Dynamic Factor Weights Red Devil Partners Joon Seong Choi, Youngjun Yoo, Richard Park, YK Kim
Overview Our purpose is to develop a stock selection strategy in order to outperform S&P 500. Our analysis includes both fixed and dynamic factor weights.
Source Data *Universe * FormulaSP_500 *Benchmark Same as Universe *Time Series *Start Date01/01/2000 *End Date12/31/2005 *CalendarUS COMPOSITE *Universe1 - Month *Main Returns1 - Month *Factors1 - Month *Weights1 - Month *Return Sources *Universe Return SourcesCompustat; *Benchmark Return Sources Compustat; *Risk Free Rate Return Sources US - Disc. Rate 91D T- bill; *Include DividendsYes *CurrencyU.S. Dollar
Steps 1)Specify list of factors 2)Univariate screens 3)Identify 5 fractiles for each factor 4)Choose significant portfolios 5)Optimize weights for portfolios with S&P500 volatility 6)Compare fixed weight strategy and dynamic weight strategy
Factor Returns
Identified factors Factors (1m lagged) - Cashflow to Price - Debt to Equity - Market Capitalization - Price to Book
Factor Screen Cashflow to Price(5) : value weighted Debt to Equity(5) : value weighted Market Cap(1) : equal weighted Price to Book(5) : equal weighted
Optimization: fixed weights Form a portfolio with same volatility of S&P500
Dynamic weight strategy Add dummy variables 3 months S&P500 momentum In negative momentum, buy more portfolio with negative correlation with S&P500 (Price to book (5))
Optimization: dynamic weights Form a dynamic portfolio with same volatility of S&P500
Results
Conclusion Multi-factor model strategy outperforms universe return (e.g. S&P500) Dynamic weight strategy outperform fixed weight strategy Future consideration: Transaction cost should be considered to evaluate strategies