Earnings Quality Analysis and Equity Valuation Richard G. Sloan Professor of Accounting Ross School of Business, University of Michigan Ann Arbor, Michigan

For years, a widely held belief was that markets are efficient and that fundamental analysis is futile. But after the discovery of several market anomalies, the stage was set for exploring the use of fundamentally based techniques. A simple quantitative trading strategy based on the quality of earnings analysis has worked remarkably well. And although this particular opportunity will inevitably be arbitraged away, it does support the concept that fundamental analysis works.

I

have long been interested in fundamental analysis, but I was originally told that markets are efficient and thus it is futile to engage in fundamental analysis. Then, academics started to discover what seemed to be violations of market efficiency, such as the size effect and the January effect. These violations struck me as being fairly technical and mechanical. Thus, the door had been opened to investigate whether fundamentally based techniques might work. I began by looking at a simple quantitative screen based on the quality of earnings analysis (discussed more fully later). When I first back tested this screen against U.S. equity data, I thought I must have made a mistake. I obtained what could be described only as hedge fund returns. Returns averaged about 10 percent a year above the market over the 30 years tested; in only 2 of the years were the returns less than the market. Predictable returns of that magnitude just had not been documented before; returns from strategies based on the size effect and the January effect were a lot smaller. Interestingly, in the 10 years that have passed since I did the original research, this screen has continued to work well. In fact, this technique has been refined to obtain even bigger returns. Needless to say, it has generated a great deal of interest in the quantitative investment community.

This presentation comes from the Equity Research and Valuation Techniques conference held in Boston on 1–2 December 2005.

52 x SEPTEMBER 2006

To me, this technique demonstrates the viability of the concept that simple fundamental analysis combined with mechanical trading rules based on quality of earnings analysis works. Moreover, I believe that this is just the tip of the iceberg in terms of what good fundamental analysis can do. Although over time simple quantitative trading rules will get arbitraged away, the important point is that gains may always be found by digging deeper—by understanding the accounting and the company’s business model and how they fit together.

Why Analyze Earnings Quality? A natural question to ask is why earnings quality should be analyzed. Certainly, everyone is aware that earnings numbers can be manipulated. Thus, some practitioners have taken the position that they will focus only on the cash flows and perform their intrinsic value analysis based on those numbers. Earnings, and the underlying subjective assumptions, are irrelevant. I think earnings quality analysis works so well because, despite the shortcomings of earnings, the bulk of the market is looking at earnings. If you had a crystal ball that allowed you to look at the value of one variable (other than stock price) one year or one quarter from now, your best pick would be earnings. By comparing that value with the current consensus expectations of earnings, you could generate a powerful trading strategy. Even if your crystal ball told you the true intrinsic value of a stock, knowing that

©2006, CFA Institute x cfa pubs .org

Earnings Quality Analysis and Equity Valuation

number would not necessarily help you generate trading strategies in the short run because it could take many years, if ever, for the stock price to revert to that intrinsic value. Knowing future earnings is better than knowing intrinsic value for those interested in forecasting stock returns over the next 3–12 months. Targeting earnings quality is also important because it forces the investor to focus on what I call the “continuous evolution” of business operations and accounting assumptions. The right accounting for a company depends critically on the business it is engaged in, the way it is conducting its business, and the things it is doing relative to its competitors. For example, consider Krispy Kreme Doughnuts, whose stock has taken significant losses in the past few years. Krispy Kreme invested heavily in doughnut making equipment and used the straight-line method to depreciate this equipment over a 10–15 year useful life. For Krispy Kreme, this was the wrong way to depreciate. I do not have any qualms with the useful life in terms of how long the machines will physically make doughnuts. But to me, this product seemed like a fad. Remember that when Krispy Kreme first opened new stores, people would form lines outside to purchase the doughnuts. Krispy Kreme would broadcast this phenomenon in press releases, and in the early days, the stock price would go higher. But clearly, it was going to wear off fairly quickly. Certainly, people would not be forming lines outside two or three years after a store opened, which is what happened. Krispy Kreme would never have booked a profit had it used an accounting policy that frontloaded the depreciation and matched it to the cash flows that the doughnut machines generated. Following is another example. Netflix, which rents DVDs through the mail, seems to be doing quite well. It is giving Blockbuster tough competition. But most of its profits over the past 18 months have come from extending the depreciation period from one year to three years for its DVD library, what Netflix calls its “back catalog”—the DVDs that are not new releases. Netflix has done so because it thinks these DVDs will physically last a long time and will be in demand three years from now. This practice looks similar to what AOL did with its subscriber acquisition costs in the late 1990s, which it eventually had to write down. One could make the argument that Netflix faces a similar situation. Three years from now, if video on demand and internet downloading of movies do indeed become available, little demand will exist for Netflix’s mail-order service. Thus, getting the accounting right involves understanding the business, which is why no quantitative analyst will ever be able to come up with ©2006, CFA Institute x cfa pubs .org

a mechanical formula to screen stocks. A fundamental analyst will always be needed to understand the business and look at the accounting.

Impact of Earnings Surprises As is widely known, stock prices do indeed respond to earnings surprises, as shown in Figure 1. Quarterly abnormal returns are plotted on the vertical axis as a function of quarterly forecast error (realized earnings versus the I/B/E/S consensus forecast adjusted by the stock price) for a large sample of companies. The result is an S-shaped nonlinear curve. For highgrowth stocks, a 1 percent earnings surprise as a percentage of price triggers a 12 percent increase on the upside and a 15 percent decrease on the downside. If one could perfectly forecast these earnings surprises, one could generate a 27 percent quarterly hedge portfolio return. Earnings quality analysis is powerful because it is very good at forecasting these earnings surprises. I would like to emphasize the asymmetric nature of this graph for high-growth stocks. Earnings quality analysis is most useful in picking “earnings torpedoes” in growth stocks (i.e., growth stocks that are trading at high valuations but are doing it through the use of aggressive accounting). These valuations get addictive to the managers of these companies. They are great when managers are raising capital. They are also great when managers have stock options in the company, so they tend to be the ones where management pushes the envelope with respect to the accounting assumptions. When the company is finally forced to take a write-down, the price crashes, even if the company just misses the consensus by a small amount. Thus, analyzing growth stocks is one area where these techniques are particularly helpful.

Defining Earnings Quality To develop a technique for assessing earnings quality, I first had to understand what accounting would look like in a “perfect” world. In my view, the way to think of a company is as a portfolio of investment projects. If I, as a company manager, knew the future cash flows of these investment projects, I could compute the internal rate of return (IRR). Rather than discounting the cash flows at an appropriate discount rate, I could estimate the implied rate of return based on the investment required in these projects and the future cash flows they will pay out. In my perfect accounting system, I would then set book value equal to invested capital and the accounting rate of return would equal the IRR. SEPTEMBER 2006 x 53

CFA Institute Conference Proceedings Quarterly

Figure 1.

Earnings Surprise Response Functions for High-Growth and Low-Growth Stocks: Quarterly Abnormal Returns vs. Quarterly Forecast Error

Abnormal Return 0.20 0.15

High Growth

0.10 0.05

Low Growth

0 −0.05 −0.10 −0.15 −0.20 −0.07

−0.05

−0.03

−0.01

0

0.01

0.03

0.05

0.07

Forecast Error

Simple Accruals Example An illustration will show how accruals can be used to discover poor accounting. This simple example begins with an investment opportunity, initially just $100. I am not going to grow my investments, so I will invest only $100 every year thereafter. And I will track this investment for five years. This investment only generates a return one period from now. The return comes in the form of cash revenues equal to 165 percent of the investment and a cash cost equal to 55 percent. So, I invest $100 cash today. I then get $110 at the end of the period, and my investment expires. Thus, this investment generates a return of capital plus a 10 percent return in the next period, and so the IRR on this investment project is 10 percent. The correct accounting should be obvious. Clearly, I should capitalize the whole amount in the period I make the investment and then expense the whole amount in the next period. If I do so, invested capital is equal to book value and the accounting rate of return is equal to my IRR of 10 percent. I will now show how, in turn, conservative and aggressive accounting assumptions alter the results. In my conservative accounting scenario, I initially capitalize only 80 percent of the investment. Ironically, this bad accounting is what the Financial Accounting Standards Board requires that companies do for all R&D (and marketing) expenditures. This is a real investment that generates future benefits, but I have to expense it all immediately. On the flip side, I have the aggressive accounting scenario where I capitalize too many costs—120 percent of the investment. In the real world, I would do this by capitalizing some operating costs, which is exactly what WorldCom did. 54 x SEPTEMBER 2006

Table 1 shows the neutral accounting scenario, the “perfect” accounting scenario. The investment is $100 each year. I capitalize 100 percent of that investment and then expense it in the income statement in the very next year as an amortization expense. Starting in Year 2, I get an accounting profit equal to $10, which is equal to the economic profit. The net operating assets (NOA)—operating assets minus operating liabilities—equal the capitalized assets. Each year, I amortize the previous year’s investment and capitalize the new year’s expenditure. The accruals, the key to this analysis, are just the change in the NOA. After the first year, the investment hits a steady state and the accruals are always zero. So, the company is in a steady state with respect to growth and has a constant business model. The accruals should be zero if the accounting is perfect. The statement of cash flows is at the bottom of the table. It shows that the free cash flow (FCF) after the first period is always $10. (Another way to obtain accruals is to take the difference between earnings and FCF.) High accruals are indicative of a company with potentially high earnings but no cash flow to back up those earnings. When the accounting is perfect and the firm is in steady state, there are no accruals. Table 2 shows the conservative accounting scenario. Note that the conservative accounting assumptions are used only in the third year and are reversed in the following year. Because I am failing to capitalize $20 of the investment, operating income drops by $20 and I report operating income of –$10, which is to be expected. But look what happens in the very next year, Year 4, when the conservative assumptions are reversed. I have to amortize only $80 instead of ©2006, CFA Institute x cfa pubs .org

Earnings Quality Analysis and Equity Valuation

Table 1.

Financials for Neutral Scenario Year

Item

1

2

Growth rate in investment

3

0.0%

4 0.0%

5 0.0%

0.0%

Investment

$100

$100

$100

$100

$100

Capitalized operating costs

$

$

$

$

$

Total Sales

0

0

0

0

0

100

100

100

100

100

$100

$100

$100

$100

$100

$

Capitalized investment costs

0

$165

$165

$165

$165

Operating expense

0

55

55

55

55

Investment expense

0

0

0

0

0

Amortization expense

0

100

100

100

100

0

$ 10

$ 10

$ 10

$ 10

$100

$100

$100

$100

$100

100

0

0

0

0

Net operating income Net operating assets (NOA) Accruals

$

Return on beginning net operating assets (RNOA) Cash inflows

10.0% $

10.0%

10.0%

0

$165

$165

$165

100

155

155

155

155

–$100

$ 10

$ 10

$ 10

$ 10

Cash outflows Free cash flow (FCF)

10.0%

$165

Note: In neutral accounting, 100 percent of investment costs are capitalized in Year 3.

$100. As a result, I get $30 of net income, which directly affects return on net operating assets (RNOA). In Year 3, the RNOA is –10 percent, and then in Year 4, it is 37.5 percent. Thus, conservative accounting drives the accounting rate of return down in the year that the conservative accounting is Table 2.

applied, and then in the subsequent year, it has the opposite effect and the accounting rate of return goes up. Therefore, if I can pick a company that is using conservative accounting, I know that in the subsequent year its earnings will increase, which is why earnings quality analysis is useful.

Financials for Conservative Scenario Year

Item

1

2

Growth rate in investment

3 0.0%

4 0.0%

5 0.0%

0.0%

Investment

$100

$100

$100

$100

$100

Capitalized operating costs

$

$

$

$

$

Total Sales

0

0

0

0

0

100

100

80

100

100

$100

$100

$ 80

$100

$100

$

Capitalized investment costs

0

$165

$165

$165

$165

Operating expense

0

55

55

55

55

Investment expense

0

0

20

0

0

Amortization expense

0

100

100

80

100

0

$ 10

–$ 10

$ 30

$ 10

$100

$100

$ 80

$100

$100

100

0

–20

20

0

–10.0%

37.5%

Net operating income Net operating assets (NOA)

$

Accruals Return on beginning net operating assets (RNOA) Cash inflows Cash outflows Free cash flow (FCF)

10.0% $

10.0%

0

$165

$165

$165

100

155

155

155

$165 155

–$100

$ 10

$ 10

$ 10

$ 10

Note: In conservative accounting, 80 percent of investment costs are capitalized in Year 3.

©2006, CFA Institute x cfa pubs .org

SEPTEMBER 2006 x 55

CFA Institute Conference Proceedings Quarterly

Aggressive accounting, shown in Table 3, is the flip side of conservative accounting. In this case, I am capitalizing too much in Year 3—$20 of my operating costs. So, my operating costs drop from $55 to $35. Compared with the neutral scenario, my income becomes $30 instead of $10, and as a result, I overstate my income on my accounting rate of return in Year 3. In Year 4 when it reverses, I have to expense through amortization the extra $20. As a result, net operating income becomes –$10, the mirror image of the conservative scenario. That is, at first earnings are too high, and then predictably, earnings drop in the next period. Figure 2 graphically illustrates these three accounting scenarios. Therefore, by going out and looking at whether accruals are positive (signaling aggressive accounting) or negative (signaling conservative accounting), I can predict earnings reversals in the next period, which is all there is to my simple method. If the world were this simple—if companies did not grow and their business models stayed constant— this analysis would lead to an easy strategy. Analysts would just have to look at the size of the accruals for a perfect measure of earnings quality. Unfortunately, the world is not that simple. Legitimate growth in investment will also obviously cause the amounts capitalized on the balance sheet to increase and cause accruals to be positive. Conversely, with negative growth, accruals will be negative. For example, Figure 3 shows three growth scenarios—that is, what happens if I increase my Table 3.

investment from $100 to $200. My capital assets, of course, go from $100 to $200. If the return on my investments stays constant, a change in accruals does not create any predictable changes in RNOA. So, I have to be able to discriminate earnings management or bad accruals from legitimate growth in the underlying business. Keep in mind that if a company legitimately changes its business model or if the competitive landscape changes, these changes can affect accruals in unpredictable ways, which is another example why this analysis can never be automated.

Decomposing Earnings The idea of earnings decomposition comes directly from the 1934 edition of Benjamin Graham and David Dodd’s Security Analysis. Broadly, earnings can be decomposed into the change in book value (retained earnings) plus the dividend, which I define broadly as the sum of cash dividends, including any cash paid out for stock repurchases and less any cash brought in through a capital transaction, such as a secondary equity offering. The change in the book value can be further broken down. Remember that the change in book value is equal to the change in assets less the change in liabilities—or the change in NOA (accruals) plus the change in cash holdings less the change in debt. As a result, earnings can be decomposed into the change in NOA (accruals) plus FCF. The result is two pieces to earnings: a hard piece, the FCF, and a soft

Financials for Aggressive Scenario Year

Item

1

2

Growth rate in investment

3 0.0%

4 0.0%

5 0.0%

0.0%

Investment

$100

$100

$100

$100

$100

Capitalized operating costs

$

$

$

$

0

$ 20

100

100

100

100

100

$100

$100

$120

$100

$100

$

Capitalized investment costs Total Sales

0

0

0

0

$165

$165

$165

$165

Operating expense

0

55

35

55

55

Investment expense

0

0

0

0

0

Amortization expense

0

100

100

120

100

0

$ 10

$ 30

–$ 10

$ 10

$100

$100

$120

$100

$100

100

0

20

–20

0

Net operating income Net operating assets (NOA) Accruals

$

Return on beginning net operating assets (RNOA) Cash inflows Cash outflows Free cash flow (FCF)

10.0% $

30.0%

–8.3%

10.0%

0

$165

$165

$165

100

155

155

155

$165 155

–$100

$ 10

$ 10

$ 10

$ 10

Note: In aggressive accounting, 120 percent of investment costs are capitalized in Year 3.

56 x SEPTEMBER 2006

©2006, CFA Institute x cfa pubs .org

Earnings Quality Analysis and Equity Valuation

Figure 2.

Three Accounting Scenarios

A. Accruals Accruals ($) 30 20 10 0 −10 −20 −30 2

3

4

5

Period

B. RNOA RNOA (%)

accruals had a gradual run-up in earnings in the previous years. The metric plateaued in the high year and then collapsed in the following year. Thus, companies with high accruals today have predictably much lower earnings tomorrow; conversely, companies with low accruals today have predictably higher earnings tomorrow. Just as suggested by the example in Panel B of Figure 2, this metric works well in predicting future earnings changes. But, of course, the important question from an investment perspective is whether the market sees these predictable earnings changes coming (i.e., is the market indeed efficient)? Figure 5 plots stock returns relative to the accrual ranking year. As the graph shows, the market does not seem to see these changes coming. High-accrual stocks performed well in the years leading up to the high-accrual year. In the high-accrual year, their performance was average or slightly better. Then, they

40 30

Figure 3.

Neutral Accounting with Three Growth Scenarios

20

A. Accruals

10

Accruals ($)

0 −10 −20 2

3

4

5

Period Aggressive

Neutral

Conservative

piece, the change in NOA, which as shown earlier results from the accounting assumptions made (this is also the area where the accounting envelope tends to be pushed). Thus, I look at the aggregate magnitude of the accruals, the change in NOA, to evaluate the quality of earnings. More specifically, my rule of thumb is that if the change in NOA divided by the average level of NOA exceeds 5 percent, a red flag goes up. Enron Corporation and WorldCom were way above 5 percent for the two years before their situation became public knowledge. This simple metric would have picked up both of them. To further demonstrate the usefulness of this metric, using 30 years of historical data I selected the top 10 percent and the bottom 10 percent for this accrual measure. (Year 0 marks the year that companies fall into the extreme 10 percent.) From this sample, I formed a high-accrual portfolio and a low-accrual portfolio. As Figure 4 shows, companies with high ©2006, CFA Institute x cfa pubs .org

25 20 15 10 5 0 −5 −10 −15 −20 −25 2

3

4

5

Period

B. RNOA RNOA (%) 40 30 20 10 0 −10 −20 2

3

4

5

Period Positive Growth

No Growth

Negative Growth

SEPTEMBER 2006 x 57

CFA Institute Conference Proceedings Quarterly

Figure 4.

Accruals and Accounting Rates of Return, 1972–2002

Figure 5. Annual Size-Adjusted Returns for Extreme-Accrual Portfolios, 1972–2002

RNOA

Annual Size-Adjusted Return (%)

0.20

30

High Accruals

20

0.15

10 High-Accrual Portfolio 0

−10

0.10

Low Accruals

−20

Low-Accrual Portfolio

−30

0.05 −5

−4

−3

−2

−1

0

1

2

3

4

5

−5

−4

−3

did predictably badly. On average, they underperformed the market by roughly 600 bps in the following year and by about 300 bps and 100 bps in the following two and three years, respectively, because some of these accruals can take more than one year to reverse. Low-accrual companies, of course, demonstrate the opposite behavior. Their earnings jumped, and their stock prices increased in response. The spread between the two lines in the figure is the predictable hedge portfolio return. Clearly, stock prices were acting as though investors did not see these predictable earnings changes coming. This graph presents convincing evidence regarding market efficiency. Moreover, this “anomaly” has persisted for the 10 years since I first documented it. As a check on this metric, I gathered a sample of about 100 companies that had been targeted by the U.S. SEC for manipulating earnings upwards. Year 0 is the year that the SEC alleged that these earnings abuses occurred, not the year that the SEC announced the enforcement action, which typically did not occur until several years after the abuses became common knowledge. Figure 6 shows that the accruals peaked in Year 0 and then plummeted in the following year. The thin solid line in this figure mirrors the solid line in Panel A of Figure 2 for aggressive accounting almost perfectly. Thus, an accruals analysis picks up incidences of aggressive accounting. I want to emphasize that rising accruals can do no more than raise a red flag. For example, if a company’s inventory is rising, that could be a bad sign of 58 x SEPTEMBER 2006

−2

−1

0

1

2

3

4

5

Year Relative to Accrual Ranking Year

Event Year

obsolete inventory or it could be a good sign because the company is growing. Perhaps the clearest way to help disentangle these two possibilities is to analyze accounts payable. If inventory is rising to meet increasing sales, then accounts payable should be rising accordingly. When inventories are rising but payables are not increasing at the same rate, an earnings quality problem caused by obsolete inventory is most likely occurring. In this way, the liability side of Figure 6.

Accruals and SEC Enforcement Actions, 1987–2002

Accruals 0.40 Mean

0.35 0.30 0.25

Median

0.20

Average for All Other Companies

0.15 0.10 0.05 0 −0.05 −0.10 −5

−4

−3

−2

−1

0

1

2

3

4

5

Year Relative to the SEC Enforcement Action Note: Accruals are measured by the annual change in NOA divided by the beginning level of NOA.

©2006, CFA Institute x cfa pubs .org

Earnings Quality Analysis and Equity Valuation

the balance sheet can be used to help judge the quality of the asset side of the balance sheet, where most of the earnings quality problems arise.

Conclusion The simple technique of earnings quality analysis shown here would have done wonderfully over the past half century in terms of generating superior stock returns. Although this quantitative method has

©2006, CFA Institute x cfa pubs .org

worked well, it will probably get arbitraged away. Still, I believe that using good fundamental analysis to detect accounting distortions by understanding the accounting and the company’s strategy and how they fit together will always be an incredibly important source of value added for the investment management community. This article qualifies for 0.5 PD credits.

SEPTEMBER 2006 x 59

CFA Institute Conference Proceedings Quarterly

Question and Answer Session Richard G. Sloan Question: In the world of insurance companies and banks, half their books are accruals. Can this type of analysis be used there? Sloan: They do have assets and liabilities, but it is more difficult to distinguish what I call “proprietary assets” and “proprietary liabilities” from the marketable securities and debt that represent a standard source of financing, such as bonds or commercial paper issued by a bank. What you must do is identify the proprietary assets and liabilities and focus on them. For insurance companies, you must focus on the spread between the reinsurance receivables and the reinsurance payables and also the liability for claims reserves; these are the proprietary assets and liabilities. One of the important areas we’re focusing on lately concerns companies in the financial sector that are issuing loans, particularly residential loans and consumer loans, and then securitizing and selling them while holding a residual interest in these loans. These residual interests hold most of the risk because they are the first not to be paid in the event of a default. They are a small chunk of the balance sheets of these companies because the residual interests are small in dollar magnitude, but the risk is enormous. So, you could take $1 billion of receivables, hold a residual interest in $20 million, and basically have the risk of $1 billion of loans concentrated in $20 million of receivables.

60 x SEPTEMBER 2006

If we see a downturn in the housing market and people defaulting on loans, these residual interests will evaporate fairly quickly. You can perform the same analysis with banks and insurance companies. It just becomes a little more important to do the detailed financial analysis and fundamental analysis to discriminate between the high-risk assets and liabilities, where the judgment exists, and the low-risk ones, which are just the marketable securities and the plain vanilla debt. Question: Do mergers and acquisitions lend themselves to greater accrual subjectivity? Sloan: We’ve tested for this by going to the statement of cash flows where we can exclude stock deals and strip out the acquisitions as a separate line item. By purging mergers and acquisitions (M&A) activity, we can get a cleaner measure of earnings quality. It turns out that the measure actually works slightly worse. I think the reason for this is that most of these stock acquisitions engage in a stock-for-stock acquisition when the company’s stock is the most overvalued. Hence, the goodwill on the balance sheet associated with that acquisition tends to be extremely subjective and often overstates assets. On the face of it, it might seem like something you want to strip out, but for practical purposes, we find that leaving the M&A activity in actually helps the measure.

Question: Which areas should we watch the most for accruals in the future? Sloan: This is why a fundamental analyst has such a competitive advantage over a quantitative analyst because new business comes along, accounting rules change, and it’s just a case of looking at things differently. One of my favorite examples when I first got interested in this was Boston Chicken and its receivables. In many cases, receivables are not a problem. In this particular case, however, receivables were over half of Boston Chicken’s total balance sheet—over half of its assets. The franchisees were making huge losses, and the only reason they were surviving was because Boston Chicken was lending them money and they were paying it back in the form of a royalty fee. Normally, when I analyze a company, I look at the company and see which is its biggest asset. If it is a proprietary asset that involves subjective judgment, then that is where I do my homework. In the case of Enron, its equity investments and its nonequity investments were huge numbers because much of the action was taking place on these off-balancesheet entities. If they showed up at all, they showed up on an equity one-line consolidation-type entry. So, that is one situation I would not normally have focused on, but in Enron’s case, that was the one that flagged the company’s problems.

©2006, CFA Institute x cfa pubs .org