Original data study · 6J / JPY-USD
U.S.-Japan Yield Spreads and 6J Futures: What the Data Actually Shows
The U.S.-Japan 10-year yield spread is useful context for Japanese yen futures, but it is not a mechanical forecast. A 318-month open-data study finds a meaningful inverse relationship, large regime shifts and no defensible basis for claiming perfect accuracy.
- Sample
- 318 months
- Level correlation
- −0.54
- 12-month changes
- −0.44
- Study period
- 2000–2026
Each series normalized independentlySources: Fed, OECD via FRED
Direct answer
The Spread Is a Regime Variable, Not a Trade Trigger
Wider long-term yield differentials often align with stronger relative-return and policy expectations for the dollar. Because 6J is quoted as U.S. dollars per yen, that environment often aligns with lower 6J prices. The 10-year spread is a macro-regime proxy, not an actual carry-return calculation; the word often matters.
What the sample supports
Directionally useful macro context
The full-sample relationship is negative, and the most extreme spread-widening observations had materially weaker median yen-proxy returns than the strongest narrowing observations.
What it does not support
Guaranteed timing or a fixed coefficient
The rolling relationship weakened to nearly zero in one 36-month window. A spread level cannot reveal whether expectations are already priced, whether a policy surprise is coming or where a futures order will fill.
spread = U.S. 10Y − Japan 10Y
→
JPY/USD proxy = 100 ÷ USD/JPY
CME quotes standard Japanese Yen futures in U.S. dollars per yen. The study scales the cash-market inverse by 100 for readability; that scaling changes the displayed level, not returns or correlations.
Reproducible method
Three Monthly Series, One Honest Proxy
The analysis uses only public monthly observations available through FRED. It does not splice expiring futures contracts or represent a cash-market transformation as a continuous 6J settlement series.
Study construction
For each month from January 2000 through June 2026, subtract Japan’s 10-year benchmark government-bond yield from the U.S. 10-year Treasury yield. Convert the Federal Reserve’s monthly USD/JPY spot average into JPY/USD direction using 100 ÷ USD/JPY. Align by month, then calculate Pearson correlations in levels and in 1-, 3-, 6-, 12- and 24-month changes.
- Frequency
- Monthly
- Window
- Jan. 2000–Jun. 2026
- Complete rows
- 318
- Accessed
- Aug. 7, 2026
FRED series GS10: monthly 10-year Treasury constant-maturity yield from the Federal Reserve Board.
FRED series IRLTLT01JPM156N: OECD monthly 10-year benchmark government-bond yield for Japan.
FRED series EXJPUS: Federal Reserve Board G.5 monthly average Japanese yen per U.S. dollar, constructed from available daily noon buying rates.
Spot and futures share currency direction, but a dated 6J contract also reflects the interest-rate basis, time to delivery, contract selection and roll. Therefore, these results describe the macro relationship relevant to 6J; they are not a futures-return backtest.
Measured findings
A Real Inverse Relationship—With Plenty of Noise
Negative coefficients mean that wider U.S.-Japan spreads tended to align with a lower JPY/USD proxy, which is the same broad direction as lower 6J.
Spread level versus the scaled JPY/USD cash proxy across all 318 months.
Contemporaneous monthly spread changes versus monthly proxy returns.
Year-over-year spread changes versus year-over-year proxy returns.
The strongest measured horizon in this descriptive comparison.
The −0.54 level correlation is descriptive. This study does not test or establish stationarity, cointegration or a stable long-run equilibrium coefficient.
Horizon comparison
The relationship does not become a clean short-term signal
| Change horizon | Observations | Correlation | Reading |
|---|---|---|---|
| 1 month | 317 | −0.41 | Moderate inverse |
| 3 months | 315 | −0.46 | Moderate inverse |
| 6 months | 312 | −0.44 | Moderate inverse |
| 12 months | 306 | −0.44 | Moderate inverse |
| 24 months | 294 | −0.55 | Stronger, still incomplete |
These are descriptive Pearson correlations. Overlapping multi-month returns are not independent observations and should not be treated as separate trades.
Twelve-month change buckets
Extreme widening and narrowing produced different medians
Each tail contains 77 observations. The center contains 152. Returns are for the scaled JPY/USD proxy.
These are overlapping 12-month windows, not 306 independent episodes or trades.
Tail cutoffs were a 12-month spread change of −0.544 percentage point or less and +0.334 point or more. Buckets describe this historical sample; they do not define future thresholds.
Regime dependence
The Same Model Did Not Work Equally Well in Every Era
A full-sample coefficient hides the most important fact: the relationship strengthened and weakened as monetary frameworks, term premiums, risk conditions and carry positioning changed.
| Era | Months | Level correlation | 12-month-change correlation | Evidence-based reading |
|---|---|---|---|---|
| 2000–2007 | 96 | −0.05 | −0.07 | Near-zero fit in this split |
| 2008–2012 | 60 | −0.69 | −0.26 | Strong level relationship, weaker changes |
| 2013–2019 | 84 | −0.28 | −0.25 | Present but modest |
| 2020–Jun. 2026 | 78 | −0.81 | −0.75 | Exceptionally strong recent regime |
Rolling 36-month monthly-change correlation
From −0.85 to −0.03
The strongest window ended December 2023. The weakest ended March 2016. The latest window ending June 2026 measured −0.52.
Why recent fit can mislead
2020–2026 should not be universalized
The post-2020 period combined large policy divergence, Japan’s transition away from yield-curve control, substantial yen-funded carry activity and official intervention. A model calibrated only to that period risks assuming the regime will persist.
The Bank of Japan ended its negative-rate and yield-curve-control framework in March 2024. BIS research also describes carry-trade transmission as state-dependent rather than fixed.
Transparent worksheet
Yield Spread and 6J-Direction Converter
This worksheet performs only the arithmetic used in the study. It does not estimate a fair value, target, probability or trade direction.
Mechanical output
June 2026 source-data example
- U.S.-Japan spread
- 1.80 pp
- JPY/USD per yen
- 0.0062201
- JPY/USD per 100 yen
- 0.62201
- Quote direction
- Inverse of USD/JPY
A dated 6J future will not equal spot inversion exactly because futures basis and time to delivery matter.
Failure modes
Six Reasons the Spread Can Stop Explaining 6J
A useful macro relationship can still fail as a trading model. These are not edge cases; they are part of the market being measured.
Exchange rates respond to surprises in the expected policy path. A visible yield level may already be priced.
The 10-year spread is not a pure policy-rate differential. Supply, inflation risk and bond term premiums matter.
Carry unwinds, deleveraging and safe-haven demand can strengthen the yen without a matching spread move.
Official dollar sales and yen purchases can move the currency abruptly. See the verified intervention event study.
6J includes basis, expiry and roll. Contract selection and execution can differ from the cash proxy.
Both yields and the yen react to common macro news. The study is descriptive and does not identify a structural causal coefficient.
Use the spread to describe the macro backdrop, then separately verify policy expectations, risk conditions, event timing, contract month, liquidity and position risk. The evidence does not justify using one spread threshold as an automatic 6J entry or exit.
Frequently asked questions
6J Yield-Spread Questions
Do wider U.S.-Japan yields always make 6J fall?
No. They aligned with lower JPY/USD values on average in this sample, but the relationship ranged from strong to nearly absent across rolling windows. Other information can dominate.
Why use 100 divided by USD/JPY?
The Federal Reserve series reports yen per U.S. dollar. Standard 6J is quoted in U.S. dollars per yen, so inversion puts the cash-market series in the same direction. Multiplying by 100 makes the small number easier to read without changing returns or correlation.
Is the −0.54 correlation a prediction score?
No. It is a full-sample contemporaneous level correlation. The study does not establish stationarity, cointegration or a stable equilibrium coefficient, and the statistic does not measure out-of-sample forecast accuracy, fill quality, profitability or risk-adjusted returns.
Why not claim this is a continuous 6J futures backtest?
Because it is not one. A continuous futures series requires an explicit contract-selection and roll method plus licensed or otherwise authorized settlement data. This study deliberately uses a reproducible public cash proxy and labels that limitation.
Sources, calculations and editorial disclosure
- Federal Reserve Bank of St. Louis: 10-Year Treasury Constant Maturity Rate (GS10), monthly, sourced from the Board of Governors of the Federal Reserve System.
- FRED / OECD: Japan 10-Year Benchmark Government Bond Yield (IRLTLT01JPM156N), monthly.
- Federal Reserve Board G.5: Foreign Exchange Rates and FRED series EXJPUS, monthly averages of available daily noon buying rates.
- CME Group FX Product Guide 2026 for the 6J code, 12,500,000-yen contract size, USD-per-JPY quotation, tick and physical-settlement specification.
- Federal Reserve IFDP Note: The Sensitivity of the U.S. Dollar Exchange Rate to Changes in Monetary Policy Expectations for evidence that surprise effects vary across currencies and time.
- Bank for International Settlements: Monetary Policy Transmission to Exchange Rates—The Role of Currency Carry Trades for state-dependent carry-trade transmission.
- Bank of Japan: Changes in the Monetary Policy Framework, March 19, 2024 for the end of the negative-rate and yield-curve-control framework.
Source series were accessed August 7, 2026. Grizzly Parrot Trading aligned the monthly observations and calculated the transformations, correlations, quartiles, medians and era splits shown above. Values are rounded for display; calculations used unrounded source observations. This is original editorial analysis, not sponsored research, and there is no paid relationship with the cited institutions.