Once per quarter I update a Graham Value Stock Portfolio (the most recent portfolio update can be viewed here). It draws inspiration from the work of a well-known investor, Benjamin Graham. The portfolio cannot precisely mimic how Graham would invest today, but it strives to remain philosophically consistent with his emphasis on value and company fundamental strength and an emphasis on stability.
Graham may be one of the best known investors of all-time, but there are myriad examples of historical and modern-day “all-star” investors. Recently, a reader asked me to investigate a Piotroski model. Joseph Piotroski is a modern academic, perhaps best known for his research paper, Value Investing: The Use of Historical Financial Statement Information to Separate Winners from Losers. His paper documented a systematic, accounting-based fundamental analysis strategy that historically outperformed equity indexes. He emphasized price-to-book valuation, supplemented with fundamental analysis.
Fortunately, we can easily run a “Piotroski-eque” screen using Portfolio123. The Portfolio123 Piotroski model is built on the concepts identified in his 2002 paper, although the details of implementation are Portfolio123’s. In other words, as with the Graham model, this screen/portfolio may not exactly mimic how Piotroski would invest, but it does stay true to the philosophy.
The screen uses the following rules:
Eliminate companies classified in the Miscellaneous Financial Services Industry
Trailing 12 month “Business Income,” defined as Sales minus Cost of Goods Sold minus Selling, General & Administrative expenses (unusuals that are often included in “operating Profits” are eliminated) must be in the black
Trailing 12 month Cash from Operations per share must be in the black
Trailing 12 month Cash from Operations per share must exceed trailing 12 month EPS
Trailing 12 month Gross Margin must exceed Gross Margin for the prior 12 months
Debt-to-assets in the latest quarter must be less than Debt-to-assets in the prior-year quarter
Current Ratio in the latest quarter must be less than Current Ratio in the prior-year quarter
Trailing 12 month Asset Turnover must exceed Asset Turnover for the prior 12 months
Trailing 12 month Return on assets must exceed Return on assets for the prior 12 months
Average shares outstanding in the trailing 12-month period must be less than average shares in the prior 12-month period
Among the companies that pass the above screen, the top 15 stocks are selected based on the Piotroski ranking system, which uses the following factors:
Price-to-Book, latest quarter – 50% of total
Fundamentals – 50% of total
Trailing 12 month Gross Margin minus Gross Margin for the prior 12 months (14.29% of category)
Trailing 12 month Cash from Operations per share minus trailing 12 month EPS (14.29% of category)
Debt-to-assets in the latest quarter minus Debt-to-assets in the prior-year quarter, lower is better (14.29% of category)
Current Ratio in the latest quarter minus Current Ratio in the prior-year quarter (14.29% of category)
Trailing 12 month Asset Turnover minus Asset Turnover for the prior 12 months (14.29% of category)
Trailing 12 month Return on assets minus Return on assets for the prior 12 months (14.29% of category)
Average shares outstanding in the trailing 12-month period minus average shares in the prior 12-month period, lower is better (14.29% of category)
Using the above screen and rank filters, one can backtest this strategy using Portfolio123.
The tests below were run from 1/2/1999 – 3/31/2013. The portfolio was rebalanced every four weeks. Slippage was assumed at .25% and the portfolio was 100% long at all times (no hedging or moving to cash). Return include dividends and the benchmark used was SPY. No assumption was made for taxes or commissions.
The 15-stock Piotroski Portfolio performed as follows:
We can compare these results to the Graham Portfolio. Using the same assumptions and timeframe as the Piotroski backtest, the Graham Portfolio performed as follows:
Both models have performed well since 1999, with annual returns in excess of 20%. However, the Piotroski 25.04% annual return and .78 Sharpe Ratio well outpaced the 20.07% annual return and .59 Sharpe Ratio for Graham.
Both models suffered large drawdowns in 2008, with the Piotroski model suffering a drawdown of 61.81% (ouch!). A 61.81% drawdown is more than a footnote, especially for the investor suffering through it. The Graham model did not fare much better, with a 56.31% drawdown, and index investors (via SPY) experienced a 55.42% drawdown!
To help reduce these drawdowns, I added a “Hedge” rule to the Piotroski model. When the benchmark (SPY) was below its 200 day simple moving average on the model’s rebalance date, the model still went long the top 15 stocks (same rules as above) but simultaneously shorted SPY until the next rebalance date.
No assumption for carrying costs (margin interest) was assumed in this test. These costs would reduce returns in the real world. However, the present-day accessibility of inverse ETFs could help mitigate these costs, since an investor could go long a 1x inverse ETF and avoid margin interest.
The hedge rule increased the Piotroski Sharpe Ratio to .98 and decreased the max drawdown to -37.01% (still not a pleasant experience, but significantly improved from -61.81%). Annualized return increased in this test to 26.55%:
(Test results courtesy of Portfolio123)
The future could bring very different results for these models. However, the application of fundamental and value-oriented criteria, combined with simple momentum based hedging rules, has generated strong returns since 1999.
4/22/13 clarification: The amount of the SPY short was equal to the amount of the 15 holdings, or, a 100% hedge.