10.6 What drives gold

10.6.1The WGC framework (GRAM)#

The World Gold Council's Gold Return Attribution Model (GRAM; https://www.gold.org/goldhub/tools/gold-return-attribution-model) is a multiple regression of monthly or weekly gold returns on four groups of drivers:

1. Economic expansion: growth proxies (equity returns, PMIs), which raise jewellery and industrial demand.
2. Risk and uncertainty: equity volatility, gold implied volatility, geopolitical-risk indices (Caldara–Iacoviello GPR; see M12).
3. Opportunity cost: the US dollar (DXY and EM currencies) and interest rates (nominal and real Treasury yields).
4. Momentum and trend: past returns, futures positioning (CFTC Commitments of Traders), ETF flows.

Formally, with $r_t$ the gold log return and $X_{k,t}$ the driver changes,

and the "attribution" for month $t$ is the vector $(\hat\beta_k X_{k,t})_k$ plus the residual. WGC publishes the methodology on the tool page (free login). WGC acknowledges a large unexplained residual since 2022, which it attributes to central-bank buying and an event-risk premium (WGC market commentaries on https://www.gold.org/goldhub/research).

Two technical warnings from M02 and M04 apply. The drivers are collinear (the dollar and real yields move together), so individual $\hat\beta_k$ are unstable even when the fit is good. And the attribution depends on the regression window; a rolling window that ends in 2021 gives a different 2026 story from one that ends in 2025.

10.6.2Real yields: the opportunity-cost argument and its breakdown#

The argument Holding gold means forgoing the real yield on an inflation-protected bond. By Proposition 10.8 the real price of a durable asset with no cash yield falls when the real discount rate rises. The observable proxy is the US 10-year TIPS yield (FRED series DFII10). Before 2022 the gold price tracked the inverse of DFII10 closely.

The breakdown In 2022 the 10-year real yield rose from about −1.1% to about +1.7% (FRED DFII10: −1.06% on 30 Dec 2021, 1.74% on 3 Nov 2022; see M11 · 11.16), and it reached about +2.5% in October 2023. Gold fell from about $2,050 to about $1,620 during 2022, much less than the pre-2022 relationship implied (reported by Bloomberg via InvestmentNews). From 2023 to 2025 real yields stayed near +2% while gold roughly doubled. The usual explanation is price-insensitive central-bank buying after the 2022 freezing of Russia's reserves, plus Asian demand and a geopolitical premium. In the language of (10.17), a large omitted variable (official-sector demand) entered the residual.

The 2026 re-coupling The 10-year real yield rose from 1.76% on 2 Mar 2026 to 2.13% on 27 Mar, and then to a 2026 high of 2.95% on 5 Oct (FRED DFII10, from US Treasury data); it was 2.64% on 4 Nov. As it rose, the opportunity-cost channel reasserted itself. The oil shock pushed up inflation expectations, the Fed turned hawkish (and hiked on 16 Sep 2026), and gold fell 11.8% in March 2026 despite a war in the Gulf (Reuters report, republished by MarketScreener; secondary). 10.11.4 analyses the episode.

How to test it yourself Regress $\Delta\ln(\text{gold})$ on $\Delta\text{DFII10}$ and $\Delta\ln(\text{DXY})$ with 52-week rolling windows from 2006 to 2026 (FRED series through toolkit.data.fred; gold and the dollar index from the app's lab or a cached CSV read with toolkit.data.load(..., source="cache")). Plot the betas. Add a Chow or Bai–Perron break test at 2022 (M13A · 13A.10). Report HAC standard errors (M04). Expect the real-yield beta to shrink toward zero after 2022 and partly recover in 2026; check whether your data say so before believing it.

10.6.3The other drivers (dated numbers)#

Central banks WGC data: central-bank net purchases were above 1,000 t a year in each of 2022, 2023 and 2024 (exact figures are revised; the 2022–24 run was the highest since 1967), 863 t in 2025 (WGC FY2025 press release), and 321 t in Q3 2026 (+42% y/y, strongest since Q4 2024), with 675 t year-to-date 2026 (WGC Gold Demand Trends Q3 2026, published 3 Nov 2026). Only about 120 t of the Q3 2026 purchases were reported to the IMF; about 63% was unreported and estimated by WGC.

ETF flows Q3 2026 global gold ETF inflows were 211 t, taking holdings to a record 4,256 t (same WGC report). In India, gold ETF net inflows were ₹4,451 cr in September 2026, with about ₹3,000 cr of outflows in May 2026 after the highs (AMFI data as reported in secondary summaries; the monthly figures are on https://www.amfiindia.com).

Aggregate demand 2025 demand was a record 5,002 t at an average $3,431/oz (WGC FY2025). Q3 2026 demand was 1,369 t (−1% y/y), with jewellery −17% (the weakest Q3 on record) at an LBMA average of $4,262/oz (WGC Q3 2026). High prices destroy jewellery demand; investment and official demand must offset it.

The dollar Gold is priced in dollars, so a stronger dollar lowers the dollar gold price mechanically for non-US buyers and through the opportunity-cost channel. The ICE dollar index (DXY) was about 101.5 on 6 Nov 2026 (secondary quote; ICE does not publish a free history). The Federal Reserve's nominal broad dollar index, a free primary alternative, was 121.67 on 30 Oct 2026 (FRED DTWEXBGS).

Momentum Gold trends: time-series momentum is documented across futures markets including gold (Moskowitz, Ooi and Pedersen 2012). Trends end abruptly when positioning is crowded, as in January 2026 (10.11.4).

Geopolitical risk Gold reacts to geopolitical events mainly through the dollar and rates response, not through the event itself (M12). The 2026 war is the cleanest counterexample to the "war means gold up" heuristic.