The crypto collapse chronicles: Decoding cryptocurrency exchange defaults

This research explores the factors contributing to the failure of cryptocurrency exchanges by analyzing a sample of 845 exchanges. Using logit and probit models, it identifies key variables affecting cryptocurrency exchange defaults. The results show that cryptocurrency exchanges that are centralized, located in countries with high transparency indices, and offer fewer peer cryptocurrencies are more likely to default. Additionally, exchanges that impose high withdrawal fees and have no restrictions on clients from the United States are also positively associated with defaults. Moreover, the absence of referral schemes and having lower ratings each contributes marginally to defaults. Machine learning (ML) models including random forest, support vector machine, stacked ensemble confirm the robustness and high predictability of cryptocurrency exchange defaults. • This study uses statistical and ML models to predict cryptocurrency exchange default. • Centralized exchanges from high-transparency-index nations are more prone to default. • Limited coin listings, high fees, and U.S. client access increase the default risk. • Lacking referral programs and low ratings both marginally contribute to default.