Comparative Multidimensional Risk Analysis of Stock and Cryptocurrency Returns Using Factor Analysis and Logistic Regression

Mohammad Sufa Ammar Habibi, A. Muhaimin, Muhammad Nasrudin

Abstract

This study presents a comparative multidimensional risk analysis of IDX30 stocks and CMC100 cryptocurrencies to examine whether both asset classes exhibit fundamentally different empirical risk-return structures beyond conventional single-indicator measures such as volatility or average return. Daily log returns of 30 IDX30 stocks and 30 CMC100 cryptocurrencies from July 2023 to February 2026 were analyzed using eleven statistical indicators representing distribution moments and extreme tail behavior. Factor analysis was employed to reduce these indicators into latent risk dimensions, followed by binary logistic regression to classify the two asset classes. Portfolio optimization was conducted using the Mean Absolute Deviation approach, while extreme downside risk was evaluated using Historical Value at Risk. The analysis extracted two dominant latent factors, namely the Tail Factor and the Moment Factor, explaining 85.70% of the total variance. Logistic regression identified the Tail Factor as the main distinguishing variable, with confusion matrix evaluation correctly classifying 56 of 60 assets and achieving an accuracy rate of 93.33%. Portfolio optimization results show that a diversified portfolio combining stocks and cryptocurrencies achieved the highest Sharpe ratio, indicating superior risk-return efficiency compared with single-asset portfolios. However, the portfolio remained exposed to significant tail risk under extreme market conditions.

Source: semanticscholar · PDF

Read the AI summary, key takeaways and discussion on WOBR Quant Research.