Analyzing Non-GAAP Line Items in Income Statements

By Wanda A. Wallace

In Brief

More Numbers Don’t Equal Greater Understanding

Arriving at a comparable figure for results from operations is, in the absence of a GAAP standard, an impossible task. There are as many ways to approach the subject as there are companies reporting financial results. The wide disparity of practice makes it difficult to determine whether a certain company is indeed providing greater transparency or is improperly managing earnings. The author examined the variety of line items found on financial statements to paint a picture of current practice. Comparing the magnitude and historical trends of such line items can provide some insight into their use and their usefulness to analysts. In the end, however, such an analysis is only a benchmark; analysts will want to further examine the specific context that a company’s industry and history supply.

FASB added a project on “Financial Performance Reporting by Business Enterprises” to its technical agenda on October 24, 2001. Its stated objectives were to enhance the quality of information displayed on financial statements and ascertain whether sufficient information is contained in financial statements to permit the calculation of key financial measures. A FASB staff paper, “Summary of User Interviews” (www.fasb.org/project/interviews) reports that user interviews identified the following commonly used measures: “operating” free cash flow or free cash flow; return on invested capital; and adjusted, “normalized,” or “operating” earnings; but likewise stated that these key measures are not necessarily well defined.

GAAP prescribes that income from continuing operations will be an all-inclusive number and has avoided the complexity of prescribing disclosure of operating, nonoperating, unusual, abnormal, and nonrecurring items affecting income [as discussed in the dissents to APB 30 (1973)]. This leads to the question of whether and how such terms are being incorporated into the line item captions of public companies.

The Study

The author analyzed the actual captions on the face of quarterly income statements found in 70 10-Q filings issued from May 2001 to January 2002.

Exhibit 1 lists the relative frequency of types of line items found on the income statements in 10-Q filings and quantifies the mean and median proportion of sales that they represent. Since comparative financial statements are presented, the captions sometimes correspond to quantities of a different time period; all magnitudes are based on the most recent timeframe for which net income was reported in each 10-Q. When the count exceeds 70, as for “earnings before” captions, it means that multiple captions appeared in a single 10-Q.

Of interest are the variety of captions used, the number of “other” line items reflected, and the diverse levels of aggregation. An income statement had, on average, 14.8 line items. About two-thirds of the quarterly presentations included net sales and revenues. About a quarter of the companies did not characterize a line item as operating income (loss), and almost half chose not to reflect a gross profit or gross margin. Single line items often corresponded to substantial proportions of sales and net income.

An evaluator of financial statements who wishes to run a performance measure such as “normalized earnings” will be faced with a plethora of line item captions from which to choose. For example, the revenue items included investments, gains to interest revenue, and transactions with related or unrelated parties. Similarly, cost captions ranged from goods sold to exit costs and delivery expenses. Selling, general, and administrative-related captions (SG&A) encompassed various marketing, engineering, stock compensation, and warehouse operations. R&D was disaggregated as a line item in 31% of the statements. Depreciation and amortization-related captions ranged from the general to the specific. Depreciation-related line items were reported in 71% of the statements. Amortization and write-downs were combined within a single line item. Interest-related captions often appeared as “net” quantities, offsetting interest income and expense. The “other” line item captions appeared at the top of the income statement, in the body of the income statement, and in the lower section of the income statement, relating to both income and expenses, as well as net quantities. The reported proportions of net profit or loss that the line items represent in some cases exceeded 1,000%.

The “assorted” line item captions in Exhibit 1 encompassed: gains and losses from various asset sales and dispositions (as well as abandonments), impairments, litigation, franchise agents’ share of gross profit, noncash compensation, severance expenses, systems, time brokerage fees, loan losses, unusual items, nonrecurring charges, and special items. Line items found in the lower section of the income statement included minority interests, equity in unconsolidated affiliates, joint ventures and dealer transactions, noninterest expense associated with personnel, appliance exit costs, occupancy and equipment costs, pending merger costs, accounts receivable securitization costs, and gains and losses associated with investments, sales of real estate, cancellation of debt offerings, and interest rate swaps.

Diversity of Opinion

A bit of a furor has erupted over Standard and Poor’s (S&P) decision to change its system for examining corporate “operating earnings” used to calculate quarterly price-to-earnings ratio (PE). Specifically, S&P intends to treat stock options as an expense, to include restructuring charges from continuing operations as an element of operating earnings, and to exclude pension-plan investment gains. In addition, S&P favors including write-downs of assets which are depreciable and the cost of buying outside research and development services, while excluding goodwill impairment charges as well as unrealized gains and losses from hedging activities and merger-and-acquisition–related expenses. This controversy is, at root, over what is in fact “operating” or “normal” in nature. Of interest is a comparison of S&P’s Research Insight Data Base’s ratio of EPS from operations to basic EPS over the past thirteen years, as depicted in Figure 1.

Theoretical Underpinnings

Financial reporting involves judgments, which can be motivated by a number of conflicting factors. In 1998, then SEC Chairman Arthur Levitt gave a speech about the judgments called “earnings management” in which he proposed enhancements to corporate governance, suggested requirements for auditors’ reporting on “accounting quality,” and expressed a need for overall improvement of the quality of earnings. “Smoothing,” “aggressive accounting,” and “cookie jar reserves” are among the practices he criticized. The interplay of earnings management and earnings quality is a subject of ongoing debate.

The quality of reported earnings has been evaluated extensively in the research literature from various perspectives, such as: whether results are positive or negative; whether the previous year’s performance has been sustained; and whether analysts’ forecasts or expectations have been met. Earnings quality reflects the usefulness of earnings in evaluating a reporting company’s future prospects. The relation between reported earnings, estimated earnings, and actual earnings is a focal point in theoretical discussions of earnings quality. Since 1991, the director of research at Thomson Financial/First Call, Charles L. Hill, has been attempting to obtain all estimates calculated in the same way. However, “the same way” tends to be defined by the majority of the analysts, and differences of opinion arise on what constitutes one-time gains and losses. A 2001Wall Street Journal article by Jonathan Weil discussed diverse earnings measures and reported that as much as 60 cents of every dollar of earnings reported by S&P 500 came from non-GAAP earnings. The implication is that where on the face of the income statement an earnings effect occurs is relevant to the evaluation of earnings quality.

Benchmarking

Comparison to benchmarks of components of earnings can be relevant when assessing the quality of earnings. The average of an earnings component for a cross section of companies at different points in time is not asserted to be either the true or predicted value of earnings for any particular company. The expected ratio of an earnings component to revenue can be more finely tailored for any company to a set of so-called control companies in the same industry. Nonetheless, evaluators of financial statements should have an understanding of the profile of income statement components for the population of widely used data sets, as well as for subsets defined by whether they generate a profit or loss in a given year. The role of key economic events (frequently defined as a change in return on assets in excess of 10%) can also be considered.

Organizations such as S&P have been active as information providers, offering the Compustat/Research Insight service with defined data fields. Such companies use public filings to collect and “translate” line captions and associated disclosures. To illustrate the manner in which line item captions from the face of income statements are categorized by S&P’s Compustat, consider the described components of the field named “XSGA: Selling, General, Administration Expense” (SG&A). Specifically, when the following expenses are broken out separately on a company’s income statement, they are included in SG&A, but these same expenses are not included if they were allocated to the cost of goods sold:

There are further exceptions. Related expenses of sales from companies with software development operations are also included. Dry-hole expenses for companies using successful-efforts accounting for oil assets are included unless combined with another item classified by S&P as depreciation. Moreover, the SG&A field is not coded by S&P for banks, utilities, or property and casualty companies.

Such detail in the manuals of proprietary data fields demonstrates the considerable effort required to provide benchmark data. Exhibit 2 displays the key line items on the income statement that are generally available from databases such as Research Insight/CompustatPC, drawn from 10-K filings by public companies. Statistics are provided for companies generating a profit versus those incurring a loss, permitting historical comparisons to Exhibit 1.

Patterns over Time

Figure 2 displays these 11 key data fields corresponding to the income statement. The proportion of sales each field represents is depicted for all companies, broken down by net income and loss. Observed variations in Figure 2 stem in part from disparate definitions of specific line items presented on the income statement which, in turn, are classified into the same CompustatPC data fields.

Line items’ proportions relative to sales remained relatively stable for over 7,000 public companies for a 10-year period. Very large charges appear periodically in the nonoperating, special item, and extraordinary captions of the income statements. A balance is always sought between information overload and transparency. Note that some circularity arises in how net income and loss are defined, since this result could easily be driven by specific line items, such as the special charges or extraordinary items. Because mean values or averages can be affected disproportionately by larger entities, frequency distributions can help explain both the relative incidence and direction of the income statements’ line items (see Exhibit 3).

Loss companies. Companies experiencing losses took a larger dip in gross profits in 2000 and remained at a similar level in 2001. SG&A expenses increased markedly in 1996, with a slight dip in 1997, a growth in 1998 and 1999, and a dip again in 2000. Companies with reported losses had lower 2000 and 2001 EBITD figures, with particular dips in 1996.

Loss companies had a larger interest burden—especially during 1996 to 1998—and nonoperating items were far more volatile for these companies, declining from 1992 through 1995 and then seesawing thereafter. Special items had negative spikes in 1992, 1994, 1996, 1998, 2000, and 2001, with the downward spike in 1998 exceeding all of the others for these loss entities. Tax levels were flat except for a large spike downward in 1998. Discontinued operations were substantially lower in 1992 and 1998, and turned positive in 2000. Extraordinary items spiked upward in 1999 and demonstrated some volatility over time.

Profitable companies. Companies with positive results exhibited comparatively flat trends in line items, except for an interest dip in 1997, nonoperating items spikes in 1993, 1999, and 2001, an increase in special items in 1994, and a dip in extraordinary items in 1999.

Proportion of Positive (Negative) Line Item Captions

Exhibit 3 reports the relative proportion of positive (negative) quantities per Compustat grouping, and shows substantial stability. It appears that, prior to 1997, EBITD over sales were more often positive and special items over sales were less often negative, leading more companies to report positive net income. After 1997, these results shifted more than 5%. This would seem to be the reverse of the pejorative earnings management claims, unless some economic event actually diminished entities’ profitability relative to sales after 1997. The relationship between economic cycles and benchmarks over time means that patterns in inflation rates, interest rates, stock market indices, and industry trends should be considered when performing analyses.

Industry Variation

An analysis of 14 industry groupings in the database in 2000 identified no significant differences from the total sample except for special items, taxes, and extraordinary differences.

The pharmaceutical and banking industries were distinctly different in their average special items, with the latter being uniquely positive and the former significantly more negative than other industries. The pharmaceutical industry had the largest negative average value of extraordinary items. For taxes, the significant difference was found in the engineering industry, which was much more negative than other industries. This suggests that benchmarking at an industry level would enhance an analysis, particularly for those items and groupings differing from the norm.

Media Attention

The Wall Street Journal carried a front-page story by Jonathan Weil (2001) titled “Moving Target: What’s the P/E Ratio? Well, Depends on What Is Meant by Earnings.” As the title suggests, the article describes how companies highlight unusual, unimportant, and one-time charges. Yet, more skeptical market participants question whether ordinary items are being inconsistently reported, creating confusion and distortion. An interesting aspect of the article is its acknowledgment that First Call reports a consensus earnings target for a stock that reflects the accepted view of the overall investment community and can exclude not only extraordinary items, accounting changes, and discontinued operations, but also other consensus items.

Diverse Nomenclature

Companies report other than the usual nomenclature of line items on income statements. How might the information in Exhibit 2 be used to evaluate whether components of operations suggested as “adjustments” in press releases should indeed be characterized as one-time, unusual, noncore, or nonrecurring?
As one example, consider the following press release: Alamosa PCS Holdings Inc., The 1999 Sprint PCS (NYSE:PCS–news) Affiliate of the Year, today reported strong sequential subscriber growth and service revenues for the third quarter ended September 30, 2000. … Subscriber revenues for the third quarter were $14.9 million, up 39% over the second quarter of 2000.

Total revenue for the third quarter of 2000 was $22,983,000 including subscriber revenue of $14,941,000, roaming revenue of $5,813,000 and product sales of $2,229,000. Subscriber revenue represented a 39% increase over the second quarter of 2000, while roaming revenue grew 35%. Earnings before interest, taxes, depreciation and amortization, or EBITDA, excluding non-cash compensation was negative ($11,094,000) for the third quarter of 2000. Excluding selling and marketing expense, pre-marketing EBITDA was a positive $2,848,000 for the quarter. …

For the nine months ended September 30, 2000, Alamosa reported total revenue of $51,902,000, comprised of $32,772,000 of subscriber revenue, $13,129,000 of roaming revenue and $6,001,000 of product sales revenue. EBITDA excluding non-cash compensation for the nine-month period was negative ($25,310,674).

This press release calculates an EBITDA figure that excludes noncash compensation, as well as one that excludes selling and marketing expense. The ratio of these quantities to sales can be compared to the EBITD reported by CompustatPC. For example, the total revenue for the third quarter of 2000 is reported as $22,983,000 and the EBITDA figure with exclusions was ($11,094,000), for a –.483 ratio; the nine-month comparison is $51,902,000 total revenue and ($25,310,674) EBITDA with exclusions, for a –.488 ratio. The press release also excludes selling and marketing expense for a premarketing EBITDA of $2,848,000 in the third quarter, which converts to a ratio of .124. The 2000 EBITD for Research Insight/CompustatPC contained 26.5% that were negative in direction and 68.8% that were positive. The loss companies in 2000 had an EBITD-to-sales ratio of –9.43; profitable companies had a ratio of .22.

The implication of this benchmarking is that the EBITDA-adjusted numbers in the press release are far more favorable than the average 2000 EBITD figure for loss companies as analyzed in Exhibit 2. On the other hand, if Alamosa PCS Holdings were deemed more comparable to the profitable entities, then the premarketing proportion is not as positive as that derived for EBITD from Research Insight/CompustatPC. The selling and marketing expense being set aside for the premarketing calculation is implied to be $13,942,000, which is .607 of total revenue. This number far exceeds the figures for profitable companies but is below the SG&A average for loss companies, which might be expected because the selling and marketing expense would presumably be a subset of SG&A. This analysis suggests that if the company is viewed as a loss company, its SG&A is lower than average. Moreover, such benchmarking might imply the usual nature of the marketing expenditures and the preferability of not adding them back to EBITDA in evaluating the quality of earnings.

Benchmarking against the information reported herein would require collection of similar line items from 1992 to 2000. The consistency of the profile of a company to the average suggests its comparability to the data set; however, it does not necessarily mean that the company’s context warrants its similarity to that average for the Research Insight/CompustatPC companies. In a similar vein, if the company differs from the average, particular facts associated with the company could explain observed divergences. As an example, EITF 00-10 (2000) explains how shipping and handling fees have historically been a component of net sales at some companies, while in others they have been reported in cost of sales, and in still others have been depicted in SG&A expenses.

Key Observations

The data depict 1998 as a year in which nonoperating items peaked, largely attributable to loss companies. The fact that the peak is positive in direction may have been responsible for regulators’ well-publicized concern regarding earnings management around this point in time. Perhaps ironically, the average value for special items likewise peaked for both the full sample and the loss companies in 1998, but in the negative direction, at three times the magnitude of nonoperating items. This observation would seem inconsistent with earnings management from a “bottom-line perspective.”

If analysts set aside special items but not nonoperating items, then the evidence in 1998 might be considered consistent with managers’ relegating negative news to a caption more likely to be deemed nonrecurring and noncore by Wall Street. The influence of such actions on quality of earnings depends upon the propriety of the categorization. Exhibit 3 provides additional insight concerning these observations regarding 1998. Specifically, the proportion of companies that had nonoperating positive line items in 1998 is 67.2%, which is similar to other years. How similar, however, is a concept that ought to be explored in tandem with materiality implications (see “Probability and Materiality,” The CPA Journal, June 2001). The special items tended to be negative in 1998 for almost one-third of the companies. This is about 3% more often than the prior year and 10% more than 1994, which could be described as consistent with either a “big bath” or poorer performance depicted as special items.

Caveats

Caveats apply to any proposed methodology. For example, two implicit assumptions underlie an approach to evaluating earnings components benchmarks. The first is that a company’s asset base at the beginning of the year produces the revenues, expenses, and cash flows earned during the year. The second implicit assumption is that the denominator of sales used to compute all of the ratios is matched in time to the expense, gain, or loss line item, precluding the need to separately consider the potentially differentiated effects of sales increases on each line item from year to year. Moreover, this research relies on the definition of income statement classifications applied to financial statements’ captions by Research Insight/CompustatPC.

Benchmarks are not tailored to individual circumstances. Instead, they are analogous to road signs to which comparisons can be made on an individual route. This does not mean that the road sign lacks usefulness; it merely means it needs to be evaluated in the proper context, considering industry type, age of company, public status, and economic events.

Even as the market increasingly calls for transparency, the best way to communicate the results of operations on the income statement remains a matter of debate. Which line item captions should be presented, how much disaggregation is desirable, and how might the presentation be useful to decision makers? This research is intended to provide a profile of current practice, as well as a historical perspective, that can be coupled with guidance on choices regarding aggregation within a financial statement (Statement of Financial Accounting Concept 5).

Components of earnings are of interest for at least four reasons:

Terms such as “nonrecurring,” “special,” and “unusual” blur into each other in practice. Now may be the right time for standards setters to specify a presentation approach with well-defined language, along with both guidance and proscriptions.