Tokenized Gold Under Stress: Divergent Transaction Costs in Thin Gold Markets During Crypto-to-Gold Rotations

We examine whether tokenized gold remains cost-effective to trade during crypto market sell-offs, when investors are most likely to shift from Bitcoin to the relative safety of gold. A safe haven that becomes expensive to transact in a crisis, is only a partial safe haven, yet we know little about how these instruments behave under stress. Using 1,574 daily observations (September 2019-December 2025) on tokenized gold (PAX Gold, PAXG; Tether Gold, XAUT), a deep gold ETF (GLD), and junior gold miners (GDXJ), we set a tokenized thin market against a conventional but equally thin one to separate what is specific to tokenization from what merely reflects a small market. Our dependent variable is the daily high-low range, which we treat throughout as a proxy for the cost of transacting under stress. Our primary result comes from the difference-in-differences (DiD) specification that absorbs a full set of 1,574 date fixed effects and so removes the common gold-price, volatility, and information shock that hits every gold vehicle on a given day. On Bitcoin to gold rotation days, the two thin vehicles move in opposite relative directions: the cost of transacting tokenized gold rises against GLD while for the junior-miner it falls. To understand why the spreads diverge, we estimate them jointly in a seemingly-unrelatedregression system. We find that the one force that formally separates the two is equity-market stress (VIX): the miner spread depends heavily on the VIX while the token spread does not. We read this as relative insulation of tokenized gold from equity fear, not immunity, because the ETF and the token co-move with the VIX by similar amounts and their difference nets out. Every other stressor, bond stress included, moves the two spreads alike. An event study around 31 candidate Bitcoin to gold rotation episodes then validates the timing and levels: both thin market vehicles (token and miner), but not the deep market GLD, deteriorate; the raw token and miner magnitudes look alike only because they share the same gold-market shock, and the resemblance dissolves once that shock is removed which is evident from DiD results.