Price discovery and volatility spillovers in the interest rate derivatives market

The interest rate derivatives market promotes the development of the bond market by contributing to the formation of benchmark interest rates and making risk-free rate pricing more effective. With the deepening of interest rate marketization, China’s interest rate derivatives market has developed rapidly, with products such as interest rate swaps, forward rates, bond forwards, and options. The current market structure features exchange-traded treasury bond futures and over-the-counter interest rate swaps. Given growing investor demand for hedging interest rate risk, this study examines whether China’s interest rate derivatives effectively perform their intended roles and provide hedging channels. As a core financial market function, price discovery is the process by which new information is impounded into security prices. An effective price discovery mechanism enhances trading and investment decision-making. Additionally, analyzing volatility spillovers between markets assists in asset pricing and portfolio construction. Therefore, this paper investigates price discovery and volatility spillovers in China’s interest rate derivatives market, focusing on treasury bond futures and interest rate swaps, using the information share (IS) model and Diebold–Yilmaz (DY) spillover index. The study evaluates whether futures and swaps effectively contribute to price discovery and explores the dynamic spillover structure among markets. The main contributions are: (1) constructing an IS model across treasury bond futures, interest rate swaps, and treasury bond spot to study price discovery; (2) employing the DY spillover index to examine static and dynamic spillovers for insights into information transmission; thereby offering a comprehensive view of price discovery efficiency and volatility spillovers in China’s interest rate derivative markets. The remainder is organized as follows: Section 2 reviews the literature; Section 3 describes data and methodology; Section 4 presents empirical results; Section 5 concludes and provides policy implications.

The transaction cost hypothesis posits that markets with lower transaction costs and higher liquidity incorporate information more rapidly. Empirical work suggests that well-functioning futures markets often lead spot markets in price discovery. Studies on government bond futures across various countries (e.g., Korea, US, Germany, UK) typically find futures lead cash markets, with lead–lag relations influenced by factors such as foreign participation and trading hours. Research using multifractal and other time-series methods finds persistent relationships between certain futures and spot tenors. For interest rate swaps, literature has examined pricing and relationships with other rates. On volatility spillovers, diverse econometric techniques (e.g., HAR, EGARCH, BEKK-GARCH, nonlinear Granger causality, TVP-VAR) have been used to study cross-market connectedness among equities, commodities, energy, and carbon markets. The Diebold–Yilmaz spillover index offers directional and time-varying measures that are invariant to variable ordering, facilitating assessment of dynamic connectedness. Prior work has applied DY to analyze links between energy and food markets and other asset classes. Specific to China’s interest rate markets, fewer studies jointly analyze treasury bond futures, spot, and interest rate swap markets for both price discovery and volatility spillovers. Some evidence indicates interactions among these markets and varying efficiencies. Building on the literature, the paper hypothesizes that interest rate derivatives guide the spot market in price discovery and that volatility spillovers exist among the markets.

Data: The spot market is represented by the China Securities Index (CSI) Aggregate Bond Index, capturing interbank and exchange bond markets. The interest rate swap (IRS) market is proxied by the daily average fixed rate of five-year FR007-based swaps. Treasury bond futures are represented by the daily closing price of the five-year treasury bond futures (TF) main contract from the China Financial Futures Exchange. The sample spans 2016-01-04 to 2020-12-31 with 1,215 observations. Natural logs of prices are taken (InF for futures, InS for spot, InIS for swaps). Returns are first differences of logs multiplied by 100 (RF, RS, RIS). Descriptive statistics, Phillips–Perron unit root tests, and correlation analyses are conducted; returns are stationary at the 1% level. Structural break analysis: Parameter stability is assessed via CUSUM tests; futures and IRS show stability, whereas spot indicates instability. Bai–Perron multiple breakpoint tests are applied to identify structural breaks in the spot series. Information Share (IS) model: Hasbrouck’s IS measure is used within a Vector Error Correction (VEC) framework to quantify each market’s contribution to the common efficient price. The VEC model is transformed to VMA, long-run impacts are derived, and when residual innovations are correlated, Cholesky decompositions provide upper and lower bounds of IS; their average serves as the estimate. Johansen cointegration tests verify long-run relationships between pairwise (InF, InS) and (InIS, InS). Component Share (CS) model: Following Gonzalo and Granger, CS decomposes permanent and transitory components in a cointegrated VEC setting, using adjustment coefficients (alpha) to infer contributions to the common stochastic trend. VAR and DY spillover models: With stationary returns, a VAR(p) is estimated; lag order is chosen via information criteria (VAR(2) selected). Stability is checked via AR roots within the unit circle. Impulse responses and variance decompositions are computed. For spillovers, the Diebold–Yilmaz framework uses generalized VAR-based forecast error variance decomposition (FEVD) that is invariant to ordering. Total, directional, net, and pairwise spillover indices are constructed using H=10-step-ahead FEVD. Dynamic connectedness is analyzed using a 100-day rolling window to obtain time-varying total and directional spillover indices.

- Descriptive and stationarity tests: IRS returns exhibit higher volatility than futures and spot. PP tests show log prices are non-stationary; first-differenced returns (RF, RS, RIS) are stationary at 1% significance. Correlations: InF–InS = 0.464 (p<0.01), InF–InIS = -0.900 (p<0.01), InS–InIS = -0.611 (p<0.01). - Structural breaks: CUSUM indicates stability for futures and swaps, instability for spot. Bai–Perron tests detect three structural breaks in spot on 2016-10-25, 2018-01-22, and 2020-04-03 (reported as April 3, 2022 in the table), while futures and swaps show none. - Cointegration: Johansen tests find one cointegrating vector for (InF, InS) and for (InIS, InS) at the 1% level, supporting a long-run relationship enabling VEC-based IS analysis. - Information Share (full sample): For the futures–spot pair, mean IS is 77.43% (futures) vs 22.57% (spot). For the IRS–spot pair, mean IS is 71.19% (IRS) vs 28.81% (spot). Upper/lower bounds indicate futures and IRS dominate price discovery under alternative Cholesky orderings. - Information Share by subperiods (based on spot breakpoints): • 2016-01-04 to 2016-10-25: Futures 89.34% vs Spot 10.66%; IRS 73.76% vs Spot 26.24%. • 2016-10-26 to 2018-01-22: Futures 15.90% vs Spot 84.10%; IRS 13.18% vs Spot 86.82%. • 2018-01-23 to 2020-04-02: Futures 88.88% vs Spot 11.12%; IRS 83.09% vs Spot 16.91%. • 2020-04-03 to 2020-12-31: Futures 50.20% vs Spot 49.80%; IRS 31.25% vs Spot 68.75%. The derivatives’ price discovery leadership is time-varying and declines relative to spot during stages 2 and 4. - Component Share (CS): For futures–spot, CS indicates stronger futures contribution (85.27% vs 14.73% for spot). For IRS–spot, results do not support IRS dominance in CS terms (insignificant/negative indication), suggesting mixed evidence for swaps’ permanent component leadership. - VAR(2) dynamics and variance decomposition: The spot return (RS) is increasingly influenced by lagged futures (RF), becoming the most important driver over horizons; RIS has a smaller but non-negligible effect. For RIS, RF’s influence grows with horizon, surpassing own shocks. Stability is confirmed by AR roots within the unit circle. - Spillover indices (static): Total spillover index = 49.0%, indicating that nearly half of forecast error variance is due to cross-market spillovers. Contribution to others: RF 51.1% (largest), RIS 48.9%, RS 47.0%. Contribution from others (received): RS 54.5% (largest), RIS 48.9%, RF 43.6%. Net spillovers: RF +7.5%, RIS 0%, RS -7.5%. Spillovers between IRS and spot exceed those between futures and spot. - Spillover indices (dynamic): Total spillover rose after late-2016, indicating strengthening linkages, with V-shaped behavior around late-2016. Directional spillovers from futures and IRS to others strengthened over time; spot increasingly received spillovers. Net indices show futures as a persistent net transmitter and spot as a persistent net receiver. Pairwise net spillovers indicate futures and IRS both transmit to spot; futures also transmit to IRS, with futures’ impact on spot strongest.

The findings address the core questions of whether China’s interest rate derivatives lead price discovery and how volatility risk transmits among futures, swaps, and spot markets. Full-sample IS and CS evidence shows treasury bond futures dominate price discovery over the spot, consistent with transaction cost and liquidity advantages of futures markets. IRS also leads spot in IS on average, though CS offers mixed support for IRS dominance, suggesting that swaps contribute to informational efficiency but may not consistently anchor the common trend. Structural break analysis reveals that derivatives’ leadership is not constant: during specific periods (notably stages 2 and 4), spot dominates, indicating that external shocks, liquidity shifts, and policy changes can temporarily reduce derivatives’ informational leadership. Volatility connectedness is economically meaningful: the total spillover index of 49% demonstrates substantial cross-market risk transmission. Futures are a persistent net transmitter, aligning with their role as central venues for rapid information incorporation and hedging. Spot is the main net receiver, highlighting its vulnerability to shocks originating in derivatives markets. Dynamic analysis indicates strengthening inter-market linkages since 2016, likely reflecting market development (e.g., product expansions, broader participation). These results underscore the need for market participants to account for time-varying leadership in price discovery and evolving spillover structures when hedging and managing interest rate risk.

The study provides a comprehensive assessment of price discovery and volatility spillovers across China’s treasury bond futures, interest rate swaps, and treasury bond spot markets. Key contributions include: (1) demonstrating that futures and, on average, swaps exhibit stronger price discovery than spot in the full sample; (2) revealing the time-varying nature of derivatives’ price discovery leadership across structural subperiods; and (3) quantifying substantial volatility spillovers, with futures as a persistent net transmitter and spot as a net receiver. Policy implications include promoting synergy among futures and IRS markets, expanding product varieties (e.g., different maturities) to improve the yield curve and pricing references, enhancing cross-market supervision and information integration to mitigate systemic risk, and improving transparency. Practitioners should adapt portfolio and hedging strategies to the dynamic spillover structure. Future research should incorporate additional maturities, macroeconomic drivers, high-frequency data, and broader derivative instruments to deepen understanding.

The study focuses on five-year treasury bond futures and five-year FR007-based swaps, potentially limiting generalizability across maturities. Macroeconomic and policy variables that may affect price discovery and spillovers are not explicitly modeled. The sample period ends in 2020, and structural dynamics beyond this window are not captured. Future work could integrate high-frequency data, additional derivative products and maturities, and macro-financial covariates to assess robustness and extend insights.