This study develops a financial-statement framework for identifying financing fragility at the firm level during thematic investment booms. We define financing fragility as persistent financing–fundamentals decoupling: repeated reliance on external capital accompanied by weak operating cash generation, limited liquidity support, and slow fundamental improvement. Using quarterly observations for publicly traded renewable-energy firms, we estimate a four-state Gaussian hidden Markov model on 2010–2018 data from five measures of external-financing dependence, operating-cash-flow weakness, liquidity stress, share dilution, and financing-conversion weakness. The fitted benchmark is held fixed for out-of-sample validation on 2019–2025 renewable-energy observations and for historical-analogue mapping of AI-related firms during 2023–2025. The model identifies Financially Supported, Intermediate Transition, Financing-Led Decoupling, and High Financing and Cash-Burn Stress regimes. Filtered state probabilities strongly predict external-financing dependence over the following eight quarters and are associated with subsequent volatility and selected downside outcomes. In the latest AI cross-section, 40 firms map to the supported regime, 9 to intermediate transition, and 4 to financing-led decoupling; the high-stress state is empty under the primary transport rule. The framework identifies evolving, bubble-like corporate financing configurations before terminal failure and provides a dynamic measure of firm-level financing fragility.
We investigate whether Bitcoin hedges trade policy uncertainty (TPU) shocks across different investment horizons. A TPU shock is defined as a day when the TPU index exceeds its 95th percentile. We combine wavelet decomposition with quantile regression, quantile-in-quantile (QQ) regression, and rolling quantile regression to examine how Bitcoin responds across return states and time scales. For investors, a positive total TPU effect means Bitcoin returns rise as trade policy uncertainty increases during shock periods, which is consistent with hedging behaviour. A negative effect means Bitcoin moves adversely with the shock and offers little protection. The standard quantile regressions produce effects that are mostly close to zero, indicating weak average hedging ability. The QQ estimates reveal stronger differences across horizons and market states. Short-horizon effects frequently change sign, while medium-horizon are largely negative. The long-run component becomes positive over parts of the middle and upper return distribution but remains negative in the lower tail. Rolling estimates at the 10th and 90th quantiles are also generally close to zero and statistically weak. Generally, Bitcoin does not provide a stable hedge against TPU shocks; any protection is conditional on market state and investment horizon.
This study examines how investor-facing cryptocurrency fraud varies with market conditions, incident technology, and major market disruptions using daily data from the De.Fi REKT database over February 2018 to October 2026. The analytical sample contains 3,096 recorded incidents, with rug pulls, honeypots, and phishing classified as investor-facing fraud and technical exploits examined separately. Negative Binomial count models show that investor-facing fraud is more frequent when investor sentiment is stronger, markets are more liquid, Ethereum network activity is more intensive, and during the COVID period. These relationships remain comparatively stable across alternative outcome definitions, sentiment orthogonalization, and controls for temporally concentrated incidents. By contrast, the positive baseline volatility relationship weakens once concentration effects are addressed, while market-regime effects are not robust across alternative cycle filters. Formal cross-type tests reveal substantial heterogeneity: rug pulls are particularly responsive to sentiment and network activity, whereas technical exploits exhibit little systematic sensitivity to investor sentiment. Crisis-period interaction models further show that the pre-Terra sensitivities of fraud to sentiment and volatility weaken materially after the 2022 market disruptions. Monetary-loss models reveal a distinct incidence-severity trade-off: stronger sentiment is associated with more frequent fraud but smaller conditional losses, whereas volatility is more strongly related to loss magnitude once losses occur. Overall, cryptocurrency fraud is state-dependent, technology-specific, and multidimensional, implying that surveillance based solely on aggregate scam counts or broad market cycles can overlook important changes in fraud composition and financial severity.
