Rethinking Gold’s Rally: Real Rates, Reserve Diversification, and the Limits of the Inflation-Hedge Story — The Full Research Note

Ask an investor why gold is rising and you’ll usually get one of two answers: inflation is coming, or a recession is coming. Both answers are frequently correct at the same time, for the same rally — which should be strange, since the two conditions are supposed to call for opposite trades.

Gold has no yield, no earnings, and no industrial use large enough to explain its price on its own. Yet it has spent the past four years setting record after record, and more than fifty years after gold stopped backing any currency on earth, central banks are buying it at the fastest pace since 1950. This note works through both puzzles, section by section, with the reasoning and the numbers behind each one.

01 — FramingThe Paradox

Textbook portfolio theory treats inflation risk and recession risk as different problems requiring different insurance. An inflation hedge should be an asset that rises when the price of everything else rises, ideally something tied to real economic output, like commodities or property. A recession hedge should be an asset that holds up when growth slows and corporate earnings fall, typically a government bond, whose fixed payments become more attractive as riskier assets wobble. These are meant to be different trades, sometimes opposite ones, because a hot economy and a cold one call for different portfolio tilts.

Gold has spent the past few years defying that division, though not in a straight line. It gained roughly 64% over 2025 alone, its best calendar year since 1979, then kept climbing into a record near $5,600 an ounce in late January 2026 on a mix of safe-haven demand and tariff-driven uncertainty. It then gave back more than a quarter of its value, falling to roughly $3,975 by late June, its worst quarter since 2013, as a more hawkish Federal Reserve and fading geopolitical risk pushed real yields higher. From there it rallied hard through the summer on softening labor-market data and rising odds of Fed rate cuts, the classic signature of recession fear, touching a three-month high near $4,700 in late August before settling back toward $4,375–$4,430 in the first days of September. Bank forecasts for where it goes next are unusually spread out: JPMorgan’s year-end call sits near $4,500, Goldman Sachs’s near $4,900, and Wells Fargo’s outlier target above $6,000. Whatever the specific catalyst in any given week appears to be, gold’s direction over the past four years has been up far more often than down, even if the path there has been considerably rockier than a “record after record” headline suggests. That volatility is itself part of the puzzle this note sets out to resolve.

02 — MechanismThe Real-Rate Mechanism

The resolving idea is a single variable: the real interest rate, meaning a safe bond’s yield minus expected inflation. Because gold pays no interest and costs money to store and insure, holding it instead of a bond carries an opportunity cost, and that cost is precisely the real rate. When real rates are high and positive, government bonds hand investors a guaranteed increase in purchasing power, and gold’s zero yield looks like a bad trade. When real rates fall toward zero or turn negative, that advantage evaporates, and gold’s central weakness, that it produces no income, stops being a weakness at all.

This single mechanism is what lets two “opposite” fears push gold in the same direction. An inflation scare raises expected inflation, which mechanically compresses the real rate even if the nominal, posted interest rate doesn’t move an inch. A recession scare raises the odds that a central bank will cut its policy rate, which compresses the real rate from the nominal side instead. Different newspaper headlines, same arithmetic result: bonds become less attractive relative to gold, from two entirely different directions.

The inverse relationship between gold prices and real interest rates is one of the most consistently observed dynamics in financial markets — not theory, but a pattern supported by decades of data.

A concrete example makes the arithmetic easier to hold onto. Suppose a 10-year Treasury yields 3% and expected inflation over that period is 2.5%. The real rate is a positive 0.5%, meaning a bondholder locks in a small but guaranteed increase in purchasing power, and gold, which offers no such guarantee, has to compete against that. Now suppose the same nominal yield falls to 2% while inflation expectations stay at 2.5%. The real rate flips to negative 0.5%: anyone holding that bond to maturity is now locking in a guaranteed loss of purchasing power. At that point, gold’s lack of yield stops being a drawback, because the alternative isn’t actually risk-free either, it’s just a slower, more certain way to lose ground. This is the switch that flips during both an inflation scare and a recession scare, just approached from opposite sides of the same equation.

That relationship isn’t just a rule of thumb. In “The Golden Dilemma,” their widely cited 2013 study, Claude Erb and Campbell Harvey ran the numbers directly: measuring gold’s real price, its nominal price divided by the CPI, against the real yield on 10-year Treasury Inflation-Protected Securities from the start of TIPS trading in 1997 through early 2012, they found a correlation of -0.82, about as tight a relationship as cross-asset finance tends to produce. But Erb and Harvey were careful not to oversell their own finding. Running the same test on a much longer run of UK data, stretching back to the early 1980s, the correlation fell to just -0.31, explaining perhaps 9% of the variation in the UK’s real gold price. Their own conclusion was refreshingly blunt on the point: it is entirely possible to disagree that low real yields “cause” high real gold prices, since both could just as easily be responding to some third, harder-to-measure factor, like a shared fear of inflation spiraling out of control. Real yields north of roughly 1.5–2.0% have historically been a headwind for gold; real yields near zero or negative have historically preceded some of gold’s strongest multi-year runs. Real yields stayed stubbornly elevated through most of 2026, hovering near 2% in the spring and climbing further, toward roughly 2.35%, by August, yet gold still managed a sharp rally into early autumn. That’s a genuine near-term tension worth flagging: gold kept climbing through a period when the traditional opportunity-cost mechanism, on its own, would have argued for the opposite.

That tension points to something worth stating plainly: the real-rate relationship is the dominant mechanism, not the only one. Central bank buying, covered later in this note, has increasingly operated as an independent source of demand that doesn’t wait for the rate math to turn favorable.

03 — Myth-BustingThe Inflation-Hedge Myth

Here the popular narrative and the academic evidence diverge sharply. The most cited study on the question, Claude Erb and Duke finance professor Campbell Harvey’s “The Golden Dilemma,” tested the claim that gold reliably tracks inflation across the time horizons that matter to actual investors, and found it largely doesn’t hold up. Gold, the paper concludes, may function as an inflation hedge if the investment horizon is measured in centuries. Over the 1-, 5-, 10-, 15-, and 20-year horizons realistic for a human portfolio, the variation in gold’s nominal and real returns has not been meaningfully driven by realized inflation. Their own regression makes the point with a single number: running gold’s monthly price against the US CPI from 1975 onward, the “inflation-implied” price of gold, as of their writing in early 2012, came out to roughly $780 an ounce. Gold was actually trading above $1,650 at the time, more than double what the CPI alone would have predicted.

The paper’s broader finding is worth sitting with: gold’s price swings widely around what a strict inflation-tracking model would predict, and the two lines, actual gold price and an inflation-implied “fair value,” have rarely lined up for any sustained stretch since the 1970s. Erb and Harvey also flag a second, related caution, this one about valuation rather than inflation directly. By their preferred measure, gold’s real price (its dollar price divided by the CPI index) stood at 7.3 in early 2012, against a long-run average since 1975 of 3.2 and a prior extreme of 8.7 set at the January 1980 blow-off top. Historically, periods when that ratio ran well above its average have tended to be followed by below-average real returns over the following decade, plain mean reversion. Taken literally, their own historical relationship implied something close to a negative real return for gold over the following ten years.

History gave that forecast a split verdict. Gold did fall hard in the years immediately after the paper’s publication, dropping from roughly $1,650 to around $1,050 by the end of 2015, a decline of more than a third that looked, for a while, like textbook mean reversion arriving on schedule. Then the pattern broke. Rather than continuing to grind lower toward its long-run average, gold found a floor, and by the 2020s it was setting records that pushed its real price past even the 1980 extreme Erb and Harvey had flagged as the historical ceiling. Their mean-reversion warning was not wrong, exactly, it played out almost precisely as described between 2012 and 2015. It was simply incomplete.

What fills in the rest of the story is a second, less-quoted part of the same paper. Erb and Harvey didn’t just run the mean-reversion numbers backward, they ran a version of them forward. Looking explicitly at official gold holdings by country, they calculated that Brazil, Russia, India, and China held a combined 2,457 tonnes of gold as of 2010, and asked what would happen if those four countries simply closed the gap with developed-market ownership norms. Matching the US’s gold-to-GDP ratio alone would have required a fairly modest increase, to roughly 6,233 tonnes. Matching the US’s gold-to-population ratio was a different order of magnitude: 77,811 tonnes, more gold than every central bank on earth held in combination at the time. The authors were careful to frame this as an illustration of scale, not a prediction. But it is hard to read that passage today, written more than a decade before the buying wave detailed later in this note, and not see it as an unusually precise anticipation of exactly the mechanism, emerging-market central banks closing a gold gap with the developed world, that has since become gold’s dominant source of demand.

More recent scholarship has continued to probe this relationship without fully resolving it. A 2019 study using a Fourier-based statistical approach found gold had maintained purchasing power reasonably well across a 39-year window, arguing for a more favorable read than Erb and Harvey’s. A 2026 paper examining time-varying correlation across multiple frequencies concluded that gold’s effectiveness as an inflation hedge is regime-dependent, shifting with interest rate conditions and the broader economic backdrop rather than holding as a constant relationship.

None of this actually contradicts the real-rate story in Section 02. It refines it. Gold does not mechanically track the CPI. It responds to how inflation fear changes the relative attractiveness of yield-bearing assets, a more indirect and less tidy relationship than “inflation rises, gold rises in proportion” — but one that still produces a strong directional correlation, even if the magnitude is unreliable. Erb and Harvey tested, and largely dismissed, one more clean story along the way: that gold is a straightforward currency hedge, appreciating in lockstep with the decline of any given currency. Across seven major currency pairs, they found an average “gold beta” of just -0.15, with low explanatory power throughout. Gold, in other words, resists almost every attempt to reduce it to a single, tidy relationship. That resistance is arguably the paper’s real finding, buried under all the individual test results.

04 — The CaveatNot a Perfect Safe Haven Either

Gold’s reputation as crisis insurance carries a genuine asterisk, and the clearest illustration is the 2008 financial crisis. Gold touched roughly $1,011 an ounce in March 2008, during the Bear Stearns rescue, then fell about 28% to around $730 by October, in the panicked weeks surrounding Lehman Brothers’ collapse. The mechanism was not a loss of faith in gold. It was a liquidity crunch: as margin calls and redemption requests hit funds and institutions simultaneously, managers sold whatever could be sold quickly to raise cash, and gold, among the most liquid assets on earth, was sold early and hard alongside equities and nearly everything else. Silver, for comparison, fell roughly 57% over the same stretch, from about $20.90 an ounce in March to under $9 by late October.

What followed validated the longer-term thesis even as it complicated the short-term one: gold rose roughly 78% from its October 2008 trough within two years, ultimately peaking near $1,918 in August 2011, a gain of about 163% from the crisis low. A similar, smaller pattern played out in March 2020, when gold briefly traded below $1,500 during the most acute phase of the COVID-19 market panic before recovering within weeks. Academic research examining crisis periods has found gold’s safe-haven behavior to be genuinely inconsistent rather than automatic. Multiple studies of the early COVID-19 period specifically found that gold did not reliably protect investors during the sharpest days of that selloff, and a broader 2025 study tracking gold’s volatility from 1975 to 2024 characterizes the more recent 2005–2024 period as markedly more unstable than the relatively placid 1980–2005 stretch that originally built gold’s safe-haven reputation.

The practical takeaway isn’t that gold’s safe-haven role is fictional. It’s that the role operates with a lag during acute, forced-selling liquidity events, and investors expecting gold to move inversely to risk assets on every single bad day have occasionally been disappointed at exactly the moment it mattered most.

05 — HistoryThe Nixon Shock

None of the market mechanics above explain why national governments still hold gold at all. To understand that, the story has to go back to August 15, 1971, when President Richard Nixon announced the suspension of the dollar’s convertibility into gold, an event now generally referred to as the Nixon Shock.

The arrangement Nixon ended had been running since the 1944 Bretton Woods Conference, where 44 Allied nations agreed to peg the US dollar to gold at $35 an ounce and peg every other major currency to the dollar in turn, making the dollar, in practice, “as good as gold” for foreign central banks that could redeem their dollar holdings at that fixed rate. By the late 1960s the arrangement was under severe strain: heavy US deficit spending, driven partly by Vietnam War costs, meant more dollars were circulating abroad than America’s gold reserves could plausibly back, and foreign governments increasingly tested that promise by asking to redeem dollars for gold. Something had to give. Nixon’s announcement, framed at the time as a temporary measure to deter speculators, in fact permanently closed the gold window; a common misconception holds that this ended gold convertibility for everyone, but US citizens had already lost the right to redeem dollars for gold back in the 1930s; what closed in 1971 specifically was the channel available to foreign central banks. The fixed exchange rate system did not collapse overnight, either, surviving in weakened form until the transition to fully floating exchange rates in March 1973.

Since that peg was abandoned, gold has moved from $35 an ounce to roughly $4,400, an increase of more than 12,000%, even as the dollar itself lost the vast majority of its purchasing power over the same stretch. Central banks, having watched the last formal link between paper money and a tangible asset get severed, did not stop wanting to hold gold. If anything, the years since 2008, and especially since 2020, when central banks expanded their balance sheets at a pace that would have been structurally impossible under any gold-backed system, appear to have reinforced the instinct.

06 — InstitutionsThe Institutional Case: Basel III

There’s a persistent claim, repeated across a great deal of financial media, that global banking regulators formally reclassified gold as a top-tier reserve asset in 2019, sparking the current buying wave. The reality is more precise, and worth getting right: gold’s treatment as a Tier 1 capital asset with a 0% risk weighting, meaning banks can count it at full value without setting aside extra capital against it, actually dates to the original 1988 Basel I Accord, not to Basel III. The London Bullion Market Association has explicitly and publicly corrected the more sensationalized version of this story, noting that nothing has changed on that specific point since 1988, and that gold has still not been classified as a High-Quality Liquid Asset, a related but distinct liquidity designation that the LBMA and World Gold Council continue to lobby for.

What is genuinely new is not the rule but the rumor mill that grew up around it. Basel III’s liquidity framework, the piece that actually determines which assets count as HQLA, took effect internationally beginning January 1, 2019, and gold has never qualified under it, before or since. A specific and quite different claim circulated widely online through 2025 and into 2026: that gold was due to be formally reclassified as a Level 1 HQLA effective July 1, 2025. The LBMA moved to correct that story directly in a May 2025 notice, stating plainly that no such announcement had been made or was expected, and attributing the confusion to commentators conflating the 1988 Tier 1 capital rule with the separate, and still unresolved, HQLA liquidity question. Separately, and with no direct bearing on gold at all, the broader US “Basel III Endgame,” a package of capital rules for the country’s largest banks, has followed its own long and unfinished road: first proposed in 2023, once targeted for July 2025, delayed amid heavy industry pushback and a change in Federal Reserve leadership on bank supervision, and re-proposed by regulators in March 2026 with the comment period closing that June and full implementation not expected before 2027. Neither strand of Basel III, in other words, changed anything about gold’s regulatory treatment on the date widely reported. What’s true, and considerably less dramatic, is that gold’s favorable capital treatment is old news wearing a new headline: the rule that makes it attractive to hold on a bank’s balance sheet was written in 1988 and has not moved since.

07 — The CatalystThe Russia Catalyst

If Basel III explains why gold sits comfortably on a modern balance sheet, a single 2022 decision explains why central banks suddenly wanted so much more of it. In February 2022, in response to Russia’s invasion of Ukraine, the United States, European Union, United Kingdom, Canada, and Japan jointly blocked the Russian central bank’s access to its foreign exchange reserves. Roughly $300 billion of Russia’s approximately $640 billion in gold and foreign currency reserves was frozen, according to the Congressional Research Service, with the bulk of that sum, managed through the Belgian depository Euroclear, held in euros, dollars, pounds, and yen.

The lesson was not lost on the rest of the world’s central banks. Reserves held in a foreign jurisdiction, denominated in a foreign currency, ultimately depend on that foreign country’s continued cooperation, and can be switched off by political decision rather than market forces. Gold held domestically, in a country’s own vaults, cannot be frozen by another government’s sanctions regime; it has no issuer, settles no counterparty risk, and answers to no foreign court. Net central bank gold purchases hit 1,136 tonnes in 2022, the highest annual total since 1950, including the immediate post-1971 period when reserve managers might have been expected to rebuild gold holdings after Bretton Woods collapsed. Buying stayed above 1,000 tonnes for two more years before cooling to 863 tonnes in 2025, a meaningful pullback of roughly a quarter from the 2022 peak, yet still nearly double the 2010–2021 average of 473 tonnes.

The buying is not confined to countries with adversarial relationships with Washington. Poland has been the single largest national buyer for two consecutive years, adding 102 tonnes in 2025 and lifting its reserves to 550 tonnes, or 28% of its total holdings, a share the National Bank of Poland has signaled it intends to keep raising, toward a longer-term goal of 700 tonnes. Kazakhstan was the year’s second-largest buyer, adding 57 tonnes, its biggest annual increase since records began in 1993. Brazil re-entered the gold market after a four-year pause, buying 43 tonnes between September and November. Azerbaijan’s sovereign wealth fund added 38 tonnes through the first three quarters of the year, and Turkey bought a further 27 tonnes. This is a useful corrective to a common oversimplification: gold accumulation is frequently framed purely as a China-and-Russia de-dollarization story, but the buyer list is dominated by countries with no particular grievance against the dollar, simply central banks responding rationally to a demonstrated new category of reserve risk.

08 — The MilestoneThe Crossover

The cumulative effect of several years of this buying, layered on top of gold’s own sharp price appreciation, produced a real milestone that made headlines across financial media when it was confirmed. According to a European Central Bank report published in June 2026, gold’s share of global central bank reserves rose to 27% by the end of 2025, up from 20% a year earlier, while the share held in US Treasuries fell to 22% from 25% over the same period. That is the first time gold’s share has exceeded Treasuries’ share in nearly three decades, since 1996.

The ECB was careful to specify how much of that shift is buying and how much is simply price. Re-running the same reserve figures using the prices that prevailed at the end of 2023, the bank noted, would leave gold and the euro each at roughly 16% of reserves and Treasuries closer to 26%, essentially the old ordering. In other words, the crossover happened this fast mostly because gold got dramatically more expensive, not because central banks suddenly rotated a large share of their portfolios into it within a single year. Both forces are real and mutually reinforcing, sustained buying provided the tonnage, a record price provided the leverage, but the balance between them matters for anyone trying to judge how durable the 27% figure is if gold’s price were ever to give back a large share of its recent gains.

The value of gold held by foreign central banks is now approaching $4 trillion, against roughly $3.9 trillion held in US Treasuries, according to World Gold Council estimates cited alongside the ECB’s figures. It’s worth being precise about what this milestone does and does not signal. Dollar-denominated assets overall, not just Treasuries specifically, still account for the largest single share of global reserves, at roughly 42%, essentially unchanged as a broad category even as its composition shifts internally. The ECB’s own framing was explicit on this point: gold overtaking Treasuries does not mean the dollar’s standing as the dominant reserve currency has suddenly collapsed.

It’s also worth revisiting the mean-reversion caution raised in Section 03. Erb and Harvey’s own real-price ratio, 7.3 against a long-run average of 3.2, already looked stretched in 2012; gold’s real price today has pushed well past even the 8.7 extreme set at the 1980 blow-off top, into territory their historical sample never actually reached. Their warning was not wrong, exactly. It played out almost precisely on schedule between 2012 and 2015. It was incomplete: the other half of their own paper, the observation about emerging-market central banks closing a gold gap with the developed world, has done more to explain the decade since than the valuation warning did on its own. None of the structural drivers in this note, the real-rate mechanism, the sanctions-driven reserve diversification, the Basel III capital treatment, guarantee that gold’s price keeps climbing at anything like its recent pace. What they do explain is why demand has proven unusually durable across very different macro conditions over the past four years, and why a growing share of the institutions setting that demand are not especially price-sensitive, short-term traders but central banks executing a multi-decade reserve strategy. What this milestone signals, more than anything, is a structural reweighting, a decade in the making and sharply accelerated since 2022, toward an asset whose appeal has nothing to do with yield and everything to do with the fact that no government can print more of it, freeze it on another country’s instruction, or default on an obligation it never issued in the first place.