1. How to use GuruFocus - Tutorials
  2. What Is in the GuruFocus Premium Membership?
  3. A DIY Guide on How to Invest Using Guru Strategies

What worked in the market from 1998-2008? Part II. Under-Valued Predictable Companies and Buffett-Munger Screener

October 14, 2008

This is the Part II of the back testing study we conducted for the stocks that have been continuously traded from Jan. 1998 to Aug. 2008. The roles of market valuation on investment returns over the past 10 years have been analyzed. A Buffett-Munger screener has been developed to find good companies at fair or undervalued prices.

In Part I of the study we reported the role of business predictability in investment returns. We found that the more predictable the business is, the higher return of the stocks has to shareholders, even if valuation is not considered. The key factor here is that the probability of investment loss is much smaller for predictable companies if the stocks are held for long period of time.

We also introduced the concept of Predictability Rank. Regardless of the stock valuation, 5-star ranked stocks had an annualized median gain of 12.1%, while non-predictable companies had an annualized median gain of 1.1%. As a reference, all the 2403 stocks under study had an annualized median gain 3.1%.

As value investors, we always discuss the relationship between price and value. Valuation certainly plays a very important role in the investment returns. The Part II of this study focuses on the roles of market valuation on the investment returns over the period from Jan. 1998 to Aug. 2008.

Some Questions to Answer

Before we discuss the role of valuation, we would like to address some questions users had for Part I of the work: the correlations between Predictability Rank and Investment Returns.

These are the questions from users and our answers:

Bakerkn:Your back testing is assuming time-travel is possible. You are using the predictability over the last 10 years to decide what stocks to have bought 10 years ago”.

Answer: What you said is certainly valid, just as past performance is not a good indicator of future performance.

But on the other hand, past underperformance is a good indicator of future underperformance. If a business did poorly at a bad economic times before, chances are it will do poorly again in future bad economic times.

Business predictability is mainly determined by the nature of the business, less dependent on who runs it.

From this we just proved what we already know: buying businesses with “proven and predictable” track record will reward investors.

In order to further validate the idea, we used the data before Jan. 2005 to test the performances of stocks in 2005, and data before Jan. 2006 to test those for 2006, and so on. In this way, time does not travel back. The results are reported below in this article.

Acedomaine1:If we had the actual data, the antithetical reply of the cautious skeptic would consist in pointing out the wide dispersion in the results and the relatively high value of some of the samples; as a consequence I would refuse to make any inference from the data. Is the conclusion based on "predictability" sound? Is the reasoning cogent? Could not such a result occur by chance even when no real differences exist? Are the data too widely dispersed to make any sensible inference?”

Answer: The distribution of the market returns for the higher-ranked business is actually the least dispersed. The distribution for the non-predictable companies is much more dispersed, as shown in the chart and table below. If you only work with the highest ranked companies, you are working with a group of good companies. You will miss some multi-baggers in the non-predictable companies, but you also miss the large percentage of loss there, too. The chance of having permanent loss is small.

5-Star 4 and 4.5-Star 3 and 3.5 Star 2 and 2.5 Star Unpredictables All
# of Samples 80 160 160 170 1833 2403
Median 12.1% 10.2% 8.7% 6.5% 1.1% 3.1%
Standard Deviation 8.1% 8.3% 12.3% 11.8% 13.9% 13.7%


P/E Expansion vs. P/E ratio

P/E ratio is one of the most widely used and important measures of stock valuation. If we value stocks based on their P/E ratios, investment returns come from two factors. The first is the growth of earnings from the growth of business, the second is P/E expansion.

The chart below shows P/E expansion vs. P/E ratios of the stocks on Jan. 1, 1998. P/E expansion is defined as the P/E ratio on Aug. 31, 2008 divided by that the P/E ratio on Jan. 1, 1998, of companies. A P/E expansion ratio of 1 means no P/E expansion. Based on the charts below, we can clearly see that there is a statistical correlation between P/E expansion and P/E ratio itself. If a stock had a P/E ratio of less than 10 in 1998, it has much higher chance of getting P/E expansion over the holding period. By contract, if a stock had a higher P/E, e.g. higher than 25, the chance of P/E expansion is much smaller. If the P/E is higher than 30, P/E expansion is almost impossible.

This chart also tells us that what we all already know: if everything else is the same, a stock with lower P/E ratio has better potential gains. If a stock is bought at a high P/E ratio, the potential return is smaller.

The Roles of Market Valuation

In order to analyze the roles of market valuations on investment returns, we divided the 2403 stocks under study into three groups: under-valued, fair-valued, and over-valued. We use a very simple indictor that we call PEPG to value the stocks. PEPG is P/E ratio over Past Growth, defined as P/E ratio divided by average EBITDA growth rate over the past 5 years.

The reason we used PEPG instead of P/E to measure the valuation of business is because P/E ratio fails to take into account the “growth” part of the business. Widely used valuation ratios such as P/E, P/S, P/B, earning yield or discount cash flow model all have their pros and cons. This will be discussed in detail in another article

For the current discussion, if PEPG of a stock is between 0 and 1, we consider it under-valued; if it is between 1 and 2, we consider it fair-valued; if PEPG is higher than 2, it is considered over-valued. This is a very rough measure of stock valuation, however, as we can see later, it makes a big difference on the investment returns.

With the simple valuation ratio of PEPG as the indication of stock valuations, the returns of different valuation groups are complete different, as show in the following table:

Top 100 Most Predictable Companies

Second 100 Most Predictable Companies







Total Number of stocks







# of Losers








Average gain








Median gain







We can see that for the top 100 most predictable companies, there were 25 under-valued stocks in Jan. 1998. This group of stocks gained about 20% annually if held for 10 years and 8 months. The fair-valued group has an annualized median gain of 12.1%, even the over-valued group has an annualized median gain of 9.5%. For the second 100 most predictable companies, the gain is lower, as expected. The undervalued group has an annualized median gain of 13.8%. The over-valued group has 7.6%. All these numbers are much higher than the 3.1% of the annualized median gain of all 2403 stocks. The S&P500 gained 2.7% annually over the same period.

Furthermore, we like to point out that the probability of loss on any of the stocks is reduced if one buys undervalued stocks with highly predictable underlying business. For the undervalued stocks in both groups, the probability of losing on any the stocks is reduced to 4% or less if one holds if for 10 years. This is in strong contrast of the 37% of stocks that are in loss among all 2403 stocks studied, and 45% among the stocks whose business are unpredictable according to our definition.

What does this tell us? It again tells us nothing new. It just reaffirms that the safest way to invest is to buy undervalued stocks with highly predictable underlying business. By doing that you will avoid most of the losses (Rule #1: Never Lost Money), and in the meantime, you achieve the highest returns.

Back Testing of 2005-2008

So far we have been using the financial data of Jan. 1998 to Aug. 2008 to back test the stock performances over the same period. However, financial information is not known until the period has passed. In order to further test the idea, as the second step of the back test, we only used the financial information that had become available before the beginning of the period that the stock performances are tested.

Therefore, for the performances of stocks in 2005, we only use the financial data that is available before Dec. 2004. We use the same algorithm to rank the predictability of business, and observe the stock performances of a portfolio equally invested in the top 25 most undervalued stocks among the top 100 most predictable companies. The portfolio is rebalanced after 12 months. Then for 2006 stock performances, we use the financial information before Dec. 2005 to determine the predictability of business, and invest in the top 25 most undervalued ones among the top 100 most predictable companies, and so on. We back tested for the period from Jan. 2005 to Aug. 2008. The results are shown below.

  2005 2006 2007 2008 to Aug. Cumulative
S&P500 3.8% 12.3% 4.4% -10.1% 9.4%
Top 25 Ranked 36.8% 13.6% 18.2% 2.3% 87.9%


As we can see, for the 3-year and 8-month period, the portfolio of the top 25 undervalued predictable companies beat the S&P500 by great margins except for 2006, when it almost matched S&P500. Over the 3-year and 8-month period, the portfolio had a cumulative gain of 87.9%, while S&P500 just gained only 9.4%. (All the numbers do not include dividends).

Competitiveness of Business

So far we have not discussed the “moat” Warren Buffett talked about in the business he likes. How do you measure moat? If a business has no moat, over time others will get into the same business, the business cannot maintain its competitive advantages and has to lower prices of its products and services, and the business profit margin will shrink. Therefore if a company can maintain or expand its profit margin while growing its business without incurring excess debt, we will consider the business has moat.

Business moat can be from the nature of the business and business operations. If a business can maintain or even expand its profit margin while growing, it must have something that may not necessarily be understood by others.


An interesting comparison here is two companies that are in almost identical business: Circuit City (CC) and Best Buy (BBY). If we look at the gross margins of the two companies, as shown in the charts below, Best Buy has a gross margin at a consistent level around 5.4%, while Circuit City’s gross margin has been in steady decline, and entered in negative territory last year. Apparently, without studying the business operations of the two companies, we can safely say that Best Buy has a moat that Circuit City does not.

Not surprisingly, during the past 10 years, Best Buy rewarded its investors with a 600% gain, while Circuit City is one of the companies with which you are still in loss even if you bought it 10 years ago hold it until today. By the way, Best Buy has a 4.5-star Business Predictability Rank, while Circuit City is 1-star (unpredictable) by our definition.

Buffett-Munger Investing vs. Cigar-Butt Investing

Warren Buffett has famously said: “I would rather invest in good companies at fair prices than invest fair companies at good prices.” It was said that Warren Buffett’s investing philosophy was influenced by Charlie Munger. Before he met Charlie Munger, Warren Buffett liked more to invest in fair companies at good prices, i.e. cigar-butt investing.

If one has to decide between Circuit City and Best Buy, a cigar-butt type of investor may as well consider Circuit City. Why? Circuit City is sold below its liquidity value on its balance sheet. Even if Circuit City is liquidated, at $0.41 a share the cigar-butt investor seems not at the position of losing money. However, for a long term Buffett-Munger investor, Best Buy is certainly a better choice because it is much better run, it can consistently grow its business without compromising its profit margin, and it has little debt. The business has grown its earnings at about 18% a year over the past 10 years, and it is sold at a P/E ratio of just 9. Of course, with Best Buy you won’t get the bargain prices that you get with Circuit City.

Charlie Munger once said: “If you buy something because it's undervalued, then you have to think about selling it when it approaches your calculation of its intrinsic value. That's hard. But if you buy a few great companies, then you can sit on your $%@. That's a good thing. We want to buy stocks to hold forever.”

Therefore, Buffett-Munger type of investors will first look at the quality of the business. They only invest in good companies that can grow its business consistently and maintain its price power (profit margin) without incurring excess debt.

One mistake that value investors often make is falling into value traps. A value trap is a company that seems to be sold at discount, but its business is in the process deterioration. If a value investor pays too much attention to the business valuation (quantity), and overlooks the quality of the business, he/she may well fall into value traps. As our study shows, investing in a low quality has a much higher chance of the permanent loss of capital. Investing in a high quality business at over-valued prices will reduce investment returns, but over time the growth of the business will compensate the high prices investors paid, and the chance for permanent loss of capital is small.

By investing in highly predictable and competitive businesses, a Buffett-Munger investor can avoid value traps.

Buffett-Munger Screener

From the above discussions we have developed a screener that we call Buffett-Munger Screener. The stocks that can make into this screener are:

  1. Companies that have high Predictability Rank, that is, companies that can consistently grow its revenue and earnings.
  2. Companies that have competitive advantages. It can maintain or even expand its profit margin while growing its business
  3. Companies that incur little debt while growing business
  4. Companies that are fair valued or under-valued. We use PEPG as indicator. PEPG is the P/E ratio divided by the average growth rate of EBITDA over the past 5 years.

The screener is here; it is for Premium Members only. If you are not a Premium Member, we invite you for a 7-day Free Trial.

How to make 20% a year?

We did another back test of the top 25 stocks generated from Buffett-Munger Screener for the period from Jan. 2005 to Aug. 2008. Similar to the test of undervalued stocks, the 25 stocks for 2005 were generated from the financial data before Dec. 2004, and those for 2006 were generated from data before Dec. 2005, and so on. The results are shown below.

  2005 2006 2007 2008 to Aug. Cumulative
S&P500 3.8% 12.3% 4.4% -10.1% 9.4%
Buffett-Munger Screener 43.4% 11.2% 17.2% 7.2% 100.3%


Similar to the test with undervalued stocks, the Buffett-Munger Screener greatly outperformed the market, although it did not beat the S&P500 every year. The cumulative gain for the 3 years and 8 months period is 100.8%, making it 21% annualized, without counting dividends.

Is This a Magic Formula?

You may call this a magic formula, as Joel Greenblatt calls the study in his book “The little book that beat the market”. The difference is that the Buffett-Munger Screener has much stronger requirements on the quality of businesses. The quality of business is measured by the long term growth of the business. The business must be able to stand the test of good times and bad times. The business must have “predictable and proven” earnings even in distressed economic times; it must have competitive advantages that cannot be easily taken by others, and may not incur debt in the course of growth.

A Buffett-Munger Screener will only select high quality companies. After that, the stock valuation is checked. Here the growth rate of the business is taken into account using PEPG ratio, instead of the earning yield used by Joel Greenblatt, as earning yield does not take into account the growth of the business.

Joel Greenblatt said that with his Magic Formula investors may have losses sometimes for two to three consecutive years, as witnessed by the large loss of magic formula portfolios in 2007 and 2008, right after the magic formula was published. The back test of portfolios generated by Buffett-Munger Screener beat the S&P500 by great margins over the same period.


Just as Warren Buffett has said multiple times, buying companies with “predictable and proven” earnings can be very profitable in stock market investing. Investors are rewarded with consistent business growth. Permanent loss of capital can also be largely avoided.

Buying undervalued predictable companies is even better, as shown in Part II of the study. This strategy can greatly outperform the market averages. Back testing shows that a gain of 20% annualized can be achieved while the market is gaining only 3% a year.

From this study we have developed a concept called Business Predictability Rank (Part I). We also develop a screening tool called “Buffett-Munger Screener”, which can be used to screen companies with high quality business at undervalued or fair-valued prices.

Also see: What worked in the market from 1998-2008? Part I: Predictability Rank

Buffett-Munger Screener;

Take a Free Trial of GuruFocus Premium Membership.

Rating: 3.1/5 (153 votes)



Jonwu - 12 years ago    Report SPAM
interesting from the fundamental view point. Why pick 10-years frame though. Stock Predictor http://www.ashkon.com/sp.html can do technical backtesting.
Wiglebot - 12 years ago    Report SPAM
Nice article.

This is a correlation model that returns the lowest probability of losing money and the highest probability of making 15% returns from 2005 to present using trailing data from 1998. The 1998 to 2008 study was flawed from what Bakerkn said. Analysis of the variables before the period in question is a must.

Its a good model if things remain the same. But 1995 to 2008 have been a super bubble. To collect data from the ramp of a hockey stick chart and assume things are predictable from that data is a little scary.

I know the model works from business fundamentals, but I would like to test it from 1959 to present.

Personally I think it works, but are the 5 star companies a bargain today?
Rozer - 12 years ago    Report SPAM
A few questions regarding methodology. How often is the model rebalanced (portfolio turnover)? In related question, any rules as to taking profits or cutting losses?

The back-tested models are interesting, but its the execution of the model that will make a difference.
Rozer - 12 years ago    Report SPAM
Also, could you define how you quantified your predictability rank.

Thanks very much.
Gurufocus premium member - 12 years ago
Business predictability based on the consistency of both revenue and EBITDA. The combination of revenue of EBITDA is used. The more smooth the revenue and EBITDA grow, the higher predictability rank the business is.

Sometimes revenue can grow smoothly, but EBITDA is not. For an example: Zale Corp.

Max7777 premium member - 12 years ago
I respect the effort of Gurufocus on Predictability and I visit this site everyday as it is truly a great site full of relevant information, this predictability could be a great addition but I need a lot more data if we are to take this information seriously. I think you can find a 3 year time period that would correlate to market returns for almost anything, even tire size on my kid's bicycle has a longer correlation to the market than 2005 to 2008. Using only 2005 to 2008 for back-testing is not serious, we need at least 50 years or even 70 years to take this study seriously.
Gurufocus premium member - 12 years ago
We call the back testing from 1998-2008 "what worked", the effort was trying to find out "what worked". As we said, we did not find anything new, what we found is we already know, i.e. buying good companies at fair prices.

The number is really not three years, since the predictability for 2005 was based on the business performances of 1997-2004, and so on.

it is very hard to do a 50 year backtesting, we do not have the financial data earlier than 1997. Also the number of stocks that survived 50 years would be very small.
Madhusudanvr - 12 years ago    Report SPAM
If one had tried to identify "predictable businesses" i am not quite sure what they might end up with. AIG, I felt was in a predictable business 10 years ago(insurance!!). You might not find data for these companies when the screen is run from 2005 to 2015. And this will certainly be the key, as while finding stocks for investment, we might end up with an AIG that will soon go bankrupt.

Gurufocus premium member - 12 years ago
It is true AIG business was quite predictable before. But the Buffet-Munger screener also look at the profit margin of the business. The profit margin of AIG has been in declining as shown in this chart:

Waveheatin - 12 years ago    Report SPAM
I think the methodology is valid and the concept of finding company with great business fundamental at a bargain price is best investment startegy, and no body would argue againt it. The difficulty leis in identifying comapies with great business fundamental in the FUTURE (not the past). Data will never tell you if a business has great fundamental in the future. If you look at Buffett's great purchases, most of them were bought at a time it looked terrible, but somehow Buffett was able to see the potential in Future. When Buffett bought Geico, KO, PTR, PKX, few saw the point. And now he bought WFC, BNI, KFT, and invested in GE, GS, and again we could not comprehend why he did in such an agreesive fashion. The difference is we cannot see the 'Future' and he can.

He has already pointed the way for us and we could esily follow, but few did.
Weiwentg - 12 years ago    Report SPAM
I like the concept. However, I think free cash flow is a better guide. It's less easily distorted than earnings.

I was pretty surprised to see AIG at 4.5 stars. I was also surprised to see 3M and American Express at one star, Coke at 2 and Pepsi at 2.5. I believe all these are stable, predictable businesses. Warren owns loads of Coke and Amex.

You might also want to adjust for currency effects. Novartis and Sanofi-Aventis are one star. I especially like Novartis, and I know Warren owns Sanofi.
Stephanus - 11 years ago    Report SPAM
Good work. Would love to see the return comparison between the S&P 500 and the Buffet-Munger screener updated for the full 2008 calendar year.
Gsrddi - 11 years ago    Report SPAM
The equities markets seem to becoming unmoored from economic reality. How could the failure of either Chrysler or GM could be other than negative for the US dollar?

Conclusion: The people who write about these things must think in terms of hour to hour, may not know who Warren Buffett is, and certainly dont understand his investment philosophy.

Rckymtns - 10 years ago    Report SPAM
I don't think EBITDA should be used for predictability as Buffett and Munger speak lowly of ebitda.
Quixote1 - 9 years ago    Report SPAM
Why you removed this article from your site?

Gurufocus premium member - 9 years ago
It is there now. Caused by a recent change. Sorry about it.
FTtrader - 9 years ago    Report SPAM
Very good article!!, both part I and part II. I hope GF continues publishing such interesting and formative articles.
Biojohn - 9 years ago    Report SPAM

Really enjoyed your article. Would like to get some clarifications though.

In your discussion regarding the competitiveness of the business, you state that "if a company can maintain or expand its profit margin while growing its business without incurring excess debt, we will consider the business has moat". How would you define profit margin? In your Best Buy/Circuit City comparison you use gross margins to identify competitiveness. In the AIG example, you use net margin. In the moat component of your screen, do you use either or both metrics in your screen? Also, how do you define business growth for measuring competitiveness (Revenue growth, EBITDA growth, book value growth)? Finally, when you mention excess debt, how do you define that? Is it incremental debt relative to some previous point in time or is it excess debt relative to some industry average or are you targeting some Debt/EBITDA ratio, etc?

Finally, is there a reason why you chose PEPG (P/E vs. EBITDA Growth) instead of EV/EBITDA vs. EBITDA growth for your valuation metric? It seems like the latter would be internally consistent. Would the output be very similar using either metric?


Gurufocus premium member - 9 years ago
Thank you for your comment, Biojohn!

Usually gross margin works well for non-financial companies. For financials, the accounting is usually complex. We use net-margin there for AIG. But you should look at both.

We define growth as both revenue growth and earnings growth.

For debt, we usually look at the debt to revenue ratio. We hope that it is not increasing constantly.

I think you can look at EV/EBITDA vs. EBITDA growth. For predictable companies, the results should be similar.

Drclark - 9 years ago    Report SPAM

What software does gurufocus use to search through these stocks?

Stephen Neumeier
Stephen Neumeier premium member - 8 years ago

Has internet retailing changed Best Buy's predictability?
Peticolas - 8 years ago    Report SPAM
You said earlier; "_ it is very hard to do a 50 year backtesting, we do not have the financial data earlier than 1997. Also the number of stocks that survived 50 years would be very small."

The CRISP & Compustat data go back farther than that. It would offer many many opportunities to test the three year, five year, etc performance of the model.
Kamyiu02 - 7 years ago    Report SPAM
What are your criteria for picking out the "
Pkv - 6 years ago    Report SPAM

good article cheers gurufocus

Tprzechodzki - 5 years ago    Report SPAM

Please comment how Karen the famous SUPERTRADER can generate so much money per year from selling naked calls and puts on the SPY, IWM or Futures

Tprzechodzki - 5 years ago    Report SPAM

Karen is making 50% per year

Tprzechodzki - 5 years ago    Report SPAM

I would love to see Karen's moves in real time if that would be possible and trade along her side any time

Kenorunagamoto - 4 years ago    Report SPAM

A trust-worthy owners cum magement of businesses in Thailand should count more than predictability and profitability calculation. I am sick and tired of children-cry-wolf-coming type of management in Thailand.. All dressed up to look good.. And a good Apple may be rotten in the core(merchants of Venice). ?????????????????????????????????

Rrurban - 3 years ago    Report SPAM

Why are we still seeing this article. The data is completely invalid. It uses 1998-2008 financial data to determine 1998 predictability scores. Come on! Either redo the study using 1988-1998 financial data or delete this article altogether.

SciDeep premium member - 3 years ago

The distribution data (charts) and data in these old 'Predictability' reports, such as Gain Distribution and P/E expansion, should be available (and posted) every quarter (TTM data) as the 'Predictability' scores for stocks appear to be updated on that frequency. The changes in statistics between 1998-2008 (value oriented market) and 2008-2018 (growth oriented market) would be insightful. I suspect the next market correction will bring on a backlash to value orientation.

Gmseegull - 1 year ago    Report SPAM

How about 1998 to 2019?

Please leave your comment:

Performances of the stocks mentioned by

User Generated Screeners

pascal.van.garsseHigh FCF-M2
kosalmmuseBest one1
DBrizanall 2019Feb26
kosalmmuseBest one
DBrizanall 2019Feb25
MsDale*52-Week Low
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)