Tuesday, January 24, 2012

Evolving thoughts: My current quantitative stock selection process

Now that the holidays are over and I have successfully passed the Level 1 of the CFA exam, I’ve gone back to one of my current investment projects. I’ve been working on developing and refining my stock selection method, specifically for non-financial companies (financial companies have distinctly unique financial statements so they won’t be covered here). Forecasting is beyond my means and, gathering from my readings on the topic from James Montier to David Dreman, probably more harmful than anything. In any case, few people can do it reliably, myself not included. In that regards, I’ve decided to focus squarely on the published financials of companies to identify and value opportunities. Like this article by Benjamin Graham points out, as well as various other observations and research from Graham again to Joseph Piotroski to Tweedy Browne Company, it is entirely possible to have objective and somewhat simple stock selection criteria based on the publicly disclosed financial statements while obtaining more than satisfactory results, provided the criteria are logical both in theory and in practice.

My goal is to identify companies that are both safe and cheap with good upside potential. The three are in general linked to each other. Bought at a cheap enough price, most companies can be a good investment opportunity with little downside. Plus, minimizing your chance of a loss is actually a precursor to achieving good returns. I’ve decided to look into four general areas of criteria: 1) financial position 2) improving fundamentals 3) valuation 4) quality. The first two areas are fairly non-controversial, though at the same time I’m sometimes surprised at how little is paid attention to them. First off, I initially screen for adequate financial position, to be more precise I look for a leverage ratio of less than about 2.5 (meaning a company has a capital structure of at least 40% equity, or conversely no more than 60% liabilities) and a current ratio (current assets to current liabilities) of about at least 1. For smaller companies, I tighten the criteria to a leverage ratio of less than 2 and a current ratio of at least 2. The next area I look at is the Piotroski F score. The F score is a convenient summary of 9 fundamental indicators. While it is a quantitative measure, there are no complicated calculations; it is instead more of a checklist of observations. To summarize the F score, it consists of the following tests:

1) If Net Income > 0, score 1
2) If Cash Flow from Operations > 0, score 1
3) If Cash Flow from Operations > Net income, score 1
4) If Return on Assets (ROA) in the current period > ROA previous period, score 1
5) If ratio of Long-term debt in the current period < ratio previous period, score 1 6) If Current Ratio current period > Current Ratio previous period, score 1
7) If number of shares outstanding in the current period < shares outstanding previous period, score 1 8) If Gross margin in the current period > Gross margin previous period, score 1
9) If Asset turnover in the current period > Asset turnover previous period, score 1

Any test that is not met gets a score of 0. The scores are added up for a maximum possible score of 9, so companies with such a rating basically have all the fundamental indicators pointing in the right direction. Some of these criteria can probably be changed as one sees fit and the F score doesn’t explicitly purport to be the be-all, end-all of fundamental summaries. Nonetheless the original observations have been proven to work very well when combined with a low valuation with the aim to avoid likely value traps.

Which leads me into my next area of selection: valuation. Not surprisingly this is probably the area where I focus the most on. In the past I would have at least taken a look at the trailing 12 month price-to-earnings (PE) ratio, or at least the PE for the most recently completed yearly period. Fairly early on in my investment practice, I had already gotten into the habit of virtually ignoring forward PE ratios. It wasn’t uncommon for me to look at stock that was supposedly trading at a seemingly “reasonable” 12 times forward earnings (that is, next year’s estimated earnings) but then I would realize, after taking a quick look at the data, that it was trading at more around 20-25 times last year’s earnings, often quite a bit more, most likely implying that the market was anticipating some bumper jump in earnings. Forward earnings are far more subjective than anything else; it depends on whatever assumptions are put into the estimation. Basically, any prediction, provided that it is optimistic enough, can yield a reasonable-looking forward PE. If a large enough amount of predictions were at least vaguely accurate, this wouldn’t be an issue but since a number of studies point out that forecasts aren’t usually accurate by even generous standards, that means the forward PE probably isn’t worth considering in most cases. One thing you can do with it though is gauge what the market seems to expect from the company and whether or not those expectations are reasonable. Another aspect to consider is that the trailing 12 months PE might be somewhat misleading for cyclical companies. For instance, in mid-2008, you could have bought shares in Posco, a Korean steelmaking company with a healthy balance sheet and decent returns, at over around $120 per share, representing about 9-10 times its 2007 earnings, which would initially look cheap. However that year mostly reflected a high point in the economic cycle. While the company remained profitable, it fell to less than $60 in late 2008 and early 2009 in the great economic crisis. To smooth out the effects of the economic cycle, Graham and Dodd proposed simply taking an average of previous years’ earnings. Taking at least 5 years average, and preferably 10 years, this should be able to encompass both boom and bust periods. Applying this to Posco would have yielded a multiplier well north of 15 times its average earnings, its highest for several years. Therefore I’ve incorporated multiyear averages of earnings into my screening.

Another idea I’ve incorporated for valuation is using Free Cash Flow instead of regular Earnings. Normally this is calculated from the financial statements as Cash Flow from Operations minus Capital Expenditures. Free Cash Flow represents the actual cash that goes to the company, as opposed to using earnings which in some ways is just an accounting figure with little indication of how much cash actually goes to the company. Free Cash Flow represents the amount left over that the company could use to, say, repurchase stock and pay dividends. Also cash flow numbers have the advantage of being less manipulatable than earnings. While thus superior to earnings as a representation of what the company has generated, Free Cash Flow, if non-adjusted, still leaves something to be desired. The capital expenditures number is basically all the capital expenditures, whether it is spent to grow the company or simply maintain its current business level. As Fairholme’s Bruce Berkowitz points out in the latest edition of Security Analysis (one book that I’m currently reading while taking copious notes), a more appropriate definition of free cash flow would be the amount of cash generated that is left over for the company to invest in growth, repurchase stock or repay shareholders. This takes away expenses needed to maintain operations at its current level. Unfortunately, this also happens to be very hard to calculate. Few companies provide detailed information on which spending is for growth and which is for business maintenance. Fortunately, as this article on the blog MagicDiligence points out, there is a way to go around that, by estimating the maintenance expenditure as the depreciation and amortization cash charges in the cash flow statement. Depreciation and amortization represent that the expensing of tangible and intangible assets over their estimated useful life. Since those assets will have to eventually be replaced it is thus entirely feasible to use depreciation and amortization as proxies for maintenance expenditures. In short, Free Cash Flow, as suggested in that article, can be estimated as Cash Flow from Operations minus Amortization & Depreciation (summarized as “Sensible Free Cash Flow” or SFCF). It can be somewhat crude but the general direction and logic is sound in my view. So putting that together with what was mentioned, my current favorite tool for putting a valuation on a company is calculating the ratio of its price compared to its multiyear average of Sensible Free Cash Flow, calculated over a period of at least 5 years, preferably 10 where available.

Finally, in the fourth major aspect of my selection, I decided to look at the quality of the company. In theory, quality companies will deliver superior investment results. As a short aside, in practice, any commitment regardless of quality can produce great results if bought at a low enough price and with enough research, but I will save this topic for another time to explore. This is partly out of convenience because this would take a whole other article, but also because if a couple of methods work well enough for you, you can probably just stick with them for the most part.

Returning to our topic, when quality has been mentioned as being identifiable through the financial numbers, it has often been interpreted that some measure of return on capital is the proxy. One measure that is seen very often is Return on Invested Capital. There are probably a few variations on the calculation but for all intents and purpose the one I am referring to is Net Operating Profit After Tax / (Short-term debt + Long-term liabilities + Equity). Basically, the Invested Capital should be the capital used to acquire the company’s income-generating assets. Starting with this calculation of ROIC, I have modified it a bit along the same lines as I explore the earnings or income issue as mentioned above. Net Operation Profit After Tax is replaced by Free Cash Flow (after depreciation and amortization) as that should better reflect the cash the company can use. I also look at the 5-year average, which should give a better indication of whether this is a quality enterprise, rather than an unusually high or low return on one specific year.

To summarize my four current stock selection criteria, I look at the financial statements for a combination of:
1) Strong financial position, measured by high current ratio and low financial leverage ratio
2) Improving fundamentals, measured by the Piotroski F score
3) Low valuation, measured by low ratio of price to average SFCF
4) Quality, defined by high average SFCF/IC

I’ve collected and processed the financial data for approximately 70 companies so far, with all the caveats that this could entail. A cursory glance at the results would seem to vindicate such a selection on strictly financial statement data. I’ll eventually be going through them more in detail but so far, they seem very robust.