Research · Dataset

Gold versus US house prices, 1987–2026

Housing is gold's most common rival: the asset people mean when they say "real assets beat inflation", and the other "hard asset outside the financial system". This page measures the two against each other over 475 months, then asks the only question that matters — when one has run ahead of the other, what actually happens next?

Last updated: 2026-08 (data through 2026-08) · Source data updated November 03, 2026

Source and method

Publisher
World Bank Commodity Price Data (The Pink Sheet) + S&P Cotality Case-Shiller U.S. National Home Price Index (via FRED)
Gold series
Monthly Prices sheet, column 70 (Gold, $/troy oz), monthly average
Housing series
FRED CSUSHPINSA, S&P Cotility Case-Shiller U.S. National Home Price Index, Index Jan 2000 = 100, Not Seasonally Adjusted, monthly
Seasonal check
FRED CSUSHPISA (seasonally adjusted) used as a robustness check only
Deflator
FRED CPIAUCSL, CPI-U All Items, used to deflate both legs
Frequency
Monthly
Coverage
475 months, 1987-01 to 2026-08
Download
gold-housing-monthly.csv (month, gold price, both house-price index vintages, CPI, monthly and real returns)
Source links
World Bank commodity data · FRED CSUSHPINSA
Reproduction
scripts/gold_housing_analysis.py — recomputes every table below from the raw workbook and the two FRED series

The housing leg is an index, not a house. There is no instrument that tracks the Case-Shiller national index, so every "housing" figure here means a passive, frictionless, unlevered, unrented exposure to it — no rental yield, no transaction costs, no maintenance, property tax or insurance, and no ability to time entry or exit. A real owner receives rent and pays all of those. The comparison is therefore favourable to housing on costs and unfavourable to it on the returns an owner actually collects.

Gold is the monthly average of the World Bank Pink Sheet, so a daily or intraday comparison would be noisier and would not reproduce these figures. Nothing here includes storage, financing, bid-ask spread or transaction cost on the gold leg either.

Where the two stand now

Gold, 12 months +32.07% $4,411/oz in 2026-08
House prices, 12 months +2.20% index 336.7
Gap +29.9pp gold minus housing
After CPI -1.45% housing vs inflation · CPI +3.71%

Over the trailing twelve months gold is +32.07% and the national house price index +2.20%. Once CPI is taken out, gold is +27.34% in real terms while the housing index is -1.45%. On this measure, house prices are currently losing to inflation while gold is comfortably ahead of it — which is close to the opposite of the usual argument for housing as the inflation hedge.

How related are they?

Monthly return correlation -0.0702 474 months · seasonally adjusted housing
Price level correlation 0.8501 mostly a trend artefact

On monthly returns the two assets are close to unrelated. The level correlation of 0.8501 looks like a strong relationship and is not one: two series that both rose over forty years will correlate in levels. The rolling 36-month return correlation has a median of -0.0537, is negative in 68.6% of windows and reaches -0.6484 at its lowest.

In other words, "gold and housing are both hard assets" is a statement about category, not about behaviour. They behave differently.

Is housing a better inflation hedge?

The test: correlate each leg's monthly change against US CPI inflation. The housing index correlates 0.1822 and gold 0.1428. Both are weak, and housing is not better — on this data the popular claim that housing is the real asset that beats inflation, and gold is not, does not hold up.

Measured over the whole sample the inflation-adjusted returns say the same thing: gold compounded at 3.30% a year after CPI, the housing index at 1.42%. The housing index lost ground to inflation over nearly four decades.

A seasonality audit, because the raw index lies

The Case-Shiller national index is published not seasonally adjusted, and its month-to-month changes carry a large seasonal signature. Averaged across the sample, the raw index rises 0.91% in May and falls 0.09% in December. That is a recording artefact, not a price signal.

Month Raw index (NSA) mean change Seasonally adjusted mean change
January -0.01% +0.36%
February +0.21% +0.30%
March +0.71% +0.39%
April +0.90% +0.35%
May +0.91% +0.35%
June +0.79% +0.35%
July +0.52% +0.35%
August +0.27% +0.33%
September +0.07% +0.36%
October -0.01% +0.36%
November -0.06% +0.37%
December -0.09% +0.36%

The consequence is not cosmetic. Using the raw series, measured housing volatility is 2.37% a year; using the seasonally adjusted twin it is 1.82%. Every return-based statistic on this page uses the adjusted series. The raw index appears only in the published index level and in long-run price totals, which are unaffected because a twelve-month window covers a full seasonal cycle.

Even adjusted, this is a smoothed repeat-sales index of a very large national stock of houses. It cannot show the single-house risk an owner carries. Volatility numbers here describe this index, not the risk of owning a house.

Decade by decade

Period Gold House prices Gold after CPI Housing after CPI Gold vol Housing vol Months gold beat housing
1980-1989 (to 1989-12) +0.2% +19.9% -11.6% +5.8% 10.9% 0.5% 13 / 35
1990-1999 -31.0% +29.8% -47.9% -1.9% 10.5% 0.8% 42 / 119
2000-2009 +299.6% +47.1% +211.3% +14.6% 14.0% 2.5% 62 / 119
2010-2019 +32.3% +45.1% +11.2% +22.0% 11.7% 1.4% 56 / 119
2020-2029 (to 2026-08) +182.6% +55.5% +119.1% +20.6% 13.4% 2.3% 40 / 78

Price totals, not annualised, and gross of any cost. Volatility is the annualised standard deviation of monthly changes within the decade, computed from the seasonally adjusted housing series.

The pattern is not one-sided. Gold lost badly to housing in the 1990s (-31.0% against +29.8%) and won heavily in the 2000s (+299.6% against +47.1%). Anyone who bought both in the right decade did well; anyone who picked one and dropped the other did badly.

Across the 38 calendar years in the sample, gold beat housing in 18 of them (47.4%) — so the long-run winner is closer to a coin flip than the return difference suggests.

The test: what happened after one ran ahead

We took the 20 months where gold had run furthest ahead of housing over the trailing twelve months and the 20 months where housing had, and looked at what each did over the following twelve months.

After gold ran far ahead gold +13.80% housing -4.26% · spread +18.66pp · gold up in 65% of them
After housing ran far ahead housing +5.75% gold -3.41% · spread -8.62pp · housing up in 100% of them

The mechanism is mean reversion, and it is strong in both directions. After the gap closed in housing's favour, the housing index rose in 100% of those months and averaged +5.75%, while gold averaged -3.41% and rose in only 25% of them.

Caveat on sample size: 20 months per tail drawn from a forty-year series. These are averages over a handful of very different environments, and they describe history rather than a rule.

Conditioning the whole sample by the trailing gap

A cleaner version of the same question: split the months with a full 12-month forward window by whether gold beat housing over the preceding twelve months, and compare the two halves.

Sample Months Gold ahead: forward spread Not ahead: forward spread Difference Non-overlapping p Bootstrap 95% CI
full sample (1988-01..2025-07) 451 +11.88pp -3.27pp +15.16pp 0.033 5.7 to 23.3
post-2000 (2000-01..2025-07) 307 +15.04pp -0.45pp +15.49pp 0.033 3.6 to 24.3
post-2010 (2010-01..2025-07) 187 +13.30pp -3.82pp +17.12pp 0.079 -0.6 to 28.3

The forward spread is the change in gold's return relative to housing's over the following twelve months, in percentage points. The non-overlapping column takes every 12th month so no two windows share a month. The bootstrap is a moving-block bootstrap with 12-month blocks, 2000 resamples and a fixed seed.

The effect is real. It is also mostly arithmetic.

On the full sample the difference is 15.16 percentage points, the non-overlapping test gives p = 0.033, and the bootstrap interval (5.7 to 23.3) excludes zero. The effect holds up in later eras too: 15.49pp after 2000 (p = 0.033) and 17.12pp after 2010, where the non-overlapping p is 0.079 and the bootstrap interval (-0.6 to 28.3) includes zero.

But the correlation between the forward spread and gold's own return is 0.9897. That is the tell. "Gold beat housing last year, so gold will do better next year" is close to a restatement of "gold's own return was high" — momentum dressed up as a cross-asset signal. The arithmetic identity explains most of what the table shows, and a reader who wants a tradable edge needs to subtract that identity out, not take the point estimate at face value.

Does one lead the other?

We cross-correlated gold's monthly return against the housing index's at lags of −6 to +6 months. The strongest reading is 0.0982 at lag -3, against -0.0702 at lag 0. For calibration, correlating gold's return against its own one-month-shifted self produces 0.1999 at the best lag — so the gold-versus-housing readings are inside the noise floor. No lead-lag structure survives the test.

Buy and hold since 1987

Gold 6.21% / year 10.8× total · max drawdown -47.3% · volatility 12.4%
House prices (index) 4.28% / year 5.2× total · max drawdown -26.0% · volatility 1.8%

Gold has won on return and lost on calmness. Gold compounded at 6.21% against housing's 4.28%, but with a maximum drawdown of -47.3% against -26.0%. After CPI the ordering is unchanged: 3.30% against 1.42%. These are historical figures on monthly average prices, gross of every cost.

Does a blend help?

Housing weight Final multiple CAGR Max drawdown
0% 10.81× 6.21% -47.33%
20% 9.82× 5.95% -35.31%
50% 8.12× 5.45% -14.82%
80% 6.35× 4.79% -11.72%
100% 5.23× 4.28% -26.04%

A 50/50 monthly-rebalanced blend returned 5.45% a year with a maximum drawdown of -14.82% — against 6.21% and -47.33% for all gold. The drawdown improvement is arithmetic, not cleverness: housing barely moves, so adding it damps the curve. The same low correlation is why the blend gives up return. These are illustrative portfolios with no transaction, spread, financing, tax or rental cost modelled, and the housing leg could not actually be held in this form.

What this dataset does and does not support

Supported by the data: gold and US house prices are close to unrelated on monthly returns (correlation -0.0702); housing is not a better inflation hedge than gold on this series, and over the long run it lost to CPI while gold did not; the raw house-price index carries a large seasonal signature that materially distorts any volatility figure computed from it; and there is a consistent tendency for the gap between the two to close over the following year.

Not supported: the reading that the gap-closing effect is a tradable cross-asset signal. The forward spread correlates 0.9897 with gold's own return, so most of the apparent edge is an identity. Nor is there any lead-lag structure: the strongest cross-correlation, 0.0982, sits below the 0.1999 noise floor set by correlating gold against itself.

Not measurable here: what owning an actual house would have returned. That would include rental yield, leverage, transaction costs, maintenance, property tax and insurance, and the fact that you cannot sell at a monthly-average index price.

Frequently asked questions

Can you actually buy the "US housing" asset this page measures?

No, and that limitation runs through every number here. The housing leg is the S&P Cotality Case-Shiller U.S. National Home Price Index — a repeat-sales index of the national stock of single-family houses. There is no instrument that tracks it. Any portfolio described as "holding housing" here means a passive, frictionless, unlevered, unrented exposure to that index: no rental yield, no transaction costs, no maintenance, no property tax, no insurance, and no ability to time entry or exit. Read the returns below as the return of an index, not of a house, and remember that an owner also receives rent and pays costs, neither of which is in these figures.

How related are gold and US house prices?

On monthly returns the two are close to unrelated: the correlation is -0.0702 over 474 months. The price LEVELS correlate 0.8501, but that is mostly a trend artefact — two series that both rose over forty years will correlate in levels, which is why the level figure is not evidence of a relationship. The rolling 36-month return correlation has a median of -0.0537 and is negative in 68.6% of windows, reaching -0.6484 at its lowest. Gold and housing are close to substitutes in the sense that both are "real assets" outside the financial system, but they do not move together.

Does housing actually hedge inflation better than gold?

Not on this data. The correlation between the monthly change in the Case-Shiller national index and monthly US CPI inflation is 0.1822; gold's correlation with the same inflation series is 0.1428. Both are weak, and housing is not clearly better. Over the whole sample the inflation-adjusted returns make the point: gold compounded at 3.3% a year after CPI, the housing index at 1.42%. The housing index lost ground to inflation over nearly four decades.

Why does this page use a seasonally adjusted house price index?

Because the published national index is not seasonally adjusted and its month-to-month changes carry a large seasonal signature. Averaged across the sample, the raw index rises about 0.9% in May and falls about 0.1% in December — an artefact of when homes are sold and recorded, not of prices. Using the raw series inflates measured housing volatility to 2.37% a year against 1.82% for the seasonally adjusted twin, and it contaminates every correlation that involves a housing monthly change. Every return-based statistic on this page therefore uses the seasonally adjusted series (FRED CSUSHPISA); the unadjusted series appears only in the published index level and in long-run price totals, which are unaffected because a twelve-month window covers a full seasonal cycle.

Does this page find a timing signal between gold and housing?

It finds a mean-reversion effect that is mostly arithmetic. Splitting the months by whether gold beat housing over the trailing twelve months, the following twelve months produced a gold-minus-housing spread of +11.88 pp in the "gold ahead" group and -3.27 pp in the other, a difference of +15.16 pp. On the non-overlapping subsample of 38 windows the difference is +14.19 pp with p = 0.033, and the moving-block bootstrap interval runs from 5.7 to 23.3 percentage points and excludes zero. The effect survives the era checks: +15.49 pp post-2000 with a non-overlapping p of 0.033, and +17.12 pp post-2010 with p = 0.079. But read the identity check before using it: the spread correlates 0.9897 with gold's own return, so this is mostly "which leg moved more last year" restated, not a discovered relationship.

What happened after the biggest gaps between gold and housing?

After the 20 months when gold was far ahead of housing on a trailing twelve-month basis (35.9 to 55.4 percentage points), gold averaged +13.8%% over the next year and the housing index -4.26%%; the spread moved +18.66 pppp. After the 20 months when housing was far ahead (-38.1 to -22.7), housing averaged +5.75%% and gold -3.41%%. The housing index rose in 100% of that second group. Caveat: 20 months per tail drawn from a forty-year series, dominated by a handful of very different environments.

Which one has the better long-run return, and how risky is each?

Since 1987, gold compounded at 6.21% a year with a maximum drawdown of -47.33% and annualised volatility of 12.4%. The housing index compounded at 4.28% with a maximum drawdown of -26.04% and volatility of 1.8%. Gold returned more and was riskier. But the housing figure understates the risk of owning a house in one specific way: the index is a smoothed repeat-sales measure of a very large stock of houses, so it cannot show the single-house risk an owner actually carries. These are historical figures on monthly average prices, gross of every cost, and they are not a forecast.

Does a mix of gold and housing beat either one alone?

On the return, no; on the drawdown, yes. A monthly-rebalanced, cost-free blend with 50% housing returned 5.45% a year with a maximum drawdown of -14.82%, against 6.21% and a -47.33% drawdown for all gold and 4.28% for all housing. The 80/20 blend in the other direction returned 5.95% with a -35.31% drawdown. The reason the drawdown falls so much is arithmetic rather than clever: housing barely moves, so adding it damps the equity curve. That low correlation also cuts both ways — it is exactly why the blend gives up return. These are illustrative portfolios with no transaction, spread, financing, tax or rental cost modelled.

Has gold beaten housing more or less often?

Less often, by a small margin. Over the 38 calendar years in the sample, gold beat housing in 18 of them, or 47.4%. The decade split is more informative: gold lost to housing in the 1990s (-31% versus 29.8%) and won heavily in the 2000s (299.6% versus 47.1%) and the 2020s so far (182.6% versus 55.5%).

Where are gold and house prices right now?

In 2026-08, gold is at $4411/oz and the Case-Shiller national index at 336.7 (January 2000 = 100). Over the last twelve months gold is +32.07%% and the housing index +2.2%% — a gap of 29.9 percentage points. After CPI, gold is +27.34%% while the housing index is -1.45%%, with US CPI up 3.71% over the year. On this measure housing is currently losing to inflation while gold is comfortably ahead of it.

Can you forecast which one wins from here?

No, and nothing on this page is intended as a forecast. It establishes what the historical relationship is: a weak one (monthly return correlation -0.0702), a mean-reversion pattern that is largely an identity, and long-run returns in which gold compounded faster both nominally (6.21% versus 4.28%) and after inflation (3.3% versus 1.42%). Whether that continues depends on monetary policy, rates, construction supply and many other things this dataset does not model.

Sources

Data accessed 2026-11-13. Prices are nominal US dollars and include no carry, financing or storage cost. Nothing on this page is investment advice or a forecast.

Educational content only. Not investment advice. Trading forex and CFDs carries substantial risk of loss.