Value Investing Forum

Graeme said:

Ok let's see if I get this. By pricing in e^(-1/2b t^2) to the cash flow you are pricing into your valuation that as time increases the probability of catastrophic failure goes up? (or, since you have a longer timeline, you have a higher probability of something bad happening eventually.) If you ran a regular DCF along side this modified disaster priced DCF, these second batches of valuations would be overall lower than the first batch right?

So for example let's say a regular DCF on $KO gave you a fair value of $65. If you had a margin of safety rule for yourself of 50% you'd be looking to buy at $32.50. Presumably the disaster DCF would value it lower due to the disaster equation. So why isn't just tacking on a bigger margin of safety to your regular DCF enough of a safety measure? Why doesn't it account for uncertainty?

Pretty much. But I think it's going to discount growth differently. Let's suppose you have two assets that are valued at $65 each. One has higher payments now (say a higher dividend yield) but the other has a higher growth rate. Standard DCF values both at $65. This model will actually value the higher dividend yield one higher. So this is all independent of an actual practicing method. It's actually saying one is worth more.

Consider this question: why do we choose a higher discount rate? For example, if I told you two bonds pay $10 per year for 10 years, what are they worth? Standard DCF would say that we have to assess the risk of each asset and use different discount rates. But to me, $10 every year for 10 years is the same as $10 every year for 10 years. What I really want to know is what is the probability that I'll actually get those payments? I think this method better reflects that.

If I'm understanding this method, you actually use the same, "risk-free", discount rate for all assets but apply different "b" values for differing risks.

From a practical standpoint, using the buy things cheap is probably a good way to go about doing it. But is more to the question on how you choose to determine the value to begin with.Why not discount all future dollars with the same discount rate but perhaps ascribe different probabilities that we'll actually receive them? That seems to make more sense to me. How to apply that I'm not sure. But then how do you choose a discount rate? Use CAPM?

In my second post, I was having difficulty in even choosing a "b" value for bonds but I don't know how to choose appropriate discount rates either. *shrug*

I think this meshes well with Graham since there is an emphasis on making inferences from the most certain information (or at least attaching a greater wait to those inferences made from more certain information). At least that's what I got from Greenwald's book.

You start at the top of the balance sheet because the value of the cash is most certain. Then you work your way down through receivables, inventories, PPE, etc. Future cash flows are of course the least certain information we have but it's those that represent present value.

As a side note, Keynes supposedly had a concept of both expectation and the weight of the expectation in his treatise on probability. I think that ties in well here.

I'm currently going through Hyman Minsky's John Maynard Keynes. I didn't get too much out of the General Theory when I read it and wanted to give Keynes another shot. Minsky is one of Keen's influences.

Post edited 11:07 pm – October 14, 2011 by somrh

I haven't found the original source on this (the book is out of print: Dynamic Economic Systems: A Post-Keynesian Approach) but Prof. Steve Keen explains it pretty well. (By the way, the entire lecture series, which I've been going through, is pretty good. He also has his updated release of his book coming out on Oct. 25: Debunking Economics.)

Keen's Lectures on Behavioural Finance. Lecture 3 Part 2

The main bit starts about 15 minutes in (the video starts off closing off his critique of CAPM). You can find links to his power points on his blog here.

Here's a summary:

Keen distinguishes between "Risk" and "Uncertainty". Risk is a "known unknown". For example, he uses the example of throwing a die which has 6 possible outcomes. You don't know what outcome will occur but you do know one of the six will. Uncertainty is an "unknown unknown". The concept he advocates here is closer to the "permanent loss of capital" discussed by value investors. He defines uncertainty as "Outcome has an unknown chance of being one of a possible infinite number of unknown outcomes".

He argues that simply choosing a higher discount rate doesn't really account for uncertainty. The sort of uncertainty that is a concern is this: What if at some point in time, some event occurs (war, disaster, obsolescence, bankruptcy, etc) that causes cash flows to simply stop?

He advocates a method developed by John Blatt. The idea is to discount rate that increases over time. In other words, since near term cash flows are more certain we discount them less so than cash flows further away.

Keen admits this is an oversimplification but contends that this is better than simply ignoring uncertainty.

The interesting thing is that comparing the two methods can give different results as to which of two alternatives is the better investment.

Typically one discounts a cash flow at time t – which we'll call C(t) – by some discount rate r (using continuous discount rate):

C(t) x e^(-r t)

You then sum up (or integrate) all of the cash flows discounted with regard to their respective times.

Blatt/Keen adds a term:

C(t) x e^(-r t) x e^(-1/2 b t^2)

The 'b' is the disaster probability and would be something like 5% (which is what Keen uses, but it can be any appropriate estimate.)

As time goes on, the last part of the term – e^(-1/2 b t^2) – will dominate the expression due to the t^2 and further out cash flows will be discounted more heavily than nearer term ones. This will have the obvious effect of valuing growth firms less than other firms. I don't know if this is desirable or not but it is what it is.