All losses are not alike: Real versus accounting-driven reported losses

We examine the value relevance of accounting-driven losses that result from the immediate expensing of firms’ internally generated intangible investments versus losses occurring irrespective of intangible investments. Contrary to the long-held view that losses are less relevant than profits for valuation, we find that once the accounting bias of intangibles-expensing is undone, earnings of firms reporting intangibles-driven losses are as informative as earnings of profitable firms. Furthermore, contrary to the view that persistent losses decrease earnings relevance, our evidence shows no decrease in the relevance of earnings for firms reporting persistent intangibles-driven losses. We also find that firms reporting intangibles-driven losses subsequently outperform other loss firms and even profitable firms in value creation from investments in technological innovation and human capital. Our evidence further shows that firms reporting intangibles-driven losses have stronger future performance than other firms. Taken together, the results of this study demonstrate the fundamental differences between losses driven by the immediate expensing of internally generated intangible investments and losses reflecting genuine business performance shortfalls. Standard accounting performance measures, however, do not properly reflect these operational differences and their implications.

Don't be seduced into thinking that that which does not make a profit is without value.
-Arthur Miller

Introduction
Loss reporting has increased significantly in the developed economies of the United States and European Union in the past two to three decades (Fig. 1).Almost 50% of all US public companies and 70% of high-tech and science-based firms reported an annual loss in the economic boom year of 2019, the last pre-Covid year.Loss reporting creates obvious valuation challenges to investors (e.g., no earnings multiple for loss firms) and researchers (the valuation implications of profits and losses are different) and therefore deserves special consideration.In this study we Fig. 1 Percentage of loss firms and the amount of loss in US, UK, and European stock markets.Data for US loss firms and tech loss firms are obtained from Compustat.Data for EU and UK loss firms and tech loss firms are obtained from Compustat Global.We identify loss firms as firms with negative income before extraordinary items (IB).Compustat and Compustat Global define income before extraordinary items (IB) as the income after all expenses, including special items, income taxes, and minority interest, but before provisions for common and/or preferred dividends.It does not, however, reflect discontinued operations or extraordinary items.We define tech loss firms as firms with negative income before extraordinary items and with three-digit SICs of 283 (drugs), 357 (computer and office equipment), 362 (electrical), 366 (communication equipment), 367 (electronic components and accessories), 381 (search, detection, navigation, guidance, and aerospace systems), 382 (lab apparatus, optical, measurement, and control instruments), 384 (surgical, medical, and dental instruments), 385 (ophthalmic products), or 737 (computer programming and data processing) 1 3 All losses are not alike: Real versus accounting-driven reported… distinguish between accounting-driven losses, which are created by the expensing of internally generated intangibles, and real losses, which occur irrespective of intangibles expensing.We show that the two groups of loss reporters, treated equally by accountants and many investors, are strikingly different in terms of valuation and implications for future performance.
We start our analysis by briefly discussing the procedure we use to adjust reported earnings for the expensing of intangibles, which is essentially achieved by replacing the accounting expense in the income statement with the amortization of the capitalized intangibles.We then show that such an adjustment turns roughly 30% of reported losses to profits.We are thus dealing with a widespread phenomenon of accounting-driven losses.Using annual cross-sectional regressions of returns on the level and change of earnings, we demonstrate that for all loss firms the coefficient on the level of earnings (losses) is statistically insignificant, consistent with previous findings.In contrast, the earnings level coefficient of the intangibles-adjusted earnings, where the intangibles expense is replaced by the amortization of the intangible capital, is positive and highly significant.Thus, our adjustment of reported losses for the expensing of intangibles restores the valuation relevance of earnings.More importantly, we show that this relevance restoration applies only to firms whose reported losses are turned into profits by our adjustment-henceforth termed GAAP losers.For firms that continue to show losses after our intangibles adjustmenthenceforth Real losers-adjusted losses are still insignificant.Thus, GAAP losers and Real losers are inherently different in earnings relevance.
Digging deeper into the rising loss phenomenon, we examine persistent losses, where firms report operating losses continuously for an extended period (e.g., five or more years) as a key element of loss reporting that accounts for a growing share of reported losses over time.When we classify loss-reporting firms according to the extent of loss persistence, we find that for GAAP losers, the coefficient (from the returns regression) of adjusted earnings is significantly positive for all loss horizons and non-declining with longer loss horizons.This evidence surprisingly indicates that for firms whose losses are due to intangibles expensing, the length of these losses is largely immaterial to investors (Amazon, with close to ten years of persistent losses in the 1990s and early 2000s, is a prime example).This evidence is important, as it contradicts earlier research documenting that investors in firms reporting persistent losses consider the liquidation option of these firms more likely, thereby lowering the valuation relevance of persistent losses.The liquidation option argument also predicts that the intangibles of firms with persistent losses are considered worthless since the value of intangibles (such as R&D) in liquidation is negligible.Our evidence, however, shows that for GAAP losers, even those with long loss streaks, investors' valuation of R&D and other intangibles is significantly positive.Thus, GAAP losers are not considered by investors as liquidation prospects.
To validate the positive valuation effects of GAAP losses versus Real losses, we examine the productivity of loss firms' intangible investments, including technological innovation and human capital.We find that GAAP losers significantly outperform Real losers in patenting frequency, patent value, and patent lead time, all of which indicate greater innovation success.Surprisingly, GAAP losers also outperform similar profitable firms in innovation outcomes.Compared to other firms, GAAP losers also have more successful human capital strategies for attracting and retaining employees and, more importantly, improving employee productivity.These differences further demonstrate that the standard accounting reports of GAAP losers do not properly reflect their performance.
We cement our evidence on the differences between GAAP losers and Real losers by examining the divergence in their subsequent performance-an issue of considerable importance to investors.We find that GAAP losers are less likely to experience future declines in value triggered by liquidation, delisting, and further deteriorating losses than Real losers.Focusing on the subsequent loss reversals-firms reporting profits after one or several years of loss-we show that GAAP losers have a substantially higher likelihood of loss reversal than Real losers.Furthermore, GAAP losers have substantially better future stock performance than Real losers.In fact, in recent decades, the investment yields generated by GAAP losers are even higher than those generated by profitable firms.This highly surprising finding suggests that investors are largely unaware of, or unduly discount, firms whose reported losses are due to intangibles expensing.
Our study expands existing research on loss reporting by considering the roles of all internally generated intangible investments, not just R&D (the focus of prior studies).We identify two different types of loss firms, GAAP losers and Real losers.For these two types, the distinctions between illusory and real losses and their implications were not fully recognized in the literature examining loss reporting.Our study makes the following contributions to this literature.First, we show that, contrary to the widely accepted view that losses are less informative than profits, the reported losses of GAAP losers are as informative as the reported earnings of profitable firms when our adjustment for capitalized intangibles is included in commonly used valuation models.Second, we show that the liquidation option hypothesis, a key idea guiding prior research on the valuation of loss firms, does not apply to GAAP losers, which are more successful innovators and create more value from intangible investments than other firms, including profitable ones.Third, we expand existing research to compare GAAP losers and profitable firms.While prior studies have compared the earnings relevance of loss firms and profitable firms, a comparison of loss firms with profitable firms in innovation outcome and future performance is missing in prior literature.(Who would have expected loss firms to be even more successful innovators than profitable firms?)We find that GAAP losers are more innovative and have better future stock performance than profitable firms, particularly over the last 20 years.
Because we examine earnings adjusted for intangibles expensing as an alternative to reported GAAP earnings, our study is also related to research examining non-GAAP earnings, also known as pro forma earnings, street earnings, core earnings, or operating earnings.According to extant research in this area (e.g., Bradshaw and  Sloan 2002; Bhattacharya et al. 2003), the items that are excluded from non-GAAP earnings are mostly associated with unusual and non-operating losses (e.g., writeoffs and income-decreasing special items) and non-cash expenses (e.g., depreciation, amortization, and stock-based compensation costs).Recurring items, such as R&D, advertising, and SG&A expenses, which are the focus of our adjustment, are not found to be excluded (e.g., Black et al. 2018).Several intangibles-related items, such 1 3 All losses are not alike: Real versus accounting-driven reported… as goodwill amortization and write-offs of goodwill and acquired in-process R&D, are among the excluded items for some firms, but these exclusions are found to be discretionary and inconsistent over time and across firms (e.g., Bhattacharya et al.  2003; Kolev et al. 2008).Thus, our adjustment of undoing the expensing of recurring and internally generated intangibles via systematic capitalization and amortization for all firms is fundamentally different from the discretionary and inconsistent exclusions of selected items under non-GAAP earnings.The restoration of earnings relevance after our adjustment indicates the usefulness of excluding the biases of expensing intangible investments from reported earnings.
Our study is distinct from research examining the value-relevance of intangible expenditures.For example, Lev and Sougiannis (1996) find that the capitalization and amortization of R&D expenditure convey value-relevant information, despite their accounting treatment as current-period expenses.More recently, Barth et al.  (2023) find that the value-relevance of intangibles-driven items, such as R&D, advertising, and recognized intangibles, has increased over time, whereas earnings relevance has declined.We focus on the effect of treating internally generated intangibles as assets versus expenses on the sign of earnings by distinguishing between GAAP losers and Real losers and considering the implications of this distinction given investors' differential beliefs about the value and prospects of loss versus profitable firms.As intangibles expenditures are recognized solely for the purpose of measuring earnings under the current accounting standards, it is critical to understand how the recognition of intangibles expenditures as assets versus expenses would affect basic earnings information, such as firms' profit/loss status, and its value relevance.
The remainder of this paper proceeds as follows.Section 2 briefly reviews prior literature on loss reporting and highlights the distinguishing aspects of our study.Section 3 describes the sample and data used in this study.Section 4 lays out the results of our examination of how investors value different types of loss firms.Section 5 compares the productivity of intangible investments across different types of loss firms.Section 6 explores the implications of GAAP losses versus real losses for the future performance of loss firms.Section 7 concludes.

Prior literature on loss reporting
Several studies have recognized the role of increased R&D spending in recent decades, particularly for technology firms, in explaining the increase of loss reporting (e.g., Joos and Plesko 2005; Darrough and Ye 2007). 1 Our study extends this research by considering all internally generated intangible investments that include but are not limited to R&D expenses.Firms' internal investments in brands, business processes, IT, and human capital are included in SG&A expenses with the same loss-inducing effect as R&D but are larger than R&D expenses by several times and are relevant for all firms, not just technology firms. 2 More importantly, unlike prior research that simply compares R&D firms with non-R&D firms or high-R&D firms with low-R&D firms, we focus on the complex dichotomous effect of undoing the immediate expensing of all intangible investments on firms' losses-it turns the losses of some firms to profits but leaves the loss status of others unchanged.This focus allows us to distinguish between the different valuation implications of GAAP losses and real losses, an important yet previously unexamined issue for loss firms. 3rior research also examined the valuation of loss firms.Hayn (1995) finds that losses are less informative than profits.This result warrants further examination, as it essentially indicates that the information content of earnings for loss firms and profitable firms is not comparable, which is inconsistent with the comparability principle defining high-quality accounting information.Subsequent research shows that R&D expense is priced more positively than other loss components (Collins et al. 1997;  Joos and Plesko 2005; Darrough and Ye 2007; Franzen and Radhakrishman 2009;  Wu et al. 2010).These studies, however, do not examine whether adjusting losses for R&D expenses can restore the value relevance of losses to the same level as profits.
Some prior research examines whether the value relevance of losses versus profits is enhanced by including other accounting variables, such as book value of equity, in the valuation model (Collins et al. 1999). 4Collins et al. find that a statistically significant and economically meaningful difference in the value relevance of losses and profits remains despite the increased value relevance of losses (and decreased value relevance of profits) when book value is included in the valuation model.(They write that "Figure 2 highlights the difference between the market's valuation 2 During 1980-2018, the ratio of SG&A expenses to R&D expenses is approximately 6.7:1 and 2.2:1 for all firms and technology firms, respectively.We define technology firms as firms with three-digit SICs of 283 (drugs), 357 (computer and office equipment), 362 (electrical), 366 (communication equipment), 367 (electronic components and accessories), 381 (search, detection, navigation, guidance, and aerospace systems), 382 (lab apparatus, optical, measurement, and control instruments), 384 (surgical, medical, and dental instruments), 385 (ophthalmic products), and 737 (computer programming and data processing).The prominence of SG&A implies that research examining only R&D substantially understates the effect of expensing internally generated intangible investments on loss reporting (Enache and Srivastava 2018). 3The phenomenon of firms converting GAAP losses to profits using alternative earnings measurement has been investigated by prior research.Several studies examine the information quality of non-GAAP profits for firms reporting GAAP losses (e.g., Black et al. 2012; Leung and Veenman 2018).Collins et al. (1999) argue that the usefulness of book value for valuing loss firms stems from the role of book value as a proxy for expected future normal earnings and for loss firms' liquidation option.However, the relevance of book value in the valuation of loss firms with high intangibles is likely to be low for two reasons.First, the accounting bias of intangibles expensing that affects reported earnings also affects book value (e.g., book value is biased downward for firms with increasing internal investments in intangibles), rendering the usefulness of book value as a predictor of future normal earnings questionable.Second, given the high uncertainty and low liquidation value of most internally developed intangibles, book value may not meaningfully inform loss firms' liquidation option, particularly when book value is measured with biases.

3
All losses are not alike: Real versus accounting-driven reported… of negative versus positive earnings even after controlling for book value of equity....The importance of earnings differs for profit firms as compared to loss firms even when using the more properly specified model of the price-earnings relation that includes both earnings and book value of equity" (p.46).)Commenting on the findings from this line of research, Ciftci and Darrough (2015) acknowledge, "The fact remains that earnings still have low value relevance for loss firms compared to profit firms" (p.139).In contrast, our results indicate that our adjustment for the capitalization of intangibles closes the earnings valuation gap between GAAP losers and profitable firms.We show that the reported losses of GAAP losers are in fact as informative as the reported earnings of profitable firms once the capitalization and amortization of firms' internally generated intangible investments are included in our valuation model.
Existing studies also examine how the valuation of losses varies with other factors, such as investor sentiment (Riedl et al. 2021) and loss persistence (Joos and  Plesko 2005).Joos and Plesko and subsequent research (e.g., Li 2011) focus on transitory versus persistent losses, defined as losses from the highest and lowest quartiles of the loss reversal probability distribution, respectively, excluding firms from the two middle quartiles of the distribution, or half of the available sample.They find that, over time, the distinction in valuation between transitory and persistent losses has been blurred by the increasing R&D expense as a driver for persistent losses.Our classification of losses-GAAP losses versus real losses-is based on whether adjusting for the capitalization and amortization of R&D and SG&A expenses turns firms' reported losses into profits, which is a key to distinguishing illusory losses from real losses.Moreover, our classification scheme is applicable to all loss firms and is not affected by changes in firms' investment trends over time. 5 key argument in the research on the valuation of loss firms is the liquidation option hypothesis, namely the possibility of liquidating loss firms and redeploying their funds elsewhere (Hayn 1995). 6Consistent with this argument, several studies find that losses and loss components less relevant for the liquidation option are more informative (Collins et al. 1997, 1999; Darrough and Ye 2007; Ciftci and Darrough  2015).Lawrence et al. (2018) conclude that the liquidation option explains why losses are less persistent than profits.Contrary to the prediction of the liquidation option hypothesis, our evidence shows no decrease in the value relevance of frequent and uninterrupted GAAP losses.Linking the liquidation option to firms' real actions, Pinnuck and Lillis (2007) find that loss firms are more likely to lay off employees.On the contrary, we find sustained increases in employee hiring and retention at firms with intangibles-driven losses.These results further demonstrate the fundamentally different nature of illusory GAAP loss and real loss, which is documented in the literature for the first time by our study.
We also depart from prior research focusing on R&D expenditure, which is subject to two limitations.First, R&D expenditure is a measure of the input of intangible investments rather than the output.Given the high uncertainty of R&D, it is important to distinguish between the input and output of R&D.Second, the positive pricing of R&D is ambiguous, as it can reflect either investor expectations of the future benefits of R&D or the risk premium associated with R&D.Our study extends prior research by examining the output and productivity of firms' investments in R&D.We find that GAAP losers create greater value from technological innovations than even profitable firms with similar characteristics, confirming the expected benefits of intangible investments.This evidence is new to the literature and explains why the liquidation option does not apply to GAAP losers, which are mostly dynamic and successful innovators.
Our study also aims to shed more light on the future performance of loss firms.There is limited prior evidence on the future performance of loss firms (e.g., ex ante estimates of the likelihood of loss reversal in the year (Joos and Plesko 2005) and one-year-ahead abnormal returns related to investors' underreaction to persistent versus transitory losses (Li 2011)).Given the growing frequency of loss reporting even in a booming economy, it is crucial to understand the implications of increases in loss reporting.We provide evidence on loss reversal over longer horizons (three to five years) and find that GAAP losers are more likely to become profitable in the future than Real losers.This difference in future performance among loss firms has not been examined by prior research.Moreover, we examine how the future performance of loss firms compares with that of profitable firms, an issue not examined by prior research.We find that GAAP losers have better future stock performance than profitable firms (and, of course, than Real losers).Our evidence thus indicates that widely reported and analyzed GAAP earnings do not properly reflect the intrinsic value and performance of GAAP losers, a finding that should be of concern to investors, managers, and accounting standard setters.

Sample data, variable measurement, and methodology
Our sample includes all Compustat firms with required data for 1980-2018.We define loss (profitable) firms as firms reporting negative (positive) annual income before extraordinary items (IB). 7Our measure for firms' unrecorded intangible assets is based on information on R&D and SG&A expenses.SG&A expenses contain many investments in intangibles (e.g., brands, IT, human capital, and organizational capital). 8We capitalize and amortize intangibles following prior research on 7 Compustat defines income before extraordinary items (IB) as the income of a company after all expenses, including special items, income taxes, and minority interest, but before provisions for common and/or preferred dividends.It does not, however, reflect discontinued operations or extraordinary items. 8In its 2019 annual report, Pfizer indicated that "selling, information and administrative costs are expensed as incurred.Among other things, these expenses include the internal and external costs of marketing, advertising, shipping and handling, information technology and legal defense."Advertising expense accounts for approximately 20% of Pfizer's selling, information, and administrative costs.Pfizer's marketing, advertising, and information technology expenses are likely potent drivers of the firm's intangibles.

3
All losses are not alike: Real versus accounting-driven reported… the pattern of future economic benefits attributed to intangible investment. 9We also consider the practices of intangibles amortization observed in the financial reporting of public companies.
To obtain our estimates of R&D amortization expense, we follow the model of Lev and Sougiannis (1996) (hereafter, LS), which infers industry-specific useful lives and amortization rates of R&D from the estimated coefficients of a distributed lag model.This model regresses operating income, adjusted for current period R&D expenditure and advertising expense, on current and lagged R&D variables obtained from a first-stage regression on industry-average R&D spending rate to address the simultaneity issue.A straightforward application of LS's approach to all firms in our sample, however, is not feasible because its estimation model requires at least ten annual lags of R&D data, which are not available for many firms in our sample.LS's approach is particularly restrictive for young firms with a higher likelihood of loss reporting, a large and key component of our sample that drives the phenomenon of increasing loss reporting in recent decades. 10To circumvent this problem, we implement LS's estimation procedure for all firms in our sample that meet their data requirements and obtain industry-specific coefficients on R&D lagged variables.We then determine industrywide amortization periods based on the number of R&D lags with positive and statistically significant coefficients found in industrylevel regressions (e.g., three-year amortization period for software industry based on three significantly positive R&D variables for years t, t-1, and t-2).We then apply these industrywide R&D amortization periods to all firms in our sample and estimate firms' annual R&D amortization expense using straight-line amortization.
Our approach for estimating the R&D amortization amount is different from the method of LS, which uses year-specific amortization rates derived from the coefficients of R&D variables (e.g., yearly amortization rates of 13.5%, 20.7%, 24.0%, 24.4%, and 17.4% for scientific instruments with a five-year amortization period, Table 3 in LS).We choose straight-line amortization over the method of LS for two reasons.First, the inverted U-shaped pattern of yearly amortization rates (i.e., smoothly increasing and then decreasing amortization rates over time) estimated by LS reflects the Almon lag structure, an assumption adopted by LS in estimating regressions with distributed lags rather than actual yearly variations of R&D amortization rates.Second, in practice, the straight-line method is the most commonly 9 Economic research offers a variety of estimates of R&D amortization rates.Many studies assume a uniform amortization rate, such as 15% (e.g., Griliches and Mairesse 1984; Corrado et al. 2009) or 20% (e.g., Chan et al. 2001), across all industries.Pakes and Schankerman (1984) infer R&D amortization rates from the patterns of patent renewal over time, yielding an estimate of 25%.Using a market value approach, Hall (2007) finds R&D amortization rates ranging from 20 to 40% for manufacturing firms.Industry-specific R&D amortization periods estimated by Li and Hall (2018) range from roughly three years for software firms to five to six years for biotech and pharmaceutical firms.Despite the lack of consensus, available estimates of R&D amortization rates seem to reflect the variation of innovation cycles across industries and underlying the economic process of R&D amortization (e.g., higher annual amortization rates for industries and technologies with shorter innovation cycles).
10 LS also restrict their sample to firms with a ratio of annual R&D expenditure to sales of no less than 2%.During our sample period of 1980-2018, a total of 23,626 firm-years, representing less than 14% of our total sample, meet all the requirements of LS. used amortization method for intangibles.11Therefore, we amortize R&D using the straight-line method, and our industry-specific amortization rates are commensurate with the R&D amortization periods estimated by the LS model.
There is less available research on the amortization of SG&A expense.Some studies add a fixed percentage of current SG&A expense (e.g., 30%) that is expected to create valuable intangibles to their estimates of organizational capital (e.g., Hulten  and Hao 2008; Eisfeldt and Papanikolaou 2014; Peters and Taylor 2017), implicitly amortizing some portion of capitalized SG&A expense.Banker et al. (2019) follow the approach of LS to estimate the future value of current and past SG&A expenses.Like LS, Banker et al. restrict their estimation to firms with at least eight years of SG&A expense data, which is not feasible for the majority of our sample firms.For several industries, the approach of Banker et al. yields estimates that are not economically useful (e.g., significantly negative or statistically insignificant coefficients on current and past SG&A expenses).
We nonetheless implement the approach of Banker et al. to our sample firms with the required data to obtain industry-specific amortization rates wherever appropriate.For industries with no valid estimates available, we supplement the approach of Banker et al. with our own estimates of industry-specific amortization rates based on available Compustat data on firms' amortization expense for balance-sheet or purchased intangibles other than goodwill.12This data is available for a relatively large number of firms in our sample and covers all major industries.Assuming that most firms use the straight-line method to amortize purchased intangibles, we can infer amortization rates from the amount of purchased intangibles and firms' annual amortization expense.For example, the ratio of annual amortization expense to the amount of purchased intangibles reflects firms' annual amortization rates and other factors, such as acquisition, divestiture, and impairment.When aggregated across all firms in an industry, the average ratio is expected to approximate industry-specific amortization rates.Amortization periods based on these ratios range between five and eight years, consistent with observed practices by real companies.For example, CISCO Systems Inc. reported useful lives of five and eight years (five, six, and seven years) for the company's purchased customer lists (relationships), amortized using straight-line method, in its 2015 (2017) annual report.
Existing methods using regressions with distributed lags to estimate the amortization rates of R&D and other intangibles, however, may not do a particularly satisfactory job, because firm-level accounting expense data, such as R&D and SG&A (as a percentage of sales), have low intertemporal variations.The existing models also rely on strong, possibly unrealistic, and difficult-to-verify assumptions about the distribution of coefficients on lagged variables (Mead 2007).Thus, it is 1 3 All losses are not alike: Real versus accounting-driven reported… important to assess the robustness of the empirical results derived from such models.One approach used by some prior studies is to examine whether their results are sensitive to alternative amortization rates (e.g., 15% in place of 25%).In Section 4.4, we use a different approach to assess the robustness of our main results.This approach is based on how the interaction between firms' intangible investment intensity (e.g., high versus low) and the reporting choice of intangibles (expensing versus capitalization) affects firms' reported and economic profitability (we elaborate on the details of this test and its results in Section 4.4, where we also assess less estimation-intensive and easier-to-apply approaches for capturing the phenomenon of GAAP-driven versus real losses examined in this study).
To adjust earnings for the bias of expensing intangibles, we add back to reported earnings (IB) the amount of current year's R&D and SG&A expenditures net of annual R&D and SGA amortization expenses, estimated using the approaches described above.Using this measure of earnings adjusted for intangibles (AIB), we classify all loss firms into two categories: GAAP losers and Real losers.GAAP losers have negative reported earnings (IB) but positive adjusted earnings (AIB).Real losers are firms whose reported earnings and adjusted earnings are both negative. 13Thus, GAAP losers would have been profitable had they recognized intangibles as assets in their financial statements, whereas Real losers would still be losers regardless of how they accounted for their intangible investments.
In our tests of earnings relevance, we regress firms' annual stock returns, measured over the 12-month period commencing three months after the beginning of a fiscal year, on both the level and change of earnings.Following prior literature on the returns-earnings association, we interpret the sum of the regression coefficients on earnings level and change as the earnings response coefficients (ERCs), a standard measure for the value relevance of earnings information.For each type of loss firms, we compare the ERCs of reported earnings and adjusted earnings to examine whether the adjustment for intangibles improves the value relevance of earnings.

The rise of loss reporting and fall of earnings relevance
We begin our tests with a replication of the main results of Hayn (1995), which documented the lack of information content for losses.Table 1 reports the results of this replication and shows that for the period examined by Hayn, 1962-1990, our results are consistent with Hayn's that the coefficient on earnings level and the associated adjusted R 2 s for loss firms are both indistinguishable from zero.

Table 1
Summary statistics from regressions of stock returns on earnings levels and earnings changes for 1962-1990 and 1991-2018   The dependent variable is the return over the 12-month period commencing three months after the beginning of the firm's fiscal year for which earnings and earnings change are measured.Earnings and earnings change are deflated by the firm's stock price at the end of the prior fiscal year Early sample period: 1962-1990   Recent sample period: 1991-2018 The results from the regression of earnings changes are also similar, with a regression coefficient of 0.50 and adjusted R 2 of 4.1% (versus 3.7% in Hayn's study).Our results for profitable firms are also similar to Hayn's, showing that profitable firms have larger regression coefficients and adjusted R 2 s than loss firms.
The results for the years subsequent to Hayn's sample period, 1991-2018, however, show significant changes in the relevance of earnings information over the past three decades.While the level of earnings continues to be irrelevant for loss firms in the more recent years, the value relevance of earnings changes for loss firms has declined considerably, from 0.50 to 0.32 (4.1% to 1.0%) for the regression coefficient (the R 2 s of the model).Similar decreases in the relevance of earnings for profitable firms are also observed in the recent results.These recent trends echo the key findings from Lev and Gu (2016) and Gu and Lev (2017), which show a broad decline in the usefulness of financial accounting information in the last 30 years, particularly for GAAP earnings.More importantly, the results in Table 1 show that the percentage of loss firms increased by more than 60%, from 19% in the pre-1990s period to 31% in the post-1990s period.With close to half (47%) of all publicly traded firms currently reporting losses that convey negligible useful information, the increased prevalence of loss reporting is obviously a major contributor to the erosion of accounting relevance in recent decades, raising important questions about the drivers and consequences of this phenomenon.In the remainder of our study, we explore these drivers and consequences for the period 1980-2018, during which the proliferation of loss reporting has accelerated.

An anatomy of losses
We examine two distinct drivers for GAAP losses: losses driven by firms' internal investments in intangibles that are expensed under current accounting rules versus real economic losses due to underperforming businesses.While both types of losses are reported identically on firms' financial reports, we argue that they are associated with fundamentally different circumstances and implications.The former reflects the effect of GAAP-mandated expensing of growth-oriented and competitivenessenhancing investments, whereas the latter is simply a reflection of firms' money-losing businesses.We empirically distinguish between GAAP-induced losses and real losses with our procedure of adjusting reported earnings for capitalized intangibles (described in Section 3).We expect adjusted earnings free of the bias of intangibles expensing to be more informative than reported earnings.Table 2, Panel A reports the descriptive statistics of our sample.Over the entire sample period of our primary tests (1980 to 2018), profitable (loss) firms account for 71% (29%) of sample firms.Approximately 27% of loss firms are classified as GAAP losers, for which losses would be reversed once the economic value of intangibles is taken into account via capitalization and amortization.GAAP losers have larger sales and higher sales growth than Real losers.The industry compositions of GAAP Losers and Real losers are similar, with both samples including firms from all types of industries.Compared to profitable firms, a higher percentage of loss firms are from technology (1) (2) (1) ( (1) (2) 1 3 All losses are not alike: Real versus accounting-driven reported… The dependent variable of the annual regressions is the return over the 12-month period commencing three months after the beginning of the firm's fiscal year for which earnings and earnings change are measured (RETURN).IB is reported income before extraordinary items, ΔIB is the annual change of IB.AIB is firms' adjusted earnings, computed by adding R&D expenditure (RDEXP) and SG&A expenses (SGAEXP) and subtracting the amount of R&D amortization and SG&A amortization to reported income before extraordinary items (IB).ΔAIB is the annual change of AIB.Independent variables are deflated by the stock price at the beginning of the return period.ERC for reported earnings (adjusted earnings) is the sum of the regression coefficients on IB and ΔIB (AIB and ΔAIB).Because the variable for SG&A expenses in Compustat often contains R&D expenditure (Peters and Taylor 2017), we subtract the amount of R&D expenditure from SG&A expenses wherever appropriate a Difference in ERC of adjusted earnings (2) and reported earnings (1) for the same type of firms (statistical significance for one-tailed test) b Difference in ERC of adjusted earnings (2) between GAAP losers and profitable firms or Real losers (statistical significance for one-tailed test) (1) (2) (1) ( (1) (2) industries, such as biotech and pharmaceutical, semiconductor, and software, reflecting the significant role of intangibles expensing in loss reporting.
In Panel B, we report the average estimates from 39 annual regressions of stock returns on the level and change of reported earnings and adjusted earnings, respectively. 14Each regression includes four-digit SIC industry dummies.For comparison purposes, we report the regression results for all firms, profitable firms, and loss firms, and we further separate loss firms into GAAP losers and Real losers.For each group of firms, we report, in Panel C, the results for examining the difference in the ERCs of reported earnings (IB) and intangibles adjusted earnings (AIB) to gain further insights into the usefulness of undoing the expensing of intangibles.Confirming the results of Table 1, Panel B shows that the reported earnings of loss firms are less informative than those of profitable firms.The adjustment for intangibles does not improve the earnings relevance of profitable firms.As shown in Panel C, the ERC of adjusted earnings, 0.268, is not statistically different from that of reported earnings, 0.305.
For GAAP losers, Panel B shows that the ERC of intangibles-adjusted earnings, 0.314, is more than double that of reported earnings, 0.153 (the difference is highly significant as shown in Panel C). 15 For Real losers, however, adjusting for intangibles decreases earnings relevance from 0.092 to 0.067, a statistically insignificant change.Surprisingly, the ERC of GAAP losers, 0.314, is significantly larger than that of profitable firms, 0.268.Thus, our results demonstrate that prior evidence showing that losses are less informative than profits is relevant only for firms with real economic losses.For firms reporting losses as a result of the expensing of investment in intangibles, adjusted earnings for intangibles expensing are highly informative-at least as informative as the earnings of profitable firms.
To shed more light on how the treatment of intangibles affects the value relevance of earnings, we expand the regressions for reported earnings to include the level and change of R&D capital (NetR&D and ΔNetR&D) and intangibles reflected in SG&A expenses (NetSG&A and ΔNetSG&A) such as IT and brand enhancement.The results of these regressions are presented in Table 3. Panel A shows that once the intangibles variables are included, reported earnings become highly informative for all firms, including loss firms.This is in sharp contrast to earlier evidence showing that earnings of loss firms are not informative.We find that GAAP losers have the highest ERCs, 0.490, followed by profitable firms (0.313) and Real losers (0.110).The coefficients on intangibles, particularly for GAAP losers, are mostly positive and significant, confirming the value-enhancing effect of intangibles and the importance of including them in the regression.Panel B shows that GAAP losers have significantly higher ERCs than profitable firms (0.490 versus 0.313).The intangibles of GAAP losers are also valued more highly by investors than those 1 3 All losses are not alike: Real versus accounting-driven reported… Table 3 Annual regressions of return on R&D and SG&A intangibles after controlling for earnings and earnings change The dependent variable of the annual regressions is the return over the 12-month period commencing three months after the beginning of the firm's fiscal year for which earnings and earnings change are measured (RETURN).NetR&D is the amount of reported R&D expenditure minus the amount of R&D amortization.ΔNetR&D is the annual change of NetR&D.NetSG&A is the amount of reported SG&A expense minus the amount of SG&A amortization.ΔNetSG&A is the annual change of NetSG&A.IB is firms' reported income before extraordinary items.ΔIB is the annual change of IB.Independent variables are deflated by the stock price at the beginning of the return period.ERC is the sum of the regression coefficients on IB and ΔIB.In Panel B, we compare the coefficients on intangibles-related variables and ERC between GAAP losers and profitable firms or between GAAP losers and Real losers.Specifically, R&D refers to the sum of the regression coefficients on NetR&D and ΔNetR&D; SG&A refers to the sum of the regression coefficients on NetSG&A and ΔNetSG&A; and INTANGIBLES refers  of profitable firms-an important insight, given the common belief that loss firms waste money on intangible investments.Taken together, the evidence in Table 3 demonstrates that prior evidence showing different value relevance of profits and losses is driven by the omission of intangible assets in the return model specification.Once intangibles are included in the model, we find that reported earnings of GAAP losers and profitable firms are similarly informative.We also find that GAAP losers are unique in that their reported earnings (losses) are the most informative among all firms when they are assessed in the context of intangible investments.When intangibles are included in our models, earnings (losses) of Real losers also become informative, though less informative than the earnings of GAAP losers.

Persistent losses and earnings relevance
A key element in the continuing rise of loss reporting over time, as shown in Fig. 1, is the growing prevalence of persistent losers, namely firms reporting operating losses continuously for an extended period of time.Indeed, we find that the percentage of firms reporting five (eight) or more consecutive years of losses grew from 8% (2%) in 1991 to 20% (9%) in 2018.Prior research argues that persistent losses are less informative than temporary losses, as the likelihood of exercising a liquidation option tends to increase with the frequency of losses (Hayn 1995).Since most internally-developed intangibles are expected to have negligible value when firms are in liquidation (Lev 2001), the liquidation-based argument predicts lower valuation for intangibles and lower value relevance of earnings adjusted for intangibles, for firms with more persistent losses.We examine whether these predictions are valid for firms reporting persistent losses due to continuing investment over time in intangibles versus continuously underperforming businesses irrespective of intangible investment.We divide each type of loss firms into four groups by loss persistence: firms with one to two, three to four, five to six, or seven or more consecutive years of losses.Firms with fewer consecutive years of losses are more likely to be temporary losers.
Table 4 reports descriptive statistics for GAAP losers (Panel A) and Real losers (Panel B) in each group of loss persistence.We find that GAAP losers with seven or more years of losses have higher R&D investment level and change and higher increases of SG&A expenses than GAAP losers with fewer years of losses, which clarifies the main reason for the persistence of GAAP losses-namely, the firms' commitment to investment in intangibles despite more frequent losses.
Table 5 reports the ERCs from our pooled regressions of returns on earnings (IB), earnings change (ΔIB), adjusted earnings (AIB), and adjusted earnings change (ΔAIB), for GAAP losers and Real losers with different loss persistence.For each type of losses, we examine the change in earnings relevance when the loss horizon grows and losses become more persistent.We assess the statistical significance of this change by regressing the estimated ERCs on a loss persistence variable that takes the value of 1 for firms with one to two years of loss, 2 for firms with three to four years of loss, and so on.These regression results are reported below the estimates from the main regressions of returns on earnings and earnings change.

3
All losses are not alike: Real versus accounting-driven reported… IB is firms' reported income before extraordinary items.ΔIB is the annual change of IB.AIB is firms' adjusted earnings, computed by adding R&D expenditure and SG&A expenses and subtracting the amount of R&D amortization and SG&A amortization to reported income before extraordinary items.ΔAIB is the annual change of AIB.NetR&D is the amount of reported R&D expenditure minus the amount of R&D amortization.ΔNetR&D is the annual change of NetR&D.NetSG&A is the amount of reported SG&A expense minus the amount of SG&A amortization.ΔNetSG&A is the annual change of NetSG&A.RETURN is firms' return over the 12-month period commencing three months after the beginning of the firm's fiscal year for which earnings and earnings change are measured.SALES is firms' reported annual sales revenue.Sales growth is the percentage growth rate of SALES.MV is firms' market value at the fiscal year-end.ROA is firms' return on assets, computed as the ratio of income before extraordinary items to average total assets.Loss years is the number of consecutive years with losses.IB, ΔIB, AIB, ΔAIB, NetR&D, ΔNetR&D, NetSG&A, and ΔNetSG&A are deflated by the stock price at the beginning of the return period 1 3 All losses are not alike: Real versus accounting-driven reported…

Table 5
Value relevance of earnings for firms with varying loss persistence The dependent variable of the regressions to estimate ERCs is return over the 12-month period commencing three months after the beginning of the firm's fiscal year for which earnings and earnings change are measured (RETURN).IB is firms' reported income before extraordinary items, ΔIB is the annual change of IB.AIB is firms' adjusted earnings, computed by adding R&D expenditure and SG&A expenses and subtracting the amount of R&D amortization and SG&A amortization to reported income before extraordinary items.ΔAIB is the annual change of AIB.ERC for reported earnings (adjusted earnings) is the sum of the regression coefficients on IB and ΔIB (AIB and ΔAIB).Panel A shows that the ERCs of reported earnings of GAAP losers decline significantly from 0.111 for firms with one to two years of loss to 0.037 for those with seven or more years of loss.A statistically significant decline in the ERCs of reported earnings, from 0.185 to -0.003, is also found for Real losers, as shown in Panel B. Thus, as predicted by the liquidation-based argument, reported earnings of all loss firms become less informative as losses become more frequent and persistent.The results for adjusted earnings, however, show a different pattern.In Panel A, the relevance of adjusted earnings increases significantly for GAAP losers when their loss horizon grows.The ERCs of GAAP losers with seven or more years of losses are 0.324, compared to 0.280 (0.247) [0.240] for firms with five to six years (three to four) [one to two years] of losses.Hence, contrary to the prediction of the liquidation-based argument, the relevance of adjusted earnings is not lower for GAAP losers with more consecutive years of loss reporting.This result also contrasts the result for Real losers (Panel B).For them, a relatively large and significant decrease in the relevance of adjusted earnings, from 0.126 to -0.016, is found as the loss horizon increases from one to two years to seven or more years.

Robustness check
Our classification for GAAP losers and Real losers relies on earnings adjusted for intangibles capitalization and amortization.Following prior research, we use distributed lag regression models to estimate intangibles amortization rates.Given the limitations of these regressions (e.g., they are applicable to only firms with long time-series of data and require strong assumptions to operationalize), it is important to examine the robustness of our results using alternative methods of identifying GAAP losers and Real losers, particularly methods that do not require the use of restrictive regressions.By definition, GAAP losers are firms that report losses but would have been profitable had their intangibles been recognized as assets.Real losers, on the other hand, would not be profitable regardless of the treatment of their intangibles due to their weak performance.Accordingly, we identify GAAP losers and Real losers based on a joint consideration of their intangible investment and reported losses.
Specifically, for each industry-year, we simultaneously sort all loss firms into five portfolios by their combined R&D and SG&A expenses as a percentage of sales, ranked from low to high, and 11 portfolios by the ratio of reported earnings to sales, ranked from the lowest ( the most negative), to the highest (the least negative).We identify Real losers from the upper-left region of this 5 × 11 matrix, which is populated by firms with relatively large reported losses and low intangible investment. 16or example, all firms in the leftmost column (firms with the highest losses regardless of their intangible investment) and firms from the second leftmost column except for those in the bottom row (firms with the second highest losses but not the 1 3 All losses are not alike: Real versus accounting-driven reported… largest intangible investment) are classified as Real losers.Similarly, firms from the intersections of the third, fourth, and fifth leftmost columns and the top three, two, and one rows are also classified as Real losers.Reversing the direction of this classification scheme would identify GAAP losers as firms from the lower-right area of the matrix, which is populated by firms with relatively small reported losses and large intangible investment.
This classification approach is objective and can be easily applied to identify loss types.While this approach may still misclassify some loss firms, we minimize the misclassification error by focusing on firms with the most extreme values of reported earnings and intangible investment.These firms likely have consistent and unambiguous loss classifications under most intangibles amortization rates.More importantly, this approach is based on directly observable accounting signals and does not require the use of estimates that may introduce measurement errors and classification biases.The tradeoff, however, is the reduction of sample size for both types of loss firms due to the exclusion of firms with medium levels of reported earnings and intangible investment.The loss status of these firms is potentially ambiguous and likely sensitive to the choice of intangibles amortization rates. 17sing this alternative approach, we identify 15,219 GAAP losers and 15,315 Real losers.We find essentially the same pattern for the earnings relevance of these GAAP losers and Real losers.In tests similar to those reported in Table 3, we regress returns on reported earnings, earnings change, and intangibles variables.Because our approach does not provide intangibles capitalization and amortization, we use reported R&D and SG&A expenses, including their levels and changes, as proxies for intangibles.These proxies are likely noisy and are thus expected to bias against finding our results for earnings relevance.In untabulated tests, we find that, consistent with the results shown in Table 3, the reported earnings of GAAP losers have substantially higher relevance than those of profitable firms and Real losers when intangibles are included in the regression (the ERCs are 0.334, 0.262, and 0.110 for GAAP losers, profitable firms, and Real losers, respectively).Thus, our conclusion that GAAP losers have higher earnings relevance than Real losers and profitable firms continues to hold under this alternative approach to classifying loss firms.
We also examine the sensitivity of our results to simplified schemes of capitalizing and amortizing intangibles that do not require the use of data-constrained and assumption-dependent methodologies for estimating intangible amortization rates.We choose industry-specific amortization rates for R&D suggested by prior research (e.g., Li and  Hall 2018) examining the cycle of innovation in different industries (e.g., three years for software firms and five years for biotech and pharmaceutical firms).We capitalize one-third of SG&A expense and add it back to earnings, based on prior studies finding that, on average, the investment portion of SG&A spending accounts for about 30% of firms' total SG&A expenses (e.g., Hulten and Hao 2008;Eisfeldt and Papanikolaou 17 Including these unclassified firms in the sample of either GAAP losers or Real losers does not change the tone of our results.The ERCs of GAAP losers (Real losers) are modestly lower (higher) with these firms included but are still significantly higher (lower) than the ERCs of Real losers (GAAP losers).
2014; Peters and Taylor 2017; Enache and Srivastava 2018; Ewens et al. 2022) and that the lives of many of these investments, such as brands, are long.To the extent that this approach does not fully capture some of the industry-specific timing of future benefits attributed to intangibles and hence misclassifies loss firms, the resulting errors are expected to work against finding systematic differences between GAAP losers and Real losers.Yet the results based on this alternative approach are very similar to our reported results and fully support all of our current conclusions. 18Given its simplicity and intuitiveness, this approach to identifying GAAP loss versus real loss is particularly useful to analysts and investors who need to distinguish accounting-driven losses by firms making value-enhancing intangible investment from real economic losses by firms whose underperformance is unrelated to intangible investment (Morgan Stanley 2022).

Value relevance of adjusted earnings for profitable firms
The results in Table 2, Panels B and C show that, in contrast to GAAP losers, adjusted earnings of profitable firms are not more informative than reported earnings, suggesting a lack of value relevance for the intangibles of profitable firms beyond the information contained in reported earnings.The evidence in Table 3 indicates that R&D-related intangibles of profitable firms are not priced favorably by investors (the coefficients on NetR&D and ΔNetR&D, -0.116 and 0.077, are statistically insignificant).Thus, the lower value relevance of adjusted earnings for profit firms seems to be driven by the less favorable pricing of these firms' R&D-related intangibles.This is consistent with the evidence of Franzen and Radhakrishman (2009) showing that R&D expense is negatively (positively) associated with stock prices of profit (loss) firms when earnings information is accounted for.They attribute this result to reported earnings of profit firms (loss firms) being informative (not being informative) of the future benefits of R&D activity, hence the relevance of R&D for loss firms but not for profit firms.
The pricing of R&D for profit firms may also reflect investor concerns about managerial opportunism in the decision-making of intangibles spending-a form of real earnings management-and its adverse effects on firms' long-term success.Prior research has shown that firms opportunistically reduce R&D expenditures to boost reported performance and meet short-term earnings targets at the expense of long-term competitiveness (e.g., Dechow and Sloan 1991; Baber et al.

3
All losses are not alike: Real versus accounting-driven reported… 1991; Bushee 1998; Bens et al. 2003). 19 Burgstahler and Dichev (1997) suggest that earnings management via reduction of discretionary expenses such as R&D is more prevalent among profit-reporting firms, as a disproportionately high number of firms are found to avoid reporting negative earnings.In contrast, real earnings management by GAAP losers seems less likely.By construction, GAAP losers are firms that commit to high intangible investment intensity and likely forgo the opportunity to report profit by underinvesting in intangibles. 20Consistent with the view that profit firms are more likely to engage in real earnings management than GAAP losers, we find that the percentage of firms underinvesting in R&D relative to industry peers is higher among profit firms than GAAP losers.Investor awareness of profit firms' high propensity to opportunistically reduce R&D expense is likely to dampen market expectations for the future benefits of R&D relative to what is already reflected in current earnings, hence the insignificant coefficient on R&D-related intangibles of profit firms once current earnings are accounted for.
To shed further light on the value relevance of adjusted earnings for profit firms versus GAAP losers, we compare the future performance of profit firms and GAAP losers that are matched on adjusted earnings (AIB), size, industry, and year.We focus on the growth rates of comprehensive performance indicators, such as future earnings and cash flows from operations, for the subsequent three years (year t + 1 through year t + 3), because unlike the levels of earnings and cash flows from operations, the growth rate is less affected by the accounting bias induced by the immediate expensing of intangibles.In untabulated tests, we find that for the subsequent three years, GAAP losers have substantially higher growth rates of earnings and cash flows from operations than profit firms. 21These results further indicate that adjusted earnings have different implications for the future performance of GAAP losers and profit firms.
19 Roychowdhury (2006) shows that real earnings management includes firms manipulating SG&A and R&D expenses.Managers' ability to manipulate these expenses is constrained by the extent to which investors can see through the manipulation.Many SG&A spending decisions, such as advertising and business systems, are relatively more observable, rendering managers' opportunistic reductions of SG&A expenses difficult to defend (e.g., it is not easy for managers to justify the decision to reduce the spending on advertising and online sales channels when competitors are maintaining or increasing their such spending).Due to the secretive nature of firms' R&D efforts, it is easier for managers to justify the decision to reduce R&D spending by citing factors that are hard to verify (e.g., changed direction of future innovation).Moreover, the adverse effects of reduced SG&A expenses can be reversed by future increases in spending that yield immediate benefits (e.g., hiring new employees with more updated skills and deploying improved business and information systems).The consequences of R&D spending reduction, however, are more serious and less reversible, due to the nature of technological innovations (e.g., a firm improperly reducing R&D spending faces the risk of missing a great opportunity of developing a key technology that can be used to blunt other firms' competitiveness).The relative ease in hiding the true motive of R&D spending reduction and the greater economic damage associated with the reduction suggest that investors have greater concerns about the adverse consequences of R&D spending reductions. 20Given the closeness of GAAP losers' reported earnings to profit (e.g., median earnings of -0.047) and their relatively high intangibles expenditures, a considerable number of GAAP losers could have been profitable had they chosen to moderately reduce their intangibles spending. 21For example, for year t + 1, the median growth rates for GAAP losers' earnings and cash flows from operations are 1.99% and 0.9%, whereas the median growth rates for profit firms are 0.16% and -0.3%.
Because a key differentiator for the adjusted earnings of GAAP losers and profit firms is firms' intangible investment, we next examine the outcome and performance of firms' investment in the two intangibles included in our earnings adjustment: technological innovation, which directly benefits from firms' R&D expenditures; and human capital, one of the intangibles supported by firms' SG&A expenses.This examination focuses on the actual value created by firms' intangible investment and thus complements our approach of capitalizing intangibles expenditures, which is based on the expectation that these expenditures produce valuable assets.Given the inherently high risk and the large irrecoverable sunk costs associated with intangible investments (e.g., Lev 2001; Haskel and Westlake 2018), confirming the actual productivity of the investment is particularly important.

Technological innovation
We measure the productivity of firms' intangible investments related to technology innovation with patent data from Kogan et al. (2017).For each firm-year, we identify a pertinent patent portfolio based on patent filings submitted by the firm over the past 18-month period, ending on the firm's fiscal year-end date.We measure the size of the firm's patent portfolio (i.e., the number of patents), the number of "forward citations" (i.e., citations by subsequent patents of other firms to a given patent granted in an earlier year), and the value of the patents.The number of forward citations is adjusted for the average forward citation intensity across peer patents issued in the same technological area and the same year. 22We also examine patent lead time, measuring firms' success in patenting innovations earlier than others, which confers significant competitive advantages to the leading firms (e.g., firstmover advantage).The lead time of a patent is the difference between the patent's application year and the starting year of significant patenting activities attributed to all types of patenting entities in the respective technological area.For each technological area defined by the United States Patent and Trademark Office (USPTO), we identify the starting year of active patenting as the second year with at least five patent applications filed by all inventors. 23able 6, Panel A reports the summary statistics of our patenting measures for GAAP losers and Real losers that have patenting data available.It shows that GAAP losers are substantially more successful in patenting their innovations than Real losers.The mean number of patents per firm for GAAP losers is 32.94, 22 By construction, this measure can be interpreted as a citation-weighted patent count.Prior research indicates that compared to simple patent counts, citation-weighted patent counts have a stronger economic association with firms' contemporaneous market value (e.g., Hall et al. 2005) and future earnings performance (e.g., Gu 2005). 23For example, the lead time for a 2003 (2009) patent in an area for which 2005 was the second year with five or more patent applications would be -2 (+ 4) years relative to the onset of active patenting in the area of the patent.

3
All losses are not alike: Real versus accounting-driven reported… Table 6 Patent count, patent citation, and patent value for GAAP losers, Real losers, and matched profitable firms Data on patent-related variables are obtained from Kogan et al. (2017) and the USPTO patent databases.Number of patents per firm, number of forward citations per firm, and total value of patent portfolio per firm ($ million) are measured for the firm's patent portfolio consisting of all patents with filings submitted over the past 18-month period ending on the firm's fiscal year-end date.The number of forward citations per patent is adjusted for the average forward citation intensity across peer patents issued in the same technology area and the same year.Patent lead time is the difference in years between the patent's application year and the starting year of significant patenting activities in the technological area of the patent.For each technological area defined by the USPTO's two-level technology classification system, we empirically identify the starting year of active patenting as the second year with at least five patent applications filed by all inventors.For example, the lead time for a 2003 (2009) patent in an area for which 2005 was the second year with five or more patent applications would be -2 (4) years relative to the onset of active patenting in the respective technological area.Market value is the firm's stock market capitalization at the end of the fiscal year.In Panel B, GAAP losers and profitable firms are matched by year, market value, industry membership, the amount of R&D expenditure, and R&D intensity (the ratio of R&D expenditure to sales) using the procedure of propensity score matching a, b, c Indicate statistical significance of the differences between GAAP losers and Real losers in Panel A (matched profitable firms in Panel B) at the 0.001, 0. which is 39% higher than the mean of 23.69 patents per firm for Real losers.The patents of GAAP losers are also more valuable than those of Real losers.To put in perspective the magnitude of the mean difference in the average value per patent, $6.78 million (= 16.15-9.37),consider an average GAAP loser reported in Panel A that has a patent portfolio with 32.94 patents and a market value of $2.042 billion.For this firm, the additional value of $6.78 million per patent implies a total added value of $223.33 million (= 32.94 × $6.78) for its entire patent portfolio, which is equivalent to 10.9% of the firm's market value.Thus, GAAP losers' superiority in patenting more valuable innovations is economically significant and meaningful.
Note that the distribution of the total patent portfolio value for GAAP losers is more positively skewed, indicating the presence of a greater number of blockbuster patents-patents covering innovations that create industry dominance and substantial firm value-in the patent portfolios of GAAP losers.This difference is economically important, as it is well documented that the value of innovation is concentrated among a relatively small number of blockbuster patents while the majority of other patents have negligible value.
Panel A also shows that GAAP losers have superior patent lead time.On average, patenting by GAAP losers occurs within 3.29 years after the start of significant patenting activities in a technological area, compared to 5.55 years for Real losers.Thus, GAAP losers lead Real losers in patenting by more than two years, a substantial advantage in areas of fast-paced technological changes.
Our earlier results show that the R&D investment of GAAP losers is valued more favorably than that of profitable firms.To corroborate these results, we compare the innovation output of GAAP losers and profitable firms.Given the vast differences in size and related attributes (e.g., the amount of resources available for innovation) between these two groups of firms and the presence of the world's largest and most successful innovators in the sample of profitable firms (e.g., Alphabet Inc., Apple Inc., and Pfizer Inc.), we match GAAP losers with profitable firms in the same year by market value and industry membership.To control for other factors responsible for the cross-sectional variations in R&D productivity and innovation success, we further match on the amount of R&D expenses, which captures the overall scale of firms' innovation activities, and by R&D intensity, measured by the ratio of R&D expenses to sales.We operationalize this matching procedure with multiple criteria by using propensity score matching without replacement.
We report the summary statistics of patenting activities by the matched GAAP losers and profitable firms in Panel B. We find that although GAAP losers and profitable firms have similar levels of patenting frequency and citation intensity, the patents of GAAP losers are more valuable than those of matched profitable firms, on average.The mean (median) value per patent of $17.25 million ($4.32 million) for GAAP losers is also significantly higher than the mean (median) of $15.09 million ($3.62 million) for profitable firms.To put the magnitude of the difference in perspective, consider an average GAAP loser with 38.98 patents and a market value of $2.243 billion.For this firm, the mean difference of $2.16 million per patent implies a total difference of $84.20 million (= 38.98 × 2.16) in its patent portfolio, which is equivalent to 3.75% (= 84.20/2243*100%) of the firm's market value.Thus, the 1 3 All losses are not alike: Real versus accounting-driven reported… advantage of GAAP losers over profitable firms in patent value is economically significant and quite meaningful.Panel B also shows that, on average, GAAP losers file patent applications 12 months earlier than profitable firms, an advantage for GAAP losers in areas of fast-paced technological changes.
In untabulated tests, we compare firms' patenting choices in technological areas with heavy patenting (more than 100 patent applications per year) versus areas with low to modest patenting (less than 100 patent applications per year).We find that in heavy patenting areas (e.g., communications and information technologies), which attract intense competition and offer substantial rewards for efficient innovators, GAAP losers patent more innovations but obtain lower individual patent value than profitable firms.The overall patent portfolio value of GAAP losers, however, exceeds that of profitable firms, as GAAP losers' advantage of higher patenting frequency more than offsets the effect of lower individual patent value.GAAP losers also innovate more quickly and dominate profitable firms in patent lead time by more than one year.In contrast, in areas with less competition for patenting, GAAP losers succeed by patenting more valuable innovations with the same patenting frequency as profitable firms.These differences suggest that GAAP losers are more nimble in adjusting their innovation strategies in response to patenting competition.

Human capital
We examine two human capital metrics, employee turnover and employee productivity.There is growing consensus on the relevance of human capital metrics (e.g., Lev and Schwartz 1971; Romer 1990; Crook et al. 2011; SEC 2020). 24Prior studies suggest that information on employee headcount is useful for assessing the value and performance of firms' human capital, consistent with the central role of employees in the creation of human capital at corporations (e.g., Huselid 1995; Belo et al.  2014).In August 2020, the Securities and Exchange Commission (SEC) released key amendments and updates to modernize corporate disclosure (SEC 2020).In this release, the SEC highlighted specific information on employee headcount and employee turnover, as important human capital metrics for corporations to disclose to investors. 25Recent research confirms that the intensity of employee turnover (employee outflow) is associated with firms' future performance, indicating the usefulness of employee outflows as an indicator for the loss of firms' human capital (e.g., Agrawal et al. 2021; Li et al. 2022).
Another important aspect of employee turnover is employee inflows, which measure new employee recruitment and reflect the appeal of firms' job positions and employers' ability to attract employees with the needed skills and competencies.Organizational theories recognize that business success relies on employee recruitment, particularly the recruitment of employees who possess competencies critical for achieving firms' objectives (e.g., Aldrich 1979; Schneider 1987). 26mployee recruitment, however, can be a challenging task, given the fierce competition in many industries for limited supplies of highly qualified employees (the "war for talent") (Axelrod et al. 2001; McKinsey 2017; Duke CFO Survey 2019), hence the importance of measuring firms' success in hiring. 27Reflecting the importance of hiring, the SEC's proposed human capital measures include information on how firms attract personnel as a focus area in managing their business (SEC 2020).Accordingly, we examine employee inflows to measure firms' success in attracting employees. 28e also examine net employee addition, the amount of inflows net of outflows, or net hiring.Holding firms' hiring needs constant, inflows reflect firms' success in attracting new talent, whereas outflows indicate employee departure due to dissatisfaction or termination by employers.Net employee addition indicates the joint effects of employee hiring and departure and thus more comprehensively measures firms' success in employee recruiting, replacement, and retention-the foundation for effective human capital strategies.Belo et al. (2014) find that annual net hiring rates are informative about firms' future value, consistent with the role of net employee addition as a forward-looking indicator of the contribution of human capital to firms' future value.Following prior research, we deflate the amounts of employee inflows, outflows, and net additions by the average number of employees during the year.
26 Aldrich (1979) indicates that organizations in environments with particular technologies need people with particular kinds of competencies.On the importance of recruitment, Schneider (1987) explained that "it follows that organizations further restrict the range of types of persons in them through the recruitment and hiring of people with the kinds of competencies needed for effectiveness....The major way organizations can actively determine the pool of candidates from which they will choose their members is through recruitment activities" (p.444-448).Anecdotal evidence suggests that successful human capital strategies include a sustained focus on the recruitment of employees with specific competencies.At Apple Inc., Steve Jobs created a culture to constantly involve top-performing employees in recruiting more top-performing employees, with the objective of maximizing the creativity of Apple's top-performing employees.Jobs explained this approach by saying, "I found that when you get enough A-players together; when you go through the incredible work to find these A-players, they really like working with each other.Because they have never had the chance to do that before" (Smart and Street 2008).Popular employment services, such as LinkedIn and Indeed, actively promote the value of their services for businesses by emphasizing the strategic importance of recruitment ("Why Recruitment is the Most Important Business Strategy" by Lybrand (2018); "Why is Recruitment Important?" by Indeed Editorial Team (2021)). 27Highlighting the strategic importance of successful hiring, the management guru Jim Collins indicated that "the single biggest constraint on the success of my organization is the ability to get and to hang on to enough of the right people" (Collins 2001). 28Examining employee inflows is also useful because prior studies suggest that the implications of employee outflows for human capital are sometimes ambiguous.For example, the benefits of relatively small turnover may outweigh the costs under certain conditions (e.g., Hausknecht and Trevor 2011;  Hancock et al. 2013).Compared to employee outflows, the implications of employee inflows seem more straightforward.

3
All losses are not alike: Real versus accounting-driven reported… We expect firms with higher (lower) employee inflows (outflows) and higher net employee additions to have greater increases in the value of their human capital.
We also consider firms' employee productivity net of related costs, consistent with the Lev (2001)'s view that given the high risk associated with human capital, "expenditures (on human capital).... do not necessarily create assets.Only when the benefits from such expenditures-in the form of increased employee productivity-exceed costs is an asset created" (p.74). 29We measure net employee productivity as the ratio of sales to employee expense.This ratio captures both the input (costs) and output (sales) of firms' human capital activities and is equivalent to sales per employee (a common measure for employee productivity) divided by average employee expense.By measuring the benefits of human capital activities relative to their costs, net employee productivity quantifies the economic value added of human capital.Thus, an increase in the ratio of sales to employee expense unambiguously indicates improved net employee productivity and enhanced human capital value.However, the number of firms that separately report expenses directly related to employees, such as staff expense, is relatively small, particularly among loss firms (e.g., less than 8% of GAAP losers).To circumvent this problem, we use a proxy based on the amount of firms' SG&A expenses.We justify the choice of this proxy by confirming the high (greater than 83%) correlation between staff expense and SG&A expense for firms reporting staff expense. 30Given the lead time it takes for firms' human capital activities to benefit future employee productivity improvement, we focus on one-year-ahead change in the ratio of sales to SG&A expense.We expect greater increases in the one-year-ahead ratio of sales to SG&A expense for firms with more improved human capital value and performance.
More importantly, we provide evidence of the association between employee turnover and firms' net employee productivity.This association informs the implications of employee turnover for human capital value and performance.For employers, high employee turnover brings both costs (e.g., the cost of replacing departed employees) and benefits (e.g., the greater value of hiring more productive employees).In particular, employee turnover can lead to knowledge spillovers in both directions, with firms gaining valuable knowledge and ideas by hiring experienced employees from other firms but losing a competitive edge when their key employees leave for other firms (including competitors).Thus, the dynamic effects of employee turnover are not always one-directional and can potentially separate winners-firms that are able to capitalize on high employee turnover to maximize the benefits of turnover (e.g., replacing underperforming employees with more productive new employees, exploiting knowledge spillovers, and improving overall employee cost efficiency)from others. 31We expect firms with more successful human capital strategies to benefit more from employee turnover and achieve more improved net employee productivity subsequent to employee turnover, hence a positive association between employee turnover and one-year-ahead change in net employee productivity.This positive association, however, is not expected if firms' human capital activities, such as hiring, replacement, and retention, do not produce positive economic benefits.
We report the descriptive statistics of firms' employee turnover and productivity in Table 7. 32 Panel A shows that GAAP losers have significantly higher mean and median levels of employee inflows than Real losers, indicating GAAP losers' greater success in hiring.Although GAAP losers have higher employee outflows than Real losers, the amount of outflows relative to inflows for GAAP losers, 70% and 76% for the mean and median, respectively, is lower than that for Real losers, for which the mean and median are 82% and 92%, respectively.Thus, GAAP losers have significantly higher net employee additions than Real losers.The mean differences in net employee additions between GAAP losers and Real losers-0.037,or 3.7% of firms' total headcount (more than three new employees added per 100 employees)are statistically significant and economically meaningful.Thus, GAAP losers are more successful than Real losers in attracting and retaining employees, a significant advantage for GAAP losers' human capital strategies.Importantly, Panel A shows that GAAP losers have significantly greater improvement in net employee productivity in the one-year-ahead period than Real losers, suggesting that GAAP losers' employee hiring and retention initiatives produce greater future economic benefits.
Panel B compares GAAP losers and profitable firms.To control for confounding factors for employee turnover, we match GAAP losers with profitable firms that have similar size and employee headcount and are from the same industry in the same year. 33We find that GAAP losers have significantly larger employee inflows than matched profitable firms.The differences in the mean and median employee inflows, 6.5% and 5.5%, respectively, are statistically significant and economically meaningful.Thus, despite their loss status, GAAP losers are stronger job magnets 1 3 All losses are not alike: Real versus accounting-driven reported…

Table 7
Employee turnover and human capital for GAAP losers, Real losers, and matched profitable firms  and employee outflows (all three variables are deflated by firms' average employee headcount across all months of the year).In Panels B and C, GAAP losers and profitable firms are matched by year, market value, industry membership, and employee headcount using the procedure of propensity score matching.In Panel C, the dependent variable is the one-year-ahead change in net employee productivity, measured as the ratio of sales to selling, general, and administrative expenses.The independent variables include the current year's employee inflows, outflows, firm size (measured as logarithm of sales), number of employees (measured as logarithm of employee headcount), R&D expenditure (deflated by average total assets), cash flows from operating activities (deflated by average total assets), and capital expenditure (deflated by average total assets)

a, b, c
Indicate statistical significance of the differences between GAAP losers and Real losers in Panel A, the differences between matched GAAP losers and profitable firms in Panel B, and the differences of "Real loss → GAAP loss" relative to the other two groups, "GAAP loss → Real loss" and "GAAP loss → Profit" in Panel D, at the 0.001, 0.01, and 0.05 levels, respectively 1 3 All losses are not alike: Real versus accounting-driven reported… for employees than profitable firms.We also find that the mean and median differences in net employee additions between GAAP losers and profitable firms, 2.9% and 1.5%, respectively, are statistically significant.Thus, compared to profitable firms, GAAP losers have higher employee inflows and net employee additions.Panel B also shows that GAAP losers have significantly higher one-year-ahead growth in net employee productivity than similar profitable firms, demonstrating GAAP losers' greater advantage in improving human capital value and performance.
In Panel C, we examine the association between employee turnover and future improvement in employee productivity by regressing one-year-ahead change in net employee productivity on current employee inflows and outflows, respectively.We control for firm size and employee headcount, as these two factors are expected to be associated with future employee productivity growth (e.g., smaller firms are more likely to experience improvement in future employee productivity than larger firms).We also control for R&D expenditure, cash flows from operations, and capital expenditure, as these inputs are expected to benefit future employee productivity growth.In the regressions for GAAP losers, we find positive and statistically significant coefficients on employee inflows (0.293) and outflows (0.320), indicating greater future improvement in net employee productivity for GAAP losers with higher employee turnover rates.This is consistent with GAAP losers' ability to benefit from employee turnover by leveraging turnover to improve human capital value and performance (e.g., replacing underperforming employees with more productive new employees, exploiting knowledge spillovers facilitated by the hiring of experienced employees, and improving overall employee cost efficiency).In contrast, the coefficients on employee inflows and outflows are statistically insignificant in the regressions for matched profitable firms, suggesting that profitable firms' net employee productivity does not benefit from human capital activities, such as hiring and departure, that are reflected in employee turnover.
To solidify our evidence on GAAP losers' human capital advantage, we examine firms experiencing changes in their loss status.We expect firms switching from Real loss to GAAP loss ("Real loss → GAAP loss") to be more successful in hiring new employees and improving net employee productivity in the future than deteriorating firms ("GAAP loss → Real loss").The results in Panel D confirm these expectations.We find that these improving firms see more net employee additions and have larger increases in net employee productivity in the subsequent year than deteriorating firms.We also examine firms changing from GAAP loss to profit-an economically ambiguous and likely non-substantive change had the value of intangible investments been properly recognized.We expect these firms to have smaller changes in human capital metrics than firms moving up more meaningfully from real loss to GAAP loss.The results in the column titled "GAAP loss → Profit" confirm our expectation, showing that these firms have significantly lower net employee additions and smaller employee productivity gains in the subsequent year than firms switching from real loss to GAAP loss.
All in all, our patenting and human capital tests establish that GAAP losers are inherently different from Real losers and should be treated differently by both investors and researchers.

Future performance of loss firms
We next examine whether intangibles-driven losses and real losses have different implications for the future prospect of firms.Specifically, we define declining firms as subsequently delisted and/or liquidated firms and firms with greater losses during the three-year period after year t, and consider other firms to be non-declining, following Yung et al. (2008). 34Given GAAP losers' advantage in intangibles, we predict that GAAP losers are less likely to decline in the future than Real losers.Consistent with our prediction, Fig. 2 shows that for each year during 1980-2015, the percentage of declining GAAP losers is smaller than that of declining Real losers.Overall, 27.1% of GAAP losers decline in the future, compared to 34.2% of Real losers.
To confirm the statistical significance of the association between loss types and firms' future performance, we run a logistic regression with a dependent variable that takes the value of one for non-declining firms and zero for declining firms.The key independent variable is a dummy variable that takes the value of one for GAAP losers and zero otherwise.We control for firm fundamentals (e.g., firm size) and variables for firms' operating (e.g., cash flows from operations and sales growth), Fig. 2 Annual percentage of declining firms among GAAP losers and Real losers.Declining firms are defined as delisted or liquidated firms and firms with greater losses during the three-year period after year t.Thus, non-declining firms are those that are acquired or achieve future increases in income before extraordinary items.Information on the events of delisting, liquidation, and acquisition is obtained from CRSP files 1 3 All losses are not alike: Real versus accounting-driven reported… investing (e.g., capital expenditures and intangible investments), financing activities (e.g., stock issuance), loss reporting history (i.e., the number of consecutive years of loss reporting), and current earnings (ROA).Table 8 reports the estimates of our logistic regression.We find a significantly positive coefficient (0.078) on the indicator variable for GAAP losers, supporting our prediction that GAAP losers are less likely to decline in the future than Real losers.
We also examine future loss reversal over one year (year t + 1), three years (year t + 1, t + 2, and t + 3), and five years (year t + 1 through t + 5).For the three-year and five-year periods, we use two measures for loss reversal, one based on annual earnings in each year within the window and the other based on combined earnings over the entire period.Figure 3 shows that compared to Real losers, GAAP losers are substantially more likely to reverse their losses in the future.Approximately 44% of GAAP losers have positive combined earnings for the three-year period, whereas only 24% of Real losers achieve this improvement.The results for the most discriminating measure for future loss reversal, which is based on positive annual earnings for five consecutive years, show that nearly 20% of GAAP losers do not report a loss in any of the subsequent five years, compared to only 9% for Real losers.
To formally test the difference in the likelihood of future loss reversal between GAAP losers and Real losers, in Table 9, we run a logistic regression of future loss reversal on a dummy variable for GAAP losers versus Real losers while controlling for firm fundamentals and activities included in the regressions of Table 8.Panel A and Panel B show the results for the three-year and five-year periods, respectively (we omit the results for the one-year period, as they are very similar to those for three and five years).In all regressions, we find a significantly positive coefficient on the indicator variable for GAAP losers.The magnitude of this coefficient is stable across model specifications with or without stock returns and over different time horizons.Thus, GAAP losers are significantly more likely to become profitable in the future than Real losers.
To solidify our results on the superiority of GAAP losers, we compare the future stock returns of GAAP losers and other firms.Specifically, we examine the growth in the value of one dollar invested in the portfolios of GAAP losers, Real losers, and profitable firms.These portfolios are constructed and reassembled annually using the most recent accounting information from 1980 to 2017.For each portfolio, we compute annual value-weighted average returns and use these returns to estimate the portfolio value each year, assuming the proceeds from liquidating the portfolios in year t are reinvested in the portfolios of year t + 1.For example, one dollar invested in the portfolio of 1980, which is constructed using the accounting data of 1980, with annual returns of 10% would generate $1.10 to be reinvested in the portfolio of 1981, and so on. 35This approach simulates firms' future value with actual annual returns to provide a realistic view of how firm value changes over time.
Figure 4A portrays the stock performance of firms using our approach of firm value estimation.The three bars on the left, "All years: 1980-2017," show that one dollar invested in the 1980 portfolios of GAAP losers, profitable firms, and Real losers would grow to $30.97, $53.14, and $3.83, respectively.Thus, there is more than 8 times difference in future value between GAAP losers and Real losers.The superb performance of profitable firms is not surprising, given the presence of the The dependent variable of the logistic regression is an indicator variable that takes the value of one for non-declining loss firms and zero for declining loss firms (delisted or liquidated firms and firms with greater losses during the three-year period after year t).GAAP-loser is an indicator variable that takes the value of one for GAAP losers and zero for Real losers.Firm size is the logarithm of firms' total assets.Firm age is the number of years for firms' coverage on Compustat.Visibility is the number of analysts following the firm.Loss years is the number of consecutive years of loss reporting.ROA is return on assets.Positive sales growth is a dummy variable that takes the value of one for firms with positive sales growth and zero otherwise.Cash flows from operations is a dummy variable that takes the value of one for firms with positive cash flows from operations over the past three years.R&D expenditure and SG&A expense are scaled by total assets.Capital expenditure is the three-year average ratio of capital expenditures to total assets.Acquisitions is a dummy variable that takes the value of one for firms with acquisition activities over the past three years and zero otherwise.Leverage is a dummy variable that takes the value of one for firms with more long-term debt than total assets and zero otherwise.Current ratio is a dummy variable that takes the value of one for firms with current ratio greater than one and zero otherwise.Long-term debt issuance is a dummy variable that takes the value of one for firms with long-term debt issuance over past the three years and zero otherwise.Stock issuance is a dummy variable that takes the value of one for firms issuing stock over the past three years and zero otherwise  Figure 4B, C shed further light on the reversal of fortune for GAAP losers and profitable firms by comparing the performance of one dollar invested in GAAP losers and firms in each year from 1980 to 1996 and 1997 to 2017, respectively.They show that profitable firms had a substantial advantage over GAAP losers early on (e.g., one dollar invested in profitable firms of 1981 growing to $60.50 versus $33.26 for GAAP losers).The earlier advantage of profitable firms, however, declined steadily over time and was barely visible by the end of 1996.Since 1997, GAAP losers have outperformed profitable firms every year, except for 1999. 36The inflection year of 1997 coincides with the timing of rising intangible investment rate permanently surpassing tangible investment rate in US private industries (Corrado  and Hulten 2010), a trend that further exacerbates the distortion of reported accounting information for intangibles-intensive firms.
Fig. 3 Percentage of future earnings reversal by GAAP losers versus Real losers 36 Our finding that some loss firms have higher future stock returns than profitable firms is consistent with the evidence of Fama and French (1992).Using sample data for 1962-1989, Fama and French report that, on average, firms with negative earnings tend to earn higher future returns than 70% of the profitable firms (Table IV,   The univariate comparison of firms' stock performance does not take into account the roles of firm size, growth/risk, and other factors driving future returns.We address this shortcoming by matching firms by size, book-to-market, industry membership, and year (i.e., matching GAAP losers with profitable firms and Real losers, respectively).The three bars on the right of Fig. 4A, "Matched firms: 1980-2017," show that one dollar invested in matched GAAP losers at the beginning of 1980 would grow to $23.72, compared to $9.05 for matched profitable firms. 37This is, of course, a very surprising finding, but it is consistent with GAAP losers' superiority in intangibles.It further adds to our evidence showing that standard accounting reports do not properly reflect GAAP losers' performance and value.
To shed more light on the future stock performance of profitable firms, GAAP losers, and Real losers, we examine their future abnormal returns using various benchmarks.In Table 10, Panel A, we report the mean values of one-year-ahead size-adjusted annual returns and the number of years with positive and negative returns, respectively.For all years, 1980-2017, we find that GAAP losers earn significantly higher size-adjusted returns than profitable firms (3.73% versus 2.46%).As expected, Real losers' mean one-year-ahead size-adjusted returns across all years are negative (-1.59%).Real losers earned negative returns for 29 out of 38 years, or 76% of the time, an overwhelming losing record.In contrast, GAAP losers and profitable firms earned positive returns for 30 and 31 years, or 79% and 82% of the time, respectively.Similar to the patterns shown in Fig. 4B and C, the second and third rows of statistics in Panel A indicate that while profitable firms had a substantial advantage over GAAP losers prior to 1997 by earning significantly higher returns (3.44% versus 1.77%), the reverse has been true since 1997, when intangible investment surpassed tangible investment to become the most dominant form of The dependent variable of the logistic regression is an indicator variable that takes the value of one for loss firms with future earnings reversal and zero otherwise.GAAP-loser is an indicator variable that takes the value of one for GAAP losers and zero for Real losers.Firm size is the logarithm of firms' total assets.Firm age is the number of years for firms' coverage on Compustat.Visibility is the number of analysts following the firm.Loss trend is the number of consecutive years with loss.ROA is return on assets.Sales growth is a dummy variable that takes the value of one for firms with positive sales growth and zero otherwise.Cash flows from operations is a dummy variable that takes the value of one for firms with positive cash flows from operations over the past three years.R&D is the firm's R&D expenditure scaled by total assets.SG&A is the firm's SG&A expense scaled by total assets.Capital expenditure is the three-year average ratio of capital expenditures to total assets.Acquisitions is a dummy variable that takes the value of one for firms with acquisition over the past three years and zero otherwise.Leverage is a dummy variable that takes the value of one for firms with more long-term debt than total assets and zero otherwise.Current ratio is a dummy variable that takes the value of one for firms with current ratio greater than one and zero otherwise.Long-term debt issuance is a dummy variable that takes the value of one for firms with long-term debt issuance over the past three years and zero otherwise.Stock issuance is a dummy variable that takes the value of one for firms issuing stock over the past three years and zero otherwise.Stock return is the firm's stock return for the year Table 9 (continued) 37 Our results show that one dollar invested in matched GAAP losers and Real losers at the beginning of 1980 would grow to $4.66 (not shown in Fig. 4A) and $2.21 (shown in Fig. 4A), respectively.

3
All losses are not alike: Real versus accounting-driven reported… investment in US private industries.For 1997-2017, GAAP losers earned significantly higher returns than profitable firms (5.02% versus 1.79%) and also had more years with positive returns (17 versus 15).
In Panel B, we report the estimates of one-year-ahead future abnormal returns for the portfolios of profitable firms, GAAP losers, and Real losers using the Fama-French five-factor model (Fama and French 2015).This model regresses firm-specific excess returns on five distinct factors-market returns, firm size, value, operating profitability, and investment patterns.We report the regression intercept in Panel B, which is the estimate of average monthly abnormal returns (i.e., alpha).We find that GAAP losers earn significantly higher future abnormal returns than profitable firms for the period since 1997 (0.0045 versus 0.0015 per month), consistent GAAP losers having an advantage over profitable firms in an era increasingly dominated by intangible investment. 38Panel B also shows that GAAP losers consistently earn higher future abnormal returns than Real losers in all periods, underscoring the fundamental differences between GAAP losers and Real losers.
Taken together, the patterns of future abnormal returns for profitable firms, GAAP losers, and Real losers documented in Table 10 are consistent with the evidence in Fig. 4 and lend further support to our conclusion that GAAP losers have consistently stronger future performance than Real losers and even profitable firms for the last 20 years.More importantly, the evidence of GAAP losers' Fig. 4 Future stock value of GAAP losers, profitable firms, and Real losers.A: Growth of one dollar invested in GAAP losers, profitable firms, and Real losers.B: Earlier years for GAAP losers and profitable firms (1980-1996).C: Later years for GAAP losers and profitable firms (1997-2017).We calculate future stock value of GAAP losers, profitable firms, and Real losers based on the growth of one dollar invested in each portfolio over time, where the growth rate is the annual value-weighted average returns for the portfolio.These portfolios are reassembled at the beginning of the fourth month of each year using the most recent accounting data for the prior year to identify GAAP losers, Real losers, and profitable firms.We use the annual portfolio returns to estimate the change in firm value in the portfolio each year, assuming the proceeds from liquidating the portfolios in year t are reinvested in the portfolios of year t + 1.For example, one dollar invested in the portfolio of 1980 with annual returns of 10% would generate $1.10 to be reinvested in the portfolio of 1981, and so on.The three bars on the left of Fig. 4A, titled "All years: 1980-2017," portray the terminal value of one dollar invested in GAAP losers, profitable firms, and Real losers, respectively, from 1980 to 2017.The two groups of three bars in the middle of Fig. 4A, titled  "Earlier years: 1980-1996" and "Later year: 1997-2017," portray the terminal value of one dollar invested over 1980-1996 and 1997-2017, respectively.The three bars on the right of Fig. 4A, titled "Matched firms: 1980-2017," portray the terminal value of one dollar invested in (1) matched GAAP losers ($23.72) and profitable firms ($9.05), and (2) matched GAAP losers ($4.85, not shown in the figure) and Real losers ($2.21, shown in the figure).Figure 4B and C portray the terminal value of one dollar invested in each year from 1980 to 1996 and from 1997 to 2017, respectively.For example, the two bars for 1990 (2010) in Fig. 4B (4C) portray the terminal value of $13.44 and $14.10 ($2.40 and $2.25) obtained from investing in GAAP losers and profitable firms, respectively, from 1990 to 2017 (2010 to 2017) 1 3 All losses are not alike: Real versus accounting-driven reported… persistent and positive abnormal returns in recent decades suggests that as the economic importance of intangible investment continues to rise over time, investors are increasingly less able to recognize and undo intangibles-driven distortions in the reported earnings of GAAP losers.The increasing difficulties for investors to "see through" accounting numbers that are biased by the expensing of intangibles attest to the need for reforming current accounting rules for intangibles.
Table 10 Future stock returns of profitable firms, GAAP losers, and Real losers Panel A reports the equal-weighted one-year-ahead size-adjusted returns for each fiscal year that starts three months after the end of the fiscal year and ends in 12 months for the portfolios of profitable firms, GAAP losers, and Real losers Panel B reports the abnormal returns estimated as the intercept of the following regression based on the Fama and French five-factor model: R it − R Ft = a i + b i (R Mt − R Ft ) + s i SMB t + h i HML t + r i RMW t + c i CMA t + e it , where R it is the monthly return for the portfolios of GAAP losers, Real losers, and profitable firms, respectively; R Ft is the risk-free return; R Mt is the return on the market portfolio; SMB t is the return on a portfolio of small stocks minus the return on a portfolio of big stocks; HML t is the difference between the returns on portfolios of high and low book-to-market (B/M) stocks; RMW t is the difference between the returns on portfolios of stocks with robust and weak profitability; CMA t is the difference between the returns on the portfolios of stocks of firms with low and high investment; and e it is a zero-mean residual (Fama and French 2015).We assess the statistical significance of the difference in the abnormal returns between any two types of firms (e.g., GAAP losers versus profitable firms) by regressing the monthly return difference between the portfolios of the two types of firms (e.g., GAAP losers and profitable firms) on the five factors and then testing the statistical significance of the regression intercepts.For each fiscal year, the return period starts from three months after the end of the fiscal year and ends in 12 months.Data on the five factors used in the regression are downloaded from Kenneth French's online research data library at https:// mba.tuck.dartm outh.edu/ pages/ facul ty/ ken.french/ data_ libra ry.html

Conclusions
We examine, in this study, the value relevance of losses by distinguishing between GAAP losses caused by the immediate expensing of firms' internally generated intangibles and real losses occurring irrespective of the intangibles of the firm.Our study is timely and relevant, given the increasing frequency of loss reporting in recent decades and the need to understand its implications.Contrary to the widely held view of the irrelevance of losses, we find that losses attributed to the expensing of intangible investments are highly informative and are in fact as informative as profits.In contrast to earlier results showing that persistent losses decrease earnings relevance, we find that the value relevance of reported earnings of GAAP losers does not decrease with the persistence of losses caused by the expensing of intangible investments.
Our results clearly show that the relevance of GAAP losses is related to investors viewing the intangible investments of GAAP losers as a potent source of shareholder value rather than a loss driver.Supporting this view, we find that GAAP losers are more successful in creating value from technological innovation and human capital than Real losers and even profitable firms.We further demonstrate that GAAP losses have more favorable implications for firms' future performance than real losses: GAAP losers are less likely to decline and more likely to reverse their losses in the future.GAAP losers also have better future stock performance than Real losers and even profitable firms.In sum, our results in this study indicate that standard accounting reports of GAAP losers seriously distort the intrinsic value and performance of these firms.This is an alarming consequence for a group of highly dynamic and innovative firms that are the driving force behind the growing intangible revolution in our economy.

==
Industry and year fixed effects are included in all regressionsPanel A. Earnings response coefficient (ERC) of GAAP losers with short versus long loss persistenceReported earnings on loss persistence (Loss persistence = 1 for 1-2 loss years, 2 for 3-4 loss years, and so on; t-statistics in parentheses):Reported earnings: ERC = 0.160 -0.029 × Loss persistence + e, Loss persistence = 1-4, Adj.R 2 on loss persistence (Loss persistence = 1 for 1-2 loss years, 2 for 3-4 loss years, and so on; t-statistics in parentheses):Reported earnings: ERC = 0.267 -0.069 × Loss persistence + e, Loss persistence = 1-4, Adj.R 2 Regressions of one-year-ahead change in net employee productivity on employee turnover (one-sided p-value Data on employee inflows and outflows for the period of 2008-2018 are provided by Revelio Labs.Net employee additions are the difference between employee inflows largest and best-performing firms of each decade in this sample (e.g., General Electric in the 1980s, Microsoft in the 1990s, and Apple in the 2000s).The two groups of three bars in the middle of Fig.4A, however, show that the performance difference between GAAP losers and profitable firms has reversed recently.While profitable firms outperformed GAAP losers by a wide margin, $12.81 to $7.11, in the earlier period of 1980-1996, GAAP losers have appreciated slightly more in value than profitable firms, $4.35 versus $4.15, in the more recent period of 1997-2017.

3
Future loss reversal of GAAP losers versus Real losers 1 All losses are not alike: Real versus accounting-driven reported… Table 9 Panel A. Mean values (number of positive annual values/number of negative annual values) of one-year-ahead size-adjusted annual returnsFirm type Mean difference (p-value) All losses are not alike: Real versus accounting-driven reported…

Table 2
Annual regressions of return on earnings and earnings change: reported earnings versus adjusted earnings Panel A.

Panel A. Mean statistics from 39 annual regressions (one-sided p-values in parentheses)
to the sum of R&D and SG&A.ERC is the sum of the coefficients on IB and ΔIB

Table 4
Descriptive statistics for firms with persistent losses

Panel A. Compare GAAP losers and Real losers
01, and 0.05 level, respectively based on one-tailed test

Table 8
Future non-decline/decline status of GAAP losers versus Real losers