[NKS, Business modelling, Intrinsic value and Buffett] - A New Kind of Science: The NKS ForumA New Kind of Science: The NKS Forum
Pages:1
NKS, Business modelling, Intrinsic value and Buffett
(Click here to view the original thread with full colors/images)
Posted by: gregrez
One area which I feel is worthwhile investigating is the connection between NKS and business rules and behaviour.
Has anyone investigated the connection between NKS and corporate valuation, and instrinsic value and growth?
I feel they are very closely related.
For example, you start with a rule of how a single store, say Starbucks/McD's etc..the profit that is collected from this is used to open more stores.
At some abstracted level, the rules for this growth could probably be modelled as something very simple. But how?
How could NKS be used to "price" a company? An asset?
Thoughts?
Posted by: Jason Cawley
The basic drivers of valuation are well enough known from traditional economics. There is a substantial gap between regular industry practice in finance (which relies instead on essentially stochastic models of price) and what economists and accounting principles already understand. That is where fundamental analysis of the Graham-Dodd or Buffett variety operates, basically. The rub is that modeling or forecasting future earnings is an art not a science, and they drive valuations.
There are several pitfalls to avoid when analysing such things. It doesn't take NKS to avoid them, just common sense, some basic economics, and accounting realism. Some people use the uncertainty they can create to make valuations stretch, or make unrealistic combinations of assumptions and get nonsense answers back from the math, treated too statically.
A typical example is using too low a discounting rate, like cash returns, which will tend to make even average companies look almost infinitely valuable if applied mechanically. Instead one needs to use a robust, business-like target rate of return assumption, like 15%, and keep it constant.
Another is failing to distinguish between excess returns on old capital, and the true marginal return on new capital. A company that truly has an opportunity to invest any desired amount of new capital at any rate above the rate of discount, is mathematically worth infinity. Which is nonsense economically - if any desired amount of new capital really had a chance to earn such returns, then the rate of discount of the whole society would be correspondingly higher.
The confusion arises from taking accounting figures like return on equity as measures of the incremental earnings of new additional capital. Take a company like Microsoft. Its earnings on its stated book capital are 20.7%. But the company also has about $40 billion in cash and short term investments on its balance sheet, earning (these days) a few percent at best. If you add a dollar of capital at the margin, is it likely to earn 20.7 cents or 3-4 cents? If the real answer was 20.7, think they'd sit on that much cash?
The reality is, some portion of their old capital is now effectively worth many times its book, and thus generates excess returns when measured by book. That old capital or the associated market position and business opportunity etc, may well generate future earnings in a strongly growing stream. But stuffing another dollar into the company will not generate commensurately strong returns. The basic function is not an exponential on invested capital, but some meandering line from the business position, plus a modestly inclined (low rate) exponential bit from actual capital invested.
Thing is, figuring out how much of it is one and how much of it is the other is a modeling problem that is as little a science as forecasting future earnings.
What G-D fundamentals tell one one should do, is come up with any independent projection of the likely future course of earnings, including any uncertainty or spread they seem likely to have - and then calculate discounted present values of representative chunks of that distribution, weighted by their expected likelihood. Earnings going this-a-way roughly x percent chance, that-away roughly y percent chance, etc. Discount each of them, get a distribution of values now.
What modern portfolio theory does instead, is just assume there is enough randomness involved that the valuation is going to be some normally distributed possible spread around the price now. Which doesn't help one spot differences between prices now and values now, and only tells one to expect diffusion in the future. For enough independently varying random things, that will follow by the law of large numbers - many independent chances add up to Gaussian statistics. But it need not be remotely true of individual instances, for that to be the case overall or on average. In other words, there is no assumption about correctness of present prices involved.
To carry out the G-D valuation program, then, one needs an independent way of modeling the likely future course of earnings, without assuming at the outset it must be normally distributed randomness. There NKS might have something to say. But so might simple rules of thumb, such as those G-D looked for or practitioners in various lines of business generally use in practice. Retailers don't forecast the likely earnings of a new store by running random walk simulations. They look at what their other stores do and assume it will be locally similar. Will that always be right? No, forecasting earnings is an art not a science. But there is no reason to expect a literal random walk, either.
What one doesn't need to do, though, is reinvent Gaussian statistics by making some complicated model with lots of randomness in it at multiple levels, which eventually produces, surprise, normally distributed expected values. If you are going to go to the trouble of building some micro-model of earnings, and think it is telling you anything you wouldn't already know from just assuming a random walk, your micro-model can't be a simple average over something completely random. It might have a few cases that "tree out" as discrete possibilities. It might have some deterministic economic logic, that you take from what you know of the actual business. If it greys everything out with random this and that everywhere, it will just laboriously reproduce modern portfolio theory of normally distributed asset returns.
I do think there is a project here worth doing. But one has to come at it knowing what there already is, both on the fundamental valuation side (the Graham-Dodd tradition, basically), and the existing portfolio theory side (statistics applied to finance). You will get meaningful deviations from the last, only if you independently model earnings with some deterministic (though potentially complex, perhaps "programmably" complex) micromodel.
Forum Sponsored by Wolfram Research
© 2004-2008 Wolfram Research, Inc. | Powered by vBulletin 2.3.0 © 2000-2002 Jelsoft Enterprises, Ltd. |
Disclaimer
vB Easy Archive Final - Created by Xenon and modified/released by SkuZZy from the Job Openings