How My Understanding of Margin of Safety Evolved

by Geoff Gannon


Quan here.

I was first attracted to value investing mainly by the idea of buying something for less than it’s worth. That’s why my first stock analyses were all about estimating intrinsic value. But the more I learn, the more irrelevant intrinsic value becomes to my research. I no longer look for buying 1 dollar for 60 cents. My analysis moved from calculating intrinsic value to estimating margin of safety.

 

From Intrinsic Value to Margin of Safety

I was stupid to not realize what Ben Graham wanted to convey in his definition of investing:

An investment operation is one which, upon thorough analysis, promises safety of principal and an adequate return. Operations not meeting these requirements are speculative.

I think my conception of value investing as buying something at less than its intrinsic value and confirmation bias made me miss the important message of this definition. I even overlooked margin of safety when I read The Intelligent Investor. But I eventually got the idea. Investment analysis is all about margin of safety.

The first milestone in my realization of margin of safety was when I bought Nutrisystem (NTRI). I asked Geoff about this stock not long after I met him. And he guided me through the analysis by asking key questions.

Before NTRI, I usually valued stocks based on earnings, and the stocks I bought were basically those that I thought had some competitive advantage and that were trading at less than 6 times normal earnings. Analyzing NTRI helped me realize one important thing: competitive advantage is not enough, predictability of earnings is necessary.

NTRI sells weight-loss diet packages to customers on average for 10-12 weeks. This is a one-time purchase. Once customers get their target weight, they stop. How good are this year’s earnings as an indicator of next year? Not very. Even though some customers may gain weight and go back to NTRI in the future, its earnings are definitely not as predictable as GEICO’s, whose customers rarely switch to other insurers.

That’s my first lesson on predictability. I cannot evaluate NTRI based on earnings although its position in its niche is great. But I found a measure to decide whether to buy NTRI. I looked at how many customers I need a year to get an adequate return. I thought the chance to get enough customers is high. I still do. (You might disagree on this issue.)

 

Predictability of Earnings

After NTRI, I paid more attention to predictability. I thought about Western Digital (WDC), a stock in my portfolio. I realized that end-customers (PC buyers, external drive user, enterprises) buy Western Digital’s products as one-time purchases. One-time purchase sounds not predictable. Then I moved on to other stocks. I analyzed News Corp (NWSA), and I face the same problem. How can I value its studio?  What about J&J Snacks Foods (JJSF)? This company sells soft pretzels and snack foods at stadiums, gas stations, grocery stores. It sells low profile products. Most people don’t remember or care about the brand name of its product. I don’t know how people’s preference for snack food will change. I don’t know how preference for specific products within snack category would be.

But I found the answer in chapter 17 of Ben Graham’s Security Analysis:

In order for a company's business to be regarded as reasonably stable, it does not suffice that the past record should show stability. The nature of the undertaking, considered apart from any figures, must be such as to indicate an inherent permanence of earning power.

...

It is possible, on the other hand, that there may be considerable variation in yearly earnings, but a reasonable basis nevertheless for taking the average as a rough index at least of future performance. States Steel Corporation may be cited as a leading case in point.

...

If compared with those of Studebaker for 1920-1929, the above earnings show much greater instability. Yet the average of about $8 per share for the ten-year period has far more significance as a guide to the future than had Studebaker's indicated earning power of about $6.75 per share. This greater dependability arises from the entrenched position of United States Steel in its industry; and also from the relatively narrow fluctuations in the annual output over most of this period, which thus affords a basis for calculating "normal earnings" of United States Steel.

...

The Average earnings for the 1923-1932 decade are thus seen to appropriate a theoretical figure based on a fairly well defined "normal output." While a substantial margin of error must be allowed for in such a computation, it at least supplies a starting point for an intelligent estimate of future probabilities.

WDC’s competitive position within the HDD industry is strong. Without technology risk, people buying PC, external drive, etc. today will buy again after 2, 3 or 4 years. Therefore, the average of past earnings can provide a good guess for the future (leaving the solid state drive issue aside.)

Fox's studio business is similar to the United States Steel example Ben Graham gave. It is a major studio, and Hollywood has always been an oligopoly controlled by major studios. Fox studio’s market share is pretty stable. So, average of past earnings can be dependable.

People buy JJSF’s products simply because it’s there. You stop at a gas station. You see an ICEE brand frozen beverage and suddenly want it. And you buy it. The number of people going to movie theaters, stadiums, gas stations, grocery stores, etc. won’t change a lot. The number of times JJSF’s products are seen by people is pretty constant. I suspect the conversion rate from being seen to being bought is stable. Past results say that. JJSF is more predictable than most companies. Again, the average of past earnings is helpful.

 

More Margin of Safety for Less Predictable Earnings

I still care a lot about earnings power. But it’s just something I look at to examine margin of safety. I’ll need less margin of safety for very high predictability companies. I’ll need more for companies that I feel uncertain about my judgment of their earnings power. In other words, I need a cushion for my ignorance. Predictability depends on a variety of factors. I think the three most important factors are moat, stable demand, and cost structure. I don’t have a general tool to judge predictability. I just know that inherent stability is something to care about. NTRI is a special case. It’s too hard for me to estimate its earnings power. But I see the margin of safety from a different point of view.

So, requiring more cushion for less predictable companies is my first application of margin of safety.

To be continued…

Talk to Quan about How His Understanding of Margin of Safety Has Evolved