REIT Correlation: 0.74, and 0.86 When Stocks Fall

Key takeaways

• Across 521 weeks to 4 September 2026, US REITs and the S&P 500 tracker moved together with a correlation of 0.74. The level a 10% sleeve had to beat to cut portfolio volatility was 0.78.
• Adding 10% US REITs to an all-equity portfolio moved annualised volatility from 17.30% to 17.28%. The same 10% in Treasuries moved it to 15.56%.
• In the 52 worst weeks for the S&P 500 tracker the REIT correlation was 0.86. In the 303 weeks it rose, the correlation was 0.63.
• The market factor alone explained 52.7% of weekly REIT returns over 516 weeks. Size and value lifted that to 57.5%, and adding interest rate duration reached 65.6%.
• Real Estate is 2.09% of the FTSE All-Share and 1.83% of the S&P 500 tracker, so a plain index fund already holds a slice of it.

The short answer: REIT correlation ran 0.74, and the bar it had to clear was 0.78

You hold a global or US equity fund. Someone tells you a REIT fund adds property, which is a different asset, so it diversifies. You want to know whether that's true in the numbers.

Over the ten years of weekly returns to 4 September 2026, it's technically true and practically almost nothing. Here's the arithmetic that decides it.

A new holding lowers a portfolio's volatility only if its correlation with what you already own is below a specific threshold: the volatility of the existing portfolio divided by the volatility of the thing you're adding. The Vanguard Real Estate ETF (VNQ) ran at 22.27% annualised volatility over the sample. The SPDR S&P 500 ETF Trust (SPY) ran at 17.30%. That puts the threshold at 0.78.

The measured REIT correlation with the S&P 500 tracker was 0.74. It cleared the bar, by 0.04. That is the entire REIT diversification case, and the next section prices it.

A 10% REIT sleeve moved portfolio volatility by 0.02 percentage points

Clearing a threshold by a fraction buys you a benefit of the same size. Run the weekly returns through a portfolio and the numbers come out like this.

All of it in the S&P 500 tracker: 17.30% annualised volatility. Move 5% into US REITs: 17.27%. Move 10%: 17.28%. Move 20%: 17.39%, which is higher than where you started. The volatility-minimising REIT weight over this decade was 6.5%, and holding exactly that weight would have saved 0.03 percentage points of volatility.

For scale, moving 10% into the iShares 7 to 10 Year Treasury Bond ETF (IEF) took the same portfolio to 15.56%. That's a reduction of 1.74 percentage points, against 0.02 for the REIT sleeve, from an asset nobody calls exotic.

The chart puts the six portfolios side by side. Every REIT bar sits on top of the all-equity bar. The bond bar is the only one you can see move.

The correlation matrix puts REITs nearer utilities than the S&P 500

A single pairwise number hides where REITs actually sit. Here is the full weekly correlation matrix for the sample, with each pair measured over the same 521 weeks.

Read the first column. REITs correlate more tightly with utility shares (0.77) than with the S&P 500 itself (0.74), and more tightly still with large-cap value stocks (0.80). Those are two equity sectors, not two asset classes. The only line in the table that looks genuinely different is Treasuries, at 0.20 against REITs and -0.03 against the S&P 500 tracker.

One fund doesn't decide this. The iShares U.S. Real Estate ETF (IYR) tracks a different index from a different provider, and its weekly returns correlate 0.99 with VNQ and 0.74 with the S&P 500 tracker. Two vehicles, one answer.

Rolling REIT correlation swung from -0.06 to 0.92, and sits at 0.32 now

A single decade-long figure averages over regimes that felt nothing alike. Measured in overlapping 52-week windows, there were 470 of them, and the REIT correlation with the S&P 500 tracker ranged from -0.06 in the window ending 12 January 2018 to 0.92 in the window ending 25 September 2020.

The median window was 0.65. Some 27.9% of windows sat above 0.80, where REITs were effectively a leveraged equity sleeve. Another 17.4% sat below 0.50, where they genuinely did something else. The window ending 4 September 2026 reads 0.32, near the bottom of the range.

So a reader who measured in early 2018 and a reader who measured in late 2020 would have reached opposite conclusions from the same fund. That instability is the reason a point estimate deserves so little weight, and it's the same problem described in our piece on correlation instability.

The correlation tightened in exactly the weeks equities fell

Average correlation is the wrong statistic if you're buying protection. What matters is the correlation in the weeks you'd want the protection.

Split the 521 weeks by how the S&P 500 tracker did. Across the 52 worst weeks, its mean return was -4.26% and the REIT fund's mean return was -3.81%, and the correlation between them was 0.86. Across the 303 weeks the tracker rose, the correlation was 0.63.

The link tightened by 0.23 going into the falls. REITs behaved most like equities precisely when behaving differently would have been worth something, which is the pattern our study of diversification in a crisis found across asset classes.

The drawdowns say the same thing more bluntly. Into the trough of 23 March 2020 the REIT fund's price fell 42.84% from its high, against 34.10% for the S&P 500 tracker. In 2022 the REIT fund fell 33.85% peak to trough, against 25.36%. In both of the decade's real tests, the supposed diversifier lost more than the thing it was meant to diversify.

Factor decomposition: the market factor alone explains 52.7% of REIT returns

Correlation says how much two series move together. A factor decomposition asks what listed property returns are made of. Regressing 516 weeks of REIT excess price returns on the Fama and French factors published in Kenneth French's data library, from 6 September 2016 to 31 July 2026, gives a market loading of 0.91 and an R-squared of 52.7%. R-squared is just the share of the ups and downs the model accounts for.

Adding the size and value factors, the two additions that define the classic three-factor model, lifts that to 57.5%. Both loadings are real rather than noise: 0.17 on size with a t statistic of 2.7, and 0.26 on value with a t statistic of 5.8. A t statistic above roughly 2 is the usual bar for saying a loading isn't zero.

The reading is unflattering for the diversification story. More than half of weekly REIT variation is the equity market wearing a hat, and the two style factors that are supposed to capture what makes property special add 4.8 percentage points between them.

Interest rate duration explains more than size and value combined

Now add a fourth variable to the same regression: the excess return on the iShares 7 to 10 Year Treasury Bond ETF, which is a plain proxy for interest rate duration. The explained share jumps to 65.6%, and the loading on it is 1.03 with a t statistic of 11.0.

That one rate variable adds 8.1 percentage points of explanatory power, against 4.8 for size and value together, on the same weeks and the same fund. The honest description of a listed REIT over this decade is an equity with a bond attached, carrying a market loading of 0.92 and roughly one full unit of Treasury duration alongside it.

The utilities sector behaves the same way for the same reason, which is why the two correlate at 0.77. Both are capital-intensive businesses that pay out most of their cash and refinance constantly.

The 90% payout rule is where the equity behaviour comes from

The structure explains the statistics. Under section 857 of the US Internal Revenue Code, a REIT keeps its tax treatment only if it distributes at least 90% of its taxable income. The UK regime says the same thing: section 530 of the Corporation Tax Act 2010 requires a UK REIT to distribute at least 90% of the UK profits of its property rental business.

A company that pays out 90% of its income can't fund growth from retained earnings. It funds growth by issuing shares or borrowing. That makes its share price unusually sensitive to the cost of both, which is the equity market and the bond market. The payout rule that makes REITs feel like property is the same rule that welds them to the two markets you already own.

Their yields don't buy much separation either. The iShares U.S. Real Estate ETF showed a 30 day SEC yield of 2.71% at 31 July 2026 against 0.96% for the S&P 500 tracker at 30 June 2026. That's a real income premium, and it is a fraction of what the payout rule implies to a reader who hasn't checked.

A UK index fund already holds 2.09% real estate

Listed property stopped being a separate thing some time ago. Standard and Poor's opened its indexes to REITs in October 2001, when Equity Office Properties Trust became the first REIT in the S&P 500. At the close on 31 August 2016, Real Estate became the 11th headline sector in the Global Industry Classification Standard, moved out of Financials by S&P Dow Jones Indices and MSCI.

The weights follow. At 31 August 2026 the FTSE All-Share carried 40 Real Estate constituents at 2.09% of the index. The SPDR S&P 500 ETF Trust held 1.83% in Real Estate at 30 June 2026, across 504 holdings; the MSCI USA Index carried 1.74% at 31 August 2026.

So the choice isn't between owning property and not owning it. A real estate allocation of 10% on top of a tracker is roughly a five-fold overweight to one sector you already hold, and the numbers above are what that overweight bought.

The strongest case against reading this as duplication

Three objections deserve a proper hearing, and one of them is strong.

The first is that the sample is rate-dominated. Rates moved enough over this decade to hand the Treasury fund an 18.43% peak-to-trough fall in its price in 2022, and a shared duration exposure mechanically pushes REIT and equity returns together. It's the standard criticism of any correlation measured inside a single macro regime, and the 0.32 reading in the window ending 4 September 2026 is consistent with it. If the rate shock was the cause, the 0.74 average overstates the steady state.

The second is geography, and it's the objection that holds up best. The Vanguard Global ex-U.S. Real Estate ETF (VNQI) ran at 18.08% volatility against 22.27% for the US fund, so its break-even correlation was far easier to clear. Its measured correlation with the S&P 500 tracker was 0.73, barely different, but the lower volatility does the work. A 10% ex-US REIT sleeve took the same portfolio from 17.30% to 16.94%, a reduction of 0.36 percentage points against 0.02 for the US version. The duplication verdict is a verdict on US REITs against a US index, not on listed property everywhere.

The third is that volatility is the wrong scorecard. A reader buying REITs for a contractual, inflation-linked income stream isn't trying to lower the standard deviation of a chart. That's a fair objection to the metric, though it doesn't rescue the diversification claim, which is a claim about volatility by definition.

What this sample cannot tell you

The series here are price returns from daily closing prices, so they exclude dividends. For REITs that omission is larger than for equities, because more of the return arrives as income: the same iShares fund showed a 12 month trailing yield of 2.15% at 31 July 2026. The fund distributes quarterly, so the omission lands in 4 weeks of 52 and is small against a weekly standard deviation of 3.09%. The effect on a correlation is minor, but it isn't zero.

Ten years is one sample and not the population, and it covers only the regimes that happened to fall inside it. The limitation runs deeper than length. At 4 September 2026 the iShares fund held 9.36% in data centre REITs and 8.08% in telecom tower REITs, against 1.82% in offices. A correlation measured on that mix is not evidence about a different mix, and the mix has been moving.

None of this is a forecast. A correlation measured to 4 September 2026 describes what happened; it doesn't say what the next decade holds. Nor does it settle what the underlying buildings did, which is a separate question we take up in REITs vs direct property, where appraisal smoothing makes the same assets look calmer than their listed equivalents.

What would change the conclusion

The finding rests on one narrow margin: a correlation of 0.74 against a break-even of 0.78. Both numbers can move, and they move for different reasons.

A sustained fall in the REIT correlation toward the 0.30s, held for several years rather than one window, would raise the margin enough to matter. So would a fall in REIT volatility toward the level of the index, since the break-even threshold is a ratio of the two. The current rolling reading is 0.32, and whether that survives the next rate cycle is the thing to watch, not the ten-year average.

Going the other way, another episode like 2020 or 2022 would push the ten-year figure back above the threshold, at which point a REIT sleeve raises portfolio volatility rather than lowering it. LedgerTouch tracks sector weights across your holdings, which is where an unintended five-fold overweight tends to surface first. And if the sector keeps drifting toward data centres and towers, the correlation has a mechanical reason to rise: the fund would increasingly be holding the tenants' business risk rather than the landlord's, and the tenants are already in the index.