Magic Formula Part 1

If this is your first visit to Three Grahamian Virtues, please take a moment to read the welcome message and site philosophy: http://3gv.blogspot.com/2006/01/welcome-and-3gv-philosophy.html

Introduction

Joel Greenblatt's Magic Formula strategy seeks to purchase baskets of "good" businesses at "bargain" prices. This presentation will explore the specific methodology Greenblatt used, the returns garnered, my reflections, suggested modifications to the plain vanilla strategy, exact steps one could follow to enhance returns through buying the best of the Magic Formula companies while avoiding the laggards, and actual portfolios generated by innovative applications of several refined Magic Formula-based approaches (these will be tracked over time to demonstrate their effectiveness; if you have the means to accurately back-test and are moved by what you read here, I would encourage testing the approaches I describe and greatly appreciate the opportunity to look over your shoulder).

Methodology

1. A set of businesses is ranked from highest to lowest trailing twelve month (ttm) Return on Invested Capital (ROC), which is computed as Earnings Before Interest and Taxes (EBIT) divided by Tangible Assets (Total Assets minus non-interest bearing current liabilities, cash, and goodwill). A score is assigned according to a company's rank in this list.

2. The same set of businesses is ranked from highest to lowest Earnings Yield, which is computed as ttm EBIT divided by Enterprise Value (share price multiplied by shares outstanding plus total debt less cash). This ranked list is scored as described earlier.

3. The two scores a company receives are summed.

4. A portfolio of the 30 lowest scoring companies is bought and held for one year (losers would be sold a little under a year for tax write-off purposes while winners would be sold after qualifying for long term capital gains tax treatment).

5. Rinse and repeat.

Results

The Magic Formula upholds Buffett's primary rule of "never" losing money (over three year periods, its worst return was 11% versus the S&P 500's worst three year period clocking in at less than negative 40%; its robustness did not dissipate during the onset of the 2000 bear market). It has also produced something in excess of Graham's goal of a satisfactory return (an annual compounded rate double to triple the market, depending on the size of the universe of companies to which it was applied).

...................MF.........................S&P500 1988...........27.1%........................16.6% 1989...........44.6............................31.7 1990...........1.7..............................(3.1) 1991...........70.6............................30.5 1992...........32.4............................7.6 1993...........17.2............................10.1 1994...........22.0............................1.3 1995...........34.0............................37.6 1996...........17.3............................23.0 1997...........40.4............................33.4 1998...........25.5............................28.6 1999...........53.0............................21.0 2000...........7.9.............................(9.1) 2001...........69.6..........................(11.9)

2002...........(4.0).........................(22.1) 2003...........79.9............................28.7 2004...........19.3............................10.9 AVG...........30.8%....................12.40%

The strength of returns for each decile of the top 100 was higher than the next (on average, the top ten were a little better than 11 through 20, which in turn were better than 21-30, and so on). Ruminations & Refinements The Magic Formula certainly lives up to the Three Grahamian Virtues of logic, practicality, and strong historical performance. It's absolute performance may suffer in the future, but its relative outperformance should persist due to the fact that investors are constantly churning prices higher and lower, meaning that some companies will always be better bargains than others.

The extraordinary past performance was accomplished while blindly buying the top 30. This tells me that there is a lot of room for improvement. While I have not seen the data, I doubt very seriously that the returns were evenly distributed across the 30 companies -- I'm sure some of them were stinkers, some did alright, and some did spectacularly. It may not be easy to differentiate between the middle and upper crust, but it should be possible to remove most of the drags (the poseurs that faked their way onto the list). It is hard to know how much this would enhance returns, but it seems worth the effort. What follows are several ideas I've had to augment expected returns.

1. Separating the Wheat from the Chaff

• In perusing the historical records of top Magic Formula companies, I noticed that three distinct groups could be categorized as the Wheat, Chaff, and Germinators.
  
• Wheat companies had extraordinary returns on capital for five or more years.
  
• Chaff companies had mediocre long-term records, experiencing uncharacteristically good results in the past year.
  
• Germinators were in between, exhibiting strong returns on capital stretching back between 2 and 4 years.
  
• My expectation is that the Wheat will, on average, experience strong revision of investor expecations and reversion to mean valuation and operating performance (if a short term factor temporarily was clouding the picture). Chaff companies, on average, are probably going to revert to their mediocrity and leave investors deflated. Germinators are an interesting grouping of generally cheap companies that may be in the early stages of building strong economic castles with wide moats...an embedded option if investor expectations have lagged positively shifting fundamentals.
  
• It would be interesting to break the Magic Formula companies into these three groupings as separate portfolios and track them to test my hypothesis.
  

2. Following Greenblatt's Advice

• In Greenblatt's book, he advised "normalizing" earnings. This could be done by taking average profit margins from several prior years and applying them to current revenues.
  
• Greenblatt has also said that it would be more meaningful to subtract estimated maintenance capital expenditures (CapEx) from reported earnings, as this more closely reflects what cash flows are available to owners and differentiates those companies that require less reinvestment to maintain and expand their operations at strong rates of return on capital.
  It should therefore be more enlightening (and profitable) to re-rank the top Magic Formula companies after altering EBIT to normalized free cash flow (FCF) and recalcuating the ROC and Earnings Yield based on these figures.
  
• Buying the top one, two, or three deciles of the top 100 after re-ranking them seems to be preferable to blindly buying the top 30.
  
• More simply, one might simply rank the top Magic Formula companies from lowest to highest Enterprise Value-to-Normalized FCF and buy a basket of the lowest.
  

3. "MF-score"

• I have been inspired by Joseph Piotroski and his creation of an FSCORE (used to ascertain a minimal level of economic performance with an improving trend) applied to low Price-to-Book Value companies. FSCOREs enable an investor to remove a lot of dead weight from the low P/B universe and dramatically increase returns (nearly 13% in excess of the market over 30 years studied).
  
• I have also been interested in Edward Altman's Z-scores. A Z-score is a refined set of accounting ratios with profound predictive power for determining bankruptcy risk. One of the striking concepts here is that Altman analyzed many ratios and found some that were especially predictive. Piotroski has not performed a similar refinement.
  
• It should be possible to identify a number of value and quality-oriented metrics (in accordance with Graham and Buffett tenets) which might help an investor differentiate between top expected performing companies and the dogs amongst them. Taking a page from Altman, one might statistically verify which metrics contain the most predictive power for finding the best Magic Formula stocks.
  

4. Magic Formula 20/20

• In my own company selection, I have opted for an amalgamation of the above, which I have named the MF 20/20.
  
• The process is close to the following: 1.) Determine those Magic Formula companies that are within 20% of their 52 week low and in the bottom 20% of their three year historical valuation range using the Price-to-Sales ratio (hence the 20/20). 2.) Determine the "quality" of earnings for the companies making it this far by applying Richard Sloan's Accrual Ratio, and taking only those whose ratio is negative (good) 3.) As a sanity check, make a rough wag of expected return from current Enterprise Value based on normalizing FCF along the lines of the method described earlier and using conservative growth rates below long-term historical rates of growth.
  
• Repeat this process every three months, easing the portfolio into a state of being fully vested in the companies of excellent quality with most depressed investor expectations (suggesting room for powerful and profitable upward re-valuation).
  

Thank you for sticking with me this far. During the course of the next week, I hope to write out exact instructions for each of these methods and produce current lists of companies rising to the top.

Relevant Resources:

http://magicformulainvesting.com/
http://dardashti.blogspot.com/2005/12/joel-greenblatt-little-essay-that_27.html
http://www0.gsb.columbia.edu/accounting/seminars/PIOTROSKI.pdf
http://www.defaultrisk.com/pp_score_14.htm
http://www.leggmason.com/funds/knowledge/mauboussin/MMonValuation_LMWWversion.pdf