Inflation Beta Is Not One Number

Four careful studies have measured the inflation beta of U.S. stocks. Vanguard’s 2019 paper put the S&P 500 at +1.07 against unexpected inflation, with a p-value of 0.28.¹ Vanguard’s 2026 follow-up, using its capital markets model, puts U.S. equities at +0.07.² A 2026 Review of Financial Studies paper finds −1.33 against headline inflation and −5.60 against core inflation, on the same stocks over the same years.³ A 2025 study of eighteen countries reports −2.93, −3.81, −4.34 or −4.60 depending on which of its four specifications you read.⁴ Every one of these is a defensible estimate of the same two words, and the spread between them is the useful fact.

Inflation beta is a regression slope, and like every regression slope it is a statement about a particular pair of series over a particular sample. Change the inflation series, the definition of a shock, the horizon of the returns or the years in the window and the number moves, sometimes across zero. Once you see why, the number becomes far more useful than the “inflation hedge” rankings it usually gets compressed into. The companion piece, Which Inflation Are You Hedging?, covers what a TIPS fund, a breakeven position and a commodity basket each protect against. The question here is narrower: what does the number itself tell you, and what does it leave out?

What the regression estimates

Inflation beta is defined the same way as market beta. Regress an asset’s return on a measure of inflation and read off the slope:

Here $R_{i,t}$ is the asset’s return in period $t$ and $\pi_t$ is whatever the researcher has chosen to call inflation in that period. If both are in percentage points, a beta of 1 means that, over the sample, the asset’s return was on average one percentage point higher in periods when the inflation measure was one point higher, with everything else the asset responds to left in the error term. Vanguard’s 2026 paper states the formula in exactly these terms, as the covariance of an asset’s return with inflation divided by the variance of inflation.²

The appeal of the slope is that it scales. Vanguard’s 2019 author Paul Bosse works the example: if unexpected inflation rises 4 percentage points, short-term TIPS with a beta near 1 might return about 4%, while commodities with a beta near 6 might return about 24%.¹ A dollar of the high-beta asset offsets inflation damage across several dollars of the rest of the portfolio, which is why institutions talk about buying “inflation beta” in a small sleeve rather than rebuilding the whole allocation. The 2026 Vanguard paper makes the same move: a 60/40 portfolio has an inflation beta near 0.07 in its model, and a 15% satellite of mostly commodities lifts the whole portfolio to a beta of 1.0.²

Everything in that logic rests on the slope being a stable property of the asset, and four separate choices in the measurement move it.

Beta, correlation and R² answer three different questions

A beta of 6 sounds like a tighter relationship than a beta of 1. It usually is the opposite. The slope, the correlation and the asset’s own volatility are linked by one identity:

A volatile asset can post an enormous beta with only a modest correlation, because its standard deviation $\sigma_i$ is many times that of inflation. Correlation answers “how reliably do these two move together?” Beta answers “by how much, on average?” R², the square of the correlation in a one-variable regression, answers “what share of this asset’s variance does inflation account for?” An inflation hedge you would rely on needs a good answer to all three.

Vanguard’s 2026 paper prints both columns for the same seven assets, and the ranking changes between them.²

Short-term TIPS have the highest correlation and a beta below 1. Commodities have a lower correlation and a beta nine times larger. Vanguard’s own gloss is that some assets “may track inflation closely but still provide only limited protection for the broader portfolio,” and that if inflation rose 3%, TIPS would be expected to gain roughly 2.6% while commodities could rise more than 20%.² Both statements are true. The first is about correlation; the second is about beta.

The 2019 paper shows the third leg. Its 1991 to 2018 regressions give gold a beta of +1.19 and REITs +2.01, both larger than short-term TIPS at +0.98. But the t-statistics are 1.40 and 1.65 against 5.52, so gold and REITs are “not significant” in Vanguard’s table while short-term TIPS are significant at any conventional level.¹ A large point estimate with a wide standard error is a guess dressed as a measurement. When you see an inflation beta quoted alone, ask for the correlation and the t-statistic before believing the size.

Expected inflation is already in the price

The second thing to notice about any published inflation beta is what the researcher used for $\pi_t$. Irving Fisher argued in 1896 that nominal interest rates embed the inflation lenders expect, so that a bond yield is roughly a real rate plus expected inflation.⁵ If the market already expects 3% inflation and 3% arrives, nothing has to reprice. The event that moves asset prices is inflation arriving above or below what was priced. Fama and Schwert made this the organizing idea of the empirical literature in 1977. Over 1953 to 1971 they found that Treasury bills and government bonds were a complete hedge against expected inflation, that residential real estate hedged both the expected and the unexpected component, and that common stocks were negatively related to expected inflation and, more weakly, to unexpected inflation as well.⁶

Vanguard’s 2019 correlations make the same split visible for cash. Treasury bills correlate about +0.75 with expected inflation over the full history and about zero with the unexpected component.¹ Bills reset to the prevailing rate every few months, so they follow inflation that has already shown up in yields, and they cannot react to a surprise until the next reset. That is the difference between an asset that keeps up with inflation and one that pays off when inflation surprises.

The trouble is that “unexpected inflation” has no observable value. It must be constructed. Bosse extracts it with a one-sided Hodrick-Prescott filter on quarterly CPI, checks it against the Survey of Professional Forecasters and a moving average, and adds the caveat that “our definition of unexpected inflation is a manufactured one based on our best attempt to build it.”¹ AQR’s 2026 paper blends two definitions, the change in year-over-year CPI from a year earlier and the gap between realized CPI and the forecast professional forecasters made twelve months before, arguing that neither is perfect and the combination reduces noise.⁷ The cleanest measurements come from the day a CPI number is released. Knox and Timmer at the Federal Reserve Board find that a CPI print one percentage point above the median forecast knocks about 0.8% off U.S. stocks that day and 3.75% in real terms over five days, and that most of the five-day move comes from a higher equity risk premium rather than from higher nominal yields. Two-year real yields fall by 24 basis points, which they read as investors treating the surprise as a supply shock the Fed will look through.⁸

Four choices that move the number

Which inflation

Fang, Liu and Roussanov separate headline CPI shocks into a core component and an energy component and estimate betas to each, for eight asset classes at quarterly frequency from 1963 to 2019. The headline column looks like the classic literature. The split looks different.³

The stock row is the one to remember. Against headline inflation the beta is −1.33 with a t-statistic of −1.38, which is the familiar “stocks are a weak, insignificant inflation hedge” result. Against core inflation it is −5.60 with a t-statistic of −3.69, one of the strongest relationships in the table. The headline number is a blend: a robustly negative response to core inflation partly cancelled by a small positive response to energy inflation. REITs show the same pattern more sharply, a headline beta near zero hiding a core beta of −6.54. Commodities are the mirror image: a headline beta of +8.59 that is almost entirely energy exposure, with a core beta indistinguishable from zero. The authors conclude that stocks, currencies, commodities and real estate look like headline hedges largely because they hedge energy inflation, and that no long position in any of the eight asset classes hedged core inflation.³ (In a separate NBER summary of the paper, commodities come out best of the real assets against core inflation, which the near-zero core beta in the table is consistent with: best of a weak field.)

The pattern reproduces on free monthly data. In a Summitward regression of the S&P 500’s trailing twelve-month return on the change in twelve-month inflation, 1960 to 2026, the headline beta is −0.87 with a t-statistic of −0.8, and the core beta is −3.23 with a t-statistic of −3.4.⁹ Different data, different shock, different frequency, same shape.

Which shock

The published literature uses at least six different objects as “the inflation shock,” and they are ordered here roughly from crudest to sharpest.

• The level of realized inflation. The original Fisher-hypothesis tests and many international comparisons regress nominal returns on inflation itself. Gultekin’s 26-country study is the classic; it found little support for a one-for-one relation in most countries.¹⁰ The level confounds what was expected with what was not.
• The change in year-over-year inflation. This treats last year’s inflation as the forecast, which is the random-walk assumption behind the first of AQR’s two measures.⁷
• Realized inflation minus a survey forecast. AQR’s second measure uses the Survey of Professional Forecasters. Kim, Ranaldi and Schularick digitized OECD Economic Outlook forecasts back to 1965 to do the same across eighteen countries.⁴
• A filtered trend. Bosse’s one-sided Hodrick-Prescott filter is a statistical stand-in for expectations that needs no survey.¹
• A VAR innovation. Fang, Liu and Roussanov model inflation jointly with other macro series and treat the forecast error as the shock.³
• A market or announcement surprise. Changes in breakeven inflation, available only since 2003, or the CPI print minus the median forecast on release day, as in Knox and Timmer.⁸

Kim, Ranaldi and Schularick show how much the choice matters while holding the data fixed. On the same eighteen-country panel from 1965 to 2023, the beta of real equity returns to an inflation surprise is −4.34 using survey-based surprises with no controls, −2.93 with country fixed effects and macro controls, and −4.60 or −3.81 using their model-based surprise instead. Ten-year government bonds range from −2.36 to −2.88.⁴ A Summitward comparison over 1991 to 2018 makes the same point with gold. Regressed on the month’s realized inflation, gold’s beta is +3.19 with a t-statistic of 4.3. Regressed by Bosse on his filtered surprise over the same years, it is +1.19 and not significant.⁹ Same metal, same twenty-eight years, opposite verdict.

Which horizon

A slope estimated on one-day returns, monthly returns, annual returns and five-year returns is four different slopes. Knox and Timmer’s one-day stock response to a one-point CPI surprise is about −0.8%; by day five the nominal response is −1.42% and the real response −3.75%.⁸ At the other extreme, Boudoukh and Richardson found that the relation between stock returns and inflation, near zero or slightly negative at one-year horizons in more than a century of U.S. and U.K. data, turns positive at five-year horizons, with expected inflation passing through nearly one-for-one.¹¹ Gorton and Rouwenhorst found the same horizon effect for commodity futures: their positive correlation with inflation, unexpected inflation and changes in expected inflation grows as the holding period lengthens from a month to a year to five years.¹²

Kim, Ranaldi and Schularick add the most instructive case. Housing has a static beta of roughly zero on impact, which would make it a perfect hedge by the usual reading. Their local-projection estimates show house prices then decline persistently over the following years after a large inflation surprise, while equities and bonds fall at once and recover only partly.⁴ The same asset is a hedge at one horizon and a casualty at another. Vanguard’s 2026 beta sits at a different extreme again: it is the slope across 10,000 simulated paths at year 30, chosen because long-horizon relationships in the model are more stable than short ones.² That is a legitimate object, and it is a different object from a regression on history.

Which sample

Bosse’s Figure 4 plots three-year rolling betas from 1973 to 2018. Short-term TIPS trace a nearly flat line near 1. The S&P 500 swings between roughly −20 and +20. Gold is “generally positive but volatile, with some significant negative spikes.” Commodities are positive but peak near 40 around 2008 before falling back.¹ Vanguard’s own headline commodity beta has moved with the sample: about 6 for the Bloomberg Commodity Index over 1991 to 2018 in the 2019 paper, 7.60 in the 2026 model.^{1, 2}

A Summitward regression of the S&P 500’s monthly return on monthly headline CPI inflation, decade by decade, shows what the rolling chart is summarizing.⁹

The 1990s and the 2010s both produce betas that are significant at conventional levels. They have opposite signs. Over the whole 1960 to 2026 sample the beta is −0.50 with a t-statistic of −1.2 and an R² of 0.002, which is the usual full-sample conclusion that stocks and monthly inflation have almost nothing to do with each other. The sixty-month rolling beta behind that average ranged from −9.59 to +11.67, was negative in 68% of windows and changed sign 23 times.⁹

The academic literature explains the instability as a change in what an inflation shock means rather than as noise. Boons, Duarte, de Roon and Szymanowska find that both the quantity of inflation risk in stocks and the price investors charge for it vary through time and changed sign around the early 2000s, and they tie the sign to whether inflation predicted lower or higher future consumption growth in that era.¹³ Campbell, Sunderam and Viceira document the same regime change in nominal Treasuries over 1953 to 2014: the covariance between bonds and stocks was strongly positive in the early 1980s, when inflation was the dominant risk, and negative in the early 2000s, when bonds had become deflation hedges.¹⁴ Fang, Liu and Roussanov find that the energy betas of stocks, currencies, commodities and REITs are significantly larger after 1999 than before, which is their explanation for the changing stock-bond correlation.³ Cieslak and Pflueger’s review organizes all of this as “good” inflation from demand shocks, which arrives with growth, and “bad” inflation from supply shocks, which arrives without it; the sign of an asset’s inflation beta depends on the mix of the two in the sample.¹⁵ Kim, Ranaldi and Schularick push back on the source of the shock mattering and point to the monetary response instead: in fixed exchange rate regimes, where the central bank could not hike, equity returns barely moved after inflation surprises.⁴ Either way, the covariance is a fact about an era’s shocks and policy, and a regression averages across eras.

There is one more reason a single slope can mislead. AQR’s 2026 paper sorts its sample into falling, stable and rising inflation and finds that equities do badly in rising inflation and below average in falling inflation, a frown, while trend-following strategies do well at both extremes, a smile, and commodities and breakevens are roughly linear.⁷ A straight line fitted through a frown or a smile reports a slope near zero for a relationship that is anything but flat.

Try it: the rolling inflation-beta explorer

The explorer below runs the regression in your browser on monthly data from 1960 to 2026: CPI-U headline, core and energy from FRED, the S&P 500 and a 10-year Treasury approximation from Robert Shiller’s series, 3-month bills from FRED, and gold and two commodity spot indices from the World Bank. Choose an asset, an inflation series, a shock definition and a rolling window. The line is the slope re-estimated over each trailing window; the dashed line is the full-sample slope; the table beneath gives the full-sample beta, t-statistic, correlation and R² for all three inflation series at once. Watch how far the estimate moves when only the definition changes.⁹

The default view, the S&P 500 against monthly headline inflation with core overlaid, shows the two rolling lines spending long stretches on opposite sides of zero. Switch the asset to commodities and the headline line stays positive in about 92% of sixty-month windows while ranging from −3 to +22, which is what a stable sign with an unstable magnitude looks like. Switch the shock to the twelve-month change and the commodity headline beta rises to +8.66 with an R² of 0.53, in the neighbourhood of Bosse’s +5.86 for the Bloomberg index and +9.28 for the S&P GSCI, while T-bills, whose monthly beta is +0.37 with a correlation of +0.45, fall to a beta of −0.19 that is indistinguishable from zero. Cash tracks the level of inflation and ignores changes in it, exactly as Fama and Schwert and Bosse describe.⁹ These are spot commodity indices and a yield-based bond approximation, so the figures are illustrations of the method rather than fund-tracking estimates; the sources note below the tool spells out the limits.

A good hedge can have a low expected return

Inflation beta describes exposure. It says nothing about whether the exposure is rewarded, and in asset pricing those are separate quantities. In a one-factor model the expected excess return is the exposure times the price of that risk:

If unexpected core inflation is a bad state of the world, an asset that pays off in that state is insurance, and investors will hold it at a lower expected return. An asset that gets hurt in that state has to offer more. Fang, Liu and Roussanov estimate the prices of risk and find that only core inflation carries a significantly negative price; the cost of hedging headline or energy inflation comes out near zero.³ Cieslak and Pflueger summarize the magnitude: one unit of core-inflation beta is worth about one percentage point a year of expected return.¹⁵ Bekaert and Wang, looking at bond and equity indices in more than forty-five countries, find that standard securities are poor hedges almost everywhere, that adding bills, foreign bonds, real estate and gold only partly helps, and that the inflation risk premium estimated in the term-structure literature they survey is both sizeable and time-varying.¹⁶ Boons and coauthors show the price itself changed sign in the early 2000s.¹³

For an investor the implication is that a reliable inflation hedge should be expected to lag over the long run, because reliability in the bad state is what you are paying for. A high-beta asset that also has a high long-run return is more likely a bet on energy prices than a hedge on core inflation. The Summitward guide to boring inflation hedges works through the instruments that accept that trade explicitly.

Asset by asset

Nominal bonds

A nominal bond promises fixed dollars, so an inflation surprise lowers the real value of every coupon and, if yields rise with it, the price as well. The longer the duration, the larger the second effect. The evidence is as consistent as anything in this literature: Fang, Liu and Roussanov put Treasuries at −2.53 against headline and −2.51 against core, both with t-statistics beyond −4;³ Kim, Ranaldi and Schularick put ten-year bonds between −2.36 and −2.88 across specifications;⁴ Bosse reports −0.60 for the U.S. aggregate at quarterly frequency;¹ and the Summitward ten-year approximation gives −1.69 on the twelve-month change, with a rolling sixty-month beta that was negative in 90% of windows.⁹ Neville, Draaisma, Funnell, Harvey and van Hemert find ten-year Treasuries returned −5% a year in real terms across the eight U.S. inflationary regimes since 1926 against +4% in other years.¹⁷ The sign is the most stable of any asset here. Even so, Campbell, Sunderam and Viceira’s regime evidence says the bond-stock covariance that rides on it has flipped before and can flip again.¹⁴

TIPS

TIPS are the one asset whose cash flows are contractually indexed to CPI, and their inflation beta still depends on duration. Bosse’s short-term TIPS index has a beta of +0.98 with a t-statistic of 5.52 and the flattest rolling line in his study; the all-maturity index has a beta of +0.66 that is only borderline significant, because a longer real duration lets real-yield moves swamp the CPI accrual.¹ 2022 is the case study. CPI-U averaged 8.00% higher in 2022 than in 2021, and the ten-year TIPS real yield went from −1.04% at the end of 2021 to +1.58% at the end of 2022 while the ten-year breakeven fell from 2.56% to 2.30%.¹⁸ The Bloomberg U.S. TIPS index returned −11.85% for the year. Across funds, duration ordered the outcome: Vanguard’s short-term TIPS ETF returned −2.84%, the iShares and Schwab broad TIPS ETFs −12.13% and −11.96%, and PIMCO’s 15+ year TIPS ETF −31.60%.¹⁹ One inflation, one indexation formula, three outcomes thirty points apart. AQR’s summary of the episode is that “gold, TIPS and breakevens disappointed as real yields soared but long-term inflation expectations remained anchored.”⁷ The companion guide’s inflation shock decomposer splits a TIPS fund’s year into CPI accrual and real-yield repricing; a TIPS bond held to maturity, or a ladder of them, has no such repricing at the horizon that matters.

Stocks

A share is a claim on a real business, and over decades that claim has beaten inflation by a wide margin: Bosse’s long-run series shows U.S. equities returning 6.75% a year in real terms from 1900 to 2018 against 1.44% for bonds and 0.28% for bills.¹ The short-run beta is a separate question and the answer has been negative since Bodie in 1976, who found that the real return on equities was negatively correlated with both anticipated and unanticipated inflation, so that a portfolio designed to hedge inflation would have had to short stocks.²⁰ Neville and coauthors find −7% a year real across U.S. inflationary regimes against +10% in other years, with positive real returns in only two of the eight episodes.¹⁷ Proposed mechanisms include discount rates rising faster than nominal cash flows, input costs leading selling prices, monetary tightening, historical-cost accounting and taxation, and Fama’s proxy hypothesis that inflation mostly carries news about weaker real activity. Knox and Timmer’s announcement-day decomposition favours the risk-premium channel.⁸

Picking individual stocks for inflation exposure is harder than the sector story suggests. Ang, Brière and Signori find large differences in inflation beta across U.S. stocks, with the best hedgers since 1990 drawn from oil and gas and technology, but as many as 20% of stocks change the sign of their inflation beta from one year to the next, and portfolios formed on past betas, on sectors or on high dividend yield did not deliver reliable hedges out of sample.²¹ A screener that ranks stocks by trailing inflation beta is mostly ranking them by the last regime’s shocks.

Commodities

Commodities have the largest inflation betas in every study because energy and food prices are components of the CPI itself. The size depends on the index: Bosse reports +5.86 for the Bloomberg Commodity Index and +9.28 for the S&P GSCI over 1991 to 2018, and notes that the GSCI was 63% energy against 31% for Bloomberg.¹ Fang, Liu and Roussanov’s split shows the exposure is to energy inflation rather than core.³ Two cautions apply to the long-run numbers. Bosse’s 0.39% real return for commodities over 1900 to 2018 is a spot-price series, and his own note says it “does not represent the return of a futures-based investment.”¹ Gorton and Rouwenhorst’s equally weighted futures index over 1959 to 2004 earned equity-like returns with a positive correlation to inflation;¹² Erb and Harvey showed in the same journal issue that individual commodity futures earned about zero on average and the index return came from rebalancing and from the term structure of futures prices;²² and Levine, Ooi, Richardson and Sasseville, with futures data back to 1877, find positive long-run index returns and good performance in rising-inflation states.²³ Neville and coauthors report +14% a year real for a broad commodity basket in U.S. inflationary regimes and +41% for energy, alongside “considerable variation within the commodity complex.”¹⁷ The beta is real. It is a beta to energy, its magnitude has ranged from about zero to above 20 in sixty-month windows, and the asset carrying it is roughly as volatile as equities.

Gold

Gold’s inflation beta is positive in most samples and unreliable in all of them. Bosse’s +1.19 has a p-value of 0.17.¹ The Summitward sixty-month rolling beta ranges from −7.27 to +14.74 and was negative in 29% of windows.⁹ Erb and Harvey’s reading is that gold “may be a good inflation hedge over many centuries but a poor inflation hedge for horizons up to 20 years,” and that the real price of gold, when high relative to its history, has predicted low subsequent real returns.²⁴ Their 2024 update notes that an influx of ETF, retail and central-bank buying has roughly doubled the real price relative to its pre-influx level and asks whether this time is different.²⁵ In 2022, with CPI averaging 8% higher, gold finished the year at $1,812 an ounce, down $3.30.¹⁹

REITs and housing

“Real estate is a real asset” is true and does not settle the beta. Listed REITs are hybrids with an income stream, equity market exposure and a property component; Bosse finds no significant relationship with unexpected inflation and volatile persistence.¹ Fang, Liu and Roussanov find a headline beta of +0.31 hiding a core beta of −6.54, the worst in their table.³ The Vanguard REIT ETF returned −26.20% in 2022.¹⁹ Directly owned housing is a different instrument again: Fama and Schwert found it hedged both expected and unexpected inflation over 1953 to 1971,⁶ and Kim, Ranaldi and Schularick find a static beta near zero that gives way to persistent declines over the following years.⁴ Rents reset at different speeds, mortgage debt changes the equity return, and appraisal-based prices smooth measured volatility. A house, a commercial property fund and a REIT ETF should not share one row in an inflation-hedge table.

How a DIY investor should use the number

Treat inflation beta as a diagnostic, not a score. A beta tells you the direction and rough size of an exposure over a stated sample. It does not tell you the exposure is reliable, that it will survive a regime change, or that it comes free. Read it with the correlation, the R², the sample dates and the shock definition attached, and distrust any table that prints the slope alone.

Match the instrument to the liability. An investor accumulating for thirty years has a purchasing-power problem, and the asset with the best record of solving it is a diversified equity portfolio with a poor short-run inflation beta. A retiree funding a known real spending stream over the next decade has a different problem, and contractual CPI-linked cash flows matched to those dates answer it directly; the real yield curve prices that stream on any given day. Neither investor needs the highest-beta asset available.

Price the hedge. Vanguard’s 2026 exercise is useful because it prints the bill. In its model, taking a 60/40 portfolio from an inflation beta of 0.07 to 1.0 requires a 15% satellite, 12.6 points of it in commodities. The projected reward over the next decade is 0.2 points of extra annual return and a lower drawdown, at the cost of 2.5% tracking error and a 50.6% probability of underperforming plain 60/40 in any given year.² Those figures are outputs of one model as of February 2026, not forecasts you should plan around, but the shape of the trade is general: a coin-flip chance of lagging every year in exchange for a payoff in the state you are worried about.

Ask which inflation you are hedging. Commodities answer an energy shock. Nothing on the long side reliably answers a core-services shock, and the assets marketed for it, REITs and “real asset” equity funds, have some of the worst core betas measured. Short-duration TIPS are the closest thing to a contractual answer, at a beta near 1 that hedges only the dollars invested in them.

When not to bother. If your horizon is long, your spending is flexible and your portfolio is mostly equities, your inflation problem is already being solved slowly by the asset with the 6.75% long-run real return, and adding a high-beta sleeve buys volatility and tracking error for protection against a short-run shock you can afford to ride through. Dimensional’s count that TIPS beat inflation in 69% of rolling twelve-month periods from 2001 to 2022 is a reminder that even the indexed asset loses the short-run race nearly a third of the time.²⁶

Bottom Line

Inflation beta is a regression slope, and its value is conditional on the inflation series, the shock definition, the horizon and the sample. Stocks carry a small, insignificant beta to headline inflation and a large negative beta to core inflation. Commodities carry a large positive beta to energy inflation and roughly none to core. Nominal bonds are negative to almost everything. Short-duration TIPS are the only asset with a beta that stays near 1 across samples, and it hedges only itself. Use the number to understand what a holding is exposed to, price the hedge before buying it, and let the liability, not the leaderboard, choose the instrument.

Key Takeaways

• Beta, correlation and R² are three measurements. In Vanguard’s 2026 model, short-term TIPS have the highest correlation with inflation (0.45) and a beta of 0.85; commodities correlate 0.34 and carry a beta of 7.60. A beta quoted alone hides which one you are looking at.
• The inflation series changes the answer. Fang, Liu and Roussanov put U.S. stocks at −1.33 against headline inflation (insignificant) and −5.60 against core (t = −3.69), 1963 to 2019. Commodities are +8.59 on headline and −0.07 on core.
• So does the shock definition and the horizon. One eighteen-country dataset yields equity betas of −2.93 to −4.60 across four specifications. Stocks respond −0.8% to a one-point CPI surprise on the day and −3.75% in real terms over five days; at five-year horizons the stock-inflation relation turns positive.
• The sample matters most of all. In a Summitward regression on monthly data, the S&P 500’s beta to headline CPI was −5.64 in the 1990s and +3.97 in the 2010s, both significant, and the sixty-month rolling estimate changed sign 23 times since 1965.
• Exposure and reward are different quantities. Only core inflation carries a significant negative price of risk in the Fang, Liu and Roussanov estimates, so a reliable core hedge should be expected to lag; a high-beta asset with high long-run returns is usually an energy bet.
• Use it as a diagnostic and price the hedge. Vanguard’s model needs a 15% satellite to lift a 60/40 portfolio to a beta of 1.0, at 2.5% tracking error and a 50.6% chance of lagging in any year. Match the instrument to the liability instead of the leaderboard.

Frequently Asked Questions

What is inflation beta?

The slope from regressing an asset’s return on a measure of inflation, usually the unexpected part. A beta of 2 means the asset’s return was, on average over the sample, two percentage points higher for every percentage point the inflation measure was higher. It is an average over a period, not a guarantee about the next shock.

Is a higher inflation beta better?

Only if it is reliable and you want that much exposure. A beta of 6 from an asset that is twenty times as volatile as inflation can come with a correlation of 0.3 and an R² near 0.1, meaning inflation explains a tenth of what the asset does. Vanguard’s 2019 study found gold and REITs with betas above short-term TIPS and t-statistics too small to distinguish from zero.

Why do stocks have a negative inflation beta if companies can raise prices?

Over decades they do, and equities have the best long-run real return of any major asset. Over months and quarters, inflation surprises have coincided with higher discount rates, tighter policy and weaker real growth, and stock prices respond to those first. The beta is most negative against core inflation and near zero or positive against energy inflation, so the sign depends on which kind of inflation the sample contains.

Do TIPS have an inflation beta of 1?

Short-duration TIPS come close, about 0.85 to 0.98 in Vanguard’s two studies, and the estimate is stable across samples. Longer TIPS funds have lower and less reliable betas because real-yield changes move their prices more than a year of CPI accrual. In 2022 a broad TIPS index lost 11.85% while CPI rose 8%.

What inflation beta does gold have?

Positive and unstable. Point estimates run from about 1 to 3 depending on the shock definition and sample, rolling estimates have swung from −7 to +15, and Vanguard’s 2019 estimate was not statistically significant. Erb and Harvey’s summary is that gold hedges inflation over centuries and poorly over any horizon up to twenty years.

How do I compute an inflation beta myself?

Take monthly CPI from FRED and a monthly total-return series for the asset, compute percent changes over the same span, and regress the asset’s return on the inflation change with an intercept. Report the slope with its standard error, the correlation and the sample dates, and repeat with core CPI and with a rolling window before drawing a conclusion. The script behind the numbers above is published and runs on the Python standard library.

Related Guides

• Which Inflation Are You Hedging? for what a TIPS fund, a breakeven position and a commodity basket each protect against, and the 2022 decomposer.
• The Best Inflation Hedges Are Boring for I Bonds, TIPS and the tax treatment of each hedge.
• Are Ben Carlson’s Best Inflation Hedges Actually Inflation Hedges? for a review of the popular version of this argument.
• Is Locking In Today’s TIPS Yields Market Timing? for pricing a real spending stream off the yield curve.
• Do You Need Managed Futures? for the strategy that did best across the historical inflation regimes.
• How to Build a TIPS Ladder for the maturity-matching mechanics.

Sources

1. Paul Bosse, “Commodities and short-term TIPS: how each combats unexpected inflation,” Vanguard Research, June 2019. Figures 2 to 7 and Appendix A-1 and A-2. vanguard.com
2. Cheng Li, Vibhor Dave and Victor Zhu, “A core-satellite approach to hedging inflation,” Vanguard Research, May 2026. Figures 3, 5 and 6; VCMM forecasts as of February 28, 2026. vanguard.com
3. Xiang Fang, Yang Liu and Nikolai Roussanov, “Getting to the Core: Inflation Risks Within and Across Asset Classes,” The Review of Financial Studies 39(3), 2026, 702–743. Table 2. doi.org. Working-paper version: NBER 30169 (2022); summary in NBER Digest, September 2022.
4. Taehoon Kim, Lorenzo Ranaldi and Moritz Schularick, “Inflation Surprises and Asset Returns: A Macrohistory Perspective,” working paper, October 2025. Table 2 and the local-projection estimates. lorenzoranaldi.com
5. Irving Fisher, Appreciation and Interest (1896); see also The Theory of Interest (1930). On where the equation appears, Robert Dimand and Rebeca Gomez Betancourt, “Irving Fisher’s Appreciation and Interest (1896) and the Fisher Relation,” Journal of Economic Perspectives 26(4), 2012, 185–196. aeaweb.org
6. Eugene F. Fama and G. William Schwert, “Asset Returns and Inflation,” Journal of Financial Economics 5(2), 1977, 115–146. doi.org
7. Pete Hecht, Antti Ilmanen, Thomas Maloney and Nick McQuinn, “Inflation Redux? Real Solutions for Real Returns,” AQR Alternative Thinking 2026 Issue 3, September 2026. Inflation news definition; Exhibits 5, 6 and 7. aqr.com
8. Benjamin Knox and Yannick Timmer, “Stagflationary Stock Returns,” Federal Reserve Board, Finance and Economics Discussion Series 2025-056, August 2025. federalreserve.gov
9. Summitward calculation. Monthly CPI-U headline, core and energy (FRED CPIAUCSL, CPILFESL, CPIENGSL) and the 3-month bill rate (TB3MS); S&P Composite price and dividends and the GS10 yield from Robert J. Shiller’s ie_data.xls; gold and the Total and Energy commodity indices from the World Bank Pink Sheet (CC BY 4.0). Sample January 1960 to August 2026; run September 20, 2026. Script, data files and printed output: github.com/engineerinvestor/summitward-research.
10. N. Bulent Gultekin, “Stock Market Returns and Inflation: Evidence from Other Countries,” Journal of Finance 38(1), 1983, 49–65. doi.org
11. Jacob Boudoukh and Matthew Richardson, “Stock Returns and Inflation: A Long-Horizon Perspective,” American Economic Review 83(5), 1993, 1346–1355. aeaweb.org
12. Gary Gorton and K. Geert Rouwenhorst, “Facts and Fantasies about Commodity Futures,” Financial Analysts Journal 62(2), 2006, 47–68. NBER 10595
13. Martijn Boons, Fernando Duarte, Frans de Roon and Marta Szymanowska, “Time-Varying Inflation Risk and Stock Returns,” Journal of Financial Economics 136(2), 2020, 444–470. doi.org
14. John Y. Campbell, Adi Sunderam and Luis M. Viceira, “Inflation Bets or Deflation Hedges? The Changing Risks of Nominal Bonds,” Critical Finance Review 6(2), 2017, 263–301. doi.org
15. Anna Cieslak and Carolin Pflueger, “Inflation and Asset Returns,” Annual Review of Financial Economics 15, 2023, 433–448. Working-paper version: NBER 30982. nber.org
16. Geert Bekaert and Xiaozheng Wang, “Inflation Risk and the Inflation Risk Premium,” Economic Policy 25(64), 2010, 755–806. doi.org
17. Henry Neville, Teun Draaisma, Ben Funnell, Campbell R. Harvey and Otto Van Hemert, “The Best Strategies for Inflationary Times,” Journal of Portfolio Management 47(8), 2021, 8–37. Exhibits 1, 2 and 4. SSRN 3813202
18. Federal Reserve Bank of St. Louis, FRED: 10-year TIPS constant maturity yield (DFII10), 10-year breakeven inflation rate (T10YIE), CPI-U not seasonally adjusted (CPIAUCNS). Year-end 2021 and 2022 values; 2022 annual-average inflation computed from CPIAUCNS. fred.stlouisfed.org
19. Calendar-2022 total returns from the “Average Annual Total Returns” tables of each fund’s 2023 statutory prospectus on SEC EDGAR: Vanguard Short-Term Inflation-Protected Securities ETF −2.84% (NAV); iShares TIPS Bond ETF −12.13%; Schwab U.S. TIPS ETF −11.96%; PIMCO 15+ Year U.S. TIPS Index ETF −31.60%; Vanguard Real Estate ETF −26.20% (NAV). The Bloomberg U.S. Treasury Inflation-Protected Securities Index (Series-L) return of −11.85% appears in the iShares, Schwab and Vanguard filings. Gold: LBMA Precious Metals Market Report Q4 2022 (year-end PM price $1,812.35, down $3.30 on the year). sec.gov; lbma.org.uk
20. Zvi Bodie, “Common Stocks as a Hedge Against Inflation,” Journal of Finance 31(2), 1976, 459–470. doi.org
21. Andrew Ang, Marie Brière and Ombretta Signori, “Inflation and Individual Equities,” Financial Analysts Journal 68(4), 2012, 36–55. NBER 17798
22. Claude B. Erb and Campbell R. Harvey, “The Strategic and Tactical Value of Commodity Futures,” Financial Analysts Journal 62(2), 2006, 69–97. doi.org
23. Ari Levine, Yao Hua Ooi, Matthew Richardson and Caroline Sasseville, “Commodities for the Long Run,” Financial Analysts Journal 74(2), 2018, 55–68. NBER 22793
24. Claude B. Erb and Campbell R. Harvey, “The Golden Dilemma,” Financial Analysts Journal 69(4), 2013, 10–42. NBER 18706
25. Claude B. Erb and Campbell R. Harvey, “Is There Still a Golden Dilemma?” working paper, 2024. SSRN 4807895
26. Dimensional Fund Advisors, “TIPS Can Help Investors When Inflation Spikes.” TIPS outperformed inflation in 69% of 264 rolling 12-month periods, 2001 to 2022. dimensional.com

Author disclosure

Summitward has no business relationship with Vanguard, AQR or any firm mentioned here and receives no compensation from them. Figures attributed to Vanguard’s 2019 Figures 2, 3 and 4 were read from the published charts because the paper does not print them as tables; its Figure 5, 6 and 7 values and all Vanguard 2026 figures are printed values. The Summitward calculations use spot commodity indices, a yield-based Treasury approximation and seasonally adjusted CPI, and are illustrations of the method rather than estimates of any fund’s sensitivity; the script, data and printed output are public. Nothing here is investment advice.