Journeys of a Bumbling Investor

Just saw some articles from Stingy Investor (http://www.ndir.com/SI/articles/1105.shtml) talking about Graham's stock selection criteria for the defensive investor.

Benjamin Graham's Criteria for the Defensive Investor
——————————————————-
P/E Ratio less than 15.
P/Book Ratio less than 1.5.
Book Value over 0.
Current Ratio over 2.
Earnings growth of 33% over 10 years.
Uninterrupted dividends over 20 years.
Some earnings in each of the past 10 years.
Annual revenue of more than $100 Million (1950).
———————————————————-
Source: The Intelligent Investor, 4th Revised Edition (pages 184-185).

As current stock screeners only provide use of 5 years of data, the approximation used by the author was:

Screening criteria used to approximate Graham's rules
———————————————————
P/E Ratio less than 15.
P/Book Ratio less than 1.5.
Book Value more than 0.01.
Current Ratio more than 2.
Annual EPS Growth (5 Yr Avg) more than 2.9186%.
5 Year Dividend Growth more than 0%.
5 Year P/E Low more than 0.01.
1 Year Revenue more than $400 Million.

The results published showed that Graham stocks gained about 36.2% annually for the past 5 years, while the S&P 500 made a small loss.

This anecdotal 5-year result is interesting, after reading the magic formula from Joel Greenblatt's book. My personal preference is still to use the all these "magic formulas" as stock screeners; significant due diligence on each stock will still be required before any purchase.

I picked up a copy of Joel Greenblatt's "The Little Book that Beats the Market" today. In it, it talks about finding cheap companies (high earnings yield) and good companies (high return on capital).

The first time I saw a discussion on the importance of high earnings yield, is in Mary Buffett's Buffettology book (1st ed). At that time, I had a lingering thought that still persists today. The earnings yield is typically compared with the long-term US treasury "risk-free" rate, e.g. if a stock's earnings yield is 10% and the risk-free rate is 6%, then you're "earning more" with the stock, compared to what you're paying.

The thing that bothers me about this comparison is that for a bond, you are guaranteed your principal, whereas for a stock, if you decide to "redeem" that piece of ownership, the money you get is the market price at that time. With a loony stock market, that "market price" can range from zero to a whole lot. Hence with a non-guaranteed principal, there is not much point comparing the earnings yield of a stock to that of a bond (even if you make the 10%, you might lose all your principal). What this also means, is that the price that you pay initially, must be justified by all the future cashflows that comes in, so that even if the price of the stock ends up to be 0, you still get the future cashflows.

Turning next to Return on Capital (ROC), an example cited in Joel's book is that Jason spent $400,000 to set up a store that ended up earning $200,000, so ROC is 50%. Again, the fact that the principal (i.e. that $400,000) must remain, is important. The ROC is 50% because we have (new value/old value) = 1.5, i.e. new value = $400,000 + $200,000 = $600,000 and old value = $400,000. Imagine, after spending $400,000 for the store, if Jason wants to liquidate the store, he can only get back $300,000. That would mean that (new value/old value) = ($300,000 + $200,000) / $400,000 = 1.25, i.e. ROC = 25%.

That leads to an interesting problem, if the accounting books record NTA at cost, i.e. $400,000, then we should do a bit of discounting at the numerator to get the ROC. If NTA is recorded at market value, then it would typically understate the cost and overstate the ROC. In that case, we need to find out how much was actually spent to generate that amount of earnings.

"How much is this business worth?" is a critical question that one needs to ask, after answering

1. Is this a good business?
2. Is there good management?

Typical techniques used to produce that answer, have been a combination of the following: Discounted Cash Flow (DCF), with varying degress of sophistication, Dividend Discount Model, Gordon Growth Model, etc. all the stuff that you find in a corporate finance textbook.

To answer the question myself, I attempted a simple exercise to value a stream of cashflows, where I know what the eventual answer should be. A company can be modeled in the simplest sense, like a savings account. You start off with some money in the account (i.e. net tangible assets), you make interest off the balance (i.e. return on capital), you can choose to withdraw some (i.e. dividends) or keep all the interest in the account (i.e. increase retained earnings, payout ratio of 0).

Hence I built a simple spreadsheet to model a savings account. The account is started off with $1, with annual interest rate of 10%, and all interest are ploughed back into the account. Hence we have a steady stream of interest cash inflows: 0.1, 0.11, 0.121, etc. Now, if I imagine that this "savings account" is in effect a company (i.e. NTA = $1, ROC = 10%, payout ratio = 0%), and given what I know about its future cash flows, will I be able to determine that the fair value of this company is, in fact, $1? (assuming that a bank does give 10% interest rate on accounts)

The short story is that, you cannot discount each of the future values, sum them up, and expect to get $1. The fact that in order to obtain the cash inflow of $0.121 in year 3, you are _forced_ to re-invest the previous year's $0.11, meant that these cash inflows cannot be freely used. Hence in a sense, the "free cash flow" is 0. How then do you value a company with the characteristics above?

What needs to be done, then, is to project future cashflows, use that to determine the value of the company at some point in future, and perform just ONE discount to determine the present value, i.e the intrinsic value of the company (I am skipping the details to do this properly, but the high-level idea is there).

This example illustrates a difficulty in using traditional DCF – its very difficult to determine what cash flow is _truly_ free. (Operating Cash Flow – Capex) is not good enough. Many companies spend this (Operating Cash Flow – Capex) to repay debt, to invest in short-term securities and to expand into new areas. Can we say that the company does not _need_ to repay debt? does not need to expand into new areas to maintain its competitive advantage? If such spending is necessary, then these money cannot be included in the DCF calculation since they are not truly free money that can be taken out (an implicit assumption when you perform a DCF).

A more accurate definition of FCF is needed: Free Cash Flow = "Truly free" cash flow that can be always taken out of the company and the company will still do well. That is, if all the projected FCFs that you have used in the DCF calculation is taken out of the company (e.g. if you project that Year XX has FCF of $YY, then we will withdraw $YY from the company in Year XX), the company must still be able to do well.

Say you have picked out a couple of companies. Done your detailed research, estimated their potential gains, estimated the time horizons in which the catalysts would occur, and finally decided that they are good buys. So how exactly should your optimal portfolio look like? Do you pick the top 8? the top 10? or maybe just the top 3?

It struck me about 1-2 months ago that modeling this basket of stocks, is very much like modeling a basket of credit default swaps.

Studies have shown that the "jump" or appreciation of a stock's price, typically happens in only a few days in a year; meaning that if you miss out say, the best performing 10 days of your stock, your return will be mediocre. This is very much like a "jump" event, that is currently used to model credit events (default, bankruptcy, downgrade, etc.) for the pricing of credit default swaps.

Hence the techniques used to model CDSes (e.g. poisson processes, copulas, etc.), can be applicable to modeling equity portfolios. The random processes can be simulated, with each jump resulting in an amplitude multiplier representing the potential gain of a stock (i.e. discount to fair value), and the intensity of a process relating to the estimated time horizon in which a catalyst would occur. The simulation would then be able to simulate different portfolio value process, get expected values, stderrs, apply variance reduction techniques and all that good stuff. Different portfolio rules (e.g. after a predicted gain is realised, shift money to highest potential gain stock among those that have not "popped" yet) can be applied in the simulation and their results analysed.

I have not seen such a thing being done before yet. Perhaps it is just rare to find someone with exposures to both fields: value/fundamental investing and mathematical finance.

They are essentially the same. In both cases, each shareholder receives a certain number of new shares free of charge, whereby the stock price is reduced accordingly. The total shareholders' equity remains the same. Cash remains the same.

In the case of a stock split, each old share is split into a number of new shares with a reduced par value, leaving the total share capital unchanged.

In the case of a stock dividend, a number of new shares are received for each share owned. The new shares have the same par value as the old shares, whereby the total share capital increases proportionally with the size of the stock dividend. This is done in the accounts by transferring retained earnings to the share capital, which has implications for the ability of a firm to pay out cash dividends in the future (some debt convenents restrict the amount of cash dividends that can be paid out depending on the level of retained earnings).

By "voluntarily" doing a stock dividend and decreasing their retained earnings, companies "signal" that they are in good financial condition and still able to increase their cash dividends (i.e. impact of retained earnings is minor).

http://academic.luzerne.edu/lmajor/AccIIChapt%2014.htm
http://www.econ.au.dk/fag/2786/e05/Splitspaper.pdf