Digital Asset Funding Rates as a Distinct Risk Factor: Evidence from a Systematic Harvesting Approach
Description
Research Hypothesis: Digital asset funding rates represent a distinct, harvestable risk premium that operates independently from traditional investment categories through delta-neutral cash-and-carry strategies. Key Findings: Performance: 10.56% annualized returns with 1.50% volatility over August 2020-August 2025 Correlation: Near-perfect decoupling with correlations bounded within ±0.10 versus traditional assets Independence: PCA shows 95.6% loading on independent third principal component Regression: Traditional factors explain only 0.13% of return variation (adjusted R²) Structural Analysis: Significant regime shift post-FTX collapse (November 2022): Pre-FTX: 13.64% returns, 1.96% volatility Post-FTX: 8.14% returns, 0.97% volatility Portfolio Impact: Mean-variance analysis shows 38bp additional return at same volatility level or 182bp risk reduction at same return level. Risk parity assigns median 91.7% allocation to the factor. Theoretical Implications: Identifies crypto-native risk-free rate statistically orthogonal to conventional risk dimensions, challenging traditional monetary policy paradigms and supporting digital assets as distinct asset class.
Files
Steps to reproduce
Strategy Implementation: Fully collateralized delta-neutral approach using 50/50 BTC-ETH spot positions with offsetting perpetual swap shorts, rebalanced thrice-daily at UTC 00:00, 08:00, 16:00. Incorporates 57bp custodian costs and 10bp trading costs per rebalancing event. Data Sources: Crypto data: Binance API (8-hour intervals, Aug 2020-Aug 2025) for funding rates, spot and perpetual prices Traditional assets: Microsoft Excel (Jan 2005-Aug 2025) for BWX, LQD, VTI, GSG, RWR ETFs AQR factors: Monthly factor premia data for alternative risk factor comparison Analysis Framework: Rolling 1-year correlations for dynamic relationship analysis PCA to extract latent dimensions of return co-movement Multivariate regression to quantify explanatory power of traditional factors Mean-variance optimization using 10,000 randomly weighted portfolios Risk parity allocation with 252-day rolling volatility windows Data Processing: Python-based pipeline converting 8-hour intervals to daily returns, handling missing values/outliers, and implementing geometric compounding for return consistency. All analysis accounts for real-world implementation constraints including transaction costs. Reproducibility: Complete framework with all code and data available for replication, ensuring transparency and verification of results through automated execution via reproduce_analysis.py.