This article is a sequel (previous episode here). It presents a dividend strategy on S&P 500 companies with no survivor bias and no look-ahead bias.
New Strategy Description:
I start in the S&P 500 universe. Only companies whose average yield for 5 trailing years is above 5% are taken into account. Stocks whose short interest is above 5% are excluded. Finally, the 10 stocks with the highest current yield are selected.
Rules of Simulation:
- The portfolio is rebalanced every quarter from 1/1/2002, to 4/1/2012.
- A 0.2% slippage is applied for every trade.
- Capital is equally allocated between the stocks.
- The maximum size of a position is limited to 1/10 of the total amount.
- Average quarterly gain (dividends included): 3.2%
- Average gain/average loss (quarterly): 1.53
- Probability of gain (quarterly): 70%
- Drawdown: 19.2%
The annualized return is about 13% (vs 15% in the previous version). The Kelly criterion (over 50%) shows a strong probabilistic robustness. The drawdown is much better. However, the Calmar ratio (= Annualized return / max Drawdown) is only 0.68 and doesn't pass the selection for my personal portfolio (more info here).
At the time this article was written, the 10 positions are filled: CTL, VZ, T, KIM, HCN, HCP, AEE, PGN, RAI, MO.
The list is more balanced than the previous version, but still concentrated in four industries: Communication Services, Real Estate, Electric Utilities and Tobacco.