The Three Conditions That Must Be Satisfied To Believe In Any Investment Process

{Euclidean Q4 2009 Letter }


We are pleased to be writing this second annual letter to our Limited Partners. In this letter, in addition to providing a review of 2009, we will discuss:

  • the limited information that can be derived from short track records;
  • and the importance of focusing on investment process when evaluating Euclidean. 

Performance

During the 12-month period ending December 31st 2009, Euclidean returned +63.2% to its investors.  During this same period the S&P 500 gained +26.5%.

During this period, all of the partnership’s realized gains have been long-term in nature.  For those of you who have taxable accounts, this is a very good thing given the preferential tax-treatment provided for long-term gains.  For those of you who are in non-taxable accounts, please take comfort in knowing that our gross returns were unlikely to have been aided by increasing the churn of our portfolio and, instead, would likely have been impaired by additional transaction costs.

Since inception, our partnership has returned +43.9% in the context of a -10.5% decline in the S&P 500.   While we were not feeling especially lucky in November 2008, in retrospect, we are fortunate that we launched our fund in mid-2008 and had money to invest during the second half of that year.

Here is a top-level view of how the concentration of our assets has changed over the year. 

Please remember that, unlike many investors, Euclidean does not invest top-down by setting target exposures to different industry sectors. Rather, we evaluate companies from the bottom up and attempt to concentrate our funds in the best company-to-price combinations that the market offers to us at a particular time. 

Our bottoms up approach had us largely concentrated in consumer discretionary firms at the start of the year and noticeably under-exposed to financial institutions (due to their high-leverage) and technology companies (as our models did not perceive that they provided the most compelling values).

Because investor pessimism sometimes creates an environment where good companies may be offered to us for less than their true worth, you will often find us investing in out-of-favor companies.  During the past year – and this will not always be the case – many of the companies we invested in were being punished less for company specific issues than for being associated with ‘themes of investor concern’ such as:

  1. The housing market is in free-fall and will take many years to recover.
  2. The consumer is dead.
  3. The new healthcare regulations are going to be very bad for providers of managed care.
  4. The global economy slowdown will create a sustained decline in demand for materials.

Opportunities created by these themes continue to play a substantial role in Euclidean’s portfolio. 

How should our 2009 performance influence how you think about Euclidean?

The past 18 months have been particularly good for Euclidean.  We benefit from investor pessimism as it creates the opportunity to buy good companies at historically attractive prices.  Pessimism has, until recently, been in great abundance.

Had we launched at a different time, our first 18 months may not have been so great.   For example, when the markets have an exuberant tone (in contrast to late 2008 through early 2009) we believe we are less likely to outperform.  Also, while we believe our approach operates with the probabilities on our side during ‘normal’ environments, chance always plays a big role in the fortunes of our portfolio companies. 

Given this, the right investor for Euclidean does not give much weight to one or two years of performance.  There are high-profile examples of investment firms with very strong multi-year track records who quickly imploded.  And, there are legendary investors who have had difficult multi-year periods where commentators wrongly wondered if they had lost their touch.  The reliable clues to Euclidean’s long-term performance come not from our short-term results but from the nature of our investment approach.  

Our Process & Where Our Confidence Comes From 

This section comes in two parts.  The second half provides perspectives we have not previously written about regarding how we validated our approach.  Please excuse the first half for repeating what you have seen in prior communications.  

We are comfortable being repetitive in part because doing so gives necessary context to the question – why do we believe in our investment process? Also, repeating the key elements of our approach will hopefully keep you connected to and comfortable with how we operate.  

A High-Level Review Of Our Approach

Our mission is to bring a quantitatively rigorous and systematic approach to long-term investing.

We operate with core beliefs that:

  1. Companies create wealth by building products, taking care of customers, and optimizing operations.  A company’s wealth creating capabilities gives it an inherent worth. 
  2. Markets do not create wealth.  Markets transfer wealth between market participants in zero-sum fashion and provide a means of exchanging cash for the wealth-creating capabilities of individual companies.
  3. Markets sometimes materially misprice the shares of individual companies. 

We seek to acquire companies’ wealth-creating capabilities at prices below their inherent worth.  To do this in a systematic, data-driven manner, we developed a suite of learning technologies to digest the quantifiable experience one could have gained by investing in domestic public companies across the past four decades. Through this machine learning process, we sought to uncover history’s lessons on how to evaluate individual companies as potential investments and how to apply these lessons in the context of an evolving portfolio.

Each day, our platform allows us to evaluate several thousand companies trading on the NYSE and NASDAQ exchanges and to apply those insights via a systematic portfolio management process.  We invest in the most attractive opportunities with the expectation of holding each position for greater than one year.  Everything about what we buy, how much we buy, when we buy more, and when we sell is systematic and avoids the emotional influences that sometimes cause investors to make bad investment decisions.

Our approach tends to favor companies with long and consistent operating histories, capabilities to weather difficult times, track records generating good returns on capital, and valuations low relative to their historical ability to generate cash.  Euclidean’s insights come in the mathematical interpretation of these qualities, the trade-offs that can be made between them, and the margin of safety in valuation required for an investment.

We define risk as primarily the risk of losing money over the long-term and secondarily as the risk of underperforming what an investor could get almost for free via buying the market index.  We manage these risks in part by employing no leverage in our operations, seeking margins of safety in valuation, and managing our portfolio to include 25 or more holdings across different industries.

How We Validated Our Approach

To gain real confidence in our (or any) investment process, three equally important conditions must be met.

  1. First, there must be strong historic evidence that the investment process worked in the past.
  2. Second, it must be shown that an understandable cause-and-effect relationship explains how the historic results were achieved.
  3. Third, there must be a reason to believe this relationship will exist in the future.

Here is a quick example of how each of these three conditions might be considered. 

Suppose a friend of yours has “beaten the house” an improbable number of times playing Blackjack in a particular casino. It is possible that your friend was very lucky, the house was not properly managing their games, or your friend was using a method that gave him better odds than the house. Either way, the first condition has been met – the particular way your friend played in this individual casino was successful. 

Now suppose your friend demonstrates that he can count cards on up to three decks (the same number this particular casino uses in Blackjack) and has the mathematical skills to use this information for calculating the odds that he has a winning hand. He explains that by proportioning his bets according to these probabilities he has an advantage over the house. This is a plausible explanation for how he was able to beat the house in the past (the second condition) but it is not sufficient to gain confidence that he will beat the house in the future. For this, the third condition must be met.

Your friend’s advantage may be at risk for the casino could increase the number decks to more than three (the limit of our friend’s ability to count) or begin periodically shuffling played cards back into the decks.  These simple changes would eliminate the cause-effect relationship underlying his strong historical results, making his past success irrelevant to his ability to perform in the future.

Let us now discuss how we went about satisfying ourselves that Euclidean’s approach could, unlike in the case of the Blackjack player above, satisfy all three conditions necessary for confidence in future success.

Condition 1 – Historic Evidence Our Process Would Have Worked In The Past

We satisfied the first condition by simulating (or back-testing) how well a fund managed precisely as we manage ours today would have performed if it had been started as far back in history as the data will allow us to go.

We differ from most quantitative funds in our desire to simulate performance over long periods of time. We are just as interested in how well our fund would have performed in 1975 as we are of its performance in 2005, and we are constantly pushing technology to allow us to go as far back in time, so to speak, as possible.

This is a crucial point. If you find that something worked very well for a short period of time (like a growth strategy might have worked in the 1990’s), there is a good chance you found a temporary trend.  If you based your investment strategy on an approach that was validated over just a few years, you would (or should!) find yourself in the frightful situation of wondering if you will wake up tomorrow and the trend will stop.

For us, it has always seemed imperative that we should seek not the best performing strategy for any given time but the most timeless investment methods possible.

Simulation Performance

When we run a simulation on data from the early 1970s through mid-2009 using our current investment process and include deductions for our fund’s fees, the simulation generates compounded annual returns in the high teens. To get a sense of the character and variability of these returns, we also measure the percentage of overlapping twelve, twenty-four, and thirty-six month periods that the simulated performance would have outperformed the S&P 500’s total return (that is, the S&P index return with dividends reinvested).  We call the 12-month consistency figure the “win-rate” of the simulated returns and, for this simulation, this number is over 70%.

The two numbers – compounded annual return and win-rate – represent the first component of how we gained confidence in our approach, but they are by no means an estimate or prediction of how the fund will perform in the future. The power of the simulation is solely its ability to tell us whether our approach has any potential for doing well.   

Condition 2 – An Underlying Cause Explaining How The Historic Results Were Achieved

With quantitative models and back-tested results, one of the real dangers is that increasingly powerful tools for mining data increase the likelihood of finding patterns that are solely the result of chance.  

This fact is the great irony of data mining.  A Caltech professor named David J. Leinweber noted a great example of the dangers here.  He explored thousands of data sources and found that, from 1981 to 1993, the rate of butter production in Bangladesh explained – better than anything else – subsequent movements of the S&P 500.  Unless Bangladesh butter producers were earning enough money to move markets and consistently pouring all their earnings into domestic equities, it seemed pretty obvious that this was a chance relationship!  Of course, that there was no cause-and-effect dynamic between Bangladesh’s butter production and the S&P 500 was proven by the relationship between the two ceasing to exist from 1994 and beyond.

In-Sample & Out-Of-Sample Testing

The process of in-sample and out-of-sample testing mitigates this risk. To understand this concept, think of a teacher in a classroom setting that is trying to evaluate students on the material they have been taught.  For the final exam, the teacher cannot use the same problems and solutions that were presented in class as the students may have simply memorized the correct answers. Instead, the teacher needs to present new problems so that the students’ understanding of the core principles – and not their ability to memorize answers – is evaluated.

As we worked to simulate performance over several decades, we needed to do the same thing as the teacher. That is, we searched across decades for what seemed to be timeless lessons on one group of companies and then tested our insights on companies that were not included in our original search. This basic idea allows us to generate simulated returns where we have greatly reduced the risk of finding a good result that exists purely by chance.

The Cause-Effect Relationships We Found

Even so, out-of-sample testing alone does not provide us with the logical foundation to proceed.  This is where the second condition of finding a cause-and-effect relationship comes in.  We need to be able to explain why a process has worked in order to explore whether that process is likely to work in the future. 

When we took an unbiased look at the data, we could have found that investing in high-growth companies at any price was a sound investment strategy.  Or, we could have found that a company’s revenue per employee is indicative of how it is likely to perform as an investment.  But, we did not find these things.

What we found in the data were two cause-and-effect relationships.   First, that companies tend perform better over time if they have a long and consistent operating history, a capability to weather difficult times, and a track record of generating strong returns on capital.  Second, these companies tend to best perform as investments when they are purchased at very low valuations relative to their historic earnings. 

While we are not the first people to note these relationships, they make sense to us.  After building a company for a decade and exploring hundreds of others in depth, we are confident that we have not identified a set of chance relationships.  It aligns with our experience (and common sense) that:

  1. Strong returns on capital are better than weak returns on capital.
  2. Strong balance sheets are better than weak balance sheets.
  3. Consistent operations are better than erratic operations.

  4. Companies generally perform better as investments when they are purchased at discounted – instead of premium – prices. 

We are staking our own capital and reputations on the basis that these relationships, which are evident in the historical data, exist today and will remain in the future.  This brings us to condition number 3.  

Condition 3 – The Cause-And-Effect Relationship Is Likely To Exist In The Future

Even promising results that can be explained by an underlying cause-and-effect relationship and are not the result of chance can fail to persist in the future.

Remember, for example, our Blackjack player would likely have continued to do very well until the moment the casino changed the rules of the game.  By moving to five decks or by shuffling cards more frequently, our player’s advantage would have been taken away. 

In investing, a pattern of outsized returns can cease to exist when it can be easily replicated and works with relative consistency over a short period of time. Once such a pattern is made popular and more people attempt to benefit from it, the outsized returns vanish because prices are inflated by the new demand for opportunities that match the pattern.

A real life example of this type of failure is the January Effect – the observation that small company stocks historically performed very well in early January.  A plausible explanation for why this has occurred is that many stocks are sold off for year-end tax planning and the proceeds are reinvested in the market in January, bidding up prices. However, this effect has become less pronounced in recent years, perhaps due its popularity and greater numbers of investors attempting to profit from it.

This risk motivates the need to meet our third and final condition. Specifically, will the cause-and-effect relationships underlying our simulated performance continue to exist in the future?

The opportunities that Euclidean depends on – good businesses selling for bargain prices – are created by investor psychology. That is, during times when a business has a temporary setback or when the conventional wisdom is that difficult times are ahead for a particular industry, investors can get very afraid.  When fear and pessimism prevail, good companies are sometimes offered at very low prices.   

Certainly the past 18 months have shown this to be true. So, when thinking about this third condition, we asked ourselves what would it take for this cause (investors sometimes acting irrationally) to go away for a long period of time?  The answer is that mispricings will become less likely when humans start acting less like humans.

Given this, we like our chances over the long-term.  We believe human nature is the one constant we can count on persisting over time.  Technology innovation will surprise us, industry structures will change, and regulations will evolve, but the basic elements of what it means to be human – fear, greed, desire for the comfort of the herd, etc. – do not seem likely to go away.  

Looking Into 2010

We are looking forward to 2010.  We have continued to slowly grow our investor base and – hopefully supported by these types of communications – believe we have a strong core of investors who are philosophically aligned with, and well informed of, our approach.

*****

We greatly value the privilege of managing a growing portion of your assets.  If you have any general questions, or specific feedback on the content or style of our communications, please call us at any time.

Best Regards,

John & Mike