rates play an important role in the slope of the yield curve. This is not surprising. Repo rates—and policy rates by extension—affect the level of short rates more disproportionately. And, to the extent that inflation affects the level of short rates—compared to the impact of growth on longer-dated yields, for example— an increase in the consumer price index (CPI) is generally associated with a flattening of the slope (that is, the slope factor turns less negative); this correlation is in line with observations in other countries. During periods of high inflation, policy rates rise, lifting short rates and flattening the yield curve correspondingly. During the global financial crisis, weakened economic activity, falling inflation, and the corresponding decline in policy rates weighed on short rates and steepened the yield curve (that is, the slope factor became more negative). In more recent periods, while inflation has remained contained, the CGB slope has flattened as regulatory tightening has pushed up short rates (Figure 3.8).

Figure 3.8.Central Government Bond Slope Factor and Consumer Price Index Inflation, 2002–17

(Percent, left scale; year-over-year percent, right scale)

Sources: CEIC; and authors’ calculations.

Linkages between Yield Curve Factors and Macroeconomic Variables

To assess the interaction between domestic macro conditions, global financial conditions, and the central government yield curve, this analysis estimated a set of vector autoregressions. Domestic macro variables include industrial production growth—a proxy for GDP growth—and CPI and producer price index (PPI) inflation. The monetary policy stance in China is proxied by the 7-day repo and benchmark lending rates. Global financial conditions are proxied by the 10-year US Treasury yield and the Chicago Board Options Exchange Volatility Index (VIX). The full data sample runs from January 2002 to December 2017 and is separated into two subsamples (2002:M1–2007:M12 and 2008:M1–2017:M12) to better distill responses in the periods before and after the global financial crisis. For ease of interpretation, the sign of the slope factor was inverted—a positive sign signals a steepening of the yield curve (that is, long rates increase more than short rates), while a negative sign indicates a flattening of the curve.

A large body of literature offers explanation for the possible channels of transmission between macro variables and the yield curve. For example, Diebold, Rudebusch, and Aruoba (2006) note that, in the United States, Treasury yields react to macroeconomic developments by anticipating the Federal Reserve’s decision, or the Fed could be setting its federal funds target by reacting to yield developments. In China, it is possible that the greater use of open market operations in recent years has increased the economy’s sensitivity to short rates—as evidenced by industrial production’s response to increases in short rates and a steepening of the slope—and by extension, changes in the slope of the government yield curve.

Several questions were considered and revealed complex dynamics between macroeconomic variables and yield curve factors:

  • Macroeconomic variables’ response to yield curve shocks, including industrial production, inflation, and the 7-day repo and benchmark lending rates

  • Yield curve responses to macroeconomic shocks

  • Yield curve responses to changes in monetary and liquidity conditions

  • Yield curve responses to shocks to the 10-year Treasury yield in the United States as well as the VIX to gauge the impact of greater market integration with the rest of the world.

Responses of the Macroeconomic Variables to Yield Curve Shocks

In general, macro variables’ sensitivity to yield curve shocks increased in the period after the global financial crisis, suggesting a growing relationship between the yield curve and economic developments since then. The most notable impacts are seen in responses to changes in the level factor:

  • Level: Increases in the level factor were associated with notable and persistent declines in CPI both before and after the global financial crisis. In the pre–global financial crisis period, a positive shock to the level factor leads to lower CPI inflation four months later; by comparison, the decline in CPI was immediately after the global financial crisis (Annex Figure 3.1.5). The significant role of the level factor is underscored by the variance of the errors in forecasting CPI. At the eight-month horizon, surprises to the level factor explain roughly 12 percent of such variance and stabilize at around that level thereafter in the period after the global financial crisis (Annex Table 3.1.5).

  • Slope: The slope factor offered limited predictive power for CPI in periods both before and after the global financial crisis. It does offer tentative evidence that increases in the slope of the yield curve lowered CPI roughly half a year later. Meanwhile, a steeper curve leads to an increase in industrial production growth three months later in the post–global financial crisis period, with persistent impacts.

Response of Yield Curve Factors to Macroeconomic Shocks

Yield curve factors did not seem to respond to macro shocks before the global financial crisis, but became more responsive to both industrial production and inflation shocks after the global financial crisis. Specifically, an increase in CPI leads to an immediate flattening of the yield curve after the global financial crisis, with the impact persistent through the one-year horizon (Figure 3.9). For PPI— to reflect the possibility that PPI may be more representative of inflationary pressure during the years of export-oriented, manufacturing-heavy production— the impact is similar on the slope factor, while somewhat larger on the curvature factor compared to CPI (Figure 3.10). Meanwhile, a positive shock to industrial production lifts the level factor with a four-month lag (Annex Figure 3.1.6).

Figure 3.9.Response of Central Government Bond Slope Factor to Inflation Shock, after the Global Financial Crisis

(Percentage points)

Source: Authors’ calculations.

Note: The period after the global financial crisis refers to January 2008 to December 2017.

Figure 3.10.Response of Central Government Bond Curvature Factor to Inflation Shock, after the Global Financial Crisis

(Percentage points)

Source: Authors’ calculations.

Note: The period after the global financial crisis refers to January 2008 to December 2017.

The findings on the slope and curvature added nuance to insights into the movements of the yield curve. The flattening of the yield curve in response to inflation increases after the global financial crisis (both PPI and CPI) was driven more by the intermediate sector, as evidenced by the steepening of the curvature, followed by short rates, evidenced by a flattening of the curve. That said, movements in the middle of the curve were relatively short lived, with the impact fading within four to six months, while a flattening of the curve was persistent out to a year.

Response of Yield Curve Factors to Monetary Policy Stance

Seven-day repo rate: The impact of the repo rate—a proxy for the monetary policy stance—on yield curve factors is immediate, significant, and more durable in the period after the global financial crisis, suggesting an increase in the linkage between the repo rate and the government yield curve. Specifically, increases in the repo rate raise yield curve levels almost immediately, though this effect fades after four months. Meanwhile, increases in the repo rate have a significant and persistent impact on the slope of the curve, with the yield curve flattening in the following 12 months. These two findings suggest that shocks to the repo rate have a durable impact on short-term yields, underscoring the growing linkage between these two rates after the global financial crisis (Annex Figure 3.1.6).

Benchmark rate: The one-year benchmark lending and deposit rate was used as the policy rate by the central bank for many years before the gradual switch to the 7-day repo rate (Box 3.1). To better capture the dynamics during these years, the analysis looked at the transmission from benchmark rate to yield curve and compared it with that of the repo rate. Interestingly, the transmission is much stronger and persistent for the repo rate in the period after the global financial crisis, with the difference most evident in the slope factor. This suggests that the 7-day repo rate has become more effective in guiding market expectations (Figure 3.11).

Figure 3.11.Response of Central Government Bond Slope Factor to Interest Rate Shock, after the Global Financial Crisis

(Percentage points)

Source: Authors’ calculations.

Note: The period after the global financial crisis refers to January 2008 to December 2017.

Response of the Central Government Bond Yield Curve to Global Financial Conditions

China’s growing financial market integration with the rest of the world suggests increasing spillover of global financial conditions into China. Using the 10-year US Treasury note as a proxy for benchmark funding conditions in the United States and the VIX as a proxy for global market volatility, findings from the period after the global financial crisis suggest that shocks to the 10-year Treasury yield have a significant and persistent impact on CGB level and slope, while the impact of VIX fades within two to three quarters (Annex Figure 3.1.7).

  • Ten-year Treasury yield: Before the global financial crisis, there is little linkage between the CGB curve and the 10-year US Treasury yield, given China’s limited integration with international markets. After the global financial crisis, increases in the 10-year Treasury yield are associated with an immediate and persistent impact on CGB levels and a steepening of the slope, suggesting greater correlation with longer-dated CGB notes. Meanwhile, in response to increases in the 10-year Treasury yield, the CGB curve steepened. Notably, the US 10-year yield can explain about 7–9 percent of variation in the CGB level and slope factors after the global financial crisis (Annex Table 3.1.7).

  • VIX: Similar to the 10-year Treasury yield, the VIX does not have a significant impact on China’s yield curve in the earlier years. After the global financial crisis, shocks to the VIX raise overall yield levels and steepen the yield curve, though these impacts fade after 8 to 11 months. These findings suggest that although an increase in volatility raises the CGB term premium, its impact is relatively short lived. Still, the VIX explains about 40 percent of variation in the slope factor in recent years, suggesting a close link between China and global financial markets.

Domestic versus External Shocks: Comparison of US Treasuries and Policy Bank Bonds

Analysis of the linkages between the CGB yield curve and macro conditions— both domestic and external—uncovered the following:

  • Sensitivity to macro shocks increased following the global financial crisis: All the curve factors evidenced greater response to macro shocks in the period after the global financial crisis. Notable increases in slope and curvature sensitivities suggest that the short and intermediate sectors of the curve are becoming more responsive to macro developments.

  • Domestic variables matter more in affecting yield curve movements: Both the level and curvature factors are more responsive to domestic developments than to external shocks.

  • External factors are increasingly more influential in CGB movements: External shocks (VIX and US Treasury yield) are increasingly driving the change in the slope factor, and are even more important than domestic factors (the sum of contributions from the 7-day repo rate, industrial production, and inflation) in the period after the global financial crisis (Figure 3.12). This likely reflects the magnitude of global financial shocks in this period, as indicated by sizable volatility in the VIX.

Figure 3.12.Contributions from External and Domestic Shocks to Variance of Forecast Error of Yield Curve Factors, before and after the Global Financial Crisis

(Percent, 12-month average)

Source: Authors’ calculations.

Nonetheless, the overall impact of yield curve movement on macro conditions remains limited compared with advanced economies. For example, the impulse response of inflation to a change in the slope factor is three times larger in the United States than in China (Figure 3.13). This likely reflects that China remains a bank-credit-driven economy, and room is significant for expanding the role of the bond market. Moreover, still-limited market liquidity—including secondary market trading—has hindered the efficiency with which CGBs incorporate new data and policy changes.

Figure 3.13.Response of Inflation to Shock in Government Yield Curve Slope Factor, after the Global Financial Crisis

(Percentage points)

Source: Authors’ calculations.

Comparisons against policy bank bonds offer greater insight into the role of trading liquidity in the information content of the bonds. Policy bank bonds are close analogs to CGBs. Both securities are considered “risk free” and are backed by the central government. The key difference between these two stems from their trading liquidity. Benchmark policy bank bonds are reopened more frequently, allowing on-the-run securities to trade with greater liquidity (Box 3.2). This level of liquidity therefore allows policy bank bonds to become more responsive than CGBs to domestic and external developments.

Applying the same vector autoregression analysis for CGBs to these policy bonds, it is found that policy bank bonds have been more responsive than CGBs to domestic shocks during the period after the global financial crisis, particularly for the level factor; this suggests these bonds offer greater monetary policy transmission than CGBs (Figure 3.14). Specifically, policy bank bonds are more sensitive to changes in industrial production and inflation, likely reflecting their deeper liquidity and greater efficiency at reflecting data and policy developments.

Figure 3.14.Contributions from Domestic Shocks to Variance of Forecast Error of Yield Curve Factors, after the Global Financial Crisis

(Percent, 12-month average)

Source: Authors’ calculations.

Conclusions

Over the years, the Chinese government has introduced reform measures to improve liquidity in the domestic bond market. The number of primary dealers and market makers that facilitate the trading of central government and policy bank bonds has increased. After the inclusion of the renminbi in the special drawing rights basket, the authorities further opened the domestic bond market to foreign investors.

The PBC also introduced changes to its monetary policy operations. For example, the introduction of new tools such as the standing lending facility and the medium-term lending facility have increased central bank flexibility in managing banking system liquidity. This has resulted in less frequent use of the reserve requirement ratio and benchmark interest rates, but both of these continue to carry strong market signals and can be too blunt for policy fine-tuning.

In part due to these developments, the linkage between CGB curve factors, macro variables, and global financial market conditions is growing. Notably, despite concerns regarding limited liquidity and opportunities for price discovery, CGBs seem to anticipate changes in macroeconomic conditions, with the correlation more pronounced and durable after the global financial crisis, though overall predictive power remains weak.

Meanwhile, changes in the 7-day repo rate have a significant and persistent impact on short-dated CGBs, suggesting a growing linkage between the PBC’s liquidity management, which affects repo rates, and CGB pricing. Also notable is the close linkage between CGB factors and external market conditions, including the 10-year Treasury yield and the VIX.

Evidence from the policy bank bond market suggests that improving trading liquidity is key to boosting the CGB market’s information efficiency and its policy transmission mechanism. In the period after the global financial crisis, policy bank bond levels have been more responsive than CGBs to changes in industrial production and the 7-day repo rate, likely a result of greater trading activities in that market. Greater market liberalization, including more active use of repo trading and hedging activities, more market-based open market operations from the PBC, and integration with the global system, can improve market liquidity for CGBs. In turn, a more robust and liquid CGB market will boost the overall efficiency of the Chinese capital market, proving the crucial benchmark for credit pricing that the country needs as it looks to further reform its economy and improve credit allocation.

Box 3.1.China’s Monetary Policy Framework and Its Market Impact

The People’s Bank of China (PBC) policy arsenal falls broadly into three categories (HE, WANG, and YU 2015; Girardin, Lunven, and MA 2014): (1) quantity-based instruments; (2) price-based instruments including lending rates, deposit rates, and open-market operations; and (3) administrative window guidance.

The PBC’s quantity-based instruments toolkit revolves primarily around reserve requirements. By changing these, the PBC can effectively add or drain system liquidity, and the central bank has used changes in required reserve ratios to signal changes in its policy stance over the years.

That said, the PBC has also changed its reliance on different policy instruments. The required reserve ratio was its dominant policy instrument from the mid-2000s until around 2011. But it has increased the use of price-based, more market-oriented instruments; notably, open market operations began rising sharply in 2015. A significant strengthening of liquidity forecasting and management has enabled the greater reliance on price-based instruments.

Background

In the fall of 2017, China’s 10-year sovereign bond yield briefly climbed above 4 percent for the first time in over three years. Market observers noted that this move signaled two developments: (1) PBC monetary policy increasingly resembles a standard interest-rate-based framework (see Harjes 2017); (2) the bond market plays an important role in the transmission of monetary policy, as actual and expected adjustments in the PBC’s monetary policy stance are immediately reflected in longer-term bond yields. As the central bank’s operations have changed, focus has shifted to the efficacy of the transmission channel of these new instruments and their impact on economic activity and inflation in China.

MA (2017) finds that changes in China’s short-term interest rates have a noticeable impact on government bond yields of various maturities, but the transmission of short-term rates to longer-term bond yields is less effective in China than in other countries (India, the Republic of Korea, the United Kingdom, and the United States). He concludes that inadequate liquidity in segments of the bond market, underdevelopment of the market for derivatives, and restrictions on market access continue to weaken the interest rate transmission channel. Nevertheless, recent analysis (Harjes 2017) suggests a significant impact of changes in short-term interest rates on economic activity and, with some lag, on prices. These results are broadly in line with other studies, including Fernald, Spiegel, and Swanson (2014), who show that changes in Chinese interest rates have a substantial impact on economic activity and inflation, while other traditional measures of monetary conditions, such as shocks to M2, do not matter much any longer.

Evolution of China’s Monetary Policy Framework

China has witnessed rapid financial innovation and undertaken significant financial market liberalization and reform, transforming its financial sector over the past few years. Banks still play a major role in financial intermediation, but other nonbank financial institutions and capital markets have become more prominent as a source of credit and in the provision of financial services to Chinese firms and households.

Overall, the financial system has become more market based but also more complex. In response, the PBC’s operational conduct of monetary policy is relying increasingly on a standard, market-based system. While quantitative instruments still play a role, interest rates have become key instruments of monetary policy.

In late 2015, the PBC formally abolished the remaining constraints on bank lending and deposit rates, in theory to allow banks to set their own lending and deposit rates after the PBC previously had eliminated mandatory loan-to-deposit ratios and put less stress on credit quotas. Although money growth (M2) remains the official intermediate target (albeit for the first time, no quantitative number was published in the 2017 Government Work Plan presented to the National People’s Congress), the central bank has clearly deemphasized its importance and indicated on several occasions that it is now using the 7-day interbank rate to send policy signals (PBC 2016). To do so, it has set a corridor around the 7-day rate, which has de facto been defined by the PBC’s repo operations bid rate as the lower bound and the rate for the PBC’s standing lending facility accessible by banks as the upper bound (Figure 3.1.1). To cap rates at the upper bound, the central bank accepts a relatively broad pool of collateral (including bank loans) for funds borrowed at the standing lending facility.

Figure 3.1.1.Interest Rate Corridor, January 2013 through October 2017

(Percent)

Source: Authors’ calculations.

Note: 7d= 7-day; DR007 = interbank 7-day repo rate for depository financial institutions; OMO = open market operations; PBC = People’s Bank of China; repo = repurchase.

Since mid-2015, the PBC has narrowed the corridor significantly and enforced a structural liquidity deficit in the banking sector that prevents the interbank rate from falling below the PBC’s repo bid rate, as banks are competing for funds provided through open market operations.1 This has kept the seven-day rate within the corridor while the central bank has nudged it up as the economy gathered pace in the second half of 2016 and as concerns about rapid credit growth and rising leverage came to the fore.

Complementing this shift has been the central bank’s use of a range of liquidity management tools. After capital inflows and PBC reserve accumulation slowed and eventually turned into outflows and reserve sales in 2014, the rising demand for domestic liquidity had to be met with other policy tools: short-term open market liquidity operations, standing lending facilities, medium-term lending facilities, and pledged supplementary lending (Figure 3.1.2). With these tools, the PBC injected and occasionally withdrew cash at different rates and for different durations and improved its ability to manage daily cash levels. High volatility in money market rates, once a regular occurrence, has been reduced significantly (Figure 3.1.3).

Figure 3.1.2.The People’s Bank of China’s Outstanding Liquidity Tools, 2013–18

(Billions of renminbi)

Source: CEIC.

Figure 3.1.3.Fixed Income Instruments: 7-Day Repurchase Rate and 2-and 10-Year Central Government Bond Yields, 2013–18

(Percent)

Source: WIND Economic Database (www.wind.com.cn).

Note: DR007 = interbank 7-day repurchase rate for depository financial institutions.

Transmission of Monetary Policy

The interest rate channel is a key channel for transmission. A policy-induced increase in the short-term nominal interest rate leads first to an increase in longer-term nominal interest rates, as investors act to arbitrage away differences in risk-adjusted expected returns on debt instruments of various maturities, as described by the expectations hypothesis of the term structure. When nominal prices are slow to adjust, these movements in nominal interest rates translate into movements in real interest rates as well. Firms, finding that their real cost of borrowing over all horizons has increased, cut back on their investment expenditures. Likewise, households facing higher real borrowing costs scale back purchases of homes, automobiles, and other durable goods. Aggregate output and employment fall, putting downward pressure on inflation.

While an interest-rate-based monetary policy reduces the importance of money demand in the transmission of policy actions to the economy, it raises the prominence of the role played by the term structure of interest rates. Under the expectations hypothesis of the term structure, long-term nominal interest rates depend on expectations of future nominal short-term interest rates. These in turn crucially depend on expectations about future monetary policy (Walsh 2010). These are influenced by various factors, including economic activity, the output gap, the exchange rate, and inflation developments. However, long-term trends, such as demographics, saving patterns, and the future rate of technological progress that affect the so-called natural rate of interest also play an important role in long-term yields. They may also reflect global factors and should therefore be less correlated with changes in the monetary policy stance.

There is a significant pass-through from the PBC’s policy rates to the average bank lending rate. The historically close link between the PBC’s benchmark rate for bank loans and the average bank lending rate reflects that the PBC’s rates used to define the floor (abolished in 2013) for bank lending rates. Regressing the average lending rate on either the PBC benchmark rate or the (smoothed) interbank repo rate reveals a significant and high coefficient, of about 70 percent, for the benchmark rate, and about 30 percent for the repo rate. It is notable that the interbank rate spike in 2013 also pushed up the average bank lending rate as banks passed on the higher funding costs to protect their margins, notwithstanding the unchanged benchmark lending rate, signifying the increasing importance of the interbank repo rate.

Changes in policy rates also have a measurable impact on short-term government yields. The two-year government bond yield is an important determinant of longer-term and corporate bond yields and can serve as a measure of market expectations of monetary policy stance over the next couple of years. All policy rates are positively correlated with the two-year government bond yield. Bivariate vector autoregression analysis reveals an almost complete pass-through from PBC benchmark rate changes to government bond yields after about six months, but a lower (25 percent) pass-through for the interbank rate.

Against that background, a forward-looking and transparent monetary policy approach, clearly communicated with the markets and public about its plans, objectives, and policy decisions, now seems the most effective and efficient way for the PBC to conduct its policy.

1 In a structural surplus (for example, in case of quantitative easing or large unsterilized foreign exchange purchases), the interest rate on excessive reserves for banks flush with cash would become the effective lower bound.

Box 3.2.The Role of Policy Banks in China’s Bond Market

Bonds issued by three policy banks—the China Development Bank, the Export-Import Bank of China, and the Agricultural Development Bank of China—have played an important role in fostering interbank bond market development in China.

These banks primarily issue bonds to fund lending to the targeted sectors with policy mandates. For example, the China Development Bank lends to support infrastructure projects undertaken by local governments; the Export-Import Bank of China lends to promote trade and support the export-oriented manufacturing sector; and the Agricultural Development Bank of China provides funding to improve water conservancy and farming facilities in rural areas.

Given that the bonds issued by policy banks are backed by the central government, investors perceive these bonds as risk free. Banks purchase policy bank bonds largely for liquidity management, as the bonds can be used as collateral in the repo markets or borrowing from the People’s Bank of China (PBC). Some prefer investing in policy bank bonds to central government bonds (CGBs) given their relatively higher yields—as domestic investors must pay tax on their policy bank bond holdings while CGB coupons are tax free (Figure 3.2.1). Given the higher yield and virtually risk-free nature of policy bank bonds, the demand for these bonds has been growing. Total market size has been catching up with CGBs in recent years (Figure 3.2.2).

Figure 3.2.1.Yield of China Development Bank Bonds versus Central Government Bonds, 2010–18

(Percent per year)

Source: WIND Economic Database (www.wind.com.cn).

Figure 3.2.2.Outstanding Central Government Bonds and Policy Bank Bonds, 2010–17

(Trillions of renminbi)

Sources: WIND Economic Database (www.wind.com.cn); and authors’ calculations.

Policy bank bonds are more actively traded in the secondary market than are other types of public sector bonds. The average daily turnover of policy bank bonds is almost twice that of CGBs (Figure 3.2.3). A crucial factor that has supported market liquidity is the high level of liquidity for benchmark issues. Unlike CGBs, recently issued on-the-run policy bank bond papers can be reopened—and often are, thus deepening the trading liquidity of these papers. Moreover, the jump in bond default cases in 2016 heightened risk aversion, resulting in flight to quality by bond investors. Portfolio shift from high-yield corporate bonds to risk-free assets boosted the demand for policy bank bonds, and trading of this paper rose markedly (Figure 3.2.4).

Figure 3.2.3.Daily Turnover of Public Sector Bonds, 2010–17

(Billions of renminbi)

Sources: WIND Economic Database (www.wind.com.cn); and authors’ calculations.

Figure 3.2.4.Ratio of Annual Turnover to Outstanding, 2010–17

(Percent)

Sources: WIND Economic Database (www.wind.com.cn); and authors’ calculations.

Figure 3.2.5.Daily Turnover of Policy Bank Bonds, by Tenor, 2011, 2014, and 2017

(Billions of renminbi)

Sources: WIND Economic Database (www.wind.com.cn); and authors’ calculations.

Another factor that supported policy bank bonds trading liquidity is the opening up of the domestic bond market to foreign investors. Appetite for policy bank bonds, particularly longer-dated, higher-yielding bonds, was strong among foreign investors. As a result, the turnover of 10-year policy bank bonds surged in 2017, surpassing the turnover of shorter-term maturities, 1-year or below, the segment with highest turnover in the past (Figure 3.2.5).

Finally, the launch of a number of bond indices by investment banks that track the return on policy bank bonds, such as China Development Bank bonds, is also driving the increase in trading liquidity. Notably, most of the surge in turnover of policy bank bonds in recent years has been largely concentrated in bonds issued by the China Development Bank, given its more regular issuance schedule and higher turnover in the secondary market (Figure 3.2.6).

Figure 3.2.6.Daily Turnover of Policy Bank Bonds, by Issuer, 2011, 2014, and 2017

(Billions of renminbi)

Sources: WIND Economic Database (www.wind.com.cn); and authors’ calculations.

Policy bank bonds, reflecting risk-free status and higher liquidity, in practice provide another benchmark yield curve in China. The China Development Bank yield curve is highly correlated with the central government curve—0.6 for the level factor and 0.9 for the slope factor. Applying the same vector autoregression analysis for CGBs to these policy bank bonds, their level factors have been more responsive to domestic shocks since the global financial crisis (2008:M1–2017:M12). The decomposition of the variance of forecast errors shows that shocks from industrial production, inflation, and the 7-day repo rate in total contributed 25 percent of the variance of the level factor of the policy bank bond versus 14 percent for the CGB (Figure 3.2.7). For the slope and curvature factors, the responsiveness of the policy bank bond is also marginally higher than the CGB.

Figure 3.2.7.Contribution from Domestic Shocks to Variance of Forecast Error of Yield Curve Factors after the Global Financial Crisis

(Percent, 12-month average)

Source: Authors’ calculations.

Impulse response results show a similar picture. A one standard deviation shock to industrial production could lead to a 0.05 standard deviation increase in the level factor of a policy bank bond, while a one standard deviation shock to the repo rate would increase the level factor by 0.1 standard deviation after six months (Figure 3.2.8). In contrast, the impact of an industrial production and repo rate shock on the level factor of the CGB yield is relatively small and not statistically significant.

For the slope factor, both policy bank and central government bonds have a similar response to an inflation shock. A one standard deviation shock from consumer price index (CPI) inflation tends to flatten the slope of both yield curves as short-term yields moved up faster than long-term yields. The greater sensitivity suggests that policy bank bonds exhibit better monetary policy transmission, likely a result of their deeper market liquidity and more efficient incorporation of new data and policy changes.

Figure 3.2.8.Impulse Response of Central Government Bonds and Policy Bank Bonds to Domestic Shocks after the Global Financial Crisis

(Percentage points)
(Percentage points)

Source: Authors’ calculations.

Note: Vector autoregression with two lags and six variables including industrial production (IP) growth, consumer price index (CPI) inflation, 7-day repurchase (repo) rate, and the three yield curve factors of central government bonds (CGBs) versus policy bank bonds. Impulse response is estimated by the Cholesky method. Panels are shown in the following sequence: response of level, slope, and curvature of CGB yield curve to one standard deviation shock from industrial production growth, CPI inflation, and 7-day repo rate during the post–global financial crisis period; response of level, slope, and curvature of policy bank bond yield curve to one standard deviation shock from industrial production growth, CPI inflation, and 7-day repo rate during the post–global financial crisis period from January 2008 to December 2017.

Annex Figure 3.1.1.Estimated and Observed Central Government Bond Yield Curves

(Percent per year)

Source: Authors’ calculations.

Note: M = month; Y = year.

Annex Figure 3.1.2.Robustness Test: Estimated Central Government Bond Yield Curves Using 3-Month to 10-Year Yields

(Percent per year, average yield curve fitting, 3-month to 10-year)

Source: Authors’ calculations.

Note: M = month; Y = year.

Annex Figure 3.1.3.Observed Central Government Bond Yield Curves, 2002–17

(Percent per year)

Source: Authors’ calculations.

Annex Figure 3.1.4.Robustness Test: Estimated Factors for the Government Bond Yield Curve Factors Using 3-Month to 10-Year Yields

(Percent)

Source: Authors’ calculations.

Annex Table 3.1.1.Model Descriptive Statistics: Average Observed Central Government Bond Yields
MaturityMeanStandard DeviationMinimumMaximum
1-month2.260.760.724.98
2-month2.370.760.794.63
3-month2.400.750.824.60
6-month2.460.730.864.21
9-month2.500.720.874.15
1-year2.550.720.934.08
2-year2.740.701.204.35
3-year2.900.641.364.38
4-year3.050.611.674.47
5-year3.160.591.984.43
6-year3.290.582.174.61
7-year3.380.562.314.67
8-year3.460.562.374.88
9-year3.530.572.435.07
10-year3.580.572.495.24
15-year3.900.572.755.82
20-year4.060.572.945.94
30-year4.030.542.765.53
Slope1.780.66-0.513.27
Curvature-0.600.40-1.481.22
Source: Authors’ calculations.
Source: Authors’ calculations.
Annex Table 3.1.2.Model Descriptive Statistics: Estimated Factors
FactorMeanStandard DeviationMinimumMaximumρ(1)ρ(3)ρ(12)ρ(60]
Level4.400.633.266.680.440.450.150.10
Slope-2.050.89-4.05-0.500.320.510.250.18
Curvature-0.970.82-2.890.670.370.220.11-0.29
Source: Authors’ calculations.Note: ρ(1), ρ(3), ρ(12), and ρ(60) are sample autocorrelations at displacements of 1, 3, 12, and 60 months, that is, relationships between movements of observations at month (n) and the factor estimates.
Source: Authors’ calculations.Note: ρ(1), ρ(3), ρ(12), and ρ(60) are sample autocorrelations at displacements of 1, 3, 12, and 60 months, that is, relationships between movements of observations at month (n) and the factor estimates.
Annex Table 3.1.3.Model Descriptive Statistics: Yield Curve Residuals
MaturityMeanStandard DeviationMinimumMaximum
1-month-0.110.24-0.870.94
2-month-0.010.07-0.230.26
3-month0.000.04-0.150.16
6-month0.010.05-0.210.20
9-month0.000.07-0.260.22
1-year0.010.08-0.370.22
2-year0.020.08-0.300.17
3-year0.010.05-0.160.14
4-year0.010.04-0.090.16
5-year-0.010.03-0.090.10
6-year0.010.04-0.050.16
7-year0.000.02-0.060.06
8-year-0.010.03-0.060.07
9-year-0.030.04-0.130.11
10-year-0.040.07-0.160.15
15-year0.040.08-0.110.25
20-year0.070.07-0.100.25
30-year-0.090.36-2.030.31
Slope0.020.48-2.070.97
Curvature-0.100.19-0.720.81
Source: Authors’ calculations.
Source: Authors’ calculations.
Annex Table 3.1.4.Robustness Test: Model Descriptive Statistics: Estimated Factors for the 3-Month to 10-Year Curve
FactorMeanStandard DeviationMinimumMaximumρ(1)ρ(3)ρ(12)ρ(60)
Level4.360.832.947.440.880.550.300.10
Slope-2.021.11–4.84-0.100.870.530.220.27
Curvature-0.471.21-3.632.440.740.310.10-0.21
Source: Authors’ calculations.Note: ρ(1), ρ(3), ρ(12), and ρ(60) are sample autocorrelations at displacements of 1, 3, 12, and 60 months, that is, relationships between movements of observations at month (n) and the factor estimates.
Source: Authors’ calculations.Note: ρ(1), ρ(3), ρ(12), and ρ(60) are sample autocorrelations at displacements of 1, 3, 12, and 60 months, that is, relationships between movements of observations at month (n) and the factor estimates.
Annex Table 3.1.5.Decomposition of Forecast Error Variance of Consumer Price Index Inflation to Central Government Bond Yield Curve Shock, after the Global Financial Crisis
MonthCGB_LEVELCGB_SLOPECGB_ CURVATUREIPCPIREP0_7D
14.031.240.006.0288.720.00
25.732.331.7315.4974.460.25
38.323.424.6119.7563.360.54
410.143.827.5223.1554.500.87
511.233.8210.0925.7047.861.30
611.803.5612.1427.7842.951.77
712.023.2113.6929.5839.282.22
812.012.8914.8031.1836.492.63
911.852.6515.5632.6234.352.96
1011.612.5516.0433.9132.683.22
1111.312.6116.3135.0331.353.40
1210.992.8316.4135.9730.273.51
Source: Authors’ calculations.Note: Cholesky ordering: CGB_LEVEL; CGB_SLOPE; CGB_CURVATURE; IP; CPI; REPO_7D. 7D = 7-day; CGB = central government bond; CPI = consumer price inflation; IP = industrial production growth; repo = repurchase.
Source: Authors’ calculations.Note: Cholesky ordering: CGB_LEVEL; CGB_SLOPE; CGB_CURVATURE; IP; CPI; REPO_7D. 7D = 7-day; CGB = central government bond; CPI = consumer price inflation; IP = industrial production growth; repo = repurchase.

Annex Figure 3.1.5.Impulse Response of Industrial Production Growth and CPI Inflation to Central Government Bond Yield Curve Factor Shocks, before and after the Global Financial Crisis

(Percentage points)
(Percentage points)

Source: Authors’ calculations.

Note: Vector autoregression (VAR) with two lags and six variables including three central government bond (CGB) factors, industrial production (IP) growth, and consumer price index (CPI) inflation. Impulse response is estimated by the Cholesky method. Panels are shown in the following sequence: response of IP growth to one standard deviation of shock from level, slope, and curvature of CGB yield curve; response of CPI inflation to one standard deviation of shock from level, slope, and curvature of CGB yield curve. GFC = global financial crisis; pre-GFC = January 2002 to December 2007; post-GFC = January 2008 to December 2017.

Annex Figure 3.1.6.Impulse Response of Central Government Bond Factors to Industrial Production Growth, Consumer Price Index Inflation, and Repo Rate Shocks, before and after the Global Financial Crisis

(Percentage points)
(Percentage points)
(Percentage points)

Source: Authors’ calculations.

Note: Vector autoregression (VAR) with two lags and six variables including industrial production (IP) growth, consumer price index (CPI) inflation, 7-day repurchase (repo) rate (REPO_7D), and three central government bond (CGB) factors. Impulse response is estimated by Cholesky method. Panels are shown in the following sequence: response of level, slope, and curvature of CGB yield curve to one standard deviation shock from IP growth; response of level, slope, and curvature of CGB yield curve to one standard deviation shock from CPI inflation; response of level, slope, and curvature of CGB yield curve to one standard deviation shock from 7-day repo rate. GFC = global financial crisis; pre-GFC = January 2002 to December 2007; post-GFC = January 2008 to December 2017.

Annex Figure 3.1.7.Impulse Response of Central Government Bond Factors to US Treasury 10-Year Yield and Volatility (VIX) Shocks, before and after the Global Financial Crisis

(Percentage points)
(Percentage points)

Source: Authors’ calculations.

Note: Vector autoregression (VAR) with two lags and five variables including US Treasury 10-year yield (UST10Y), Chicago Board Options Exchange Volatility Index (VIX), and three central government bond (CGB) factors. Impulse response is estimated by Cholesky method. Panels are shown in the following sequence: response of level, slope, and curvature of CGB yield curve to one standard deviation shock from US Treasury 10-year yield; response of level, slope, and curvature of CGB yield curve to one standard deviation shock from VIX. pre-GFC = January 2002 to December 2007; post-GFC = January 2008 to December 2017.

Annex Table 3.1.6.Forecast Error Variance Decomposition: Changes in Central Government Bond Factors to Domestic and External Shocks, before the Global Financial Crisis
Variance Decomposition of CGB_LEVEL
MonthUST10YVIXREP0_7DIPCPI
11.432.920.931.950.03
23.661.582.363.981.13
34.731.041.796.001.97
44.861.321.807.213.27
54.341.542.488.494.67
63.861.713.729.506.28
73.531.864.9110.118.05
83.342.085.9310.469.81
93.262.346.8210.7611.37
103.222.647.6311.0812.71
113.222.968.3211.3913.87
123.233.308.9011.6414.91
Variance Decomposition of CGB_SLOPE
MonthUST10YVIXREP0_7DIPCPI
11.223.210.001.710.43
24.632.680.013.040.65
36.482.190.623.570.68
46.821.900.813.760.78
56.371.730.744.190.89
65.921.610.874.601.10
75.551.521.074.841.41
85.291.431.254.961.74
95.141.371.455.102.00
105.051.351.685.292.20
115.031.381.925.522.38
125.051.462.165.772.56
Variance Decomposition of CGB_CURVATURE
MonthUST10YVIXREPO 7DIPCPI
13.150.310.420.190.09
25.950.210.910.232.44
37.680.360.840.401.91
48.580.620.960.681.61
57.980.960.961.151.96
67.211.360.841.512.40
76.551.680.801.592.62
86.062.060.831.522.59
95.682.510.931.432.45
105.373.071.111.362.29
115.103.741.361.332.15
124.874.531.651.342.05
Source: Authors’ calculations.Note: Cholesky ordering: UST10Y; VIX; REP0_7D; IP; CPI; CGB_LEVEL; CGB_SLOPE; CGB_CURVATURE. Tables of variance decomposition are shown in the sequence of level, slope, and curvature of the CGB (central government bond) yield curve to shocks from US Treasury 10-year yield (UST10Y), VIX (Chicago Board Options Exchange Volatility Index), 7-day repurchase (repo) rate (REPO-7D), industrial production growth (IP), and consumer price index (CPI) inflation during the period January 2002 (2002:M1) to December 2007 (2007:M12).
Source: Authors’ calculations.Note: Cholesky ordering: UST10Y; VIX; REP0_7D; IP; CPI; CGB_LEVEL; CGB_SLOPE; CGB_CURVATURE. Tables of variance decomposition are shown in the sequence of level, slope, and curvature of the CGB (central government bond) yield curve to shocks from US Treasury 10-year yield (UST10Y), VIX (Chicago Board Options Exchange Volatility Index), 7-day repurchase (repo) rate (REPO-7D), industrial production growth (IP), and consumer price index (CPI) inflation during the period January 2002 (2002:M1) to December 2007 (2007:M12).
Annex Table 3.1.7.Forecast Error Variance Decomposition: Changes in Central Government Bond Factors to Domestic and External Shocks, after the Global Financial Crisis
Variance Decomposition of CGB_LEVEL
MonthUST10YVIXREP0_7DIPCPI
10.042.2313.290.422.76
20.472.5319.350.242.21
31.321.8820.070.402.16
42.102.7819.930.692.18
52.694.6119.770.872.15
63.255.9519.710.972.09
73.906.4419.651.072.03
84.686.4619.471.242.00
95.526.3619.181.512.00
106.356.2518.841.892.02
117.126.1418.522.342.04
127.796.0418.242.812.04
Variance Decomposition of CGB_SLOPE
MonthUST10YVIXREP0_7DIPCPI
10.811.464.660.731.82
20.400.664.180.472.74
30.626.724.360.514.25
40.7320.154.110.385.59
50.7831.493.530.526.23
60.9437.603.091.006.45
71.4040.112.881.556.44
82.2640.772.852.006.30
93.5740.592.962.266.08
105.2340.053.122.375.83
117.1239.383.282.395.58
129.1038.653.422.375.34
Variance Decomposition of CGB_CURVATURE
MonthUST10YVIXREPO 7DIPCPI
10.500.7518.424.350.57
21.960.4826.805.392.01
33.352.2724.695.802.18
44.684.3922.095.982.01
55.475.7220.576.081.87
65.756.1719.846.101.85
75.786.1819.596.051.94
85.726.1019.595.982.12
95.676.0719.715.942.33
105.636.0719.835.962.53
115.616.0819.926.052.68
125.616.0719.966.202.79
Source: Authors’ calculations.Note: Cholesky ordering: UST10Y; VIX; REPO_7D; IP; CPI; CGB_LEVEL; CGB_SLOPE; CGB_CURVATURE. Tables of variance decomposition are shown in the sequence of level, slope, and curvature of CGB yield curve to shocks from US Treasury 10-year (UST10Y) yield, Chicago Board Options Exchange Volatility Index (VIX), 7-day repurchase (repo) rate, industrial production (IP) growth, and consumer price index (CPI) inflation during the period January 2008 (2008:M1) to December 2017 (2017:M12).
Source: Authors’ calculations.Note: Cholesky ordering: UST10Y; VIX; REPO_7D; IP; CPI; CGB_LEVEL; CGB_SLOPE; CGB_CURVATURE. Tables of variance decomposition are shown in the sequence of level, slope, and curvature of CGB yield curve to shocks from US Treasury 10-year (UST10Y) yield, Chicago Board Options Exchange Volatility Index (VIX), 7-day repurchase (repo) rate, industrial production (IP) growth, and consumer price index (CPI) inflation during the period January 2008 (2008:M1) to December 2017 (2017:M12).
References

    AfonsoA. andM.Martins. 2010. “Level, Slope, Curvature of the Sovereign Yield Curve, and Fiscal Behavior.” ECB Working Paper 1276European Central BankFrankfurt.

    BauerM. andT.Mertens. 2018. “Economic Forecasts with the Yield Curve.” FRBSF Economic Letter 2018–07 (March 5)Federal Reserve Bank of San Francisco.

    CassolaN. andN.Porter. 2011. “Understanding Chinese Bond Yields and their Role in Monetary Policy.” IMF Working Paper 11/225International Monetary FundWashington, DC.

    DieboldF. andC.LI. 2006. “Forecasting the Term Structure of Government Bond Yields.” Journal of Econometrics130 (2): 33764.

    DieboldF.G.Rudebusch andS. B.Aruoba. 2006. “The Macroeconomy and the Yield Curve: A Dynamic Latent Factor Approach.” Journal of Econometrics131 (1–2): 30938.

    FernaldJ.M.Spiegel andE.Swanson. 2014. “Monetary Policy Effectiveness in China: Evidence from a FAVAR Model.” Journal of International Money and Finance49 (Part A): 83103.

    GashaG. and others. 2010. “On the Estimation of Term Structure Models and an Application to the United States.” IMF Working Paper 10/258International Monetary FundWashington, DC.

    GirardinE.S.Lunven andG.MA. 2014. “Inflation and China’s Monetary Policy Reaction Function: 2002–2013.” BIS Paper 77Bank for International SettlementsBasel.

    HarjesT. 2017. “A Market-Based Monetary Policy Approach for China.” Selected Issues Paper IMF Country Report 17/248International Monetary FundWashington, DC.

    HED.H.WANG andX.YU. 2015. “Interest Rate Determination in China: Past, Present, and Future.” Hong Kong Institute for Monetary Research Working Paper 042014Hong Kong SAR.

    JIAOA. andJ.MA. 2017. “The Predictive Power of Nelson-Siegel Factor Loadings for the Real Economy.” Working PaperUniversity of AlabamaSeptember.

    MAJ.2017. “Interest Rate Transmission in a New Monetary Policy Framework.” In Modernizing China edited by W.Raphael LamMarkusRodlauer andAlfredSchipke. Washington, DC: International Monetary Fund.

    People’s Bank of China (PBC). 2016. Quarterly Monetary Policy Reports. Beijing.

    WalshC. 2010. Monetary Theory and Policy3rd edition. Cambridge, MA: MIT Press.

Box 3.1 details the evolution of China’s monetary policy operations and their macroeconomic impact.

This risk-free status is relative to other bonds for the same country, such as corporate bonds, which carry additional credit risk.

A fat yield curve, in which longer-dated yields are roughly the same as short rates, tends to suggest impending recession. In China’s case, the interpretation is complicated by both a fattening Philips curve and limited liquidity at the longer end of the yield curve. In particular, the latter curtailed the predictive power of the central government bond curve, as discussed later in the chapter.

For more in-depth discussions on liquidity conditions across different yield curve segments, please refer to Chapter 12 on Trading in China’s Bond Market.

For more in-depth discussion on policy bank bonds, please refer to Box 3.2.

This exercise relies on an IMF term structure estimation tool; for more information, please see Gasha and others (2010).

WIND Economic Database (www.wind.com.cn).

The parameter λt controls both the exponential decay rate and the maturity at which the loading on β3t reaches its maximum.

    Other Resources Citing This Publication