Beware of GAAP

Includes: DIA, QQQ, SPY
by: Marc Gerstein

Watch out for GAAP (Generally Accepted Accounting Principles). It can diminish your investment performance. Seriously.

One of the worst things to come out of the early-decade Wall Street scandals was the tendency it fostered, among frightened investors and crusading journalists to mistrust the adjustments analysts make to reported financials. Now, with the economy moving or preparing to move out of recession, it especially important to put a lid on the politics of accounting and get serious about how one ought to be looking at the numbers.

Who Can We Listen To?

So where does my suspicion of GAAP come from? Aggressive hedge fund operators? No (I don't actually know any.) Vintage-2000 internet-tech-telecom analysts? Not a chance. Warren Buffett? No, but we're starting to get warmer.

Actually, this sentiment comes directly from Graham & Dodd.

A major activity of security analysis is the analysis of financial statements. This analysis includes two steps: First, the financial statements must be adjusted to reflect an analyst's viewpoint, that is, the analyst changes the published numbers, eliminates some assets and liabilities, creates new ones, alters the allocation of expenses to time periods, and, in effect, creates a new set of financial statements. Second, the analyst processes the new information by the calculation of averages, ratios, trends, equations, and other statistical treatment.

Graham and Dodd's Security Analysis (Cottle, Murray, Block, McGraw Hill, 1988), page 133.

They may be able to capture a more faithful picture of reality by adding to or adjusting the in ways not permitted by accounting rules. Analysts make adjustments to accounting information for a variety of reasons. Perhaps the most important is to adjust the numbers to reflect the analysts' notions of income and capital maintenance, where those notions disagree with the notions used in accounting.

Id. at 137.

Analysis requires the ability to discriminate, to separate the ordinary from the unusual, and to detect change. These require disaggregation, enhancement, and reconstruction of accounting information to prepare it for further processing.

Id. at 138.

Analysts vs. Accountants

There's nothing at all wrong with what accountants do. And actually, I tend to like many aspects of GAAP. For example, I'm a rarity insofar as I prefer reported net income, with the associated accruals for things like depreciation and amortization because I like the way it matches revenues and expenses, which is something you don't get with the more fashionable free cash flow numbers, which can conceal as much as they reveal.

But investors cannot accept GAAP blindly. They and accountants pursue different goals.

Accountants track and explain what a company did in the past and, hopefully, present a good view of where it stands in the present. This comes under the heading of "disclosure" and is critical: Nobody can form a rational opinion about a company absent accurate information regarding what it has done and where it now stands.

Investors are not interested in disclosure per se. They assume this has been done correctly by the accountants and hope regulators and prosecutors will step in if that's not the case. (Hence the importance of accuracy in accounting as summarized by the appropriate certifications.) Instead, investors use the disclosed numbers as a starting point, as raw material for their work; "forecasting." They must look to the future and want to know where the company is going, because that's the way one makes money in the stock market.

Obviously, the future is unknowable so investors can't rely on objective facts. Instead, they are left to develop reasonable expectations and often, reasonableness can be assessed by the manner in which forecasts mesh with relevant portions of the financial statements. Hence when an investor examines financials, he or she must consider which aspects of the past are most likely to hold clues about the future.

This is why Graham & Dodd say analysts may (should!) adjust or discard certain portions of indisputably accurate statements. Accountants seek accurate depictions of the past and present. Investors and analysts seek relevant clues about the future.

Table 1 offers an illustration, a hypothetical set of income-statement items.

click to enlarge images

Table 1

Click to enlarge

What is this company's three-year net-income growth rate?

The major financial databases using GAAP accounting information would report the growth rate as negative 24.9%, based on the deterioration from $360.53 to $152.68. However, an analyst would note that the plant closing expenses and the gains or losses from sales of subsidiaries are not ordinary recurring parts of the business. Factoring these out, the analyst would assume a plus 4.64% growth rate. Which do you think provides a more realistic starting point form which you might develop expectations regarding the future?

What are the company's operating margins?

Financial databases, picking up the operating income numbers reported in the financial statements, would report the following.

Table 2

Year 1

Year 2

Year 3

Year 4





Click to enlarge

How might an observer interpret such data? Can we use this to help articulate rational assumptions about the future?

We cannot. The trend appears almost random.

But an analyst would recalculate operating income by excluding plant closing expense and gains or losses from sales of subsidiaries and come up with the following:

Table 3

Year 1

Year 2

Year 3

Year 4





Click to enlarge

This is much more informative. Rather than a seemingly random set of numbers, we see a slow trend of deterioration. An analyst seeking to project an operating margin for Year 5 now knows that he or she should start with a number that is less than 19%, perhaps something in the 17%-18% range. This won't be the final word. The analyst would want to look into reasons for the deterioration we've seen and determine whether it's likely to continue, reverse, or even accelerate. But at least we now have a good starting point for analysis. And an analyst who would project, say 20%, knows he or she will be grilled by a supervisory analyst or a client. (Actually, a written research report should explain the reason for an anticipated turnaround. This is how an analyst can zero in on issues likely to make or break a stock.) The pure accounting numbers, on the other hand, left us with accurate information about the past, but contained noting that could help us develop expectations about the future.

None of this is to say an analyst would ignore the non-recurring items. There are many ways analysts can assess the preponderance of such events and reward or penalize companies for their scarcity or prevalence. But however that's done, it needs to be separate from such basic matters as growth rates and margins.

A Front-Burner Issue

We've been through a horrendous period for the economy, one which produces a larger-than-usual helping of non-recurring items. The stock market as a whole is well aware of this and adjusts accordingly.

Figure 1 shows the results of a backtest (4-week rebalancing) of a simple screen that selected all companies for which the trailing 12 months (TTM) EPS growth rate was above zero. These are the GAAP numbers reported by the standard financial databases (the same sorts of numbers you see when you go anywhere on the web to check up on how companies are doing).

Figure 1
Click to enlarge

That doesn't seem earth-shattering. We learn that shares of companies whose GAAP EPS over the past 12 months was above breakeven closely kept pace with the stock market's recent rally.

Now look at Figure 3, which changes the screen such as to identify companies whose TTM EPS growth rate was minus 25% or worse.

Figure 2
Click to enlarge

There are plenty finger-waggers out there who will only be too happy to explain what's going on. Risk aversion is back, they'll say. Investors are bottom-fishing for garbage and if we don't watch out, we're all going to pay the piper!

That sort of explanation offers considerable emotional appeal to the many who've been burned badly by the bear market, especially those who sold at or near the bottom. However, it's wrong.

Figure 3 shows the results of a backtest based on a screen seeking to identify companies whose growth in TTM Business Income (I'll talk about this below) was above zero.

Figure 3
Click to enlarge

Figure 2 clearly shows that the market was completely disinterested in the substantial recent EPS shrinkage. But contrary to the cries of the naysayers, it seems equally clear that the market was very much interested in something else. (Of the 1,024 stocks that made the EPS shrinkage screen, 668 also made the growing Business Income screen.) This isn't to say Business Income was the only relevant factor. I assume there are many others that could be identified if I continued to build more exploratory screens like this.

But we can definitely see that the stocks that led the rally are not reflective of a mindless bottom-fishing quest for garbage. It's more likely that we've been seeing some good-old Graham-Dodd security analysis.

What Is "Business Income?"

This is a phrase I made up as a label for a number computed as follows: Sales minus Cost of Goods Sold minus Selling, General & Administrative Expense

Actually, it feels weird that I had to make up a label for this sort of thing. When I learned security analysis in the late 1970 and early 1980s, we referred to it as Operating Income. Unfortunately, however, that label has morphed into a hodgepodge that covers this as well as all sorts of unusual and non-recurring items that vary from company to company. And unfortunately, all this financial noise finds its way into pretax income, net income, and EPS too.

Business Income, as I have defined it, isn't the be all and end all. We do have to deal with interest and taxes. I've been experimenting with another item, which I refer to as Core Income, to address that. The backtest result for that a screen using it is pretty much the same as what we see in Figure 3.

This will never be foolproof. for example some companies put odd items into cost of sales or SG&A, leaving me no automated way to pull them out. Realistically, I can't even expect any automated algorithm to match what I might see if I were able to dig into each set of financials one at a time. But even the approximations I have seem to add analytic information missing from the official GAAP EPS numbers.

Going Forward

Generally, separation between recurring and unusual is always important. But now, we need to be even more sensitive to this issue. The financial crisis and recession produced more usual financial-statement items than usual, and items of larger magnitude. This distorts recent comparisons, longer-term growth rates and many valuation ratios.

It's important to reiterate what these adjustments are intended to do. I'm not trying to ignore the past and pretend the companies didn't experience the problems that did, in fact, occur. What I am trying to do is have my models evaluate companies based only on those aspects of the past that are most likely to persist into the future.

Remember the difference between accounting and analysis: I'm interested in forecasting, not disclosure. The latter is certainly critical, but it's not what I do when I create and run models to help me identify investment opportunities.

This is not just a short-term issue relevant only to an initial post-recession period. Better analysis works at all times. Even during good periods, when unusuals are less likely to be troublesome, use of streamlined computations described here can still be better, or at least no-worse than neutral. Table 4 summarizes backtest results covering 6/30/03-6/30/07.

Table 4

S&P 500
grew to

grew to

Average % change
in up periods
relative to S&P 500

Average % change
in down periods
relative to S&P 500

% Growth above zero





TTM Buss. Inc.
% Growth above zero





% Growth above zero





Click to enlarge

To reiterate another point made above, I have no intention of turning a blind eye toward companies that make too much of a habit of experiencing unusuals. These tell their own stories and they are important. For example, if a company takes huge charges all the time for closing businesses, I want to know it.

But I can accomplish that without merging the information into recurring results. For example, I might rank companies based on some sort of noise factor; i.e., noise (defined as the absolute value of business income minus operating income) divided by business income. I've had a bit of success with this sort of thing, but am still in the early stages of refining such modeling factors. I suspect I wouldn't want to get carried away with oddball periods every now and then but would be more interested in the persistence of noteworthy noise.

I'll also be working the cleaner income numbers into valuation ratios. For example one ranking system that's worked pretty well for me dropped P/E and substituted and a version of earnings yield wherein I divide Business Income by Market Capitalization.

Disclosures: None