## _wp11225

## Source details

**Canonical URL:** [_wp11225](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11225.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11225.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11225.pdf.json)

---

### Purpose and scope
- Objective: investigate the efficiency and functioning of China’s various bond markets, describe the current behavior of yields, their response to policy changes, their interaction across markets, and their impact on the aggregate economy.
- Specific research questions:
  - How does regulation of retail interest rates influence market determined bond yields?
  - Do the yield curves for different bonds interact?
  - What are the macro-financial interactions between real variables and bond yields in China?
  - Are Chinese bond yields efficient (arbitrage free)?
- Markets analyzed:
  - Interbank markets: IBCB (Central Bank paper), IBTSY (Treasury bonds), IBPBB (Policy Banks bonds), IBCAAA (AAA-rated Corporate bonds).
  - Exchange market: EXTSY (Treasury bonds traded on China’s two stock exchanges).
- Institutional context: fixed income markets large (around ¾ of GDP), relatively young, highly regulated; trading segmented between exchanges and an institutional interbank market; wholesale and foreign currency rates liberalized while retail interest rates remain regulated.

### Data and descriptive statistics
- Data sources and sample coverage:
  - Zero coupon yields from China Central Clearing and Depository.
  - Yield series: interbank and exchange traded treasury bonds (IBTSY), interbank traded PBC bills (IBCB), interbank Policy Banks bonds (IBPBB), corporate AAA bonds traded on interbank market (IBCAAA).
  - Sample coverage: month-end data from April 2006 to May 2010 for all series except PBC bills; PBC paper yields start in June 2007.
- Market shares and turnover:
  - As of end-May 2010, outstanding Treasury bonds and PBC bills accounted for 30.0 percent and 22.8 percent of outstanding bonds, respectively.
  - Transaction volumes: 9.6 percent and 40.0 percent, respectively, of cumulative bond market transaction volume (context: Treasury bonds and PBC bills cumulative transaction volume share reported in source excerpt).

### Methodology: yield-curve representation, estimation, and no-arbitrage testing
- Yield-curve model:
  - 3-factor Nelson-Siegel representation summarizing each yield curve into latent factors at each date: level (β1t), slope (β2t; reported as negative of slope factor in exposition), and curvature (β3t).
  - Lambda (λt) parameter governs exponential decay along maturities and maturity at which curvature factor maximizes impact.
- Estimation approach:
  - Two-step Diebold-Li style procedure with λ fixed over time: (1) for given λ estimate {β1t, β2t, β3t} by OLS and record residual sum of squares; (2) choose λ that minimizes residual sum of squares.
  - For US Treasuries Diebold and Li (2006) set λ = 0.0609 (30 months maximum curvature); paper estimates λ separately by curve.
- No-arbitrage testing approach:
  - Express any linear term structure model in matrix form y_t = a + B X_t + e_t.
  - Use estimated Nelson-Siegel factors to make estimation tractable, estimate a restricted affine no-arbitrage model conditional on NS factors, and test equality of statistical NS loadings with implied no-arbitrage loadings.
  - Null hypothesis tested maturity-by-maturity for four loadings: a, B^(1), B^(2), B^(3).
  - Bootstrap/block resampling procedure (Coroneo et al. (2008) style) used to generate distributions for no-arbitrage parameters and to derive confidence intervals.

### Estimated lambda and curvature timing (Table 4, preserved verbatim)
- Table 4. Estimated lambda by type of curve (rows preserved exactly)
  - Inte rbank Tre asury0.05036 Apr-06 -  May- 10123120
  - Ex change  Tre asury0.06528 Apr-06 -  May- 10106120
  - Inte rbank PBC0.2158 Jun- 07 - May- 10100.2536
  - Inte rbank Pol i cy Bank0.04540 Apr-06 -  May- 10123120
  - Inte rbank Corporate  (AAA)0.04540 Apr-06 -  May- 10106120

### Key empirical findings — pricing efficiency and arbitrage
- No-arbitrage tests and deviations:
  - The hypothesis that bond yield curves are arbitrage-free is rejected for all four markets examined; degree of deviation varies across markets.
  - Deviation from arbitrage-free conditions is largest at shorter maturities in the two largest bond markets.
- Market-specific deviations (preserved summaries):
  - Interbank Treasury bonds: NS loadings on slope and curvature deviate significantly from arbitrage-free factors at short maturities (less than two years) — weight on short-term factors too high; weight on middle-tenor too low.
  - Exchange-traded Treasury market: weight on long-run level factors and short-term factors too high across all maturities; middle-tenor (curvature) weight statistically consistent with NS.
  - Policy Banks bonds: short-term (slope) factor weight closest to arbitrage-free consistency; long-term level weight too low at short and middle maturities (less than two years); significant deviations at middle-maturities (around 4-6 years).
  - Central Bank paper (PBC bills): weights on all factors seem too high to be consistent with arbitrage-free pricing; maximum maturity 36 months with large deviations at those maturities.
- Liquidity and structural drivers:
  - Bid-ask spreads indicate interbank Treasury and Policy Bank markets are least liquid at very short maturities—consistent with deviations at those maturities.
  - Policy Bank bid-ask spreads rise faster (from a maturity of 3 years) and higher than in Treasury market—may explain deviations at those maturities.
  - Regulated retail interest rate structure (specified at 5 years) likely affects pricing efficiency at longer maturities.

### Yield-curve behavior, model fit, and factor properties
- Model performance:
  - Estimated errors (root mean squared error) typically small; largest errors at short-end PBC paper yields; corporate and Treasury largest errors at 2 year maturity; policy financial bonds largest errors at 12 month maturity.
  - Estimated errors show little significant autocorrelation beyond three months; estimated factors display significant autocorrelation over a few months.
- Shape and curvature:
  - Estimated λ coefficients imply maximum curvature at around 3 years for many curves (longer than US Treasury market); PBC and exchange Treasury markets have shorter maturities of maximum curvature.
  - Interbank and exchange Treasury curves cross around a median value of 12 months; crossings and non-concavity produce multiple crossing points at times.
- Market segmentation and arbitrage impediments:
  - Segmentation leads to inconsistent pricing and apparent profitable arbitrage opportunities between treasury bonds in the two markets.
  - Frictions in moving bonds across markets, limited liquidity, and wide bid-ask spreads impede exploitation of apparent arbitrage.

### Monetary policy transmission, VAR results, and macro-financial linkages
- VAR specifications used (preserved verbatim):
  - VAR with 4 lags and the following variables: 3 yield curve factors, monthly industrial value added growth (year-on-year), and CPI inflation (year-on-year), and regulated margins (benchmark lending – deposit rate) at 1, 3 and 5 years.
  - Alternative VARs used in analyses include two-lag VARs with six or seven variables (three IBCB factors and three IBPBB factors; or three IBPBB factors, 3-mth PBC bill rate, gryy, infyy, and corporate bond spreads).
- Transmission evidence:
  - Policy-induced changes in PBC (IBCB) yields significantly feed through to IBPBB factors: shocks to level of PBC curve have largest and most lasting impact on level of IBPBB; slope and curvature shocks to PBC affect slope and curvature of IBPBB.
  - One necessary element of a rate-based monetary transmission mechanism appears to be working/emerging in China: PBC rate changes propagate to other yields and macroaggregates.
- VAR-based impulse-response summaries (preserved timing/numeric relations where given):
  - Unexpected rises in the level factor of IBPBB yields dampen growth after about half a year.
  - A rise in the slope (lower near-term yields) raises growth after 3 months (not always significant) and raises inflation after 5 months.
  - Increase in curvature associated with temporary increase in growth which dies out after about a year.
  - Inflation strongly affected by unexpected changes in the curvature of IBPBB yields (medium-term factor).
  - Slope and curvature factors affect the 3 month PBC rate with opposite signs: slope negative, curvature positive.
  - The 3 month PBC rate responds to corporate spread shocks and inflation shocks, but little to growth shocks.
  - Unexpected changes in corporate spreads tend to depress both growth and inflation.
  - Changes in the 3 month PBC rate affect all factors in the IBPBB yields.
  - Growth does not respond significantly to unexpected changes in the PBC bill rate, whereas inflation and corporate spreads decline significantly and persistently after such changes.
- Cross-correlation findings (preserved verbatim notes):
  - Level factor leads growth by around +12 months (growth lags level by around one year); level factor leads inflation by around +20 months.
  - Slope factor: growth leads slope by around -5 months; slope leads inflation by around +2 months.
  - Curvature: growth leads curvature by around -1/-2 months; curvature leads inflation by around +6 months.
  - Corporate spread leads inflation by around +7 months and leads growth by around +2 months.

### Impact of regulated retail interest rate margins
- Regulated margins examined at 1, 3, and 5 year maturities (benchmark lending – deposit rate).
- VAR results for IBPBB:
  - Level factor: an unexpected temporary widening of margins tends to push the longer-term factor down (impact generally negative).
  - Slope factor: unexpected temporary widening of margins tends to push the slope down (flatten the yield curve) — impact generally significant and negative.
  - Curvature factor: unexpected widening of margins leads to a temporary increase in curvature.
  - Margins also affect macro: higher margins → lower growth and lower inflation.
  - Interpretation: narrowing (widening) of regulated margins associated with steeper (flatter) yield curves and lower (higher) medium-term rates; movements align with monetary loosening (tightening) effects on growth and inflation.

### Policy implications and recommendations
- Stage set for increased use of interest rate targeting as primary monetary policy tool: changes in PBC rates feed through to other yields and macro aggregates.
- Financial prices are informative signals of macro and financial conditions; further liberalization of financial prices would strengthen monetary transmission.
- Increasing disintermediation underscores urgency of moving towards indirect policy tools.
- Introducing some limited foreign participants into the interbank market may improve liquidity and pricing efficiency.
- Operational caution: exploiting arbitrage between interbank and exchange markets is unlikely to be profitable given complications in moving bonds between markets despite observed yield curve crossings.

### Concise summary of numeric and market facts preserved verbatim
- Markets covered: PBC bill market, segmented interbank and exchange traded Treasury Bonds market, Policy Financial Bonds market, AAA corporate bond market.
- As of end-May 2010: outstanding Treasury bonds = 30.0 percent; PBC bills = 22.8 percent of outstanding bonds.
- Transaction-volume shares reported: 9.6 percent and 40.0 percent (context: cumulative bond market transaction volume shares).
- Estimated lambdas and sample ranges (Table 4 rows preserved above).

*Source — _wp11225 - References (PDF chapter/section)_*

### References .............................................................................................................

### _wp11225 - References

### Major themes: purpose and scope
- Objective: investigate the efficiency and functioning of China’s various bond markets, describe the current behavior of yields, their response to policy changes, their interaction across markets, and their impact on the aggregate economy.
- Specific research questions:
  - How does regulation of retail interest rates influence market determined bond yields?
  - Do the yield curves for different bonds interact?
  - What are the macro-financial interactions between real variables and bond yields in China?
  - Are Chinese bond yields efficient (arbitrage free)?
- Markets analyzed:
  - Interbank markets: IBCB (Central Bank paper), IBTSY (Treasury bonds), IBPBB (Policy Banks bonds), IBCAAA (AAA-rated Corporate bonds).
  - Exchange market: EXTSY (Treasury bonds traded on China’s two stock exchanges).
- Institutional context: Chinese fixed income markets are large (around ¾ of GDP), relatively young, and highly regulated; trading is segmented between exchanges and an institutional interbank market; wholesale and foreign currency rates liberalized while retail interest rates remain regulated.

### Key findings and conclusions
- Pricing efficiency and arbitrage:
  - The hypothesis that bond yield curves are arbitrage-free is rejected.
  - Deviation from arbitrage-free conditions is largest at shorter maturities in the two largest bond markets.
- Impact of regulation:
  - Strong evidence that regulated retail interest rates significantly affect bond yields and the entire structure of yields.
  - Regulation of retail interest rates is identified as one likely cause of pricing inefficiencies.
- Informational content and transmission:
  - Bond yields in China are informative signals of macroeconomic developments and serve as leading indicators of growth and inflation.
  - Policy-induced changes in PBC yields significantly feed through to other yields and to macroeconomic aggregates, suggesting an emerging price-based monetary transmission mechanism despite continued non-price controls.
- Broader contributions:
  - Evidence relevant for developing a benchmark for pricing financial assets.
  - Documentation of substantial macro-financial linkages, a necessary condition for developing indirect monetary policy instruments.
  - Contribution to the study of emerging market local currency bond markets beyond sovereign yields (including quasi-sovereign and corporate bonds).

### Methodology: yield curve representation and estimation
- Yield curve model:
  - Use of a 3-factor Nelson-Siegel representation to summarize each yield curve into three latent factors at each date: level, slope (negative of slope factor), and curvature.
  - Interpretation of factors:
    - First factor (β1t): long-run level of interest rates (loading equals one as maturity → ∞).
    - Second factor (β2t): short-term factor (loading declines monotonically from 1 to 0 as maturity increases).
    - Third factor (β3t): medium-term or curvature factor (loading starts at 0, increases for intermediate maturities, then converges back to 0 at long maturities).
  - Lambda (λt) parameter: exponential decay rate along maturities for each factor; larger λ produces faster decay and governs the maturity that maximizes the curvature factor’s impact.
- Estimation approach:
  - Two-step procedure (Diebold and Li style, with λ fixed over time to simplify estimation):
    1. For a given λ estimate the three latent factors {β1t, β2t, β3t} by OLS regressing yields on maturity-dependent loadings and record the residual sum of squares.
    2. Choose λ that minimizes the residual sum of squares (procedure follows Diebold and Li (2006) recommendation to fix λ over time).
  - For US Treasury yield data, Diebold and Li (2006) set λ = 0.0609, which maximizes the impact of the medium-term factor at a maturity of 30 months.
- Use of estimated factors:
  - The three estimated latent factors at each date provide a parsimonious summary of the term structure, facilitating analysis of:
    - Interaction with regulated interest rates.
    - Cross-market and cross-maturity interactions.
    - Relationships with macroeconomic variables.
  - These representations also enable testing for arbitrage-free conditions.

### Empirical focus and structure of analysis
- Empirical analyses presented in the paper include:
  - Behavior of yields across markets and maturities.
  - Interaction between factors characterizing different yield curves.
  - Interaction between macroeconomic variables and yield-curve factors.
  - No-arbitrage tests for the estimated yield curves.
- Paper organization:
  - Statistical model and no-arbitrage testing methodology.
  - Description of PBC monetary tools and institutional features of China’s money and bond markets.
  - Data description.
  - Results on yields, factor interactions, macro-financial linkages, and no-arbitrage tests.
  - Conclusion summarizing implications for monetary policy and market development.

*Italic: Source — _wp11225 - References (PDF chapter/section)_*

### 2.      We repeat this exercise until a value of ߣ  that minimizes the residual sum of squares

### _wp11225 - 2.      We repeat this exercise until a value of ߣ  that minimizes the residual sum of squares

### B. Comparing Statistical and No-Arbitrage Yield Curves
- The Nelson-Siegel yield curve provides a convenient statistical representation of the yield curve but is not guaranteed to be arbitrage-free; arbitrage opportunities may exist across maturities.
- Procedure outlined follows Coroneo and others (2008): estimate the distribution of parameters (and factor loadings) in a (restricted) affine arbitrage-free term structure model conditional on estimated Nelson-Siegel factors; then test equality of statistical loadings with implied no-arbitrage loadings.
- Rationale: using estimated Nelson-Siegel factors makes estimation tractable and allows a direct statistical comparison under the null hypothesis that the yield curve is arbitrage free.

### Nelson-Siegel and No Arbitrage Models
- Any linear term structure model can be expressed in matrix form (equation (2)): yields vector y_t (N maturities) = a + B X_t + e_t, where X_t is a K-factor vector, a varies by maturity, and B is an N×K matrix of factor loadings.
- Nelson-Siegel representation in this matrix form:
  - a_Nelson-Siegel = [0 0...0 0]′ (equation (3)).
  - B_Nelson-Siegel specified via the Nelson-Siegel functional form for a constant ߣ (equation (4)).
- Gaussian no-arbitrage affine term structure models also take the form of (2) with factor dynamics (equation (5)):
  - X_{t+1}^{NA} = μ + Φ X_t^{NA} + ε_{t+1}, with ε_t ~ (0, Σ_ε).
  - No-arbitrage imposes restrictions on constants and factor loadings as in (6)–(9), and links market price of risk parameters ߣ_1 and ߣ_2 (denoted ߣ^{(1)} and ߣ^{(2)} in the source) with Λ_t = ߣ^{(1)} + ߣ^{(2)} X_t^{NA}.
- Difficulty: direct estimation of no-arbitrage models is hard because loadings and factors are unobservable.
- Approach used: follow Coroneo and others (2008) by using standardized estimated Nelson-Siegel factors, estimate a VAR(1) to obtain parameters of (5), estimate short-rate factor loadings from (9), and estimate market price of risk parameters by minimizing sum of squared errors between actual and fitted yields.

### No-Arbitrage Testing
- Null: factor loadings of the statistical Nelson-Siegel yield curve (expressed in (2)–(4)) are not statistically different from the no-arbitrage model loadings in (6)–(9).
- Because parameters vary by maturity, no-arbitrage conditions may hold at some maturities but not others.
- For each maturity, four hypotheses are tested (equation (10)):
  - H1: a_Nelson-Siegel = a_no-arbitrage  (notation: ܪ_{1,1} ܽ : a^{NS}_i = a^{NA}_i)
  - H2: B^{(1)}_NS = B^{(1)}_NA
  - H3: B^{(2)}_NS = B^{(2)}_NA
  - H4: B^{(3)}_NS = B^{(3)}_NA
- Bootstrap/block resampling procedure (as in Coroneo and others (2008)) used to generate bootstrap distribution over no-arbitrage model parameters because it ensures:
  - (i) exact asymmetric recovery of all eigenvalues and eigenvectors of yields;
  - (ii) correct reproduction of distribution of curvatures of the yield curve across maturities;
  - (iii) correct qualitative recovery of transition from super- to sub-linearity as maturity increases in variance of n-day changes;
  - (iv) satisfactory accounting of empirically-observed positive serial correlations in yields.
- Steps of the testing/bootstrap procedure:
  1. Construct the yield curve from the re-sampled data;
  2. Estimate the Nelson-Siegel factors;
  3. Estimate the no-arbitrage curve parameters using these Nelson-Siegel factors;
  4. Repeat steps (1) to (3) many times to build distributions for \hat{a}_i^{NA} and \hat{B}_i^{NA};
  5. Compare distributions of \hat{a}_i^{NA} and \hat{B}_i^{NA} with the Nelson-Siegel factors.

### III. MONETARY OPERATIONS AND MARKETS IN CHINA
- Policy framework and instruments:
  - The PBC has largely relied on quantity-based monetary targets and has not yet tried targeting a money market interest rate.
  - Market-based instruments: repo transactions (maturities between 7 days and 6 months; in recent years mostly to drain liquidity), PBC bill issuance (maturities of 3 months, 6 months, 1 year, and 3 years).
  - Administrative instruments: window guidance, regulated benchmark retail deposit interest rates (as ceilings) and lending interest rates (as floors), reserve requirements (differentiated by institution size and currently over 20 percent).
  - Standing facilities: lending windows (intra- and inter-day), rediscount facility, excess-reserve deposit window, term auction and FX swap facilities.
  - Government reliance on “quantitative limits to slow credit growth” (IMF, 2010).
  - Recent PBC practice: heavy use of reserve requirements and open market operations to drain liquidity resulting from reserve accumulation.
- Money markets structure:
  - China’s money markets include collateralized (repo) market and uncollateralized call-loan (CHIBOR) market.
  - Collateralized lending dominates, averaging 10 times the size of the unsecured market; almost all trading at maturities ≤ 7 days.
  - SHIBOR benchmark introduced in 2007: daily quotes up to 12-month maturity from 16 commercial banks.
  - Pricing anomalies observed: three interest rates usually tightly correlated but sometimes diverge; collateralized repo interest rate can be highest during stress despite collateral—likely reflecting market power of large banks with excess liquidity and higher relative counterparty risk of small banks during stress (including around IPOs).
  - PBC lending facility does not always provide a binding upper bound on the overnight interest rate corridor.
  - PBC’s repo rates systematically diverge from market repo rates.
- Bond market segmentation and trading:
  - Bond markets segmented since early 1990s; trading volumes vary considerably across bond types.
  - Historical notes: Treasury bond trading began in 1980s, moved from OTC to exchange floor trading after Shanghai and Shenzhen exchanges (1990, 1991); PBC prevented commercial banks from trading on exchange-based bond markets in 1997 to limit speculation; exchange-based market remained retail, interbank market limited to financial institutions.
  - Despite smaller stock of outstanding Policy Banks (Financial) bonds, their turnover is much larger than Treasury bonds; many Treasury bonds are held to maturity by banks for repo purposes.

### IV. DATA
- Data sources and coverage:
  - Zero coupon yields from China Central Clearing and Depository.
  - Yield series: interbank (IB) and exchange traded treasury bonds (IBTSY), interbank traded PBC bills (IBCB), interbank bonds issued by China’s financial banks (IBPBB), and corporate AAA bonds traded on the interbank market (IBCAAA).
  - Sample coverage: month-end data from April 2006 to May 2010 for all series except PBC bills; PBC paper yields start in June 2007.
- Empirical note:
  - Main descriptive statistics of Chinese bond yields presented in Table 1 (in source).
  - Chinese bond yields display considerably less autocorrelation relative to US Treasury yields studied by Diebold and Li (2006) and Coroneo and others (2008).
- Additional numerical details in source:
  - As of end-May 2010, outstanding Treasury bonds and PBC bills accounted for 30.0 percent and 22.8 percent of outstanding bonds, respectively.
  - Between June 2007 and May 2010, cumulative transaction volumes of Treasury bonds and PBC bills accounted for [detail truncated in source excerpt].

*Source: IMF Working Paper content provided in the supplied PDF excerpt.*

### 9.6 percent and 40.0 percent, respectively, of cumulative bond market transaction volume.

### _wp11225 - 9.6 percent and 40.0 percent, respectively, of cumulative bond market transaction volume.

### V. RESULTS — Overview
- Four principal questions addressed:
  - (i) Are bond yields distorted by the structure of administered interest rates?
  - (ii) Is there interaction between the yields on PBC paper and the yields on other bonds?
  - (iii) What are the macro-financial linkages between bond yields and macroeconomic variables?
  - (iv) Are Chinese bond yields arbitrage free and, in that sense, efficiently priced?
- General findings:
  - Yields are influenced by administratively set benchmark rates.
  - PBC yields do influence yields on other bonds, and those yields inform about macroeconomic variables.
  - Prices in Chinese bond markets are not fully efficient due to interest rate regulation, illiquidity, and market segmentation.

### A. Estimated Yield Curves and Market Behavior
- Model performance and errors:
  - Estimated errors (root mean squared error) are typically small.
  - Largest errors: short-end PBC paper yields.
  - Corporate and Treasury bonds: largest errors at the 2 year maturity.
  - Policy financial bonds: largest errors at the 12 month maturity.
  - Estimated errors show little significant autocorrelation beyond three months.
  - Estimated factors display significant autocorrelation over a few months.
- Shape and curvature of yield curves:
  - Estimated λ coefficients imply maximum curvature at around 3 years (impact of medium-term factors), longer than in the US Treasury market.
  - PBC and exchange Treasury markets have shorter estimated maturities at which curvature is maximized.
  - For relevant maturities (three- to thirty-six months) PBC bill rates are, on average, closely aligned with interbank Policy Banks bonds rates.
  - Relatively high spread between interbank Treasury bond yields and Policy Banks bonds yields likely reflects taxation differences.
- Market segmentation and arbitrage indications:
  - Interbank and exchange Treasury curves cross; crossing point varies over time around a median value of 12 months.
  - Crossing and non-concavity can produce two crossing points (example: December 2009).
  - Segmentation leads to inconsistent pricing—apparent profitable arbitrage opportunities—between treasury bonds in the two markets.
  - Limited liquidity and wide bid-ask spreads at certain maturities and frictions in transferring bonds across markets impede arbitrage.
- Money market vs. yield curves:
  - SHIBOR curve tends to converge to the PBC yield at the 12 month maturity.
  - On dates examined (end-June 2007; end-November 2007; end-December 2008) market overnight rates were below the Nelson-Siegel implied overnight Treasury/PBC rates; some longer money market rates are above the curves.
  - SHIBOR seems to converge to PBC curve toward the one-year point; PBC curve may serve as a traded benchmark alternative to SHIBOR.
  - The gap between implied Nelson-Siegel overnight rate and market overnight rate can be interpreted as a “liquidity effect”.
  - Interbank market overnight interest rate systematically below implied Nelson-Siegel overnight rate suggests aggregate liquidity surplus in the banking system relative to the PBC.

### B. Yield Curves and Monetary Policy Transmission
- Use of three Nelson-Siegel factors (level, slope, curvature) to summarize dynamics.
- Yield curve dynamics and policy:
  - Except for PBC paper, curves display little movement in the level (long-run) factor.
  - Policy loosening beginning late-2008 lowered the level of PBC’s rate structure and affected mainly the short- and medium-end of bond yield curves.
  - Policy response led to a parallel downward shift in the PBC yield curve; slope factor remained largely constant.
  - Policy loosening resulted in a significant and persistent rise in the slope of other yield curves; the curvature factor switched from positive impact to negative around the loosening.
- Impact of regulated retail interest rate margins (VAR results for IBPBB):
  - Regulated margins examined at 1, 3, and 5 year maturities.
  - Level, slope, and curvature of IBPBB yield curve are affected by regulated margins.
    - Impact on level factor: generally negative—an unexpected temporary widening of margins tends to push the longer-term factor down.
    - Impact on slope factor: generally significant and negative—an unexpected temporary widening of margins tends to push the slope down (flatten the yield curve).
    - Impact on curvature factor: an unexpected widening of margins leads to a temporary increase in curvature.
  - Margins also affect economic growth and inflation: higher margins → lower growth and lower inflation (consistent with relatively high short-term yields and lower long-term yields).
  - Interpretation: narrowing (widening) of regulated margins associated with steeper (flatter) yield curves and lower (higher) medium-term rates; these movements align with monetary loosening (tightening) effects on growth and inflation.
- Interrelation between PBC (IBCB) and Policy Banks (IBPBB) curves:
  - Changes in IBCB yield curve have statistically significant impacts on all IBPBB factors.
  - Shocks to the level of PBC curve have the largest and most lasting impact on the level of IBPBB.
  - Shocks to slope and curvature of PBC curve affect slope and curvature of IBPBB; curvature shocks have contemporaneous, short-lived impact on IBPBB level.
  - Conclusion: one necessary element of a rate-based monetary transmission mechanism appears to be working/emerging in China.
- Macro-financial linkages (focus on IBPBB factors, real industrial value added growth, CPI inflation in year-on-year terms):
  - Cross-correlation findings:
    - Level factor of Policy Banks bonds tends to lead the business cycle: significantly negatively correlated with growth and positively correlated with inflation 12 to 20 months ahead.
    - Corporate spreads lead the cycle: higher spreads related to lower growth and higher inflation in the future.
    - Short-term (slope) factor: negatively correlated with past growth and positively correlated with contemporaneous and future inflation.
    - Medium-term (curvature) factor: positively correlated with current and future growth and negatively correlated with inflation a few months ahead.
  - VAR-based results (summary bullets preserve numeric timing where given):
    - Unexpected rises in the level factor of IBPBB yields dampen growth after about half a year.
    - A rise in the slope (lower near-term yields) raises growth after 3 months (although not significant) and raises inflation after 5 months.
    - Increase in the curvature factor is associated with a temporary increase in growth which dies out after about a year.
    - Inflation strongly affected by unexpected changes in the curvature of IBPBB yields (medium-term factor).
    - The slope and curvature factors affect the 3 month PBC rate with opposite signs: slope negative, curvature positive.
    - The 3 month PBC rate responds to corporate spread shocks and inflation shocks, but little to growth shocks.
    - Unexpected changes in corporate spreads tend to depress both growth and inflation.
    - Changes in the 3 month PBC rate affect all factors in the IBPBB yields.
    - Growth does not respond significantly to unexpected changes in the PBC bill rate, whereas inflation and corporate spreads decline significantly and persistently after such changes.
  - Overall interpretation:
    - A nascent monetary policy transmission channel exists: changes in PBC open market operation rates feed through to longer-term yields and corporate spreads, influencing growth and inflation.
    - This channel operates despite regulated interest rates, with the shape of bond curves influencing macro outcomes (higher longer-term Policy Bank yields and corporate spreads depress growth and inflation; a fall in slope—higher short-term yields—similarly depresses activity).

### C. Arbitrage-Freeness and Efficiency of Yield Curves
- Method:
  - Test compares deterministic Nelson-Siegel factor loadings with distribution of factor loading estimated using arbitrage-free affine yield curve model; rejection when NS loadings lie outside the 95 percent confidence interval.
- Results summary:
  - All four markets examined show some significant deviation from arbitrage-free bond yields; degree varies across markets.
  - Policy Bank and Treasury bonds trading on the interbank market are closer to not rejecting the no-arbitrage hypothesis.
- Market-specific deviations (preserve exact maturities and descriptions):
  - Interbank traded Treasury bonds:
    - NS loadings on slope and curvature deviate significantly from arbitrage-free factors at short maturities (less than two years).
    - Result: weight on short-term factors too high; weight on middle-tenor factors too low.
  - Exchange traded Treasury market:
    - Weight on long-run level factors and short-term factors too high across all maturities.
    - Weight on middle-tenor (curvature) factor seems statistically consistent with NS factors.
  - Policy Banks bonds:
    - Weight of short-term (slope) factor seems closest to arbitrage-free consistency.
    - Weight on long-term level seems too low at short and middle maturities (less than two years).
    - Significant deviations at middle-maturities (around 4-6 years).
  - Central Bank paper market:
    - Weights placed on all factors seem too high to be consistent with arbitrage-free pricing.
    - Maximum maturity in this market is 36 months; large deviations could be consistent with deviations in other interbank bond markets largest at these maturities.
- Liquidity and other drivers of deviations:
  - Liquidity measured by bid-ask spreads suggests interbank Treasury and Policy Bank markets are least liquid at very short maturity—consistent with deviations from arbitrage-free pricing at those maturities.
  - In Policy Bank market, bid-ask spreads rise faster (from a maturity of 3 years) and higher than in the Treasury market—may explain deviations at those maturities.
  - Other considerations: buy-and-hold behavior in Treasury market (making correct pricing of longer yields important); higher short-term turnover and repo use for Policy Bank bonds (favoring accurate short-term pricing).
  - Regulated structure of interest rates (specifying long-term retail rate structure at 5 years) likely impacts more efficient pricing at longer maturities.

### VI. CONCLUSIONS — Key findings and policy-relevant implications
- Markets and instruments covered: PBC bill market, segmented interbank and exchange traded Treasury Bonds market, Policy Financial Bonds market, AAA corporate bond market.
- Statistical approach: latent factor analysis (Nelson-Siegel), conditional comparisons to an affine no-arbitrage yield model.
- Efficiency and arbitrage:
  - Interbank Treasury and Policy Banks bond yields are closest to not rejecting arbitrage-free hypothesis but deviate at shorter maturities.
  - Central bank bills deviate from efficient pricing at similar maturities.
  - Deviations correlate with market liquidity and other market structural factors.
- Macro signal value:
  - Bond yields, despite distortions, anticipate macroeconomic movements.
  - Implied long-term corporate spreads correlated with inflation up to 7 months hence and growth up to 2 months.
  - Changes in Policy Banks bond yield curve drive changes in growth and inflation; PBC rates affect all Policy Banks curve factors.
- Policy implications and recommendations:
  - Stage is set for increased use of interest rate targeting as primary monetary policy tool: changes in PBC rates feed through to other yields and macro aggregates.
  - Financial prices are informative signals of macro and financial conditions; further liberalization of financial prices would strengthen monetary transmission.
  - Increasing disintermediation underscores urgency of moving towards indirect policy tools.
  - Introducing some limited foreign participants into the interbank market may improve liquidity and pricing efficiency.
- Operational caution:
  - Exploiting arbitrage between interbank and exchange markets is unlikely to be profitable given complications in moving bonds between markets despite observed yield curve crossings.

*Italic source: _wp11225 - 9.6 percent and 40.0 percent, respectively, of cumulative bond market transaction volume.*

### 2. VAR with 4 lags and the following variables: 3 yield curve factors, monthly industrial value added growth (year-on-ye

### _wp11225 - 2. VAR with 4 lags and the following variables: 3 yield curve factors, monthly industrial value added growth (year-on-year), and CPI inflation (year-on-year), and regulated margins (benchmark lending – deposit rate) at 1, 3 and 5 years.

### VAR specifications and identification
- VAR with 4 lags and the following variables: 3 yield curve factors, monthly industrial value added growth (year-on-year), and CPI inflation (year-on-year), and regulated margins (benchmark lending – deposit rate) at 1, 3 and 5 years.
- Several figures and notes refer to alternative VARs used in the analysis:
  - "Note: VAR with two lags and six variables which are three IBCB factors and three IBPBB factors. The factors refer to the level, curvature and slope factors in the Nelson-Siegel model."
  - "Note: VAR with two 2 lags and seven variables as follows: three factors of the IBPBB curve, 3-mth PBC bill rate, real growth of industrial value added, CPI inflation and corporate bond spreads."
  - "Note: VAR with two 2 lags and seven variables as follows: three factors of the IBPBB curve, 3-mth PBC bill rate, real growth of industrial value added, CPI inflation and corporate bond spreads." (repeated across figures)

### Key impulse-response contexts reported (figure notes)
- Interaction between Policy Bank and PBC yield curves:
  - Responses of IBPBB factors to level of PBC curve, slope of PBC curve, and curvature of PBC curve reported using a VAR with two lags and six variables (three IBCB factors and three IBPBB factors).
- Yield curve factors and macroeconomic variables:
  - Responses of macro variables to impulses to level factor, slope factor, and curvature factor using VAR with two lags and seven variables (three IBPBB factors, 3-mth PBC bill rate, gryy, infyy, and corporate bond spreads).
- Growth, inflation, corporate spreads and 3-month PBC rate:
  - Responses of macro to growth shock, inflation shock, level PBC rate shock, and spread shock using the seven-variable VAR (two lags).
- Macroeconomic and yield curve factors:
  - Responses of factors to impulses to PBC 3mth rate, growth, inflation, and corporate spread shock using the seven-variable VAR (two lags).

### Selected numeric scales and confidence reporting in impulse responses
- All impulse-response plots annotate "step" on the x-axis with tick labels "0 5 10 15" and indicate "90% CI for oirfoirf".
- Many impulse-response y-axis ranges appear in the figures; examples (preserved verbatim from figure axes):
  - -.5 0 .5 (varbasic: mar1y -> gryy)
  - -1 -.5 0 .5 (varbasic: mar3y -> gryy)
  - -.4 -.2 0 .2 (varbasic: mar5y -> gryy)
  - -.2 -.1 0 .1 .2 (varbasic: mar1y -> infyy)
  - -1 -.5 0 .5 1 (varbasic: slo_ibpbb -> cur_ibpbb)
  - -2 -1 0 1 (varbasic: slo_ibpbb -> gryy)
  - 0 .5 1 1.5 (varbasic: slo_ibpbb -> infyy)
  - -1.5 -1 -.5 0 .5 (varbasic: cur_ibpbb -> gryy)
  - -1.5 -1 -0.5 0 0.5 1 1.5 2 (Figure 17 constant axes shown as x 10 -3)
- Confidence bands shown: "90% CI for oirfoirf" on all reported impulse-response panels.

### No-arbitrage tests for Chinese bond yields (Figure 17)
- Panels reported: (a) Interbank Treasury Bond; (b) Exchange Treasury Bond; (c) Interbank Policy Bank Bond; (d) Central Bank Bill.
- Plots present:
  - Constant term and distributions labeled "NS", "Median", "50%", "90%", "95%".
  - Loadings of the level, slope, and curvature plotted across time/horizon axes labelled "0 20 40 60 80 100 120" (preserved as sequences shown in figures).

### Interbank bond market bid-ask spreads (Figure 18)
- Source: Bloomberg LP and authors’ estimates.
- Panels and legend entries:
  - "Interbank  Treasury  Bond  Bid-Ask  Spreads (In basis points)" with "Median" and "Mean".
  - "Policy Bank Bond  Bid-Ask  Spreads (In basis points)" with "Median" and "Mean".
- Maturity axis labels (as shown): 2 3 4 5 0.250.51235710 and a longer set 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0.250.51235710 (preserved exactly as in figure).

### Tables — properties of the yield data, errors, and estimated factors
- Table 1. Properties of the Yield Data (columns shown in source header: Maturity Mean Median std min max 1 3 6 12 24)
  - Interbank Treasury rows (preserved verbatim fragments as in source):
    - "3 2.01.90.80.83.6 0.950.750.43 -0.15 -0.30 62.12.00.80.93.6 0.950.750.42 -0.15 -0.28 92.22.00.90.93.7 0.950.750.42 -0.16-0.28 122.22.10.91.03.8 0.950.750.41 -0.16 -0.28 242.52.30.91.24.1 0.950.760.42 -0.22-0.26 362.82.50.81.34.1 0.940.700.33 -0.26-0.20 483.02.80.71.64.2 0.930.66 0.28 -0.29 -0.14 603.12.90.71.84.3 0.910.630.28-0.29 -0.13 723.33.10.62.04.4 0.910.62 0.27 -0.32 -0.11 843.43.20.62.24.5 0.900.60 0.26 -0.32 -0.10 963.53.30.62.44.5 0.910.60 0.25 -0.32 -0.09 1203.73.60.62.94.6 0.910.61 0.24 -0.32 -0.06"
  - Exchange Treasury rows (preserved verbatim fragments as in source):
    - "62.02.00.70.93.3 0.890.710.38 -0.09 -0.29 122.32.20.81.13.7 0.920.730.42 -0.08 -0.34 242.72.50.81.44.1 0.930.750.43 -0.13 -0.30 ..." (and continuing as printed)
  - PBC Bills rows (preserved verbatim fragments as in source):
    - "0.251.91.60.80.73.5 0.900.65 0.27-0.03-0.20 0.52.01.51.00.84.7 0.830.63 0.33-0.09-0.22 ..." (and continuing)
  - Interbank Policy Financial Bonds, Interbank Corporate AAA rows similarly printed in the source table.
  - Footnote: "1/ Bold entries indicate significance at 5 percent using asymtotic standard errors. Autocorrelati on Functon 1/"

- Table 2. Properties of the Estimated Errors (header: Maturity Mean Median std min max rmse 1 3 6 12 24)
  - Interbank Treasury fragments preserved verbatim (examples from source):
    - "3-0.03-0.020.05-0.140.100.06 0.50 0.05-0.21-0.02-0.04 60.000.000.03-0.050.080.03 0.31 -0.100.25-0.02-0.09 ..." (and continuing as printed)
  - Exchange Treasury, PBC Bills, Interbank Policy Financial Bonds, Interbank Corporate AAA entries appear in the source and are preserved with the same numeric strings in the table.
  - Footnote: "1/ Bold entries indicate significance at 5 percent using asymtotic standard errors. Autocorrelation Functon 1/"

- Table 3. Properties of the Estimated Factors (selected entries preserved verbatim)
  - Interbank Treasury:
    - Level (Factor 1) 4.34.40.53.45.2 0.840.43 0.09 -0.33 0.17
    - Slope (Factor 2) -2.4-2.10.9-3.9-0.7 0.930.680.33 -0.14-0.07
    - Curvature (Factor 3) -1.5-1.31.2-4.40.5 0.880.64 0.26 -0.31 -0.10
  - Exchange Treasury:
    - Level (Factor 1) 4.44.50.63.45.4 0.880.46 0.11 -0.42 0.22
    - Slope (Factor 2) -2.6-2.70.9-3.9-0.9 0.860.62 0.21 -0.30 0.16
    - Curvature (Factor 3) -1.2-1.41.7-3.72.3 0.890.730.50 -0.07 -0.29
  - PBC Bills:
    - Level (Factor 1) 3.53.41.11.75.2 0.930.67 0.19-0.31-0.08
    - Slope (Factor 2) -1.6-1.80.5-2.2-0.2 0.780.45 -0.06 -0.41 0.09
    - Curvature (Factor 3) -0.8-0.81.3-3.12.0 0.700.580.36 -0.21-0.20
  - Interbank Policy Financial Bonds:
    - Level (Factor 1) 4.74.90.63.55.7 0.850.46 0.24-0.19-0.05
    - Slope (Factor 2) -2.3-2.01.1-4.7-0.8 0.930.700.36 -0.04-0.11
    - Curvature (Factor 3) -0.7-0.51.6-5.01.8 0.890.59 0.01 -0.38 -0.03
  - Interbank Corporate AAA:
    - Level (Factor 1) 5.65.70.84.26.7 0.880.590.39 -0.06-0.11
    - Slope (Factor 2) -2.7-2.61.3-5.0-0.9 0.940.750.45 -0.03-0.11
    - Curvature (Factor 3) -0.8-0.61.4-4.61.5 0.860.60 -0.05 -0.44 -0.04
  - Footnote: "1/ Bold entries indicate significance at 5 percent using asymtotic standard errors. Table 3. Properties of the Estimated Factors Autocorrelation Functon 1/"

### Table 4. Estimated lambda by type of curve (preserved exactly as printed)
- Table header: "Table 4. Estimated lambda by type of curve Bond Estimated Lambda Maturity of maximum curvature (months) Estimation Sam ple Number of Maturi ti e s Minimum Maturi ty Maximum Maturi t y"
- Rows (verbatim):
  - "Inte rbank Tre asury0.05036 Apr-06 -  May- 10123120"
  - "Ex change  Tre asury0.06528 Apr-06 -  May- 10106120"
  - "Inte rbank PBC0.2158 Jun- 07 - May- 10100.2536"
  - "Inte rbank Pol i cy Bank0.04540 Apr-06 -  May- 10123120"
  - "Inte rbank Corporate  (AAA)0.04540 Apr-06 -  May- 10106120"

### Appendix: Cross-correlograms of IBPBB factors and macro variables — summarized findings preserved verbatim
- Cross-correlation between Factor Y(t) and Macro variable X(t+k). Vertical bars show highest and second highest correlations in absolute value.
- Level factor (Y) and growth (X):
  - "Note: cross-correlation is highest (in abs. value) at around +12 months suggesting that growth lags the level factor by around one year; there is also some evidence of growth leading the level factor by about half-year (-5). Correlation is mainly negative meaning higher factor level indicating slowdown in growth rate."
- Level factor (Y) and inflation (X):
  - "Note: cross-correlation is highest at around +20 months suggesting that inflation lags the level factor by around two years; there is also some evidence of inflation lagging the level factor by about one- two-months (+2). Correlation is positive meaning higher factor level indicating acceleration in inflation rate."
- Slope factor and growth:
  - "Note: cross-correlation is highest (in abs. value) at around -5 months suggesting that growth leads the slope factor by around five months. Correlation is negative meaning that higher growth leads a flattening of the yield curve."
- Slope factor and inflation:
  - "Note: cross-correlation is highest at around +2 months suggesting that inflation slightly lags the slope factor by around two months. Correlation is positive meaning that higher inflation lags a steepening of the yield curve."
- Curvature and growth:
  - "Note: cross-correlation is highest at around -1/-2 months suggesting that growth leads the curvature factor by around two months. At these lags cross-correlation is positive meaning that higher inflation (slightly) lags a steepening of the yield curve."
- Curvature and inflation:
  - "Note: cross-correlation is highest (in abs. value) at around 6 months suggesting that inflation lags the curvature factor by around half year. Correlation is negative meaning that lower inflation lags an increase in curvature."
- Corporate spread and inflation:
  - "Note: cross-correlation peaks at around 7 months suggesting that inflation lags the corporate spread factor by around half year. Correlation is positive meaning that higher inflation lags an increase in spreads."
- Corporate spread and growth:
  - "Note: cross-correlation peaks (in absolute value) at around 2 months suggesting that growth lags the corporate spread factor by around 2 months. Correlation is negative meaning that higher growth lags an increase in spreads."

*Content derived verbatim from the supplied IMF PDF content unit.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11225.pdf_