*It will give you a feel for human nature. They take psychological subjects, and they show them a white square against a black background. And they ask them to estimate how far away the square is. They tell them it is between 1 and 20 feet away. And they ask them to estimate the error brackets; in other words, is it plus or minus a foot. They all give a precise estimate like 7.5 feet and they all give*

*very precise error brackets of plus or minus a foot. What they don’t tell them, of course, is the size of the square. And they vary the sizes of the square. It turns out; you can’t know physiologically the location of that square without knowing its size. So the true answer is it is 10.5 feet away on average and the error brackets are between 1 and 20 feet. People just don’t seem to grasp that reality. Faced*

*with that uncertainty they impose the idea they know where the square is. And what they do is show them the squares of different sizes. And they say look, these squares are of different sizes, and then they run the experiment again.*

**A Human Experiment in Psychology**

**The final point—and this is significant—that you would think people would learn. The statistics I showed you would become embedded in how people behave. These statistical portfolios that outperform by even three percent puts you in the top 2% of money managers. There are also experiments that show that people suppress uncertainty. They believe they know what they know with certainty. That, by the way, tends for them to exaggerate both good news and bad news—drives ****good news stocks too high and bad news stocks too low. (The market always goes too far or to extremes). But they never learn, and this is an experiment worth talking about.**

## ADDING VALUE TO YOUR ANALYSIS

But having identified where you are going to look if you are not going to employ my students, just do it with a computer. However, if you are going to someone who will do active research or active selection; among these statistical opportunities they must have a process for valuing what you are

looking at to actually add something to the statistical test. And this is the most depressing thing I am going to talk about here.

Sanford Bernstein which is a terrific operation, it is technically very sophisticated, they have 200 analysts, and they monitor their performance. They have outperformed the market for over 25 years now by 3%. They manage, by the way, 400 to 500 billion dollars. That puts them in the absolute top tier of institutional money managers. If they had just done market to book portfolios, they would

have outperformed the market by 3.9%. So for all that work by the most successful money management out there, they add negative 1% to their performance.

**The Problems with Using the Discounted Cash Flow Models for Valuation**

Companies grow at vastly different levels and rates. Growth obviously adds under certain circumstances to multiples though you will see, but it also reduces multiples. In fact, when you have done all this, ask yourself are these cash flow levels really sustainable? You are looking at errors realistically of plus or minus 100% in these estimates of multiples. In fact, of course, you can do it by computer. You don’t need to be an analyst to do them for future valuation. So it doesn’t really add anything to valuations. So this is almost an exercise in futility. Thank goodness we teach our MBAs to do something more sophisticated than that. What I hope you were all taught to do was to estimate future cash flows up to a certain year and then estimate a terminal cash flow multiple by a factor which is usually the terminal value which is all the cash flows discounted back to the present. That is a discounted cash flow (DCF) model. You discount the cash flows back to the present usually by the cost of capital and the growth rate. Anybody who has seen any serious business research or any serious Wall Street research will have seen that done. It has two advantages: first, since you are used to the ratio analysis, just doing a terminal value means that any serious diseases that this detailed approach has are also diseases that affect the ratio analysis. You can’t really do worse than the ratio analysis because you are thinking in a way more carefully about things. Second, it is theoretically the right thing to do.

** On the other hand, when you go through it and calculate these cash flows – starting with estimating revenues subtracting required investment then getting the cost of capital, there are a huge number of embedded assumptions. In particular, anybody who has done these models should have a sense that they are very imprecise. Why? It is because almost all the value is always in the terminal value for these models. And the terminal value is usually a cash flow that you may estimate reasonably well times 1/over the difference between the cost of capital—of say 10% and a growth rate of 5% which is just the growth rate of nominal world ****GDP. 1/ (0.10 – 0.05) or 1/0.05 = 20 x multiple. On the other hand, suppose this is the growth rate from say 5 years out—not today’s growth rate but from 5 years onward. Say you were off by 1% in your estimates in either direction and 1% in the cost of capital in either direction; you could easily have a cost of capital of 9% and a growth rate of 6%. 9% minus 6% is 3%. 1/3% is a multiple of 33. You could equally well have a growth rate of 4% and a cost of capital of 11% and the difference is 7%–1/.07 = 14 multiple. In the critical element
of value that is the difference of more than 2 to 1. And I believe the people who have done this have seen that happen. The range of values is too wide to be of practicable use.**

*That is not a problem that arises from the terminal value formula. That is a problem that is fundamental to a discounted cash flow approach to valuation. And this is the second thing it is important for you to know. Any investment managers who are out there who are doing discounted cash flow measures of value are using a technique that in practice is an incredibly stupid thing to do. Assumptions Required for Using the DCF Model There are three (3) reasons for that. One of which will be obvious and two are a little less obvious. The obvious reason is, and it is reflected in the terminal value problem. You take good information which is the near term cash flows, and you take really bad information which is the long-term estimation of cash flows and you add it together. (Charlie Munger used the metaphor: “When you mix turds with raisins, you get turds.”) When you add bad information to good information what do you almost always get? Well, you always get bad information because the bad information dominates. What you would like to do in a valuation approach is start with the pieces of value that you know are there and segregate out the bad information. That is the natural thing to do, and this approach doesn’t do any of that. The second problem with discounted cash flow and NPV approach is, think about what you want a valuation approach to do. It is a rule; it is a machine for translating between the assumptions that you can make reliably about the future and the present day value of the security. The **input you want for that machine are the assumptions which you can reliably make. But think about the assumptions that you make for a discounted cash flow analysis. *

*Suppose we were doing this for the Ford Motor Company or for Mercedes Benz, I guess there are no more British examples that I can use, sorry. The DCF models have to use estimates of the future profit rate, estimates of the future cost of capital, estimates of investment intensity, and estimates of growth rates. Can anyone guess twenty (20) years from now what Ford’s profit margins will be? 10 years from now, 5 years from now? But that is what the DCF model requires. How many think they know what Ford’s growth rate is going to be? 10 years from now, 5 years from now? Lots of luck! So you are using assumptions that you are not very good at making. You try to do sensitivity analysis; you try all sorts of values. Things don’t vary independently. The sensitivity analysis will*

*usually give you any outcome you want. And that is why the investment bankers love them. (The last person you want to fool is yourself—anonymous).*

*The third thing is you would like your valuation approach not to throw away information. You would like it to use all the information available. And what particular important information is never used in a DCF analysis. The balance sheet information—it just disappears. In favor of earnings which are*

*income statement projections. So what I would like is an approach to valuation. And again, I am going to describe one, today which is the Ben Graham approach to valuation but you ought to know that this ought to apply. It ought to organize value components from most to least reliable. So you can say, I know I have this much value at $8 per share there. I am less certain about the next twelve dollars, and the next $20 is pure speculation. It should organize valuations by strategic*

*assumptions. This is the valuation if the industry is non viable, this is the valuation if the industry is viable but no companies enjoy competitive advantages but the industry is viable. This is the value if there are sustainable competitive advantages. Segment information and values from most reliable to*

*least reliable. And third we would like to use all the information and cross-correlate that information. Well there is a way to do that. There may be other ways to do it that I am not aware of but one way to do it was the way Benjamin Graham did it in 1935/36, and who thought about this problem. And later as*

*it was refined by Professor David Dodd at Columbia University.*

Source : Csinvesting