What drives gold prices: real yields still matter, but no longer explain the trend
A guide built on World Bank and St. Louis Fed data, 2003 to September 2026. Updated October 5, 2026. · Written by Svetoslav Grigorov
Short answer: real interest rates still explain a good part of gold's month-to-month moves, but since 2022 they explain much less of its long-term trend. Between 2003 and 2021, each one-point rise in the 10-year real yield went with an average gold decline of about 10.8%. Between December 2021 and September 2026 the real yield rose 3.63 points, which on that relationship "called for" a fall of roughly 33%; gold rose 141% instead.
This is a reference page, not a daily news piece. The data are public series and the calculations are our own: they are labeled CALCULATED and can be reproduced with the method described at the end. Explanations we cannot prove with the data are labeled HYPOTHESIS.
Why real yields matter for gold
Gold pays no interest. When a US Treasury bond offers a high return above inflation — the so-called real yield — holding gold carries a larger opportunity cost; when that real yield falls, the cost shrinks. The most widely used benchmark is the yield on the 10-year inflation-protected Treasury (TIPS), published daily by the St. Louis Fed as series DFII10. Its monthly average was 2.64% in September 2026 and −0.99% in December 2021.
Two other mechanisms are usually mentioned alongside it: the dollar (gold is priced in dollars, so a strong dollar tends to make it cheaper in that currency) and expected inflation. This guide measures all three with public data.
What the data say, 2003 to 2026
CALCULATED Comparing gold's monthly change with the change in the 10-year real yield, the negative relationship shows up in every period, including after 2022:
| Period | Correl. | Per point |
|---|---|---|
| 2003-07 | −0.26 | −6.2% |
| 2008-12 | −0.56 | −11.1% |
| 2013-21 | −0.61 | −14.0% |
| 2003-21 | −0.49 | −10.8% |
| 2022-23 | −0.66 | −7.5% |
| 2024-26 | −0.29 | −10.1% not significant |
In 2022-2023 gold's monthly reaction to the real yield was even tighter than the historical average. Since 2024 the estimate has become imprecise: its 95% confidence interval runs from −24.2% to +4.0% per point and includes zero. In a formal test over the whole period, gold's sensitivity to the real yield does not change in a statistically significant way after 2022. What changes is something else, and it is easier to see in a chart:
The part the real yield does not explain
CALCULATED Applying the average 2003-2021 sensitivity to each month separates the part of gold's move that corresponds to the real yield from the part left unexplained:
| Period | Gold | Unexplained |
|---|---|---|
| 2003-07 | +1.37% | +1.28%/month ≈ +17% a year |
| 2008-12 | +1.24% | +0.77%/month ≈ +10% a year |
| 2013-21 | +0.06% | +0.03%/month ≈ 0 |
| 2022-23 | +0.52% | +1.79%/month ≈ +24% a year |
| 2024-26 | +2.29% | +2.56%/month ≈ +36% a year |
Two readings. First, 2013-2021 was the exception, not the rule: in 2003-2007 gold already rose far more than the real yield justified. Second, since 2022 that unexplained part is back, stronger than in any earlier period, and it has come with real yields rising against it. With robust (Newey-West) standard errors, the unexplained part rises by 1.55% a month after 2022, and that increase is statistically significant (t = 2.8).
CALCULATED Cumulatively, from December 2021 to September 2026 (World Bank monthly averages: from $1,790 to $4,319 an ounce), gold rose 141%. On the historical sensitivity, the rise in the real yield would have gone with a fall of roughly 33%. Adding the dollar and expected inflation, with their 2006-2021 sensitivities, does not close the gap: the explained part is still a 32% fall.
CALCULATED Since 2022 gold does react more than before to expected inflation and to the dollar: in a regression with all three variables, the sensitivity to 10-year expected inflation goes from not significant in 2006-2021 to about +10% per point, and the dollar's becomes more negative. That is only 57 months of data, and why it happens is an open HYPOTHESIS, not a conclusion.
A statistical reference, not a fair value
ESTIMATED Another way to look at it is to fit a simple model between the level of gold and the level of the real yield on pre-2022 data and check what it would give today. Fitted on 2006-2021, that model explained 86% of the variation in gold's level, and it was not a fluke of the fit: trained only on 2006-2015, it tracked 2016-2021 with errors of between −12% and +15%. For September 2026, depending on the fitting period (from 2003, 2006 or 2010), it would give roughly $490 to $850 an ounce, against an observed $4,319.
Read it with care: relating the levels of two trending series can produce misleading relationships, which is why this guide does not rest on this model. The robust conclusion is the earlier one, based on monthly changes. What the chart adds is the scale of the shift: gold trades at several times what would be consistent with its 2006-2021 relationship with the real yield.
Episodes: when the rule worked and when it didn't
| Episode | Gold | Real yield |
|---|---|---|
| Oct-08 → Aug-11 Fed stimulus | +118% | −2.62 pp |
| Dec-12 → Dec-13 stimulus wind-down | −27% | +1.50 pp |
| Dec-21 → Oct-22 Fed hikes | −7% | +2.58 pp |
| Oct-22 → Oct-23 | +15% | +0.82 pp |
| Oct-23 → Sep-26 | +125% | +0.23 pp |
CALCULATED In 2013 the rule worked forcefully: the real yield rose and gold fell 27%. In 2022 the real yield rose more than it did in 2013 and gold fell only 7%, well short of what the historical sensitivity implied. Between October 2022 and October 2023 both rose together, and since then gold has risen 125% with the real yield almost unchanged.
What might explain the rest
The data in this guide show that part of gold's move is not explained by the real yield, the dollar or expected inflation. They do not show what causes it. These are the explanations given by official and industry sources, with their figures:
VERIFIED Central bank buying. According to the World Gold Council (with data from Metals Focus), central banks bought more than 1,000 tonnes a year between 2022 and 2024 (1,092t in 2024) and 863t in 2025, against an average of 473t a year in 2010-2021 (World Gold Council). These figures include estimates of unreported purchases and are revised over time, which is why official data compiled by the IMF show a much smaller increase, as the European Central Bank notes. The same ECB report (June 2025) estimates that the official sector accounted for more than 20% of gold demand in 2024, against around one-tenth in the 2010s, and dates the break in the relationship with real yields to after Russia's invasion of Ukraine. The data in this guide qualify that break: it shows up in the level of gold, not in its month-to-month reaction.
VERIFIED Sanctions. On February 28, 2022 the United States prohibited transactions with the Central Bank of Russia, immobilizing its assets within US reach (US Treasury). According to the ECB, one in four emerging-market central banks cites sanctions or changes in the international monetary system among its reasons for buying gold.
VERIFIED De-dollarization? A Federal Reserve working paper concludes that central banks' gold accumulation is generally a modest diversification of reserves rather than a move away from the dollar, except in a few prominent cases, including in 2022 and 2023 (Weiss, Federal Reserve, IFDP 1420; the author's view, not the Fed's).
VERIFIED Investment. In 2025 total gold demand reached a record 5,002 tonnes, with 2,175t of investment and a net 801t added to exchange-traded funds (ETFs), according to the World Gold Council.
HYPOTHESIS Fiscal risk. Some analyses, such as one by the World Gold Council in June 2025, argue that concern about public debt has also supported gold, as one factor alongside rates and geopolitics. We have not been able to measure its effect with the same rigor.
HYPOTHESIS None of these explanations is proven as the cause of the drift. The timing coincides strongly with central bank buying, but in 2025 those banks bought less than in the three previous years and gold rose more. The most prudent reading is to treat them as compatible factors, not as a single explanation.
Gold after inflation
CALCULATED Using the US consumer price index, the January 1980 average ($675 an ounce) is equivalent to about $2,892 in August 2026 dollars. The August 2026 average was $4,411: even after inflation, gold trades well above that peak.
What this guide does not say
- It is not a forecast: it does not say whether gold will rise or fall, or what price it "should" trade at.
- Correlation is not causation: two series moving together does not prove that one moves the other.
- There are only 57 months of data since 2022; figures for that period are less precise than those for 2003-2021.
Methodology
Data. Gold: monthly average in dollars per ounce from the World Bank Pink Sheet (file updated October 2, 2026, CC BY 4.0 license). Since June 2025 the World Bank uses the average of spot quotes; before that, the London afternoon fixing. Real yield: DFII10; expected inflation: T10YIE; dollar: DTWEXBGS (from 2006); CPI: CPIAUCSL. Daily series are converted to monthly averages. Period: January 2003 to September 2026.
Calculations. Gold's monthly log change (in %) against the change in the real yield in percentage points. Least-squares regressions by period, with Newey-West standard errors (6 lags). The unexplained part is gold's change minus the 2003-2021 sensitivity times the change in the real yield. The levels model regresses the log of gold on the real yield using data up to December 2021.
Limitations. Regressions on the levels of trending series can be spurious. The periods are our own choice, and other periods give somewhat different figures. Monthly averages smooth out single-day moves.
This guide and our gold analyses
Daily analyses, such as the August 25 three-month high or the September 29 piece, cover what happened at a given moment. This guide is the big picture and will be updated with new monthly data.
Sources consulted on October 5, 2026: World Bank: Commodity Markets (Pink Sheet), FRED: DFII10, FRED: T10YIE, FRED: DTWEXBGS, FRED: CPIAUCSL, World Gold Council: Gold Demand Trends 2025, central banks, World Gold Council: Gold Demand Trends 2025 press release (Jan 29, 2026), World Gold Council: Are fiscal concerns driving gold? (Jun 24, 2025), ECB: Gold demand: the role of the official sector and geopolitics (June 2025), US Treasury: sanctions on the Central Bank of Russia (Feb 28, 2022), Federal Reserve: IFDP 1420.
Educational notice: this guide is analysis based on public data, may be wrong, and is not financial advice. It is not a recommendation to buy or sell gold or any other asset. The decision and the risk are always the reader's own.