Author: mungofitch

Subject: Can anybody test this screen?

Date: 2/8/2007

Recommendations: 8

I have an idea for a screen, but, alas, no way to test it. Rather than starting with a close approximation, I thought I`d start with posting the ideal version of the screen, then ask around for who might be able to test it. The goal is for a very conservative, value-oriented screen, based on some sort of estimate of price-to-intrinsic-value. The starting notion is low price-to-earnings. Price is no problem, but earnings (specifically, recent earnings) are a bit more problematic. They are hard to measure accurately, and hard to smooth out. Plus, I want things with a long history. So, I thought a good measure would be total return 10 years, but based on value creation, not on market returns. The real value created by a company in the last 10 years is the book value growth per share, plus the dividends per share: there isn`t really anywhere else for the money to go (assuming the per-share numbers are all corrected for the dilution between the time it happened and the present). Book value and dividends are much harder to manipulate than earnings, especially when measured over a long time frame. The figures may be imperfect metrics because of depreciation or appreciation of assets, but at least they are aren`t "mushy". So, my proposed definition of value: current price divided by trailing 10 year value creation, defined as the sum of the dilution- adjusted dividends and book value growth in the last 10 years. Let`s call this tvr10y (total value return 10 year), which is measured in total dollars per fully diluted share, not a percentage growth rate. So, the basic screen so far is simply best stocks on price/tvr10y. By construction, we are eliminating any company under ten years old. Example: for a company with no dilution which increased its book value per share by $1 per share for each of the last 10 years, and paid a 5 cent dividend for 10 years, the tvr10y would be $10.50. But, if they had an event (like an acquisition) that caused a doubling in the number of shares outstanding 5 years ago (a 2-for-1 dilution), then the tvr10y would be $7.875. The five most recent years gave $5.25, and the earlier years gave $5.25 divided by the 2:1 dilution ratio. But, there`s one more adjustment---the value creation (book value growth or dividends) for each of the 10 years would be discounted back to the present by its distance into the past before being summed--a company which created nothing but $1 of value 10 years ago isn`t as valuable as a company which created nothing but $1 of value one year ago, for a variety of reasons. Let`s assume a constant discount rate of 7% to make life simple. In the example of the company above (with the dilution), I get a final tvr10y fo $5.84. Most recent year (1 year ago): ($1.00 + .05) / (1.07^1) = 0.98 2 years ago: ($1.00 + .05) / (1.07^2) = 0.92 3 years ago: ($1.00 + .05) / (1.07^3) = 0.86 4 years ago: ($1.00 + .05) / (1.07^4) = 0.80 5 years ago: ($1.00 + .05) / (1.07^5) = 0.75 6 years ago: ($1.00 + .05) * 0.5 (due to dilution)/ (1.07^6) = 0.35 7 years ago: ($1.00 + .05) * 0.5 (due to dilution)/ (1.07^7) = 0.33 8 years ago: ($1.00 + .05) * 0.5 (due to dilution)/ (1.07^8) = 0.31 8 years ago: ($1.00 + .05) * 0.5 (due to dilution)/ (1.07^9) = 0.29 8 years ago: ($1.00 + .05) * 0.5 (due to dilution)/ (1.07^10) = 0.27 Add them up, and you get $5.84 The main hole left over is that we might be catching companies which had a good time a few years back, but are now on the way out. The discounting might not be sufficiently sensitive to this problem. So, my proposed filter would be on operating margin. A company which can maintain its margins has a future, and is probably not being dumb enough to invest capital at low rates of return. Plus, the best companies in the long term are those with long run above-average margins, so a hurdle rate here should be good. Lastly, if we`re getting the historical earning power cheaply, then we might be getting some dregs along with the gold nuggets. A good way to separate the two is a strong balance sheet. With a strong enough balance sheet, I`m happy with any really low PE. This means the earnings weren`t generated with an unwise level of leverage, which is the main way a low PE on a good history can be dangerous. So, the final proposed screen: Conservative low debt-to-equity ratio, e.g., <0.75 or <0.50 (average operating margin last 2 years) >= (average operating margin last 10 years), within noise of 2% (average operating margin last 2 years) above hurdle (top N or maybe >10%) price/tvr10y top 10 If my reasoning is sound, this should work very well for long holds. So, it should probably would very well with hold-till-drop too. Or, an annual-hold "dozens", buying top-1-not-held each month. Does anyone have a database that could test this family of screens? Jim