U.S. Monetary Policy and Central Banks' Physical Gold Reserves: Evidence from a Cross-Country Panel, 2000–2025

Abstract

This study examines the relationship between U.S. monetary policy and adjustments in central banks' physical gold reserves. Unlike the U.S.-dollar value of gold reserves, physical gold tonnage is not mechanically revalued when gold prices change. We construct a quarterly cross-country panel from raw data covering 2000Q1–2025Q4, after lagging and common-sample screening, the effective estimation window is 2000Q3–2025Q4. The empirical specification includes country fixed effects, country-specific linear and quadratic trends, quarterly seasonal indicators, and Driscoll-Kraay standard errors that allow for heteroskedasticity, serial correlation, and cross-sectional dependence. The baseline sample contains 5,832 observations for 65 economies. The coefficient on the one-quarter-lagged effective federal funds rate is −0.0167 and statistically significant at the 1% level. Estimates retain the same direction when the shadow rate is used, additional macroeconomic controls and the U.S. real yield are included, the reserve ratio is winsorized, and the sample begins in 2006. The absolute point estimate for the low-foreign-exchange-reserve group is about 1.94 times that for the high-reserve group, but the interaction is not statistically significant and does not establish a structural difference. After lagged gold holdings are included, the policy coefficient approaches zero, it is also statistically insignificant in first differences. The evidence therefore indicates a negative medium-run association with the level of physical gold reserves, but does not show that central banks actively sell gold in response to a rate increase. The results support rule-based reserve rebalancing across return, liquidity, and safety objectives rather than treating short-run U.S. policy changes as mechanical signals to buy or sell gold.