Abstract The emergence of cryptocurrencies has presented investors with novel portfolio diversification opportunities. This study investigates the interplay between precious metals and cryptocurrencies, examining their potential for enhancing portfolio returns and mitigating risk. Using daily closing prices from August 2017 to November 2022, we employ an autoregressive distributed lag (ARDL) approach to analyze comovement and causality between these asset classes. Our findings reveal a short-run linkage between Bitcoin, precious metals, and other cryptocurrencies but no long-run cointegration. Notably, gold prices unidirectionally influence cryptocurrency prices, a relationship not observed with other assets. This absence of long-term comovement suggests that precious metal investors can leverage modern portfolio theory to diversify cryptocurrency volatility risk. These results offer valuable insights for investors seeking to optimize portfolios by strategically incorporating cryptocurrencies for improved risk-adjusted returns.
Dwi Fitrizal Salim, Farida Titik Kristanti, Hosam Alden Riyadh, Mailinda Tri Wahyuni
Type of the article: Research ArticleAbstractThis study evaluates and compares risk measurement models for ten major cryptocurrencies: Bitcoin, Ethereum, Tether, Ripple, Dogecoin, Cardano, Binance Coin, Polkadot, Solana, and USD Coin. Using daily log-return data from January 2017 to October 2024, the analysis applies Modified Cornish-Fisher Value-at-Risk and standard, exponential, threshold, and Markov-switching generalized autoregressive conditional heteroskedasticity models. The main comparison is conducted at the 99% confidence level, while model reliability is assessed through out-of-sample backtesting using 500 observations and the Kupiec unconditional coverage and Christoffersen conditional coverage tests. The results reveal substantial heterogeneity in cryptocurrency risk. Modified Cornish-Fisher Value-at-Risk produces highly sensitive estimates for assets with extreme skewness and kurtosis, particularly Ripple, Cardano, and Dogecoin. However, no single model performs consistently better across all assets. Bitcoin is the only cryptocurrency for which all tested models pass both backtesting procedures. The Markov-switching specification provides acceptable coverage for Bitcoin, Ripple, and Dogecoin but does not consistently outperform conventional volatility models. Standard and asymmetric volatility models provide stronger support for Cardano, Binance Coin, and Polkadot, whereas Ethereum, Solana, and USD Coin remain difficult to model under the examined specifications. These findings demonstrate that cryptocurrency risk measurement requires asset-specific model selection based on both estimated loss magnitude and formal backtesting evidence.
Arif Billah Dar, Isha Kumari, Safika Praveen Sheikh
Abstract The financialization of gold is investigated by analyzing the dynamic return spillovers between traditional, equity-based, exchange-traded and blockchain-based gold investment means. The empirical system includes physical gold, the SPDR Gold Shares ETF (GLD), VanEck Gold Miners ETF (GDX), NYSE Arca Gold BUGS Index (HUI), First Trust Gold Miners ETF (FTGM), PAX Gold (PAXG), and Tether Gold (XAUT). The study employs the Time-Varying Parameter Vector Autoregressive connectedness framework with generalized forecast error variance decomposition to estimate the magnitude, direction and evolution of shock transmission. The findings show that the financial economy around gold is quite tightly connected, with an average TCI of 81.13%, meaning that most of the forecast error variance of the system is attributable to cross-market shocks. GLD is the biggest net shock transmitter followed by GDX and XAUT, while the biggest net receivers are physical gold and PAXG. The results indicate that the price discovery and information transmission have been shifting from the underlying physical bullion market to exchange-traded markets and digital gold markets. The dynamic analysis also reveals that connectedness remains high but fluctuates throughout the sample, whereas PAXG gradually evolves from a net receiver to a net transmitter in the later part of the sample. Using the alternate Quantile connectedness methodology, we also find that the results are robust. Overall, it is clear that the results reflect the gradual transformation of gold from a commodity asset to a highly networked financial asset. The results are relevant to portfolio diversification, hedging, market monitoring and regulation of new gold products.
Fernando Henrique Antunes de Araujo, Milena Kojić, Petar Mitić, Kerolly Kedma Felix do Nascimento · 5 authors
This study applies a prespecified dynamic MFDFA workflow across pandemic, geopolitical-conflict, and tariff-policy regimes for ten non-stable, long-history cryptoassets selected ex post from the 23 July 2026 market-cap ranking. The common sample comprises 3155 daily log returns per asset from 2 December 2017 to 22 July 2026; the final endpoint regime includes the U.S.–Iran conflict. The estimator uses q=−10,−8,…,10, linear detrending, 22 scales from 16 to N10, adjacent-secant Legendre transformation, a cubic spectrum peak, and 500-day windows stepped by 21 days plus a terminal endpoint. Full-sample IE=|α0−0.5| ranges from 0.01397 (LINK) to 0.08706 (BNB), and observed width ranges from 0.32625 to 0.74678. The controlled incremental U.S.–Iran endpoint coefficient is −0.02734 (two-way clustered SE 0.02369; p=0.2489), with asset-cluster t(9) interval [−0.08226, 0.02759]. Quantile estimates range from +0.00206 at the 0.10 quantile to −0.04769 at the 0.90 quantile; all five 999-replication asset-cluster bootstrap percentile intervals include zero. Exact rolling sensitivities are negative for the 500/14, 500/30, and 730/30 designs but positive for the 250/21 design, and every small-cluster interval includes zero. Direct spectrum-width contrasts also remain nonsignificant after Holm adjustment. None of ten observed widths survives BH correction in 200 shuffled-return surrogates per asset (2000 fits in total). The results document heterogeneous and specification-sensitive dynamic multifractal patterns, not isolated or causal crisis effects.
This paper selects the data of Bitcoin, Gold, and the S&P 500 index from 2018 to 2025, utilizing GARCH(1,1) and DCC-GARCH models to depict the dynamic conditional correlations among assets. By incorporating the Global Geopolitical Risk Index, the 10-year breakeven inflation rate, and the VIX panic index, it constructs daily and monthly cross-frequency regression models to examine their macro-driving mechanisms. The results show that whether at the high-frequency daily level or the smoothed monthly level, macroeconomic variables exhibit extremely significant driving effects on the co-movement of Bitcoin. Under the liquidity squeeze concerns triggered by intensified global panic or high inflation expectations, Bitcoin fails to act as a haven alongside gold. Instead, it exhibits a stronger synchronous crash with the US stock market. This empirical study rejects the hypothesis of Bitcoin as "digital gold," revealing its essence as a "risk amplifier" highly dependent on traditional liquidity, and provides quantitative support for international investors in asset allocation under extreme macroeconomic scenarios.
Cryptocurrency time-series forecasting is a challenging task because market data usually exhibit high noise, strong volatility, non-stationarity, nonlinear dynamics, and long-range dependencies. In addition, multivariate market indicators often contain redundant or weakly informative variables, which may reduce forecasting accuracy and model interpretability. To address these issues, this study proposes BSFinformer, a Boruta-SHAP enhanced Finformer framework for multivariate cryptocurrency time-series forecasting. The proposed framework first applies a leakage-aware Boruta-SHAP feature selection strategy to identify informative market variables and remove redundant features. To avoid temporal information leakage, feature selection is performed only on the training set, and the selected feature subset is then applied unchanged to the validation and test sets. The selected features are subsequently fed into an improved Finformer model that integrates temporal embedding, sequence decomposition, and sparse self-attention to capture local fluctuations, trend evolution, and long-range temporal dependencies. Experiments are conducted on three cryptocurrency assets, namely Bitcoin, Dogecoin, and Binance Coin, using chronological train–validation–test splits. The proposed model is compared with classical forecasting models and recent long-sequence forecasting baselines, including LSTM, Transformer, Informer, Autoformer, DLinear, PatchTST, TimesNet, and iTransformer. Experimental results show that BSFinformer achieves competitive forecasting performance in terms of MSE and MAE. Ablation experiments further demonstrate the contributions of Boruta-SHAP feature selection, temporal embedding, sequence decomposition, and sparse self-attention. These results indicate that feature-selected temporal modeling can improve forecasting accuracy and interpretability for multivariate cryptocurrency market data.
The fluctuation characteristics of financial time series have always been one of the research hotspots in the academic community. Generally speaking, financial return series have the characteristics of volatility clustering, fat tails, conditional heteroskedasticity, asymmetric shocks, etc. The above phenomena can be explained from the perspective of dynamic conditional variance by GARCH models and their extensions. This paper first introduces the basic ideas of ARCH and GARCH models, with a focus on the issue of volatility clustering of financial returns. Then, it reviews the relevant research from three aspects: model evolution, application scenarios, and practical value. It also analyzes the role of GARCH-type models in capturing volatility persistence, asymmetric impact, and risk transmission through applications in cryptocurrencies, energy assets, and high-frequency financial data. The study shows that GARCH-type models capture the volatility clustering feature of financial returns well, but there is still room to improve the modeling of extreme risk, the handling of high-dimensional assets, and model interpretability.
Volatility control converts risk estimates into portfolio exposure, yet existing approaches often rely on a fixed volatility estimator or a pre-defined control rule that may not adapt to changing market conditions. We propose VolRouter, a modular framework that formulates volatility control as state-conditioned routing over estimator-controller pairs. VolRouter first summarizes market conditions into a control-relevant state profile and then performs routing through three stages: state inference, switch review, and pair selection. The Router can be implemented using rule-based, learnable, or LLM-based decision modules, while portfolio actions remain generated by predefined control policies. We evaluate VolRouter across S&P 500, Multi-Asset, Bitcoin, and USDT volatility-control settings. VolRouter achieves the highest Sharpe ratio in three of four settings. On S&P 500, it improves Sharpe from 0.952 for RV + Naive Scaling to 1.222 while reducing maximum drawdown from 15.10% to 12.58% and daily CVaR from 1.76% to 1.32%. On Multi-Asset, it improves Sharpe from 1.498 to 1.540 and reduces CVaR from 1.56% to 1.18%. Bitcoin shows similar improvements in risk-adjusted performance, while USDT provides a boundary case where simpler state-aware selectors remain competitive. Ablation and sensitivity analyses show that the improvement comes from relative policy evaluation and selective persistent switching rather than simply expanding the policy library. These results suggest that volatility control can be viewed as a policy-selection problem when risk management requirements vary across market states.