Life simply never turns out the way it is modeled. Conflicts arise, triumphs enter out of nowhere, randomness intercedes. Models fail and yet they are still essential to the decision-making process. How do we reconcile the ultimate fallibility of models with their undeniable usefulness? I have found an answer here that makes sense to me and that I can put to use: be careful what you model.
Models, in general, and most certainly financial investment valuation models, tend to produce an answer. They often single-mindedly point towards a value based on various input factors (Monte Carlo simulations seek to add distribution robustness to the single-mindedness, but even these ultimately exist for their single-minded statistics). The single-mindedness flaw of earnings models is unavoidable but is not a disqualifer if one adheres to the warning in the preceding paragraph. In choosing situations to model, one is correspondingly choosing guesses for the aforementioned input factors. Now, of course, in any and all modelable situations, it is prudent to choose a range for the input factors and test various scenarios (this is, in fact, the strength of Monte Carlo simulation). The key, however, is the recognition that the reliability of our input factor ranges varies widely, make that very widely across different situations. To put it more simply, some situations are more predictable, in terms of both the number and magnitude of the influencing factors, than others. This is the point of this blurb so allow me to repeat: some situations are more predictable, in terms of both the number and magnitude of the influencing factors, than others. Models seduce us into not recognizing just how powerful this concept is. We figure we can use ranges for our factors to capture the unpredictability; I contend that, in complicated situations, not only will our ranges be suspect but, much more importantly, even figuring out what factors to isolate is extremely difficult, if not very close to impossible.
The antidote for me is to be very selective of what you model. For earnings power companies, I only like businesses that have been around and that, in all likelihood, will continue to be around. I do not take this consideration lightly. Has WWY been around? Let’s see, it was founded in the nineteenth century to make the same chewing gum product that really has not changed much in over 100 years (and neither has the bad breath need for the product). WWY passes: go ahead and model its earnings power. How about EXPE? Well, it’s been around for around a decade - not bad - but will it continue to be around? This, in my estimate, is not certain at all (it will probably survive, in all likelihood; however, the level of profitability of the business model is very difficult to predict). Modeling EXPE on earnings power is a mistake in my world. I could try: I’d get a nice range of numbers that may even point to an investment. Yet, the investment would be built on false premises and factors that very well may not be material at all. I’ll pass.
There are other considerations beyond earnings power that make many companies investable (especially tangible book value), and I personally spend a lot of my time studying these situations. The point of this note is to signify the flaws of earnings models. They should not be relied upon for those companies where the future is simply too difficult to define.