Do Classic Factor Models Explain Cryptocurrency Returns?

Abstract

This thesis examines whether classical asset pricing models can explain cryptocurrency returns or whether cryptocurrency-specific elements are required. Using daily data from 2017 to 2025 on a sample of over 2,300 cryptocurrencies, five factors are constructed: market, size, momentum, volatility and liquidity. The analysis compares the Capital Asset Pricing Model (CAPM), a three-factor model (market, size and momentum) and an extended five-factor model with volatility and liquidity factors. The results indicate that the CAPM performs poorly and produces large alphas. The extended five-factor model outperforms both nested models, showing higher R² with slightly reduced intercepts, however, the alphas remain large. Volatility is significantly associated with return in the 2023-2025 period, the liquidity factor is negative in later years, while the size results suggest relatively stronger performance of large cap cryptocurrencies in this sample, and momentum reverses (winners underperform). Sub-period analysis demonstrates time-varying factor importance, and leave-one-out tests confirm Bitcoin's high R² is partly mechanical. Our findings suggest that traditional single-factor models do not effectively capture the variation of cryptocurrency returns. A multi-factor framework incorporating market size, momentum, liquidity, and volatility provides a more informative benchmark for future research on cryptocurrency factors.