I finally got the book and started doing my own scans. I put them on the web with a description of how I did the calculations:
members.cox.net
I tried to stick closely to the description in the book's appendix. I choose to scan indices, mainly because I had lists of stocks readily available. Along the way, I found some problems in implementing the scan.
The biggest problem is that often data published on the Web is incorrect. Morningstar had a habit of giving incomplete data for trailing twelve information and sometimes Morningstar was just inaccurate. Hoover's doesn't have really up to date quarterly data. Yahoo seems the most consistent. It does occasionally contain an incorrect number for 'Non Recurring' expenses, and sometimes throws in other incorrect numbers. I went with Yahoo. Once you move beyond the S&P 500 there seems to be more & more incorrect values and the incorrect values matter more to the smaller capitalized stocks. An example of a bad value: finance.yahoo.com - you can see that the somehow for the 31-Dec-04 Quarter are messed up - in fact, the numbers down to the EBIT line don't even add up! Unfortunately, they don't always fail to add up. You can see an example of that in of finance.yahoo.com - someone's obviously made a 'Cost of Revenue' item into a 'Non Recurring' expense in the 31-Dec-04 Quarter. In future, I will probably try and flag such conditions when they can be automatically determined. I have no idea how these errors affect the rating of other stocks within overall scans - the problems is that if you mis-rank one stock with either attribute, it can change the overall ranking of many other stocks.
Another problem was that Net Working Capital (which is usually described as Total Current Assets - Total Current Liabilities), is occasionally negative. That's sometimes good (e.g. you receive cash from your customers faster than you need to pay your suppliers), sometimes bad (i.e. you don't have enough money to pay your suppliers or creditors), sometimes, it's simply a temporary issue. I set Net Working Capital to zero for the purpose of calculating the Return on Capital.
Another question is whether to include 'Intangible Assets' or not in the return on capital calculation. The book says to exclude Goodwill, while the Magic Formula site FAQ says 'Intangible assets are excluded as described in detail in the book'. Yahoo income statements break out 3 lines for this purpose - Goodwill, Intangible Assets, and Accumulated Amortization. Excluding Goodwill is obvious. I decided to include Intangible Assets, as this frequently seems to be more of the capitalized software purchases. I ignored Accumulated Amortization for 2 reasons: 1. It is almost always 0 on Yahoo, and 2. I can't tell what is amortized - goodwill or intangible assets.
With EBIT, I decided to back out non-recurring expenses & other income/expenses net. Usually this was reasonable - other income was very small, and non-recurring was really non-recurring & non-cash. It's not really clear in the book what should be done.
The book mentions determining excess cash and eliminating it from net working capital. I figured excess cash when total current assets are 2 times total current liabilities and allocated any cash available to that. I also used that number when calculating EV (rather than just adding raw cash). In future, I may revisit this calculation & consider cash, short term investments vs. short term or current portion of debt payable, etc. The book doesn't really recommend how this should be determined.
I can see some problems with the data selection when using the 'most recent balance sheet' and 'latest 12 months'. This is what is specified in the book, and I did the same, most recent quarter for balance sheet & summing quarterly returns for EBIT related values. There are some obvious problems with this, for example, if a company assumes a significant amount of debt during the year and the EBIT values for the earlier quarters are not really representative of the full year. Another example of this problem would be a company that has cycles during the year. I can probably flag this condition in the future.
I only filtered out Financial Sector & Health Care Plans. It might also be smart to filter out cyclics, since the ROC at either the years during the high or low points of the cycle would not be representative of the ROC that you expect when you buy. Another example of companies that should be filtered out are 'pseudo' finance companies, for example auto dealers or home builders. These companies are harder to identify. Another example would be the "one or two" big quarter companies - like movie studios or certain other entertainment companies - DWA and MVL comes to mind.
But after all those problems, it does seem like a great screening tool at least. I haven't signed up for the Magic Formula Investing site yet, but I will shortly. For the S&P 500 scan, I'd not have too much trouble with the buying top 20 or 30. And it does bring some very interesting companies that I haven't looked at lately - HRB, SHW, DGX, MAT, LIZ - all of which seem to be those Buffett-like companies that never really get super cheap, but might be cheap enough now.
Regards,
Stew |
