Free 7-day Trial
All Articles and Columns »

An Evening with Mr. Greenblatt

January 11, 2006

by Shai Dardashti  

Mr. Greenblatt, author of The Little Book that Beats the Market, treated members of the NYSSA to a special evening program on " Special Situations Investing".

Providing a wonderful glimpse into the evolution of his investment mindset, Mr. Greenblatt opened with an account of his college years - working with Rich Pzena to deconstruct the methods of Benjamin Graham. In the late 1970s, Mr. Greenblatt recalls, he "read a Forbes article about Graham" that discussed Net-Net Strategy, that is - stocks trading below liquidation value.

From my personal explorations into Grahamian techniques, I believe this is the actual article which inspired Mr. Greenblatt's value investing pursuits:

The Return of Benjamin Graham, Forbes October 15, 1979

"Think of a time when stocks of 191 important American corporations are selling for less than net working capital per share. Are we talking about 1932? No, 1979."

Continuing his story of self discovery, Greenblatt recalls that at Graduate school at Wharton, he "wrote a paper published in the Journal of portfolio management."

In his typically humble and modest nature, Mr. Greenblatt chose to leave out the findings of his early explorations - published in a 1981 study. Below are some notes I compiled from research on the actual report:


Greenblatt, Pzena, and Newberg

"How the Small Investor Can Beat the Market

Journal of Portfolio Management

The Greenblatt/Pzena/Newberg study was intrigued by Graham's writings in Security Analysis in which "he outlines in little more than a page the opportunities to be found in stocks selling below their liquidation value. In studies between 1923 and 1957, Graham reported superior results when market levels enabled him to buy a diversified list of these bargain stocks."

The study examined the performance of stocks meeting Graham's rough liquidation value (net-net) estimate:

Accounting Definition of Rough Liquidation Estimate:

"Current Assets" (cash, accounts receivable, inventory, etc.)

Less: "Current Liabilities" (short term debt, accounts payable, etc.)

Less: "Long Term Liabilities" (long term debt, capitalized leases, etc.)

Less: "Preferred Stock" (claim on corporate assets before common stock)

Divided by: Total Shares Outstanding

EQUALS "Liquidating Value Per Share"

The study "did not consider the stocks that had shown a loss over the preceding 12 months."

Research covered "15 segments of 4 months each over a six-year period in which the over-the-counter (NASDAQ) averages halved and then doubled... The period under study an from April 1972 to April 1978."

The process looked exclusively at three factors:

  • Price in relation to liquidation value
  • Price/Earnings ratio
  • Dividend Yield
  • Stocks were sold after a 100% gain or after 2 years, whichever came first (as per Graham's writings)

The Graham net-net buying process was applied within four distinct portfolio dynamics, each described below with its respective results:

  • Portfolio 1:
    • Price/liquidation value </= 1.0;
    • Price/earnings: floating with bond yields;
      • [require a P/E corresponding to twice the prevailing triple A yield in each period]
    • Dividends: no dividend requirements


"During the 15 4-month periods our constraints dictated a position in the market, we averaged an annual compounded rate of return of 20.0% before dividends, commissions and taxes. The OTC index appreciated at an annual compounded rate of 1.5% during the same period."

"We would expect higher returns to accrue to riskier investments to compensate us for taking on the additional risk. Therefore, we also studied the volatility of the returns of our selected stocks with that of the NASDAQ market average. (During this period, the NASDAQ averages significantly outperformed the S&P indexes of larger companies) A regression of our Portfolio 1 return and the OTC market return over the 15 periods resulted in the following:

Portfolio 1 return = +6.14 +.836 (NASDAQ return), (4 month period)

Portfolio 1 standard deviation = 14.15;

OTC portfolio standard deviation = 12.75."

Portfolio 2

    • Price/liquidation value </= 0.85;
    • Price/earnings: floating with bond yields;
    • Dividends: no dividend requirements


"After we limited the purchases in Portfolio 1 to stocks selling below 85% of liquidation value, the returns increased to a 27.1% annualized rate before dividends, commissions, and taxes (compared with the market's 1.3% annual performance). After taxes and commissions, this return approximated 16.5% annually.

The regression worked out to:

Portfolio 2 return = +8.54 +.752 (NASDAQ return), (4 month period)

Portfolio 2 standard deviation = 14.58;

OTC portfolio standard deviation = 12.75.

Adjusted beta = 1.14"

Portfolio 3

    • Price/liquidation value </= 1.0;
    • Price/earnings: </= 5.0
    • Dividends: no dividend requirements


"When we used a low constant P/E ratio coupled with a discount to liquidation value, our returns were significantly improved to a 32.3% annualized rate before dividends, commissions, and taxes. After taxes and commissions, our return falls to 20.1% per year, compared to the OTC market's return of 2.0% during the 14 periods when we had a position in the market.

No stocks were purchased until August 1973 using the parameter of a PE below 5. The portfolio also entered the market closer to the tough and with more conservatively valued stocks. We outperformed the OTC index by 5% or more in 9 four-month periods, while we underperformed the market by 5% in only one period. The regression analysis was:

The regression analysis was:

Portfolio 3 return = +9.9 + .753(NASDAQ return);

Portfolio 3 standard deviation = 14.35;

OTC portfolio standard deviation = 13.16;

Adjusted beta = 1.09.

Portfolio 4

    • Price/liquidation value </= 0.85;
    • Price/earnings: </= 5.0
    • Dividends: no dividend restrictions


"Our most successful screen. I t resulted in an annualized rate of over 42.2% before dividends, commissions and taxes. The result before dividend returns approximated 29.2% for the 14 periods studied, compared to the 2.0% annual returns of the OTC markets. The regression analysis was:

Portfolio 4 return = +12.83 + .671(NASDAQ return); (4 month period)

Portfolio 4 standard deviation = 14.94

OTC portfolio standard deviation = 13.17

Adjusted beta = 1.13.

Not bad, indeed.

Greenblatt discusses the study

To produce the paper, Greenblatt explained to the NYSSA, he explored the S&P stock guide - by hand - and, together with Rich Pzena, made a unique database of stock information. At the time of the study, the aspiring super investors had to calculate their findings on University of Pennsylvania's DEK 10 Digital Equipment Computer, a far stretch from the modern Compustat database and computing power of the internet age that Greenblatt explains was used to research his Magic Formula,

The downside to the "42.2% annualized rate" of the Graham Formula, Greenblatt shared with his NYSSA audience, was that investors were getting a bargain, but the bargains disappeared in 1980s.

Greenblatt clearly is aware that, as Graham teaches, "cheap works" - and Mr. Greenblatt cited a variety of studies documenting the performance of low price-to-book, low-price-to-earnings, etc. (The sources of which, I presume, are the Tweedy Browne "What has Worked" report and the various experiments documents in Haugen's "The New Finance")

In an attempt to adapt the quantitative construct to reflect his appreciation for Warren Buffett's investment techniques, Greenblatt commenting on Buffett's presumed thought-Process: "Buying cheap works, I know that... But what if I buy good companies that are cheap? And see how it would do..."

Working Towards The Magic Formula

So, Mr. Greenblatt began to study the question of "what is a good company?" The simple answer: A business with a high return on capital.

Business A)

$400,000 cost of store to build.

$200,000 earnings a year.

50% ROIC

Business B)

$400K to build.

10,000 earnings a year.

2.5% ROC

Clearly, in this simplified example, Business A is the superior business.

As per his wonderfully concise summary:

1) Greenblatt ranked the businesses by ROIC.

2) And then ranked the shares by cheapness.

"More earns relative to price... I call that 'cheap.'" - in reference to Earnings Yield.

Developing the Magic Formula

Working with the two variables, high earnings yield and high return on invested capital, Mr. Greenblatt decided simply to combine the two rankings to create a list of businesses that have the best of both components.

Logically, a business that ranks #100 for ROIC and ranks #50 ranking for Earnings Yield would rank #150 in The Magic Formula hierarchy.

To further research the effectiveness of this mechanical process, Mr. Greenblatt took his ranked list of Magic Formula results and divided the hierarchy into deciles, and simply performance of each decile. The results: The top ranked decile outperformed the 2 nd best, in turn was better than the 3 rd, etc. So, the performance of each decile was absolutely in line with the rankings from the Magic Formula.

Mr. Greenblatt, a veteran of Wall Street's inquisitive approach towards ground breaking claims, outlined complicated possible concerns with the process, and simple counter-arguments:

"But, the Magic Formula is subject to error due to... "

Look ahead bias

  • The study used the Compustat point in time database. So the data used was reflective of information available precisely at the time period under examination.

Survivorship bias

  • Again, the study used the point in time database

Small companies couldn't be purchased, transaction costs would kill you

  • The same perfectly aligned decile rankings appeared when only exploring at top 1000 companies by market capitalization.

Fama frech argument: the formula is picking riskier stocks.

  • Next question.

This is data mining

  • This was the 1 st test attempted, and the weight of "good company" to "cheap stock" was a simple 50/50%

More Advanced Considerations: Piotroski and Haugen

Mr. Greenblatt compared his Magic Formula results to the stock selection techniques of Piotroski. Generally, Piotroski's work performs very well, but only its utility is effectively limited just to companies with a market capitalization up to $700 million. So, for large cap stocks, Piotroski's work isn't all that effective.

Robert Haugen introduced a 71 factor model for superior stock selection. With monthly periodic turnover over the 10 year period, Haugen's technique demonstrated a 30% superior performance of his top ranked decile over the lowest ranked class of stocks.

Greenblatt found a 32% spread when researching the 2 factor Magic Formula. (ROIC and Earnings Yield)

To assess the long term viability of their respective approaches, and to reduce the transaction costs, Greenblatt compared his results with those of Haugen's.

In Search of The Magic Formula

Mr. Greenblatt created an experiment in which he held selections derived from Haugen's 71 factor model for a year, with annual turn over, and developed sample portfolios every month for the 10 years. (So, he created portfolios tracking 120 rolling one year periods)

  • Haugen's top decile beat the bottom decile by 5.63%
  • Greenblatt's 2 factor model recorded an 18.5% spread of out-performance.

Greenblatt repeated this experiment, looking at rolling 3 year periods (there were 169 such periods covered in the duration of his Magic Formula study)

  • Haugen's Method: The worst 3-year period return was "-35% or - 45% "
  • Greenblatt's Method: The worst 3-year period return was actually a positive return "around 10%"


Shai Dardashti is the executive editor of Shai Dardashti on Grahamian Value, a media outlet reporting news, commentary, and analysis to the value investing community. Shai is currently a senior undergraduate student at The University of Maryland, with plans to graduate in May 2006 and pursue a career in money management. He can be reached at

Rating: 4.1/5 (17 votes)


Please leave your comment:

Get WordPress Plugins for easy affiliate links on Stock Tickers and Guru Names | Earn affiliate commissions by embedding GuruFocus Charts
GuruFocus Affiliate Program: Earn up to $400 per referral. ( Learn More)
Free 7-day Trial