Trivialization of the bottom line and losing relevance of losses

When the stock market crashed in October 1929, so did public confidence in the US markets.¹ Congress held hearings to identify the cause of the crash and to search for solutions. Based on its findings, Congress passed the Securities Act (SEA) of 1933 in the peak year of the Great Depression. The act requires companies offering securities for sale to the public to “tell the truth about their business, the securities they are selling, and the risks involved in investing in those securities.” An integral part of this truth-telling exercise is to provide an audited financial report, inclusive of an income statement, on at least an annual basis.²

Academics and practitioners have since debated the purpose of the income statement and its importance relative to the balance sheet (Basu and Waymire 2010). Debate also is ongoing about the meaning and measurement of earnings, which is the bottom-line, summary metric of the income statement. For example, Black (1980) argues that earnings should be a measure of value (adjusted for price-to-earnings ratio), not of changes in value. In contrast, the Financial Accounting Standards Board (FASB) and some academics, arguably questionably, interpret Hick’s concept of personal income as the amount that a firm can pay out as dividends while leaving itself equally well off at the end of a period as it was at the beginning (Jameson 2005; Bromwich et al. 2005). This idea, consistent with the clean surplus assumption, more likely demonstrates the changes in value than the levels of value.

Dichev (2008) argues that a manager’s aim of securing revenues greater than costs, embodied in the income statement and earnings, is congruent with how firms create value. A positive earnings number would represent good news for shareholders, and a negative number would indicate bad news. A money-losing business would show expenses exceeding revenues. A reported loss, when also indicative of future losses, would imply a bad business for shareholders. Supporting this idea, Alchian (1950) argues that, in an impersonal market system, positive profits are the sole criterion for the selection of successful and surviving firms, while firms that suffer losses disappear. Graham et al. (2005) survey chief finance officers and analysts and find that earnings is the single most important output of financial statements. In the past, standard setters considered the income statement and earnings to be the primary focus of financial reporting [FASB Concepts Statement No. l (paragraph 43)].³

Given the importance of earnings, much research has been conducted on the quality of earnings (Dechow et al. 2010). Yet no straightforward way exists to calculate earnings. The purpose of this article is to illustrate the growing limitations of the current methods of determining earnings, along with the resultant decreasing relevance of earnings, particularly when earnings is a negative number. To illustrate the point, I provide data on matching with revenues for cost of goods sold (COGS) and selling, general, and administrative (SG&A) expenses—the two dominant operating costs—by cohorts of listed firms. I show that SG&A, inclusive of intangible investments, has become the dominant operating cost for newer firms, but its matching with revenues has diminished. Without matching, measured profits and profit margins become meaningless summary measures (Dichev 2008). This trend cannot be ignored, because cohorts listed in the last 30 years now constitute over 80% of the set of listed firms. Because revenue recognition and intangibles capitalization rules are unlikely to change soon, I conclude with a consideration of whether firms should be allowed to report Generally Accepted Accounting Principles (GAAP) earnings on a voluntary basis, instead of it being mandatory (Beaulieu 2014).

This paper complements Gu et al. (2022), who use an innovative measure to identify firms that report losses because of the GAAP requirement to immediately expense inhouse intangible investments. Those GAAP-loss firms would report profits had they capitalized inhouse intangible investments and then expensed them over their useful lives, similar to capital expenditures. Gu et al. (2022) find that after the accounting bias for intangibles expensing is unraveled, the earnings of GAAP-loss firms become as informative as earnings of profitable firms. Furthermore, GAAP-loss firms subsequently outperform other loss firms and even profitable firms in value creation from intangible investments. The authors caution financial statement users to consider operational differences between GAAP-loss and other loss-making firms in the evaluation of losses.

Earnings is obtained by subtracting expense line items from revenues, both of which are calculated using different principles and accounting conventions.⁴ Revenue recognition requires the fulfilment of two conditions: an objective determination of the value of goods and services that have been delivered and customer payments being probable. These conditions, however well intended, cause the first point of departure for revenues from numerous events that managers and shareholders consider value creating. For example, for a long time, Apple could not recognize any revenues upon selling and delivering an iPhone.⁵ That is, the company that became the most valuable in the world, having created the “product of the century,” could not recognize revenues even after conceiving of that product, selling it, and having collected customer payments.⁶ While that deficiency has been fixed with the new revenue recognition rules, numerous other events that could create a recurring stream of cash flows are still not recognized as revenues, such as the signing of an exclusive partnership with a tech giant, a patent approval, approval of a blockbuster drug by the Food and Drug Administration (FDA), a social influencer’s endorsement of a company’s product, and a social media company signing up a million new subscribers. None is recognized as revenues, because no delivery occurred. While this limitation applies to all companies, it is especially important for modern technology companies, which build a probable future stream of cash flows through a series of serendipitous events that increase the value and marketability of their ideas, not through investments in inventory or property, plant, and equipment (PP&E). The delivery criterion is based on a sound principle of resolution of uncertainty and applies to all businesses. Yet it increasingly causes an important departure of earnings from what investors consider to be value creating.⁷

Expenses are calculated using numerous, inconsistent conventions.⁸ Most costs associated with the procurement and production of physical goods are initially capitalized and presented in an inventory account. Inventory is then diminished via a COGS account when goods are delivered to customers. COGS is thus reported in the same period as its associated revenues and is highly matched to current revenues. COGS contributes 60–90% of total costs for brick-and-mortar firms such as Walmart and Ford. Thus, for a traditional brick-and-mortar company, profits is calculated after deducting operating costs that are highly matched to revenues.

A similar concept is applied to investments in PP&E. Purchase, delivery, and installation costs are initially capitalized in the PP&E account and then retired through a depreciation expense account over the expected period of use of PP&E. Unlike COGS, which is traced to revenues, depreciation expense is calculated mechanically, like ratable retirement over five years. Yet depreciation expense is expected to match the periods of PP&E’s benefits and thus, approximately, to revenues.

Another important core operating expense is SG&A. Numerous operating costs that cannot be capitalized in the inventory account or can be reported as cost of sales are reported in the SG&A account. Those costs include items such as head office expenses, sales commissions, warehouse rents, executive compensation, and delivery costs. Some costs, such as sales commissions and sales promotions, are recognized at the same time as the associated revenues and are highly matched to revenues. Others, such as head office expenses, do not easily adjust with ups and downs in sales and are less matched to revenues (Anderson et al. 2003).

More important, over the last 40 years or so, SG&A has increasingly included an emerging class of outlays: inhouse investments in intangibles. Those outlays include research and development (R&D), process improvement, information technology, software development, advertisement, organizational strategy, acquisition of customers, brand building, and hiring and training personnel. Except in certain industries, such as software development, oil and gas, and radio and recording, most inhouse intangible outlays are expensed as incurred. To the extent that SG&A represents investments in expectation of future benefits, it would be unmatched to current revenues.

Stock options, a principal way of paying workers in the knowledge economy, follow an altogether different convention. The Black-Scholes formula, used to calculate the value of stock options, is based on historical as well forward-looking information, but, more important, does not equal cash paid, a principle that applies to most other expenses. That value is then recognized as expense ratably over the options’ vesting period or immediately upon cliff vesting. The actual compensation received by employees, through exercise of stock options and sale of stocks, occurs only after the vesting period, when the stock price has gone up. Thus, stock option expense would likely be recognized in a period different than when the company’s fortune improves and its stock price increases.

Certain expenses, such as restructuring charges and those leading to contingent liabilities, result from current events but are based on estimates and judgment about future payouts. Other expenses, such as goodwill impairments, often pertain to events that occurred in past periods (Li and Sloan 2017). Interest expense is recognized during the period of loan outstanding and incorporates amortization of the bond premium and discount. Tax expense is related to profits, not revenues.

In sum, how net income is calculated has at least three deficiencies: (1) revenues often do not incorporate effects of events that could create recurring cash flows, (2) different line items of the expense statement are recognized using different principles and conventions, and (3) many expenses do not match with contemporaneous revenues.

A bridge built using parts that followed different engineering conventions, safety standards, and measurement systems, and with a design that ignored relevant information such as local weather and soil conditions, would likely be unstable. Conceptually, an income statement is no different than that bridge.

It is not a new idea that the income statement is built using inconsistent principles and reports earnings that would be less meaningful than what the SEA (1933) originally intended. Ball and Brown (1968, p. 160) state: “Because accounting lacks an all-embracing theoretical framework, dissimilarities in practices have evolved. As a consequence, net income is an aggregate of components which are not homogeneous. It is thus alleged to be a ‘meaningless’ figure, not unlike the difference between twenty-seven tables and eight chairs.” Ball and Brown (1968) cite Canning (1929, p. 98): “What is set out as a measure of net income can never be supposed to be a fact in any sense at all except that it is the figure that results when the accountant has finished applying the procedures which he adopts.”

Despite these pessimistic predictions about the usefulness of earnings, Ball and Brown (1968) find that net income corresponds with contemporaneous stock returns. Profitable companies have positive returns on average, and loss-making companies have negative returns. Long and short portfolios formed with perfect foresight of profits and losses, respectively, earn significant returns. Loss is economically meaningful, as it is associated with real actions such as employee layoffs (Pinnuck and Lillis 2007) and shareholders’ exercise of the abandonment option (Hayn 1995). In addition, earnings is more predictive of cash flows than cash flows are (Finger 1994; Barth et al. 2001). Accruals, which differentiate earnings from cash flows, mitigate the timing differences between economic events and their cash flows, making earnings more informative of firm performance. As a result, working capital accruals are negatively associated with contemporaneous cash flows (Dechow 1994; Dechow et al. 1998).

More recent studies point to declining usefulness of earnings. Lev and Zarowin (1999) show that the value relevance of earnings (measured by the R-squared of regression between contemporaneous stock returns and levels and changes of net income) has declined over time. Srivastava (2014) shows that declining relevance is related to progressively lower relevance of successive cohorts of listed firms. After 2008, hedged portfolios formed based on high and low earnings-to-price ratio (long and short positions, respectively) no longer earn positive results, on average.⁹ Nallareddy et al. (2020) show that earnings is less predictive of future cash flows than cash flows themselves are.¹⁰ Bushman et al. (2016) document a decline in the magnitude of negative association between working capital accruals and contemporaneous cash flows. Green et al. (2022) show that this decline is related to inadequate capitalization of intangible investments. Gu et al. (2022) find that losses, to the extent they are due to immediate expensing of intangibles (GAAP loss firms), do not mean losses in the traditional sense. Contrary to expectations, GAAP loss firms survive over long periods, increase hiring, create valuable patents, and provide positive returns to shareholders.

I examine one reason that earnings, particularly losses, has lost its relevance since Ball and Brown (1968). I build on arguments offered in prior literature (e.g., Lev and Zarowin 1999; Dichev and Tang 2008; Srivastava 2014), while offering new insights. Dichev and Tang (2008) argue that, without matching, net income is not a useful measure for valuation. For example, consider a biotechnology startup that has only one large expense: research and development outlays. With little or no revenues, it would report large losses. That company invests in R&D today with the hope of earning future revenues after the drug is developed and approved by FDA and marketed over its patent period. Almost every business operates on this principle, advancing costs in expectation of benefits (Fisher 1930; Dichev 2008). Firms invest in new factories, products, markets, brands, employees, business segments, warehouses, and inventory, anticipating that the present value of future revenues would exceed the current investments. Without comparing current investments with future revenues, a manager cannot determine the cost-benefit considerations that go into investment decisions. Similarly, an investor cannot figure out the recurring profit potential from reported earnings if expenses are not matched to revenues.

Dichev and Tang (2008) show a significant decline in matching over time. Srivastava (2014) extends Dichev and Tang (2008) by dividing firms into listing cohorts. I follow Srivastava (2014) by calling the first year in which a firm’s data are available in Compustat the “listing year.”¹¹ All of the firms with a listing year before 1960 are classified as “pre-1960s.” All other firms are divided into listing cohorts based on the decade in which they were listed. Thus, I divide all companies into seven groups: pre-1960s and cohorts for the 1960s, 1970s, 1980s, 1990s, 2000s, and 2010s.

Following Dichev and Tang (2008), I estimate the following regression on an annual cross-sectional basis for each group for each year:

I calculate total expenses by subtracting net income (Compustat NI) from revenues (Compustat SALES).¹² I scale all of the variables by average total assets. Matching is measured by the regression coefficient on contemporaneous expenses (β₃), which represents the contemporaneous revenue–expense correlation. I calculate matching for each of the seven groups by averaging their matching calculated on an annual basis. Figure 1 shows that matching for the pre-1960s, 1960s, and 1970s groups is approximately 1.0, indicating excellent matching. Each successive cohort shows lower matching. Matching for cohorts listed in the twenty-first century is noteworthy. It is only 0.25 and 0.10, respectively, for the 2000 and 2010 s cohorts, demonstrating that most of their expenses are unmatched to revenues.

I next measure matching of major components of expenses. I measure COGS and SG&A expenses by the Compustat data items COGS and XSGA, respectively, and call the remaining expenses “other expenses.” I measure the matching of COGS and SG&A expenses by β₃ and β₄, respectively, in the following equation, estimated first by group-year and then averaged over all years:

Panel A of Fig. 2 shows no decline in matching of COGS for successive cohorts. Thus, what is reported as cost of goods sold, cost of sales, or cost of revenues in the income statement is highly matched to revenues. Panel B shows that SG&A is also highly matched for traditional firms, represented by the pre-1960s, 1960s, and 1970s groups. Every successive cohort shows a progressive decline in SG&A matching. It is just 0.3 for the 1990s cohort and negative for the last two cohorts (depicted as zero in the figure). Thus, what is reported in SG&A for the firms listed in the last 30 years is practically unmatched to revenues.

I calculate the relative proportions of the three types of expenses for each firm-year by dividing them by that firm-year’s total expenses. I next examine those percentages by cohorts. To control for life-cycle stage, I consider only one year’s observation per firm in each cohort, from when that firm is three years old.¹³ Cohort-wise cost patterns, after controlling for life-cycle stages, are represented in Fig. 3.

COGS was the predominant cost for the pre-1960s, 1960s, and 1970s groups, being 68%, 68%, and 64%, respectively. SG&A was just 20%, 21%, and 24%, respectively, for the three groups. Each successive cohort has lower COGS, with the latest cohort reporting just 44%. More important, each cohort reports higher SG&A. By the time the 2010s cohort is reached, SG&A almost equals COGS. This trend is intuitive. Consider a modern business such as LinkedIn, Facebook, Moderna, or Google and contrast it with brick-and-mortar businesses such as Walmart and Ford. One would be hard-pressed to guess what items are reported in COGS for the former category of firms, while for the latter group the main cost items can be easily predicted as COGS. Table 1 lists prominent companies with the largest SG&A exceeding COGS from 2017 to 2020, indicating that SG&A being a dominant cost is a widespread phenomenon among modern firms. Figure 4 shows that the percentage of companies with SG&A greater than COGS increases with successive cohorts.

I next examine the cost trends for a sample of loss firms while controlling for the life-cycle stage. Figure 5 shows that among loss firms in the last three cohorts, SG&A exceeds COGS. This finding, when combined with the trend exhibited in Fig. 2, shows that loss firms listed in the last 30 years display SG&A as their dominant operating cost, but their SG&A is unmatched with revenues. Thus, their bottom-line numbers and profit margins are unlikely to be informative of recurring profitability. This problem is no different than what it would be for traditional firms if their investments in PP&E and inventory were immediately expensed instead of being capitalized.

The mismatch problem for traditional firms is solved by accrual accounting that scatters the investments in inventory and PP&E over the periods of their expected benefits, achieving a high degree of matching. Thus, their loss accurately represents a money-losing business with expenses exceeding revenues. For newer cohorts, losses lose their meaning because of a mismatch of their dominant costs—intangibles—even before considering the deficiencies in revenue recognition.

The deficiencies in the income statement would not matter if the firms listed in the last 30 years did not add significantly to the economy. Figure 6 shows that the composition of the set of listed firms keeps gravitating toward newer cohorts, no different than trends in any human population. Traditional firms, which represented a significant percentage of the listed population until 1980, now form a miniscule percentage of the set of listed firms. In 2020, the pre-1960s, 1960s, and 1970s cohorts provided just 4%, 4%, and 3%, respectively, to the set of listed firms (see Fig. 7). Thus, firms with COGS-dominated costs and a high degree of matching are just 11% of the listed population in 2020. In contrast, in 2020, the 1990s, 2000s, and 2010s cohorts contributed as much as 17%, 17%, and 48%, respectively, to the set of listed firms (see Fig. 7).

Stated differently, over 80% of the set of listed firms in 2020 is attributed to cohorts listed after 1990.¹⁴ These firms are characterized by deficient revenue recognition and unmatched operating costs. Hence, one can interpret their earnings and losses in the spirit of Canning (1929, p. 98), in that they “can never be supposed to be a fact in any sense at all except that it is the figure that results when the accountant has finished applying the procedures which he adopts.” This explains the results of Gu et al. (2022), that losses for GAAP loss firms do not mean the same as losses for traditional firms.¹⁵

Barth et al. (2023) examine changes in value relevance of financial statement components over time and find that the value relevance of at least some income statement components has increased, not decreased. I discuss three components of the income statement that are highly value relevant: revenues, R&D expenses, and investment portion of SG&A.

Revenue has become highly value relevant for modern technology firms (Ertimur et al. 2003; Chandra and Ro 2008; Bowen et al. 2002). Many modern companies, such as LinkedIn and Facebook, offer an altogether novel class of services and create a new market for it. They typically have low variable costs and could earn winner-takes-all profits by dominating their market and taking advantage of the network effects. Achieving revenues is their best demonstration of the feasibility of their atypical business ideas. In addition, revenue growth is a reliable indicator of progress toward achieving viable market size and dominant market position. For such companies, revenues are highly value relevant, and investors often react more strongly to meeting or beating revenue targets than to achievement of earnings targets.¹⁶

R&D is also highly value relevant because it is an essential tool for survival and growth for modern corporations and is not just a discretionary expense as considered in prior literature (Govindarajan et al. 2019; Barth et al. 2023). Furthermore, the investment portion of MainSG&A (that is, SG&A minus R&D and advertising) has become the largest category of firms’ operating investments, surpassing PP&E (Enache and Srivastava 2018). Treating it as an investment in the calculation of book value improves the returns from a value investment strategy (Lev and Srivastava 2022; Arnott et al. 2021; Iqbal et al. 2022).

In addition, investors pay attention to adjustments that firms make to provide non-GAAP earnings. Over 90% of Standard & Poor’s (S&P) 500 firms now report a non-GAAP performance measure, indicating either that a demand exists for such a metric or that firms consider a number other than the GAAP earnings number to be a better performance measure. Contracting is more likely based on a non-GAAP earnings metric than GAAP earnings. For example, firms frequently link chief executive officer compensation to a non-GAAP metric (Kyung et al. 2021). Many common debt covenants use some form of adjusted accounting earnings, such as earnings adjusted for interest, taxes, depreciation, and amortization (Li 2016).

In sum, highly value relevant line items are reported in income statements. But the relevance of the summary measure of GAAP earnings, particularly when it is negative, is becoming doubtful.

The relevance of GAAP earnings for modern technology firms is increasingly suspect. The nonrecognition of numerous value-creating events in revenues and the increasing mismatch between contemporaneous revenues and expenses are two principal reasons for this trend. FASB is unlikely to change revenue and expense recognition criteria soon, which raises the question of whether reporting a bottom-line number based on GAAP should be made an optional feature instead of being mandatory.

Mandatory reporting is supported by the arguments of consistency and comparability, but they must be weighed against the cost of preparation, audit, and user comprehension (Beaulieu 2014). Applying these considerations to data reported in this article raises the question: What is the point of reporting GAAP earnings for a biotechnology startup when no one cares about such earnings or compares them against earnings of a steel company or even another biotechnology company? If reporting GAAP earnings is made optional, only the companies that consider it valuable would make an effort to report that number. Another possibility is that firms could report what they consider the bottom-line metric, which is the best indicator of their recurring profits, while explaining how it is calculated, instead of first reporting a GAAP earnings number and then performing numerous adjustments to determine a non-GAAP number.¹⁷

It is noteworthy that by reporting a bottom-line number, accountants implicitly convey their perspective on equity valuation (either levels of value, adjusted for P/E ratio, or of changes in value). The same accountants are reluctant to convey their assessments of the value of self-developed intangibles.

Arguably, a focus on the current income statement structure and earnings distracts resources and attention away from more important and detailed disclosures that investors consider relevant. Govindarajan et al. (2018) suggest a structure for more value relevant disclosures. They argue that, first and foremost, companies should provide detailed information about their market and market prospects, broken into two main variables: total addressable market and targeted market share. Firms should report not just revenues but also the drivers behind the revenues, especially because modern companies’ operating activities often differ from their revenue-generating activities. For example, Facebook’s revenue-providing customers are the companies that pay for advertisements, not those who post personal information on Facebook’s website. How the activity and growth of Facebook’s subscribers drive the advertising revenues is highly relevant information for investors. An analyst following Facebook would thus look at variables such as the number of active users, their geographical distribution, their retention rates, the average amount of time they spend on the website, and the growth or decline in any of these factors.

As far as the costs are concerned, firms must provide a detailed description of outlays in three broad categories. First are outlays that support current operations, which could include soft expenses, such as customer acquisition, data breach and safety, regulatory fines, and product enhancement, or hard expenses, such as investments in hardware, servers, and cellphone towers. The economic life of hardware is increasingly lower than its physical life, given rapid technological obsolescence. Maintenance outlays should also include amounts spent on acquisitions required to maintain the firm’s competitive edge. Companies must also distinguish fixed from variable costs and detail the variable costs associated with a unit of activity. For example, Twitter provides “cost per ad engagement.”

Second are outlays that are expected to produce future benefits. Those outlays must be further divided into two more categories: those that improve current operations and those that have highly risky, lottery-like payoffs. More enhanced disclosures would include progress of future-oriented projects, how such projects are expected to affect (if at all) their current operations, the aggregate resources committed to the projects, and the likely launch dates.

Third are the so-called one-time or special items, which companies should identify and separate. They should be divided into two categories: those related to past investments that did not work out and those that are expected to require future payouts, such as expected layoffs and restructuring.

By making use of disclosures that follow this blueprint, a smart analyst could project a company’s future revenues, estimate the outlays required to sustain the firm’s business model, and calculate the present value of future cash flows. Analysts would then have the option to add, as they see fit, to that current operation valuation their own estimates of the option values of risky, moonshot projects.

Current revenue recognition rules ignore numerous events that are considered value creating by investors and managers of modern corporations. Expense recognition conventions increase the mismatch between revenues and expenses. Those two developments make GAAP earnings of modern companies, particularly losses, largely meaningless in their economic and valuation implications. Instead of GAAP earnings, investors seek more elaborate disclosures of items that would help them ascertain recurring profits. However, the way GAAP earnings is calculated is unlikely to change anytime soon. Hence, I recommend that companies be allowed to report a GAAP earnings number on a voluntary basis but that enhanced disclosures be mandated. Another possibility is to allow companies to report a bottom-line earnings number—one they consider value relevant—and to detail how it was calculated.