Ethereum Risk States as a Tail-Risk Switch for Art NFTs

Abstract

Non-fungible token (NFT) markets are thin and typically settle in a cryptocurrency, so stress in the settlement asset can translate into abrupt drawdowns. This letter asks whether observable Ethereum (ETH) risk states provide an ex-ante ranking of crash risk in a curated art-NFT marketplace. Using SuperRare sales aggregated to a daily price proxy (2021–2023), we sort days by (i) 7-day realized ETH volatility and (ii) the filtered high-volatility probability from a two-state Markov-switching model. Forward 30-day drawdown crashes are sharply monotone across state quartiles: for example, a 30% USD crash rate rises from 9.9% to 38.8% from the lowest to highest volatility-probability quartile. Because crash windows overlap mechanically, conventional logit inference is overconfident; we therefore report main results as conservative linear probability models with Newey–West HAC errors and a moving-block bootstrap (logit results appear in the appendix for comparison). We further confirm results using a fully real-time state proxy based on an expanding-window volatility threshold, and document that crash predictability is strongest during the 2022 market stress episode, consistent with ETH risk regimes activating precisely when tail risks materialise. The settlement asset operates as a tail-risk switch for art NFTs, with limited corresponding mean-return predictability.