Same Issuer, Two Prices? A Microstructure Analysis of China's MTN-Corp Basis

China's credit bond market has a structural quirk: the same issuer can issue bonds in two different markets simultaneously.

Huaneng Group can issue a Medium-Term Note (MTN) in the interbank market while placing a public corporate bond on the exchange. Both instruments have identical credit risk, similar tenors—but trade at different prices because they're listed on different venues.

The question I was curious about: how large is this spread? What drives it? How much dispersion exists across individual names? Systematic empirical analysis shows: the spread averages just 1.2 bps. What's interesting is the cross-sectional dispersion, and the underlying drivers.

I. Why Do Two Prices Exist?

To understand the Basis, you first need to understand China's "one bond, two markets" structure.

Interbank Market (MTNs) are primarily held by bank proprietary desks. Their mandate is hold-to-maturity (HTM)—as long as coupon income covers the internal funds transfer pricing (FTP) cost, mark-to-market volatility is largely irrelevant. This makes MTN pricing more "stable" but less liquid, with dealer bid-ask spreads of 3-8 bps.

Exchange Market (Corporate Bonds) are primarily held by mutual funds. Their mandate is mark-to-market (MTM)—NAV volatility directly drives redemptions and AUM. This makes corporate bond pricing more "dynamic" and more liquid, with centralized auction spreads of just 1-3 bps.

The Basis is fundamentally the different pricing of the same credit by two distinct investor bases.

This isn't theoretical—it's a real difference in trading behavior. Banks are sensitive to absolute yield levels but insensitive to volatility. Funds are sensitive to relative value and liquidity but highly sensitive to volatility. When the same issuer has bonds in both markets, this investor structure difference manifests as a price differential.

So the question becomes: how large is this differential, and is it tradeable?

II. Research Process

I pulled all credit bond data from Wind, covering January 2024 through April 2026.

Initial sample: 13,678 MTNs + 7,152 public corporate bonds = 20,830 total.

Screening criteria: No LGFVs, no financials, AAA/AA+ ratings, 1-7 year tenors, fixed rate, no embedded options. After screening: 1,145 bonds.

Pairing logic: Same issuer, same rating, tenor difference ≤ 0.5 years.

Pairing result: 147 pairs. But only 31 pairs had complete valuation data—these 31 pairs form my core analytical sample. While limited in number, they cover 31 issuers across 8 industries with 9,143 daily observations.

Basis Definition:

Basis = Yield(MTN) - Yield(Corp)

• Basis > 0: MTN yields more, MTN is cheaper
• Basis < 0: Corp yields more, MTN is richer

III. How Large is the Basis?

The first key number:

Average Basis is only 1.2 bps. This tells us that interbank and exchange markets largely agree on pricing the same credit. If you're hoping to "buy the basis and wait for mean reversion," transaction costs (6-15 bps) will consume the entire edge.

Basis distribution histogram: concentrated around mean, but with notable tails

But averages mask important individual differences. Cross-sectional dispersion of 2.8 bps is more than 2x the mean.

What does this mean? Some issuers have Basis as high as +6 bps (MTN significantly cheaper), others at -14 bps (Corp significantly cheaper). The alpha opportunity in name selection far exceeds the beta opportunity in market-level positioning.

Basis Index time series: overall volatility is low, but individual name dispersion is notable

IV. Which Is Cheaper?

Sorting the 31 issuers by Basis reveals clear stratification:

Issuers where MTNs are cheaper (Basis > 0, Top 5):

Issuers where Corps are cheaper (Basis < 0):

Sector effects are clear:

• Chemicals, coal, utilities → MTNs cheaper (bank allocation preference)
• Real estate → Corps cheaper (funds have higher risk appetite than banks)

This isn't coincidental. Banks have an allocation preference for central SOE utilities (stable cash flows, low credit risk), creating strong MTN demand. Real estate bonds have better liquidity on the exchange—funds under yield pressure are willing to take credit risk, while banks are constrained by risk limits. Shenzhen Metro's corporate bonds are exchange favorites, heavily held by funds, with liquidity premium dominating pricing.

Average Basis by issuer: China Energy Conservation and China Merchants Port show MTNs significantly cheaper; Shenzhen Metro shows Corps significantly cheaper

V. The Key Driver: Duration Gap

Panel regression analysis revealed a stable predictive factor: Duration Gap.

Duration Gap = Duration(MTN) - Duration(Corp)

The coefficient is approximately -6 bps/year. For each year that MTN duration exceeds Corp duration, Basis decreases by 6 bps.

Economic logic:

The -6 bps/year coefficient may reflect two mechanisms:

1. Investor behavior difference: Funds (MTM) are duration-sensitive and demand higher term premiums. When MTN duration is longer, funds tend to avoid long-end MTNs and embrace shorter-duration Corps, pushing Corp yields lower, causing Basis = MTN Yield - Corp Yield to turn negative

2. Valuation model effect: China Bond/China Securities valuations themselves calculate term premium by duration—the larger the Duration Gap, the larger the valuation difference between the two bonds. This is model input, not real market pricing

Distinguishing the two requires execution price data. Based on available data, the conservative judgment is both factors are at play. This means Duration Gap's predictive power for Basis may partially come from valuation model "auto-regression"—Duration Gap is a model input, Basis is a model output.

Strategy implication: When going long Basis (buy MTN, sell Corp), select pairs where MTN duration doesn't exceed Corp duration. When Duration Gap > 0.5 years, Basis tends to be more negative—unsuitable for long positions.

VI. How Fast Does Mean Reversion Work?

Basis exhibits mean reversion, but with an 83-day half-life—reversion is slow.

Moreover, the Basis series is non-stationary, meaning it can deviate from mean for extended periods without reverting. Passive basis buying and waiting for mean reversion is not viable. Dynamic risk management is required: 8 bps absolute stop-loss + 90-day time stop.

Cross-sectional dispersion time series: dispersion expands significantly during stress periods

VII. Backtest: Long Works, Short Doesn't

Based on Duration Gap and Basis extreme signals, I ran a simple backtest.

Long Basis (buy MTN, sell Corp):

• Entry: Basis < -5 bps (betting on Basis reverting toward 0), Duration Gap ≤ 0.1
• Stop-loss: 50 bps
• Signal deduplication: ≥5 days between adjacent signals for same pair

Results:

Win rate just over 50%, but 3.5x payoff ratio—this is a tradeable strategy.

Short Basis performed poorly: 32.5% win rate, -9.6 bps average return. The reason: when Basis is high, it tends to widen further rather than converge. Sample concentration was also an issue—short signals came from only 3 issuers.

My conclusion: go long only, don't short.

Issuer selection matters. China Energy Conservation (62 signals, 64.5% win rate) and China Merchants Port (11 signals, 63.6% win rate, 114 bps return) performed best. Issuers with strong fundamentals and international ratings show more deterministic valuation reversion.

VIII. Primary vs. Secondary Market Pricing Divergence

This is the most interesting finding in the study.

I analyzed 221 issuance pairs (same issuer, issued within 90 days, tenor difference < 1 year):

At issuance, Corps are 3.1 bps more expensive. In secondary trading, MTNs are 1.2 bps more expensive—the directions are completely opposite.

How to interpret this?

One possible explanation: at issuance, exchange underwriters have stronger pricing power and may price corporate bonds at a premium. But after listing, funds chase liquid corporate bonds, pushing prices even higher (yields lower), causing Basis to reverse.

Strategy implications:

For issuers: Corporate bonds are priced more expensive at issuance (-3.1 bps), meaning lower financing cost. The suggestion is to prefer corporate bond issuance, then buy back same-issuer MTNs in the interbank market after listing to hedge Basis risk. But note: 3.1 bps difference after deducting underwriting fees may leave little optimization space—practical gains are limited.

For investors: The primary-secondary pricing reversal suggests an "issuance arbitrage" window—subscribe to newly issued corporate bonds, while going long same-issuer MTNs, betting on Basis reversal from negative to positive. But note: the average difference across 221 issuance pairs is only 3.1 bps; after deducting bilateral transaction costs (6-15 bps), this window essentially doesn't exist. Moreover, this is a mean—individual bond dispersion is larger. Truly executable opportunities require bond-by-bond analysis.

Core conclusion: The primary-secondary pricing reversal is an interesting phenomenon, but based on available data, unlikely to translate into an executable arbitrage strategy.

IX. Counterarguments

Basis is too small to trade. 1.2 bps mean vs. 6-15 bps transaction costs—this critique is valid. My backtest doesn't deduct transaction costs; actual executable returns would be lower.

Limited sample size. 31 pairs constrains statistical robustness. Short signals came from only 3 issuers—conclusions aren't reliable.

Valuation vs. execution price. I used China Bond/China Securities valuations, which may deviate 1-3 bps from actual execution prices. True Basis may be smaller.

No full-cycle data. The 2024-2026 period doesn't include a credit crisis or sharp rate spike stress test. Basis behavior under market stress may differ significantly.

Primary-secondary divergence may have other explanations. Underwriter pricing power, issuance timing, investor subscription preferences—these factors could explain the pricing difference without implying arbitrage opportunity.

X. Open Questions

Several data points, once available, would materially affect conclusions:

Sample expansion. Scaling from 31 to 50+ pairs is necessary to move from research to live trading.

Execution price data. Using actual transaction prices rather than valuations would more accurately capture true Basis.

Stress period analysis. How does Basis behave under market stress—widen or tighten? How does dispersion change? This determines strategy risk characteristics.

Real-time monitoring. Building real-time Basis monitoring is essential to convert research into an executable trading framework.

Practical Takeaways for PMs

In one sentence:

The core of Basis trading is combining Duration Gap + rate environment analysis, holding 10-20 days, focusing on chemicals/coal/utilities, avoiding real estate.

Execution checklist:

1. Duration Gap threshold: Only go long when Gap ≤ 0.3 years; avoid if > 0.5 years
2. Holding period: Target 10-20 trading days, maximum 90 days (half-life constraint)
3. Stop-loss/exit: 8 bps absolute stop + 90-day time stop; scale out above 30 bps profit
4. Issuer selection: Prioritize China Energy Conservation, China Merchants Port—quality SOEs with strong fundamentals and international ratings; avoid real estate and local platforms
5. Rate environment: Basis more likely to widen during rate hikes—be more cautious; can be more aggressive when rates are stable
6. Position sizing: Single issuer exposure no more than 20% of total Basis strategy portfolio to avoid concentration risk

Data & Methodology

Data from Wind, covering 2024-01-02 to 2026-04-10. Methods include panel regression (fixed effects), ADF stationarity tests, autocorrelation analysis. Full research report available on my GitHub repository.

This analysis is based on public data and does not constitute investment advice. Readers should conduct their own due diligence.

Quinn | quinnmacro.com | April 22, 2026

Tags: Fixed Income · Credit Bonds · Market Microstructure · Relative Value