A nominal bond promises fixed dollars. Inflation decides what those dollars buy. The inflation risk premium (IRP) is the extra yield investors demand for taking that bet, the compensation for inflation being uncertain rather than the compensation for its expected level. It is small, it is invisible, and its sign is not fixed.
What the premium compensates for
Decompose a nominal Treasury yield and you get three pieces: a real yield, expected inflation over the term, and the IRP. The first two would price the bond in a world of certainty. The IRP is what is left because the future inflation path is a distribution, not a point. An investor who dislikes the chance that inflation runs hot and erodes a fixed coupon demands extra yield to hold the nominal bond instead of an inflation-protected one.
That makes the IRP a component of the breakeven. The breakeven is nominal yield minus TIPS yield, so it carries expected inflation plus the IRP. A positive IRP pushes the breakeven above true expected inflation. Reading a breakeven as a clean forecast double-counts the premium.
Why the sign flips
The textbook story says the IRP is positive: inflation risk is bad, so it is paid for. The data say otherwise for the 2000s. The reason is that inflation risk is only bad if it shows up when you can least afford it.
Campbell, Sunderam, and Viceira make this the center of their model in Inflation Bets or Deflation Hedges?. What matters is the nominal-real covariance, whether inflation is high or low when the real economy is weak. Two regimes:
- Countercyclical inflation (the 1970s and early 1980s). Prices spike when growth stalls. A nominal bond loses real value exactly when the rest of your portfolio is hurting. It amplifies your bad states, so it is risky, so the IRP is positive and large.
- Procyclical inflation (the 2000s). Inflation falls when growth falls, because recessions bring deflation scares. Now the nominal bond gains real value in the downturn. It is a deflation hedge, insurance that pays off in the bad state. Investors will accept a lower yield for that, so the IRP goes negative.
The premium tracks the bond-stock correlation because the same covariance drives both. Campbell et al. find the Treasury-stock return covariance was slightly positive on average over 1953 to 2009, unusually high in the early 1980s, and negative in the 2000s, with the deepest readings in the 2000 to 2002 and 2007 to 2009 downturns. The IRP follows. Inflation is priced only because it predicts real consumption growth; when that predictive sign changes, the premium changes sign with it.
The 2022 regime change
For two decades the stock-bond correlation was reliably negative and bonds were the portfolio’s recession insurance. In 2021 it crossed into positive territory and stayed above 0.5 from 2022 through 2024. A common rolling estimate moved from roughly -0.53 in 2020 to +0.50 in 2023.
The driver was inflation becoming the dominant shock again. When the post-pandemic shock was prices rather than growth, the Fed tightened, and stocks and bonds fell together. The rate-hike channel is hard to disentangle here: aggressive tightening hits bond prices and equity valuations at once, so some of the co-movement is a discount-rate shock rather than an inflation-cyclicality story. Morningstar puts the correlation flip near 5% inflation. By the procyclical-countercyclical logic, that configuration should push the IRP back up off its 2000s floor. The 2022 episode is the live test of a mechanism estimated mostly on twentieth-century data.
Why estimates disagree on even the sign
You cannot observe the IRP. You observe yields, and you need a model to split them into expected inflation and premium. Different term-structure models give different splits, and the splits disagree on sign. The Fed’s D’Amico, Kim, and Wei study, Inflation Risk Premium: Evidence from the TIPS Market, puts a representative 10-year IRP at roughly 14 to 19 basis points depending on which proxy you use for expected inflation. Other TIPS-market estimates put the full-sample average between about -16 and +10 basis points, dragged negative in the early years because TIPS were illiquid and the 2002 to 2003 deflation scare dominated. A point estimate and a sample mean are not the same object, which is the point: the literature does not even agree on whether the 10-year premium is normally positive.
So the disagreement is not noise around an agreed number. Models disagree on whether the 10-year premium is positive or negative, on magnitudes of tens of basis points. Two contaminants make it worse. The IRP and the TIPS liquidity premium push the breakeven in opposite directions and are not the same size; in a crunch the liquidity premium blows out while the IRP barely moves. And the premium is genuinely time-varying, so any single point estimate is an average over regimes that may not recur. This is why a breakeven move of 10 basis points is inside the noise: that is the scale of the premium it is trying to net out.
Try it
Reconstruct the inflation-bets vs deflation-hedges signal (2-3 hours, Python + FRED API). The IRP sign should track whether inflation is a recession hedge. Pull daily 10-year breakeven (T10YIE), the S&P 500 (SP500), and a Treasury total-return or 10-year yield series from FRED into pandas. Compute a rolling 2-year correlation between bond and stock returns. Overlay it on realized CPI. What you are looking for: the correlation sitting negative through the 2010s, then crossing into positive territory in 2021 to 2022 as inflation climbs past 5%. That crossing is the regime where the deflation-hedge property of nominal bonds disappears and, by the Campbell-Sunderam-Viceira mechanism, the IRP should be pushing breakevens up rather than down. You will not isolate the premium itself (that needs a no-arbitrage term-structure model) but you can watch the covariance that drives its sign.
See also
- Liability Matching — pensions and insurers hold nominal bonds partly for the deflation-hedge property that makes the IRP negative; the same covariance that prices the premium decides whether the hedge works.
- Anchored Inflation Expectations — the IRP is what makes a breakeven an imperfect read on whether expectations are actually anchored.
Sources
- Inflation Bets or Deflation Hedges? The Changing Risks of Nominal Bonds — Campbell, Sunderam, Viceira. The nominal-real covariance model and the sign-flip mechanism.
- Inflation Risk Premium: Evidence from the TIPS Market — D’Amico, Kim, Wei. Term-structure decomposition and premium magnitudes.
- What Higher Inflation Means for Stock/Bond Correlations — the post-2021 correlation flip and the ~5% inflation threshold.