Is it Real Interest Rates or Expected Inflation that Moves US Long-Term Interest Rates?

The US October CPI, released on November 10, 2026, rose 0.4% month-on-month, exceeding the market forecast of 0.3%. Core CPI also rose 0.3%, exceeding the forecast of 0.2% (Eurasia Business News). On the same day, the US 10-year Treasury yield rose 10bp from 4.88% to 4.98%. When long-term interest rates rise due to an upward surprise in inflation indicators, is the market pricing in future inflation, or is it pricing in a rise in real interest rates?
Nominal interest rates can be decomposed into the real yield of Treasury Inflation-Protected Securities (TIPS) and the break-even inflation rate, which is the yield spread between nominal Treasury bonds and TIPS. In this paper, using daily data from 2003 to 2026, we examined how nominal interest rates changed on the days of CPI, employment statistics, and FOMC announcements.
The analysis results show that on days of CPI and employment statistics releases, about 70-80% of the change in nominal interest rates was due to real interest rates, and this ratio was almost unchanged from normal days. On the other hand, on FOMC days, the contribution of real interest rates exceeded the change in nominal interest rates, and the break-even inflation rate tended to move in the opposite direction. On the CPI release day of November 10, 2026, almost all of the interest rate increase was due to real interest rates, and the contribution of real interest rates was larger than on past CPI release days.
Beechey & Wright (2009) divided the reaction of long-term interest rates to macroeconomic indicator announcements into real interest rates and inflation compensation, showing that real interest rates mainly react to real economic indicators such as employment statistics. Gürkaynak, Levin & Swanson (2010) reported that in the US, even long-term forward inflation compensation reacts to economic news, and that long-term inflation expectations are not necessarily stable. In this paper, we extend the analysis period to 2026 and verify the differences between CPI, employment statistics, and FOMC, as well as the reaction of long-term inflation compensation.

Data

The sample consists of daily changes for 5,967 business days from January 3, 2003, to November 9, 2026. For November 10, 2026, since it was before publication on FRED, we used the US Treasury’s published values as a case study.

• Constant maturity yields for Treasury securities (DGS5, DGS10): FRED

• Constant maturity real yields for TIPS (DFII5, DFII10): FRED

• Break-even inflation rate (T5YIE, T10YIE (Nominal – Real)): FRED

• 5-year, 5-year forward inflation expectation rate (T5YIFR): FRED

• Yields for November 9-10, 2026 (Nominal and real constant maturity yields): US Treasury Daily Treasury Par Yield Curve Rates, Par Real Yield Curve Rates

• Event days (CPI and employment statistics are ALFRED’s first release dates, FOMC is the FRB statement date): ALFRED, FRB

Confirmation of identity: The break-even inflation rate on FRED is created as the difference between nominal and real. For both 5-year and 10-year, across all 5,968 days, the error in BEI = Nominal – Real was 0bp.
Creation of constant maturity yields: According to the US Treasury, nominal constant maturity yields are values from a yield curve estimated from indicative prices (not actual trades) based on the ask side of the most recently auctioned securities, obtained by the New York Fed around 3:30 PM each business day. There are two breaks in the series. For nominal yields, the estimation method changed to a monotone convex spline on December 6, 2021, and for TIPS, starting December 1, 2008, the most recently auctioned securities rather than all securities were used as nodes for the curve.
Event calendar: The sample period includes 285 CPI releases, 286 employment statistics releases (from 2003 onwards), and 65 FOMC meetings (from 2019 onwards, including unscheduled meetings). PPI, GDP, and PCE release days were categorized as “Other.” Since FOMC dates are only maintained from 2019 onwards, FOMC days from 2003-2018 are included in non-event days (approximately 2.7% of non-event days). Three days where two of CPI, employment statistics, and FOMC overlapped were excluded from individual groups. Releases after 15:00 ET and holiday releases were assigned to the next business day.

Methodology

Let ytN(m)y^N_t(m)ytN​(m) be the nominal yield for maturity $m$ (5 years, 10 years), ytR(m)y^R_t(m)ytR​(m) be the real yield of TIPS, and πtBE(m)=ytN(m)−ytR(m)pi^{BE}_t(m) = y^N_t(m) − y^R_t(m) πtBE​(m)=ytN​(m)−ytR​(m) be the break-even inflation rate. Daily changes (bp, from the previous business day’s closing price to the current day’s closing price) can be divided as follows without using a model:
ΔytN(m)=ΔytR(m)+ΔπtBE(m)Delta y^{N}_t(m) = Delta y^{R}_t(m) + Delta pi^{BE}_t(m)ΔytN​(m)=ΔytR​(m)+ΔπtBE​(m)
Ratio of real interest rates: For each group g (CPI, employment statistics, FOMC, other, non-event days), we define the ratio of the variance of nominal changes accounted for by changes in real interest rates as follows:
βgR=Covg(ΔyR,ΔyN)Varg(ΔyN),βgBE=1−βgRdisplaystyle{beta^{R}_g = frac{mathrm{Cov}_g(Delta y^{R}, Delta y^{N})}{mathrm{Var}_g(Delta y^{N})}, qquad beta^{BE}_g = 1 – beta^{R}_g}βgR​=Varg​(ΔyN)Covg​(ΔyR,ΔyN)​,βgBE​=1−βgR​
Covgmathrm{Cov}_gCovg​ and Vargmathrm{Var}_gVarg​ are the covariance and variance calculated for days belonging to group g. By identity, βBEg=1−βgRbeta^BE_g = 1 − beta^R_gβBEg​=1−βgR​ holds strictly. βgR>1beta^R_g > 1βgR​>1 occurs when changes in expected inflation rates negatively covary with changes in nominal rates.
Testing:For the difference in βRbeta^RβR between each group and non-event days, we calculated 95% confidence intervals using bootstrapping (4,999 iterations) with replacement within groups, and p-values using a permutation test by shuffling group labels. Since FOMC data only exists from 2019 onwards, the comparison group was also limited to non-event days from 2019 onwards. Multiple comparisons were Holm-corrected for each table.
Difference-in-differences. For CPI, employment statistics, and others, the difference in βgR−βnoneRbeta^R_g − beta^R_{none}βgR​−βnoneR​ between 2022 onwards and 2021 and earlier was evaluated using the same method. The boundary date of January 1, 2022, was fixed in advance to allow comparison with the ACM article.
Long-term expected inflation. The standard deviation of daily changes in the 5-year, 5-year forward inflation expectation rate (T5YIFR) was compared between CPI/employment statistics days and non-event days (ratio = sdg/sdnonesd_g / sd_{none}sdg​/sdnone​). If the ratio is close to 1, long-term expected inflation is not particularly moved by indicator releases.
Sensitivity analysis. Estimates were also performed excluding the periods of September 2008–March 2009 and March–April 2020, when TIPS liquidity distortions were large. The exclusion periods were decided before looking at the results.

Python code (partial)

In the permutation test, it is necessary to recalculate the slope every time the group labels are shuffled. By using sufficient statistics (count, sum, sum of squares, sum of products), the slope of the other group can be obtained simply by subtracting the statistics of the selected days from the statistics of the entire sample.

import numpy as np

def beta_real(dy_nom, dy_real):
    """beta^R = Cov(Δ実質, Δ名目) / Var(Δ名目). 最後の軸に沿って計算."""
    xm = dy_nom - dy_nom.mean(-1, keepdims=True)
    return (xm * (dy_real - dy_real.mean(-1, keepdims=True))).sum(-1) / (xm ** 2).sum(-1)

def permutation_diff(x_g, y_g, x_0, y_0, B=4999, seed=0):
    """群 g と非イベント日の beta^R の差の並べ替え分布 (十分統計量で補集合を計算)."""
    rng = np.random.default_rng(seed)
    x, y = np.r_[x_g, x_0], np.r_[y_g, y_0]
    n, n1 = len(x), len(x_g)
    idx = np.argsort(rng.random((B, n)), axis=1)[:, :n1]
    xs, ys = x[idx], y[idx]
    s1 = (n1, xs.sum(1), ys.sum(1), (xs * xs).sum(1), (xs * ys).sum(1))
    tot = (n, x.sum(), y.sum(), (x * x).sum(), (x * y).sum())
    s2 = tuple(t - u for t, u in zip(tot, s1))
    slope = lambda s: (s[4] - s[1] * s[2] / s[0]) / (s[3] - s[1] ** 2 / s[0])
    return slope(s1) - slope(s2)

Result 1: Contribution ratio of real interest rates on CPI and employment statistics days

The contribution ratio of real interest rates βRbeta^RβR to daily changes in nominal interest rates was 0.75 for 5-year bonds and 0.73 for 10-year bonds on normal days. On CPI and employment statistics release days, it was also around 0.7–0.8, and the difference from normal days was not statistically significant.

The real interest rate contribution ratio is the covariance ratio to the daily change in nominal interest rates. The differences between event days and non-event days were not significant after Holm correction, and even when comparing 2022 onwards with earlier periods, no change in the contribution ratio on CPI and employment statistics release days was detected. Even with the transition to a high-inflation phase, there is no clear change in the breakdown of interest rate fluctuations.
Beechey & Wright (2009) reported that the reaction of long-term interest rates to announcements such as employment statistics is mainly due to real interest rates, while inflation compensation also reacts to price indicators. While this paper also shares the finding that the contribution of real interest rates is large, we did not obtain results showing that the contribution of inflation compensation increases specifically on CPI release days. Beechey & Wright (2009) measure surprise reactions to unexpected news from high-frequency data immediately after announcements, whereas this paper compares covariance contribution ratios of daily changes, and the two capture different aspects. With this method, no evidence is obtained that the contribution ratios of real interest rates and break-even inflation rates on CPI release days differ from those on normal days.

Result 2: Real interest rates and expected inflation on FOMC days

Unlike CPI and employment statistics, the contribution ratio of real interest rates was high on FOMC days. On FOMC release days from 2019 onwards, the real interest rate contribution ratio (βRbeta^RβR) was 1.01 for 5-year bonds and 0.93 for 10-year bonds, exceeding the approximately 0.75 for non-event days.

When periods with large liquidity distortions were excluded, the difference from non-event days was significant for both 5-year and 10-year bonds after Holm correction (corrected p-values: 0.011, 0.015). It was not significant in the main analysis.
βR>1beta^R>1βR>1 means that real interest rates tend to move more in the same direction as changes in nominal interest rates, while the break-even inflation rate tends to move in the opposite direction. This is consistent with the monetary policy credibility channel, where monetary tightening pushes up real interest rates while suppressing future inflation prospects.
However, this result depends on the treatment of periods when TIPS market liquidity declined, such as March 2020. Beechey & Wright (2009) also reported the reaction of inflation compensation to monetary policy announcements, and this paper verified that relationship with data from 2019 onwards. A bias in the contribution to real interest rates, which was not seen on CPI and employment statistics days, is observed on FOMC days.

Result 3: Fluctuations in Long-Term Inflation Compensation on CPI Days

Regarding the 5-year forward inflation compensation starting 5 years from now, we compared the daily fluctuations on CPI and employment report release days with those on non-event days.
The standard deviation of changes on CPI release days was 0.86 times that of non-event days, with a 95% confidence interval of 0.62 to 1.17. Excluding periods when TIPS market liquidity was low, this figure became 1.05 times, and in neither case was a result obtained showing that fluctuations were larger than on normal days. The same was true for employment reports.
On CPI release days since 2022, it was actually smaller at 0.76 times, but this was not significant in a permutation test.
Gürkaynak, Levin & Swanson (2010) report that U.S. long-term forward inflation compensation reacts significantly to economic news. In this paper, using daily data from 2003 to 2026, we did not obtain a result showing that fluctuations in long-term inflation compensation expand on CPI release days compared to normal days. While previous studies examined reactions to news immediately after release, this paper compares the magnitude of daily fluctuations.
At least in daily data, no tendency for long-term inflation compensation to fluctuate more than on normal days was confirmed, even on days when CPI or employment reports were released.

Recent Case: CPI Upside Surprise on November 10, 2026

On the CPI release day of November 10, 2026, almost all of the rise in nominal interest rates was due to a rise in real interest rates. Comparing the values published by the U.S. Department of the Treasury with the previous day, the results are as follows.

For the 5-year maturity, real interest rates rose by 14bp, the same as nominal interest rates, and for the 10-year maturity, they rose by 12bp, exceeding the 10bp rise in nominal interest rates. Meanwhile, the breakeven inflation rate remained flat or fell slightly. The 2-year nominal government bond yield also rose by 17bp. This movement is consistent with the interpretation that the upside surprise in inflation indicators was priced in not as a rise in long-term inflation compensation, but as a rise in real interest rates reflecting future policy rate outlooks, etc.
Even if inflation indicators exceed expectations, long-term inflation compensation does not necessarily rise. November 10 is a concrete example of this.

Robustness and Limitations

The breakeven inflation rate includes not only expected inflation but also inflation risk premiums and TIPS liquidity premiums. The decomposition in this paper targets inflation compensation observed in the market, not pure expected inflation rates.
The result that the contribution of real interest rates increases on FOMC days became statistically significant excluding March–April 2020, when TIPS market liquidity was low. Therefore, this result depends on the treatment of the excluded period. Also, the analysis period for the FOMC is limited to 2019 onwards.
Since this paper uses daily data, it includes not only market reactions immediately after release but also reactions to other news on the same day. This is not an analysis that identifies the causal impact of individual economic indicator surprises on interest rates.

Conclusion

In daily data from 2003 to 2026, the contribution ratio of real interest rates on CPI and employment report days was approximately 70–80%, with no significant difference from normal days. This relationship remained unchanged even during the high-inflation phase since 2022, and the long-term breakeven inflation rate did not show fluctuations greater than on normal days on CPI days. On the other hand, on FOMC days, the contribution ratio of real interest rates was around 1, and a tendency was observed for the breakeven inflation rate to move in the opposite direction to nominal interest rates. However, this difference became statistically significant only when excluding periods when TIPS market liquidity was low.
The CPI upside surprise on November 10, 2026, also saw almost all of the rise in nominal interest rates driven by real interest rates. Even if inflation indicators exceed expectations, long-term inflation compensation does not necessarily rise.
ACM Analysis in the Decomposition of U.S. Treasury Yields in the Previous Article showed that on employment report and FOMC days since 2022, the change in the 10-year yield had a greater contribution from the expected short-term interest rate path than from the term premium. In this TIPS analysis, the contribution ratio of real interest rates on employment report days was no different from normal days, while the contribution to real interest rates increased on FOMC days. Although ACM captures different aspects—expected short-term interest rate path and term premium—and TIPS captures real interest rates and inflation compensation, both results indicate that fluctuations in long-term interest rates accompanying the release of economic indicators and monetary policy cannot be explained solely by changes in expected inflation.
When long-term interest rates rise, it is necessary to consider changes in real interest rates and future policy rate outlooks together, rather than immediately concluding that inflation expectations have risen.

*This article is for informational purposes only and does not recommend investment in any specific asset. Please make investment decisions at your own responsibility.

Sources

• Federal Reserve Bank of St. Louis, FRED (DGS5, DGS10, DFII5, DFII10, T5YIE, T10YIE, T5YIFR), ALFRED (Initial release dates for CPIAUCSL, PAYEMS)

• U.S. Department of the Treasury, Daily Treasury Par Yield Curve Rates, Daily Treasury Par Real Yield Curve Rates (Accessed November 12, 2026)

• Board of Governors of the Federal Reserve System, FOMC Meeting Calendars

• Eurasia Business News, Stock Market Today: Dow Jones, Nasdaq, S&P 500 Fall on November 10, 2026

• Holm, S. (1979). A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics, 6(2), 65–70. JSTOR 4615733

References

• Beechey, M. J., & Wright, J. H. (2009). The high-frequency impact of news on long-term yields and forward rates: Is it real? Journal of Monetary Economics, 56(4), 535–544. doi:10.1016/j.jmoneco.2009.03.011, ScienceDirect

• Gürkaynak, R. S., Levin, A., & Swanson, E. (2010). Does inflation targeting anchor long-run inflation expectations? Evidence from the U.S., UK, and Sweden. Journal of the European Economic Association, 8(6), 1208–1242. doi:10.1111/j.1542-4774.2010.tb00553.x