Gold and Real Yields: An Empirical Read of the Inverse Correlation by Sub-Period, 1997-2026

The inverse correlation between gold and the 10-year TIPS yield is empirically documented since 2003, the first year of the DFII10 series, but it has never been a stable coefficient: it deforms and breaks according to macro regimes.

This article does not restate the theoretical opportunity-cost mechanic, covered elsewhere. It deploys a strictly empirical reading, sub-period by sub-period, of data available since the DFII10 series began in 2003.

The relationship between gold prices and the real yield on 10-year Treasury Inflation-Protected Securities is among the most cited in market literature, sometimes to the point of masking its real empirical variability. The correlation coefficient is not a constant. Over the window covered by the DFII10 series, since January 2003, seven sub-periods of distinct statistical behavior are clearly identifiable, each illuminated by its own macro reading. This logic is brought out in the critical reading of linkers against inflation. For the theoretical opportunity-cost mechanic underlying the relationship, why a non-yielding asset depreciates when the alternative real yield rises, the generic framework is covered separately. The present article distinguishes itself by angle: here, we read regimes, not theory.

10-Year TIPS (DFII10): What the Series Actually Measures

Before any empirical reading, the perimeter of the DFII10 series deserves to be stated without ambiguity. Launched in January 1997 by the U.S. Treasury Department, Treasury Inflation-Protected Securities are bonds whose principal is adjusted daily to BLS-published CPI-U inflation. The published yield, the FRED DFII10 series, is the market yield on adjusted principal, hence an expected real yield by construction. The 10-year breakeven, calculated as the spread between nominal 10Y Treasury (FRED DGS10) and DFII10, gives by construction the market-expected average inflation over the 10-year horizon. Underlying data: the 10-year breakeven record.

The DFII10 series begins on January 2, 2003: TIPS yields for 1997-2002 are not in it, so any statistical measure of the relationship starts in 2003. The correlations cited below are computed on monthly averages of DFII10 and of the gold price published by the World Bank (Pink Sheet), in levels, window by window (Eco3min calculations). Beyond the technical limitations, DFII10 properly reflects the real opportunity cost of a risk-free bond investment over the 10-year horizon, making it the reference proxy for gold-real-yields analysis. A related question: why real yields matter more than nominal.

2003-2007: Before the Crisis, a Positive Correlation in Levels

Over the first five years of the series, the inverse relationship does not show up in levels. The DFII10 yield ranges between 1.36% (March 17, 2004) and 2.83% (June 12, 2007). Gold goes from a monthly average of $357 in January 2003 to $806 in November 2007 (World Bank). The correlation of monthly levels comes out at +0.51 over 2003-2007: gold advances while the real yield, after a trough in March 2004, climbs back to mid-2007. In depth: the structure of gold demand from ETFs to physical bars.

The period is also one of gradual tightening: the Fed raises its policy rate from 1% to 5.25% in seventeen hikes, from June 2004 to June 2006, and year-on-year CPI inflation ranges between 1.4% and 4.7% (2.9% on average). The dollar weakens markedly from 2002, which supports gold through a channel other than the real yield. The inverse gold-real-yields relationship, often presented as a constant, only takes hold in levels with the 2008 crisis.

2008-2011: GFC and Three QE Rounds, the Correlation at Its Extreme

The global financial crisis produces the most spectacular sequence of the empirical TIPS history. Between January 2008 and August 2011, the DFII10 yield goes from 1.47% to 0.13% in monthly average, after a 3.15% spike on November 21, 2008 at the height of the liquidity crisis, and turns negative for the first time on August 10, 2011 (-0.13%). The decline reflects the combined effect of flight-to-quality, three Fed QE rounds (QE1 November 2008, QE2 November 2010, Operation Twist September 2011, QE3 in September 2012) and the brutal collapse of long-run inflation expectations.

Gold prices trace the mirror trajectory: a monthly average of $761 in November 2008, then $1,772 in September 2011 (World Bank). The correlation of monthly levels reaches -0.92 over 2010-2011 and -0.86 from January 2008 to August 2011. Over this period, the DFII10 trajectory accounts for most of the gold move. Also relevant: the opportunity cost of holding gold.

The September 2011 peak closely coincides with the first S&P downgrade of the U.S. sovereign AAA (August 5, 2011) and the entry into the eurozone sovereign debt crisis. This concomitance provides a robustness test: even in the most acute systemic stress peak since the GFC, the real-yields channel remains analytically sufficient to account for the observed price. The “systemic safe haven” factors invoked in contemporary press are then largely redundant with the real-yields channel itself.

2011-2020: Fed Normalization and a Stable but Weakened Correlation

The 2011-2020 period sees a consolidation of the relationship: the correlation of monthly levels comes out at -0.86 from September 2011 to December 2020. Three main sub-windows succeed each other.

2011-2015: classical mean-reversion phase. The taper tantrum (May 2013 QE3 tapering announcement), the end of QE3 in October 2014, and the first Fed Funds hike in December 2015 organize a DFII10 climb from -0.87% (December 6, 2012) to a 0.73% average in December 2015. Gold retreats in parallel from $1,772 to $1,076 in monthly average between September 2011 and December 2015, a 39% loss. The correlation of levels comes out at -0.83 over this window.

2016-2019: macro-stable zone. The Fed delivers nine hikes between December 2015 and December 2018, DFII10 oscillates between 0.04% and 1.11% in monthly average, gold between $1,098 and $1,511. The correlation of levels weakens to -0.49, still negative. This phase signals that under “calm” monetary regimes, the real-yields channel loses some explanatory power, perhaps in favor of idiosyncratic factors (speculative ETF flows, trade-policy expectations under the Trump administration, etc.).

2020: Covid shock. In March 2020, the Fed deploys unlimited QE, brings Fed Funds to zero, and floods the system with liquidity. DFII10 plunges to -1.08% on August 6, 2020; the series low comes a year later, -1.19% on August 3, 2021. Gold reaches a record monthly average of $1,969 in August 2020 (World Bank), above the September 2011 peak. The correlation of monthly levels reaches -0.98 over 2020. The real-yields channel dominates without contest.

2022-2024: The Major Empirical Break and What It Reveals

The 2022-2024 period constitutes the most instructive break in the empirical TIPS history. Over this three-year window, the DFII10 yield performs a climb of unprecedented amplitude: from -1.19% on August 3, 2021 to 2.52% on October 25, 2023, 3.7 points in just over two years, the largest rise in the series over a comparable span.

On the regression line estimated over 2003-2021, a 2.4% real yield, its October 2023 average, corresponded to gold around $500. Yet gold does the opposite: it goes from $1,948 in March 2022 to $2,026 in December 2023 in monthly average (World Bank), a 4% gain while opportunity cost climbs. The correlation of monthly levels turns positive: +0.47 over 2022-2024 and +0.46 from January 2022 to August 2026, against -0.91 over 2003-2021.

This decorrelation is not transient statistical noise. It lasts three years, and its amplitude falls outside the usual deviations: at end-2023, gold trades about $1,300 above the level implied by DFII10 on the 2003-2021 line, whose deviations have a standard deviation of $185. The reading that imposes itself is that an alternative channel, different from real yields, numerically dominates price formation over this period. Three plausible readings coexist to explain this channel, of which the first, structural override from central bank buying, is treated specifically in the central bank channel and the 2022-2024 decorrelation. The other two, de-dollarization expectations and perceived monetary debasement, are synthesized in the analytical framework for gold as a monetary signal of this cluster.

2025-2026: The Real Yield Eases, Then Climbs, and Gold Keeps Rising

The DFII10 yield eases in 2025, from a 2.23% average in January to 1.75% in September, then climbs in 2026: 2.04% on average in May, 2.93% on September 30, its highest since November 2008. Gold crosses $3,000 at the London fixing in March 2025 (LBMA). It reaches a monthly average of $5,020 in February 2026, then $4,411 in August (World Bank). From January 2025 to August 2026, the correlation of monthly levels is only -0.18, far from the -0.91 of 2003-2021.

This near-zero correlation suggests that the real-yields channel still operates, but with a much reduced relative weight. Central bank net purchases, above 1,000 tonnes a year in 2022-2024 and lower in 2025 according to the World Gold Council, remain a source of price support independent of the real yield. For the attentive reader, this configuration imposes a two-channel simultaneous reading grid: real yields for the short-to-medium cyclical component, central banks for the structural floor. The correct superposition of these two readings is the condition to interpret the $3,000+ regime and its empirical rupture without falling into mono-causal projection.

A Live Mechanic, Never a Stable Coefficient

The central lesson of this empirical reading sub-period by sub-period is that the gold-real-yields correlation is neither a stable regularity nor a fiction. It is a live mechanic whose relative weight varies with dominant macro regimes. When the real-yields channel dominates (2008-2011, 2011-2015, 2020), the correlation of monthly levels sits between -0.83 and -0.98. When another major channel activates (official buying in 2022-2024, perhaps soon de-dollarization expectations) the real-yields channel persists but becomes one channel among several, and the rolling correlation weakens or temporarily inverts. Directly related: the tax bill behind exiting gold.

To situate this empirical reading in the broader macro panorama, the complete inflation guide documents the conceptual framework in which breakeven expectations and the TIPS-CPI relationship sit. The CPI-adjusted gold price dataset supports the empirical inspection of the real-terms claims made here. The page on the broader physical commodities ecosystem aggregates connected analyses on commodities and structural macro signals.

Last updated — 8 October 2026