Gold Price Transmission and Tail Risk in a Frontier Commodity Market: Evidence from Vietnam

Vietnam’s domestic gold price has persistently exceeded the world price by a wide margin, even as recent reforms have begun to relax the state’s historical monopoly over gold-bar production and imports. This paper asks why the gap persists, and whether it is confined to normal market conditions or extends into periods of extreme price movement. Using daily data spanning 2 January 2019 to 31 July 2026 (1856 trading days), covering the reform introduced by Decree No. 232/2025/ND-CP we decompose the domestic premium into a currency component and a pure physical-gold component, and use a copula-based framework to separately assess average price linkage and tail (extreme-event) co-movement between the domestic and world markets. Domestic gold bars traded at an average premium of 16.0% over import-parity world prices, of which 13.8 percentage points reflect the physical-gold component driven by constrained arbitrage, while currency factors account for only about 2 percentage points. The average linkage between the two markets is weak, indicating persistent segmentation, and this segmentation extends into the tails of the distribution for most of the sample. The premium itself carries substantial latent risk: a reversion to price parity would imply a one-off loss of about 9.6% of value, roughly eight to ten times the historical one-day 5% Value-at-Risk. Following the reform’s effective date, however, we find early evidence of emerging co-movement specifically in extreme upside price movements, even though the physical premium itself has not yet narrowed—consistent with a reform that has been enacted in law but remains at an early stage of operational implementation. The results indicate that administrative restrictions on the physical gold supply chain, rather than currency controls, are the principal source of Vietnam’s persistent gold-price gap, with direct implications for how the ongoing liberalization process should be sequenced.