The stock market boosts its rewards for increasing earnings patterns

Abstract

Barth et al. (J Account Res 37:387–413, 1999. https://doi.org/10.2307/2491414) find that the market rewards firms showing increasing earnings patterns with larger price-earnings multiples from 1982 to 1992. This paper examines whether the rewards for increasing earnings patterns have changed since then. The declining relevance of accounting earnings for equity valuation or accounting scandals in the early 2000s would predict a lower valuation for earnings numbers. On the other hand, earnings-related disclosures and post-scandal regulatory efforts, such as the Sarbanes–Oxley Act (SOX), may boost earnings relevance and restore investor confidence in the accounting system, predicting increased rewards for increasing earnings patterns. We find that the earnings multiples for firms with increasing earnings patterns decrease temporarily in the years with major accounting scandals, but the market boosts its rewards to firms showing increasing earnings patterns in the reshaped reporting environment after 1993, and SOX is associated with this boost.

1 Introduction

The financial reporting environment has changed dramatically in the past three decades. These changes include the shift in investor focus from GAAP to non-GAAP earnings in the late 1990s (Bradshaw and Sloan 2002; Lev and Gu 2016) and the accounting scandals and subsequent Sarbanes–Oxley Act (SOX) reforms in the early 2000s (Coates and Srinivasan 2014). Investors, regulators, and researchers are interested in the effects of these market changes and associated regulations on financial reporting practices (Bartov and Cohen 2009; Cohen et al. 2008; Gilliam et al. 2015; Graham et al. 2005; Iliev 2010) and market rewards for firms’ reporting behaviors (Bartov et al. 2002; Gipper et al. 2020; Kasznik and McNichols 2002; Koh et al. 2008).

This paper reexamines Barth, Elliott, and Finn (BEF) (1999). As an influential paper with over 1200 citations, BEF (1999) was motivated by an observation that many managers focus on steadily increasing earnings. Firms with patterns of increasing earnings for more than 5 years have significantly higher price-earnings multiples than other firms.¹ The earnings multiples associated with increasing earnings patterns increase almost monotonically with the length of the increasing pattern and decrease when firms have a pattern of decreasing earnings after a previous pattern of increasing earnings. BEF interpret their findings as showing that the earnings patterns capture dimensions of value-relevant firm characteristics, such as the path of growth over time, which are not captured by the growth or risk proxies incorporated in their model. Overall, BEF’s results support capital market incentives for pursuing increasing earnings streaks.

Several papers investigate cross-sectional factors associated with higher rewards for increasing earning patterns, such as earnings quality (Francis et al. 2003; Kao and Na 2013; Liu and García Lara 2015) or business strategy (Ghosh et al. 2005), but this paper focuses on intertemporal variations. Since the 1990s, the market has reacted more strongly to non-GAAP earnings than GAAP earnings news in a single period (Bradshaw and Sloan 2002). However, this does not imply that the market would reward firms less now than in the past for increasing GAAP earnings patterns across multiple periods. Besides, the accounting scandals in the early 2000s are associated with decreased market reactions to earnings news in a single period (Koh et al. 2008), but subsequent regulatory reforms, such as SOX, may alter how the market prices earnings news. The market may change its pricing of GAAP earnings information from a short-term to a long-term orientation. The market’s changes in rewards for long-term firm performance would offer managers different implications and incentives to balance short-term and long-term goals.

Given the changes in the financial reporting environment, such as non-GAAP disclosures, accounting frauds, or regulatory reforms, it is unclear how investors change the degree to which they value increasing earnings patterns over time. The decreased relevance of GAAP earnings (Lev and Gu 2016), increased focus on non-GAAP earnings (Black et al. 2018), and decreased management credibility following accounting scandals (Koh et al. 2008; Koonce and Lipe 2010) would predict diminished rewards for increasing earnings patterns. On the other hand, post-scandals regulatory reforms, such as SOX, may reduce managerial opportunism in financial reporting (Cohen et al. 2008; Pincus et al. 2022) and enhance the predictive value of earnings information (Altamuro and Beatty 2010; Gipper et al. 2020). Thus, the way that market shifts and regulatory efforts in the past decades affect the market rewards to firms exhibiting patterns of increasing earnings turns out to be an empirical question. This paper replicates and extends BEF’s (1999) work to determine whether the market rewards to firms with increasing earnings patterns have changed in the reshaped environment and investigates factors associated with the changed rewards.²

This paper first reaffirms BEF’s finding that the market assigns higher earnings multiples to firms with increasing earnings patterns than other firms. BEF’s results are reasonably replicable.

This paper then documents the larger premiums for increasing earnings patterns. On average, the market rewards firms that report at least 5 years of increasing earnings streaks with earnings multiples that are almost twice the multiples for continuously profitable firms as shown in BEF’s work in the past. The boosted rewards for increasing earnings patterns exist in various pattern lengths, and the price-earnings multiple changes as the pattern develops. Overall, the market premiums for increasing earnings patterns get stronger.

Both regulatory reforms and accounting frauds can play roles in the changes in market premiums over time. Investors sharply but temporarily cut the premiums to firms with increasing earnings patterns in the years when well-known frauds hit the market (Koh et al. 2008). However, the average market premiums for increasing earnings patterns rise significantly after SOX.

The changing relationship between GAAP and non-GAAP earnings information over time and across SOX can be another factor. The incremental rewards for increasing GAAP earnings patterns are present in subsamples with and without non-GAAP earnings history. For firms that have a non-GAAP earnings history, the market offers increased (decreased) premiums for increasing non-GAAP (GAAP) earnings patterns after 1994, consistent with investors’ shifted focus at that time (Bartov et al. 2002). The findings regarding decreased premiums around frauds and increased premiums after SOX appear to be mainly driven by firms with non-GAAP earnings history. Given that GAAP and non-GAAP earnings are reconciled after Regulation G under SOX, the findings that the rewards for increasing non-GAAP earnings patterns contribute to the rewards for increasing GAAP earnings patterns and that the rewards increase after SOX suggest that SOX can play a role in enhancing the information value of earnings and their patterns.

This paper contributes to the replicability of accounting research (Hail et al. 2020; Zeff 2019) and may be useful to regulators, researchers, investors, and managers. The evidence offers regulatory implications. Unlike prior research (Liu and García Lara 2015; Koh et al. 2008), this paper supports that SOX effectively restored investor trust in the accounting system by showing increased market rewards associated with increasing earnings patterns after SOX. While Koh et al. (2008) imply that the market offers fewer premiums for single-period meeting or beating earnings targets after SOX, this paper provides another insight that the market tends to boost its rewards for multi-period earnings increases since then. Thus, the stock market appears to have changed how it views and rewards earnings information after regulatory reforms.

This paper contributes to research on accounting quality by showing how intertemporal variations in accounting quality at the market-wide level are associated with varying pricing of earnings patterns. The evidence complements studies about cross-sectional variations.³ Despite not offering an exhaustive list, this paper explores SOX as a positive factor and fraud as a negative factor in intertemporal variations. This paper offers an estimate of the cost of fraud in terms of reducing price-earnings multiples (Coates and Srinivasan 2014) and facilitates understanding the interplays between GAAP and non-GAAP earnings information across regulatory regimes (Black et al. 2018). Given the wide use of multiples for valuation, investors may find it relevant to learn the changing market pricing of increasing earnings patterns and associated factors in the reshaped environment. In addition, larger rewards for increasing earnings patterns can inform managers about adjusting for market-based incentives to balance and sustain short-term and long-term performance goals.

This paper proceeds as follows: Sect. 2 discusses the research question. Section 3 outlines the design. Sections 4 and 5 show and discuss the results. Section 6 concludes.

2 Does the market change its rewards for increasing earnings patterns?

The US stock market changed after the late 1990s and early 2000s. Notable changes include investors’ shifted focus from GAAP to non-GAAP earnings, accounting scandals, and regulatory reforms. These changes could alter the information about a firm’s prospects and management credibility conveyed through increasing earnings patterns (Koonce and Lipe 2010), thereby affecting how the market rewards firms with increasing earnings patterns.⁴

2.1 Factors predicting decreased premiums for increasing earnings patterns

2.1.1 Declining relevance of GAAP earnings for firm prospects

Higher market rewards to firms with increasing earnings patterns may diminish after the late 1990s, when GAAP earnings appeared to keep losing relevance for valuation (Lev and Gu 2016).⁵ Investors now find non-GAAP earnings more informative about a firm’s prospects than GAAP earnings (Bradshaw and Sloan 2002; Bhattacharya et al. 2003; Black et al. 2018). Meeting analysts’ forecasts on street earnings became important to investors after 1994 (Bartov et al. 2002; Bartov and Cohen 2009). If investors view non-GAAP earnings as substitutes for GAAP earnings for valuation purposes, then the shift in focus matters (Ribeiro et al. 2019). Given the reduced relevance and informativeness of GAAP earnings, the patterns of GAAP earnings might offer less valuable clues about firms’ prospects, resulting in lower rewards for increasing earnings patterns.

2.1.2 Lost investor trust in management credibility following frauds

Decreased management credibility also predicts lower rewards for increasing earnings patterns. Accounting frauds and scandals, such as Enron, WorldCom, and Tyco, occurred in the early 2000s. These scandals shook investors’ trust in management’s credibility. After these scandals, investors became skeptical of firms that meet or beat analysts’ expectations and offered them thinner market rewards (Koh et al. 2008). Investors could also be skeptical of firms that meet or beat prior years’ earnings and discount the value premiums for increasing earnings patterns (Kasznik and McNichols 2002). Even after SOX, managers initiate more real activities to manage earnings (Bartov and Cohen 2009; Cohen et al. 2008), indicating potential managerial opportunism and lower ability of earnings to predict firms’ prospects (Li 2019).⁶ Management’s credibility might decrease after 2007, when attestation requirements for SOX compliance were eased, or after 2012, when the Jumpstart Our Business Startups Act (JOBS Act) allowed exemptions from certain SOX provisions (Coates and Srinivasan 2014).⁷ Investors may demand higher risk premiums for skepticism and reduce the value of the earnings patterns from an accounting system lacking credibility. The decline in earnings relevance and management credibility predicts that the rewards for increasing earnings patterns will diminish after the late 1990s (Koonce and Lipe 2010).

2.2 Factors predicting increased premiums for increasing earnings patterns

On the other hand, the finding of higher market rewards for increasing earnings patterns may hold or even strengthen after the late 1990s. After several major accounting scandals, SOX and related regulatory reforms were enacted in the early 2000s to restore investors’ confidence in the financial reporting system through increased disclosures and audit monitoring.⁸

2.2.1 Restored investor trust in management credibility under SOX

SOX is associated with decreased accrual-based earnings management (Cohen et al. 2008; Pincus et al. 2022), which allows GAAP earnings to map economic actions more faithfully onto accounting earnings. Even in the presence of real earnings management, investors may view firms’ meeting or beating prior-year GAAP earnings in streaks as a signal of management credibility and future prospects when firms’ financial reporting is overall subject to increased scrutiny under SOX (Chen et al. 2016; Graham et al. 2005; Gunny 2010).⁹ Gilliam et al. (2015) suggest that the zero-earnings (zero-earnings change) discontinuity has disappeared (almost disappeared) since the passage of SOX, supporting a reduction in the overall effect of earnings management to avoid losses or earnings declines after SOX (Burgstahler and Chuk 2017). If increasing earnings patterns are less driven by manipulation after SOX, one would expect higher premiums associated with more legitimate increasing earnings patterns (Ghosh et al. 2005; Liu and García Lara 2015).

2.2.2 Enhanced predictive value of earnings for firm prospects under SOX

Moreover, regulation on internal controls under SOX may enhance the persistence of earnings and their ability to predict cash flows and deliver robust information about earnings growth (Altamuro and Beatty 2010). Singer and You (2011) suggest that overall earnings quality in terms of reliability and relevance improves for firms complying with SOX Section 404 after its implementation. Gipper et al. (2020) show that the market prices earnings news higher as financial reporting becomes more credible after SOX introduced the Public Company Accounting Oversight Board (PCAOB). These results further support that SOX enhances the information value of earnings through increased reporting credibility.

Under SOX, the 2003 Regulation G requires firms that report non-GAAP earnings to reconcile them with their GAAP earnings. This earnings reconciliation between GAAP and non-GAAP may change their relationships from substitutes to complements. The credibility and predictive value of earnings patterns are likely to be enhanced when GAAP and non-GAAP earnings are mutually verified and informed under Regulation G (Chen et al. 2021, 2022). Bonus contracts also increase their emphasis on both earnings and increases in earnings after SOX (Carter et al. 2009). If investors can better evaluate a firm’s performance and prospects when the two earnings metrics are reconciled under Regulation G, one would predict higher rewards for increasing earnings patterns. Even if GAAP and non-GAAP earnings may be substitutes after SOX, the presence of non-GAAP earnings appears to change managers’ target-meeting behaviors from a focus on GAAP to non-GAAP earnings after Regulation G (Gilliam et al. 2015; Bauer et al. 2019; Guggenmos et al. 2022). Investors may assign higher premiums for increasing GAAP earnings patterns arising from less opportunistic meet-or-beat behavior over time. Thus, enhanced disclosures and increased scrutiny under SOX may credibly inform investors about firms’ prospects through the pattern of GAAP earnings. If these regulatory efforts work, investors may offer higher rewards to firms with patterns of increasing earnings.

Whether the market changes its rewards for increasing earnings patterns and what factors are associated with the changes are empirical questions.

3 Research design

This paper replicates and extends BEF (1999) to examine how the market changes its rewards for increasing earnings patterns over time and then examine associated factors.

3.1 Price specifications

To determine whether firms with patterns of increasing earnings have higher earnings multiples than other firms and how these multiples vary over time, we estimate Model (1), which is Eq. (3) in BEF (1999), pooled cross-sectionally and temporally, and separately by year.

$$\begin{aligned} PRICE_{it} & = a_{0} + \mathop \sum \limits_{yr} a_{0,yr} D_{yr,t} + a_{1} NI_{it} + \mathop \sum \limits_{yr} a_{1,yr} \left( {D_{yr,t} \times NI_{it} } \right) \\ & \quad + a_{1,b\& u} \left( {D_{b\& u,it} \times NI_{it} } \right) + a_{1,up} \left( {D_{up,it} \times NI_{it} } \right) \\ & \quad + a_{2} \left( {GROWTH_{it} \times NI_{it} } \right) + a_{3} \left( {EVAR_{it} \times NI_{it} } \right) \\ & \quad + a_{4} \left( {DE_{it} \times NI_{it} } \right) + \varepsilon_{it} \\ \end{aligned}$$

(1)

${D}_{up}$ is coded 1 for firm-years with at least five consecutive prior years of increasing earnings and coded 0 otherwise. The year indicator, ${D}_{yr}$, allows the intercept and earnings coefficient to vary by year. The coefficient of interest is ${a}_{1,up}$, which is the incremental coefficient on earnings for firms with patterns of increasing earnings. ${a}_{1,up}>0$ means that firms with at least five years of increasing earnings have higher earnings multiples than other firms. Model (1) allows varying multiples for firms in regulated industries (${D}_{b\&u}$), and it controls for growth (${GROWTH}_{it}$), operating risk (${EVAR}_{it}$) and financial risk (${DE}_{it}$). Based on the permanent earnings model, Model (1) tests whether increasing earnings patterns capture value-relevant characteristics not reflected in these risk and growth proxies. Variable definitions are in the Appendix.

$$\begin{aligned} PRICE_{it} & = a_{0} + \mathop \sum \limits_{yr} a_{0,yr} D_{yr,t} + a_{1} NI_{it} + \mathop \sum \limits_{yr} a_{1,yr} \left( {D_{yr,t} \times NI_{it} } \right) \\ & \quad + a_{1,b\& u} \left( {D_{b\& u,it} \times NI_{it} } \right) + a_{1,up} \left( {D_{up,it} \times NI_{it} } \right) \\ & \quad + a_{2} \left( {GROWTH_{it} \times NI_{it} } \right) + a_{3} \left( {EVAR_{it} \times NI_{it} } \right) \\ & \quad + a_{4} \left( {DE_{it} \times NI_{it} } \right) + a_{5} BVE_{it} + \varepsilon_{it} \\ \end{aligned}$$

(2)

Based on Ohlson’s (1995) valuation model, Model (2), which is Eq. (5) in BEF (1999), additionally considers the book value of equity (${BVE}_{it}$) in its estimation.

3.2 Change in price and returns specifications

Model (3), which is Eq. (6) in BEF (1999), shows two alternative specifications based on a first-difference version of the price level relationships. One specification uses annual changes in price (${\Delta P}_{it}$) as the dependent variable, and the other uses compound 12-month market-adjusted returns (${RET}_{it}$). The independent variables are first-differences of those in Model (2).

$$\begin{aligned} \Delta P_{it} {\text{or}} RET_{it} & = b_{0} + \mathop \sum \limits_{yr} b_{0,yr} D_{yr,t} + b_{1} \Delta NI_{it} \\ & \quad + \mathop \sum \limits_{yr} b_{1,yr} \left( {D_{yr,t} \times \Delta NI_{it} } \right) + b_{1,b\& u} \left( {D_{b\& u,it} \times \Delta NI_{it} } \right) \\ & \quad + b_{1,up} \left( {D_{up,it} \times \Delta NI_{it} } \right) + b_{1,dn} \left( {D_{dn,it} \times \Delta NI_{it} } \right) \\ & \quad + b_{2} \left( {GROWTH_{it} \times \Delta NI_{it} } \right) + b_{3} \left( {EVAR_{it} \times \Delta NI_{it} } \right) \\ & \quad + b_{4} \left( {DE_{it} \times \Delta NI_{it} } \right) + b_{5} \Delta BVE_{it} + \xi_{it} \\ \end{aligned}$$

(3)

Model (3) also examines whether any earnings multiples associated with firms with a pattern of increasing earnings are reduced when the pattern breaks. Given that ${\Delta NI}_{it}<0$ for firms with a breaking pattern, ${b}_{1,dn}$ is expected to be positive if returns are more negative for firms breaking patterns of increasing earnings. This paper also tests whether ${b}_{1,up}$ is different from ${b}_{1,dn}$.

4 Empirical analyses

4.1 Data and descriptive statistics

The sample comprises all firms on Compustat for 1982–2020, with data required by each estimation equation and at least 5 years of earnings history.¹⁰ The focus is on the Extension sample for 1993–2020, which covers the sample period after BEF’s study. This paper also discusses results using the Replication sample for 1982–1992. Returns data are from CRSP, and analysts’ long-term growth forecasts are from IBES. We eliminate outliers with extreme Cooks D-statistics and firms with negative debt-to-equity ratios, and we winsorize ${EVAR}_{it}$ and ${DE}_{it}$ to 100.

Table 1 presents descriptive statistics on the frequency of various patterns of increasing earnings. In Panel A, the rows identify the fiscal year in which the earnings pattern is assessed. The columns present the number of prior years in which the firm reported consecutively increasing earnings, ranging from zero to ten or more. The last column indicates the unconditional percentage of all firms that reported increasing earnings for a given year. Each cell in Panel A indicates the difference between the percentage of firms in that earnings pattern category (i.e., column) and year (i.e., row) reporting increasing earnings and the unconditional percentage for that year. A cell amount is boldfaced if the probability difference is positive. Panel B (C) shows the percentage of the 15,493 (67,127) firm-years with increases in earnings. The 59.13 (57.92) percent in Panel B (C) is similar to 57.43 percent in BEF (1999) and 58.07 percent in Ghosh et al. (2005).

Table 1 Deviations of conditional probability of an earnings increase from yearly unconditional probability by prior earnings pattern for samples from 1984 to 2020

Consistent with BEF (1999), this paper finds a tendency for positive values to cluster in the right half of Panel A. Panel B shows that columns 5 to 10 + , rather than columns 1 to 4, have more positive cells in each column. Furthermore, the average cell amounts for each column are negative in columns 1 to 3 and positive in columns 5 to 10 + . The likelihood that a firm with a pattern of increasing earnings reports increased earnings the following year increases with the length of the increased earnings patterns after 4 years. Our replication results for 1984 to 1992 are comparable to the results reported by BEF (1999). In the extension period, the pattern of increasing earnings for at least 5 years still contains predictive values for earnings increases. Thus, Table 1 justifies the focus on earnings increases for at least 5 years (${D}_{up}$).

Table 2 presents descriptive statistics for variables used in the pooled regressions and the means of variables for the subsamples of firms with increasing earnings patterns (UPEARN) or not (OTHER). UPEARN firms are firms that have at least 5 years of increasing GAAP earnings. Variable distributions in Panel A are comparable to those in BEF (1999). Untabulated Pearson and Spearman correlations support that UPEARN firms are correlated with higher growth (${GROWTH}_{it}$ and ${AN\_GROW}_{it}$) and lower risk (${EVAR}_{it}$). UPEARN firms also have significantly larger market, accounting performance, and size measures in both the Replication and Extension samples.¹¹

Table 2 Descriptive statistics for regression variables, including differences in means for UPEARN firms with at least 5 years of increasing earnings versus OTHER firms without at least 5 years of increasing earnings for a pooled replication (extension) sample of compustat firms from 1982 to 1992 (1993–2020)

4.2 Summary of replication work: BEF’s results are replicable

Based on the Replication sample for 1982–1992, this paper reaffirms BEF’s findings. Key replication results are available in the Online Appendix of this paper.

For pooled regression estimates, this paper considers raw t-statistics and robust t-statistics with standard errors clustered by firm and year (Gow et al. 2010). Regarding Models (1) and (2) on price regressions, the replication analysis reveals that the earnings multiples for firms with at least 5 years of increasing earnings are significantly higher than those for other firms after controlling for growth and risk proxies. ${a}_{1,up}$ is significantly positive (Coef. Est. = 1.51 and 2.11, t = 18.01 and 30.27, robust t = 3.41 and 5.08 for Models (1) and (2)). The other coefficient estimates are consistent with predictions and comparable to BEF’s (1999) results. Firms with increasing earnings patterns have higher price-earnings multiples than other firms.

Findings from separate-year regressions support the pooled sample findings. Regarding Models (1) and (2), the mean estimates of ${a}_{1,up}$ are 1.93 and 2.57, and the mean t-statistics are 6.14 and 10.07. ${a}_{1,up}$ is significantly positive in all eleven yearly regressions for Model (2). Tests based on Z1 and Z2, as used by BEF, indicate that UPEARN firms have significantly positive incremental coefficients on their earnings.¹² In both Models (1) and (2), the coefficient estimates of other control variables are consistent with those in BEF (1999).¹³

To investigate whether earnings multiples increase in pattern length, this paper replaces ${a}_{1,up}\left({D}_{up,it}\times {NI}_{it}\right)$ in Model (2) with ${\sum }_{N=1}^{9+}{a}_{1,up, N}\left({D}_{up, N,it}\times {NI}_{it}\right)$, where $N$ is the number of years in an increasing earnings pattern. Besides finding the effects of increasing pattern lengths, this paper and BEF (1999) document large increases in price-earnings multiples from pattern year four to five. Both studies get F-statistics rejecting the equality of the year nine and five coefficients. The slope coefficient from regressing ${a}_{1,up, N}$ on $N$ is 0.32 (t = 8.62) here, and 0.51 (t = 11.82) in BEF (1999). Thus, BEF’s evidence regarding increasing earnings pattern lengths is replicable.

This paper also examines whether firms with earnings decreases after a previous pattern of increasing earnings have decreased earnings multiples. Confirming BEF’s (1999) findings, the replication results show that earnings multiples decrease as the increasing earnings pattern breaks. The magnitudes of decreases in earnings multiples increase when the decreasing earnings patterns extend. Thus, BEF’s evidence regarding decreasing earnings pattern lengths is also replicable.

This paper follows BEF and performs a battery of specification checks. The paper replaces historical growth (${GROWTH}_{it}$) with analysts’ growth forecasts (${AN\_GROW}_{it}$) and gets robust main findings. The paper also considers change specifications. By estimating Model (3) with price changes (${\Delta P}_{it}$) as the dependent variable, this paper replicates BEF’s (1999) results. Incremental earnings multiples are positively associated with increasing earnings patterns, and there are more negative returns for firms that break patterns of increasing earnings. The main inference remains unchanged when using market-adjusted return (${RET}_{it}$) as the dependent variable. Moreover, the replication results hold after following BEF to consider alternative risk and growth proxies in model specifications, control for firms’ earnings persistence, estimate regressions by industry, or use estimated earnings to replace actual earnings.

This paper also replicates subsample tests by the year of increasing earnings innovations to better infer whether the multiples change over time as earnings patterns develop. In the Replication sample, there is an incremental earnings coefficient over the years of increasing earnings innovations. Firms with 5 years of increasing earnings innovations have incremental earnings coefficients rising almost monotonically from − 0.61 to 2.73. The slope coefficients from regressions of the incremental earnings coefficients on a time trend are significantly positive for all various pattern durations. The replication results are consistent with BEF’s (1999) results.

4.3 Price specification findings

Table 3 presents regression summary statistics from Models (1) and (2) for the Extension sample. In Panel A, ${a}_{1,up}$ is positive and higher than it was in the pre-1993 replication period (Coef. Est. = 2.82 and 3.90, t = 28.33 and 43.01, robust t = 3.82 and 6.34 for Models (1) and (2)). In Panel B, the mean estimates of ${a}_{1,up}$ are 2.74 and 3.71, and the mean t-statistics are 6.21 and 10.04 for Models (1) and (2). Tests based on Z1 and Z2 support the main findings. UPEARN firms with at least 5 years of increasing earnings have higher price-earnings multiples than OTHER firms. BEF’s (1999) results are generalizable to the extension period.

Table 3 Summary statistics from price regressions for samples from 1993 to 2020

By comparing the results between the Replication and Extension samples, the magnitude of ${a}_{1,up}$ increases drastically by over 80% in the pooled regressions.¹⁴ The market boosts its rewards to UPEARN firms after 1993. Further analysis in Sect. 5 supports that SOX is a factor associated with the increased premiums for increasing earnings patterns.

4.4 Evidence regarding pattern length

In Table 4, ${a}_{1,up}\left({D}_{up,it}\times {NI}_{it}\right)$ in Model (2) is replaced with ${\sum }_{N=1}^{9+}{a}_{1,up, N}\left({D}_{up, N,it}\times {NI}_{it}\right)$, where $N$ is the number of years in an increasing earnings pattern. The almost monotonically increasing relationships between pattern lengths and earnings multiples hold. By comparing Table 4 with replication results, this paper finds that the magnitudes of premiums increased after 1993. Untabulated t-tests find significant increases in the estimates of ${a}_{1,up, N}$ between the Replication and Extension samples, when $N$ equals 3, 4, 5, 6, 7, and 9 + at the 1% significance level.¹⁵ Again, F-statistics reject the equality of the year nine and five coefficients. The slope coefficient from regressing ${a}_{1,up, N}$ on $N$ is 0.47 (t = 3.46), which is higher than 0.32 for the Replication sample and supports boosted rewards to increasing earnings patterns.

Table 4 Regression summary statistics for increasing earnings patterns for a pooled sample of compustat firms from 1993 to 2020

By adding ${\sum }_{N=6}^{9+}{a}_{1,dn, N}\left({D}_{dn, N,it}\times {NI}_{it}\right)$ to Model (2), Table 5 examines whether firms with earnings decreases after a previous pattern of increasing earnings have decreased earnings multiples. ${D}_{dn, N}$ is an indicator variable that equals one for firms with $N-5$ consecutive years of decreasing earnings following five or more consecutive years of increasing earnings for $N\in \left[6, 9+\right]$; zero otherwise. ${D}_{oth}$ is an indicator variable that equals one for firms classified as UPEARN in one of the previous 4 years and with at least one but not consecutive years of decreasing earnings after its increasing earnings pattern; zero otherwise. ${D}_{oth}$ accounts for other pattern-breaking firms. In Table 5, two of the four incremental earnings coefficients (${a}_{1,dn, 7}$ and ${a}_{1,dn, 8}$) are lower than those associated with UPEARN firms. However, the magnitudes of decreases in earnings multiples no longer increase monotonically when the decreasing earnings patterns extend to the fourth year of losses.¹⁶ Table 5 finds no monotonically diminishing effect of decreasing earnings pattern length. Overall, this paper supports an increase in premiums associated with increasing pattern length from 1993 to 2020.

Table 5 Regression summary statistics for decreasing earnings patterns after a pattern of increasing earnings for a pooled sample of compustat firms from 1993 to 2020

4.5 Specification checks

This section follows BEF (1999) to perform specification checks. First, the paper replaces historical growth (${GROWTH}_{it}$) with analysts’ growth forecasts (${AN\_GROW}_{it}$). The main inference holds in the extension sample. In the pooled specification, the coefficient of ${D}_{up}\times NI$ is 2.80 (t = 22.94) from Model (2). In the separate-year regressions, the coefficient is significantly positive in 25 out of 28 years with a mean t-statistic of 5.24, Z1 of 27.72, and Z2 of 5.99.

Second, the paper considers change specifications. Table 6 shows statistics from estimating Model (3) with price changes (${\Delta P}_{it}$) as the dependent variable. Panels A and B corroborate price level specification findings. Supporting that incremental earnings multiples are associated with increasing earnings patterns, ${b}_{1,up}$ is significantly positive in both the pooled estimation (t = 8.09) and separate-year regressions (Z1 = 9.30 and Z2 = 5.32). Supporting more negative returns for firms that break patterns of increasing earnings, ${b}_{1,dn}$ is significantly positive in both the pooled estimation (t = 4.54) and separate-year regressions (Z1 = 5.63 and Z2 = 5.90). Untabulated F-statistics reject the equality of ${b}_{1,up}$ and ${b}_{1,dn}$.

Table 6 Summary statistics from returns regressions for samples from 1993 to 2020: changes in price as the dependent variable

Table 7 presents statistics from estimating Model (3) with market-adjusted return (${RET}_{it}$) as the dependent variable. For UPEARN firms, the results are similar to those in Table 6. Untabulated t-tests find marginally significant positive differences in ${b}_{1,up}$ estimates between the Replication and Extension samples. The main results regarding higher valuation multiples for UPEARN firms than for OTHER firms are robust to return or change specifications.

Table 7 Summary statistics from returns regressions for samples from 1993 to 2020: market-adjusted return as the dependent variable

Third, this paper follows BEF (1999) in considering alternative risk and growth proxies in price and returns specifications. Results are robust when using the firms’ annual beta to proxy for risk, the book-to-market ratio for growth, or the earnings trend estimated as the slope coefficient of regressing earnings on a 6-year time trend for growth. Results also hold when measuring earnings variability with the unexplained sum of squared residuals from the time-trend regression. Inferences about higher earnings multiples associated with earnings patterns are unaffected.

Fourth, this paper allows the coefficients of ${NI}_{it}$ and ${BVE}_{it}$ to vary with the persistence of abnormal earnings, measured as the firm-specific relationships between abnormal earnings and lagged abnormal earnings. Following BEF (1999), this paper uses both the winsorized and decile-ranked measurements for earnings persistence and reaches robust conclusions. Fifth, this paper controls earnings persistence with the moving average coefficient from a first-order moving-average, MA (1), process. Untabulated results offer similar inferences.

Sixth, this paper estimates the price and return specifications by industry. Untabulated findings suggest that the inferences are robust across industries using either Fama–French 12 industry codes or SIC 2-digit codes without yearly slopes. Seventh, this paper partitions the sample based on the medians of ${GROWTH}_{it}$, ${EVAR}_{it}$, and ${DE}_{it}$ to alleviate concern about non-linearity. This study estimates the price and returns specifications for the two groups of firms for each variable. Inferences regarding the higher earnings multiples for UPEARN firms are similar across different subsamples. Eighth, this paper follows BEF (1999) to use the instrumental variable of earnings in testing regressions to alleviate concern about earnings measurement errors. The instrumented earnings are predicted from a regression of Model (1) variables, dividend yield, sales, asset book values, and lagged earnings. The incremental coefficients on estimated earnings for UPEARN firms remain significantly positive.

Finally, this paper uses stock prices three months after fiscal year-end, performs tests only on firms with positive earnings, and runs estimations after excluding banks and utility firms. Inferences regarding the earnings multiples for UPEARN firms are robust to specification checks.

4.6 Subsample tests by years of increasing earnings innovations¹⁷

Subsample tests by the year of increasing earnings innovations help to make better inferences on whether the multiples change over time as earnings patterns develop. The replication results are consistent with BEF’s (1999) results. Table 8 presents the results for the Extension period, where it is clear that the earnings multiples change over time as earnings patterns develop. In Panel A, firms with five or more years of increasing earnings innovations have incremental earnings coefficients rising almost monotonically. In the case of eight years ($N=8$) of increasing earnings innovations, the incremental earnings coefficients rise from 1.77 in the first pattern year (${a}_{\text{1,1}}$) to 10.92 in the last pattern year (${a}_{\text{1,8}}$). Panel B presents significantly positive time-trend coefficients for firms with five or more years of increasing earnings innovations. Panel C shows that the incremental earnings coefficient for the first pattern year (${a}_{\text{1,1}}$) is different from that for the last pattern year (${a}_{1,N}, N\in [4, 8]$). These results generalize BEF’s conclusion that the multiples change over time as earnings patterns develop.

Table 8 Summary statistics from price regressions for selected samples within samples from 1993 to 2020: selected firms ultimately have $N$ years of increasing earnings innovations

So far, this paper extends BEF’s results and the results on average strength in the extended sample period: The stock market boosts its rewards for increasing earnings patterns.

5 Further evidence on earnings multiples and patterns

This section explores how notable changes in the financial reporting environment in the past decades are associated with the changes in premiums for increasing earnings patterns. Section 5.1 reports yearly regression analyses with major economic events and examines SOX (fraud) as a positive (negative) factor associated with increased (decreased) premiums for increasing earnings patterns. Section 5.2 further examines the role of non-GAAP earnings information. Section 5.3 discusses further analyses of earnings multiples and patterns.

5.1 SOX, frauds, and yearly analyses of price-earnings multiples

Despite a net market effect of boosted premiums for increasing earnings patterns, this does not imply that factors predicting decreased premiums must be false. Both regulatory reforms and accounting frauds can play roles in the premium changes over time.

5.1.1 Price-earnings multiples over time

This study explores yearly regression estimates to observe potential factors associated with changes in price-earnings multiples for increasing earnings patterns over time. Panel A of Table 9 tabulates yearly earnings multiples (${a}_{1}$) for OTHER firms and incremental earnings multiples (${a}_{1,up}$) for UPEARN firms from the pooled regression of Model (2). Panel A also shows the univariate means of ${a}_{1}$ and ${a}_{1,up}$ partitioned by major economic events. Both the mean multiple ${a}_{1}$ for OTHER firms and the mean incremental multiple ${a}_{1,up}$ for UPEARN firms are higher by over 40% after SOX than before, and the means of ${a}_{1}$ and ${a}_{1,up}$ are lower after frauds occur. Figure 1 visualizes the yearly coefficient estimates of earnings multiples, conditional on exhibiting increasing earnings patterns, together with major economic events from Table 9.

Table 9 Yearly regression statistics from price regressions and statistics from tests of the value premiums to UPEARN firms across SOX and around major frauds from 1982 to 2020

Fig. 1

Price-Earnings Multiple Coefficients for UPEARN and OTHER Firms. Based on the sample from 1982 to 2020, this figure shows the yearly results of price-earnings multiple coefficients estimated from Model (2) for UPEARN and OTHER firms. The solid line presents the earnings multiples (${a}_{1}+{a}_{1,up}$) for UPEARN firms with patterns of at least five consecutive years of increasing earnings. The dashed line presents the earnings multiples (${a}_{1}$) for OTHER firms. This figure supports BEF’s findings (1999) of higher earnings multiples for UPEARN firms before 1992. The market rewards for increasing earnings patterns are higher after 1993 than before. The market rewards for increasing earnings patterns shrink when scandals hit the market

5.1.2 Roles of SOX and frauds

Either fraud or SOX can change the information conveyed about a firm’s prospects and management credibility through increasing earnings patterns (Koonce and Lipe 2010). During the years of fraud, investors may require higher risk premiums from firms showing increasing earnings streaks (Kasznik and McNichols 2002; Koh et al. 2008). Prior research shows a contagion effect of unfavorable accounting quality among peer firms (Darrough et al. 2020; Jia and Zhao 2020). The contagion could have market-wide effects. Thus, this paper expects fraud to be associated with decreased rewards for increasing earnings patterns.

The effect of SOX is unclear. SOX and related regulatory reforms might effectively restore investors’ confidence in the accounting system through increased disclosures and audit monitoring (Cohen et al. 2008; Altamuro and Beatty 2010; Singer and You 2011; Gilliam et al. 2015; Burgstahler and Chuk 2017; Gipper et al. 2020). On the other hand, Francis et al. (2003), Ghosh et al. (2005), and Kao and Na (2013) are silent about the effect of SOX on the market pricing premium for increasing earnings patterns. Liu and García Lara (2015) find that SOX has no effect, and Koh et al. (2008) find decreased market rewards associated with street earnings meeting or beating analysts’ forecasts even after SOX. Although visual evidence in Fig. 1 and descriptive evidence from Panel A of Table 9 support a positive effect of SOX, this paper pursues statistical tests to examine conjectures regarding fraud and SOX.

To examine how investors’ pricing of earnings and increasing earnings patterns vary across SOX and around frauds, this study estimates modified Models (1) and (2) after adding interacting terms between the original independent variables and $SOX$ or $Scandal$. The variables of interest are ${D}_{up}\times NI\times SOX$ and ${D}_{up}\times NI\times Scandal$.

In Panel B of Table 9, the coefficient on ${D}_{up}\times NI$ is 2.95 (robust t = 14.74), that of ${D}_{up}\times NI\times SOX$ is 1.43 (robust t = 2.97), and that of ${D}_{up}\times NI\times Scandal$ is − 1.31 (robust t = − 2.32) from the modified Model (2). After controlling for market pricing of earnings information across SOX and around scandals, results show that the premium for increasing earnings patterns increases by 48% on average (e.g., 1.43/2.95) after SOX and decreases by 44% (e.g., − 1.31/2.95) when scandals occur. Results from the modified Model (1) offer similar inferences.

These results support the effectiveness of regulatory efforts under SOX and offer an estimate of fraud costs in terms of a multiplier associated with the incremental earnings multiples for UPEARN firms. Overall, about half of the value premiums to continuously profitable firms evaporate when frauds shake investors' trust in the credibility of the accounting system.¹⁸

5.2 GAAP and non-GAAP earnings, their patterns, and price-earnings multiples

This section explores the role of non-GAAP earnings in investors’ pricing of increasing earnings patterns. When both GAAP and non-GAAP earnings are present, their information may have a dynamic relationship in the changing reporting environment. Bradshaw and Sloan (2002) and Bartov et al. (2002) find that investors shifted focus from GAAP to non-GAAP earnings news after the 1990s. If investors view non-GAAP and GAAP earnings as substitutes, this paper would expect a shift in market rewards for increasing GAAP earnings patterns to rewards for increasing non-GAAP earnings patterns. If the rewards for increasing non-GAAP earnings patterns go up when those for increasing GAAP earnings go down, the premiums for increasing GAAP earnings patterns would be negatively correlated with those for increasing non-GAAP earnings patterns.

However, the mandatory reconciliation between GAAP and non-GAAP earnings under the 2003 Regulation G, as a reform under SOX to promote transparent earnings-related disclosures, may change the relationships between GAAP and non-GAAP earnings information from substitutes to complements (Chen et al. 2021). If so, this paper would expect a change in the correlation between increased market rewards for increasing GAAP earnings patterns and those for non-GAAP earnings patterns across SOX.

5.2.1 Descriptive statistics for subsamples by the presence of non-GAAP earnings history

Panel A of Table 10 shows descriptive statistics for the subsamples partitioned by whether a firm has at least 5 years of non-GAAP earnings history. Firms with non-GAAP earnings history have higher prices, earnings, book equity, growth, and lower earnings volatility than firms without non-GAAP earnings history. For firms with non-GAAP earnings history, the mean of non-GAAP earnings is higher than that of GAAP earnings, which is consistent with the finding that items excluded from GAAP earnings are typically income-decreasing (Black et al. 2021).

Table 10 Analyses for firms with or without non-GAAP earnings history from 1989 to 2020

5.2.2 Regression analyses for subsamples by the presence of non-GAAP earnings history

Panel B of Table 10 shows the results of three tests for the two subsamples separated by the presence of non-GAAP earnings history. Recent studies suggest that the presence of non-GAAP earnings alters managers’ reporting incentives to manage GAAP earnings to meet benchmarks (Bauer et al. 2019; Guggenmos et al. 2022). If firms without (with) a non-GAAP earnings history tend to report their GAAP earnings more (less) aggressively to meet benchmarks, one would expect lower (higher) incremental earnings multiples for their increasing GAAP earnings patterns. The tests with the subsamples separated by the presence of non-GAAP earnings history help to assess whether the market would price GAAP earnings patterns differently in the cases with or without non-GAAP earnings information.

Based on firms without non-GAAP earnings history, Test (1) finds a significantly positive ${a}_{1,up}$ (Coef. Est. = 2.78, robust t = 7.54), meaning that firms with increasing GAAP earnings patterns have higher earnings multiples than other firms. Similarly, Test (2) finds a significantly positive ${a}_{1,up}$ (Coef. Est. = 2.91, robust t = 6.45), indicating incrementally positive rewards for increasing GAAP earnings patterns among firms with non-GAAP earnings history. Tests (1) and (2) reaffirm that firms with increasing GAAP earnings patterns have higher earnings multiples than other firms. While the incremental earnings multiple for increasing earnings patterns is higher in Test (2) than that in Test (1), which supports higher market rewards for less aggressive GAAP earnings patterns when non-GAAP earnings are present (Bauer et al. 2019; Guggenmos et al. 2022), the differences in the incremental multiples between the two cases are insignificant.

Among firms with non-GAAP earnings history, Test (3) further examines how the market prices their increasing GAAP and non-GAAP earnings patterns. Given a positive ${a}_{1,up}$ from Test (2), finding an insignificant ${a}_{1,up}$ (Coef. Est. = 0.37, robust t = 1.03) and a significantly positive ${a}_{1,up}^{NG}$ (Coef. Est. = 3.32, robust t = 6.87) from Test (3) implies that the incremental rewards for increasing GAAP earnings patterns are subsumed by the rewards for increasing non-GAAP earnings patterns. Since Test (2) shows incremental rewards for increasing GAAP earnings patterns, Test (3) cannot imply that investors value only non-GAAP earnings patterns and ignore GAAP earnings patterns, especially when GAAP and non-GAAP earnings mutually inform each other. Instead, Tests (2) and (3) jointly suggest that the information value of GAAP earnings patterns can be driven and reinforced by the information value of non-GAAP earnings patterns.

5.2.3 Pricing of GAAP and non-GAAP earnings information over time and across SOX

To understand the dynamic relationship between GAAP and non-GAAP earnings information over time, Panel C of Table 10 tabulates yearly earnings multiples and incremental multiples for increasing earnings patterns for the subsamples partitioned by the presence of non-GAAP earnings history. The ${a}_{1}$ coefficient estimates and their mean differences across SOX from Tests (1) and (2) support Gipper et al.’s (2020) findings that the market values earnings information more after SOX. There is little evidence from Tests (1) and (2) that the market gives more or less value to earnings patterns after SOX.¹⁹ However, results from Test (3) suggest that the market rewards increasing non-GAAP (GAAP) earnings patterns more (less) among firms with non-GAAP earnings history after SOX.

Based on firms with non-GAAP earnings history, Panel D of Table 10 shows a significantly negative correlation, − 0.643, between ${a}_{1,up}$ and ${a}_{1,up}^{NG}$ in the pre-SOX period, but an insignificant correlation in the post-SOX period. The correlation evidence supports that GAAP and non-GAAP earnings information substituted each other before SOX, but that the substitute relationship might change after SOX when the two earnings measures are reconciled and mutually informed (Chen et al. 2021, 2022).

5.2.4 Non-GAAP earnings, SOX, frauds, and price-earnings multiples

SOX (fraud) contributes positively (negatively) to the increased premiums for increasing earnings patterns. A pattern of increasing non-GAAP earnings helps explain the premiums for increasing GAAP earnings patterns, and the market changes its pricing of increasing GAAP and non-GAAP earnings patterns after SOX. Does non-GAAP earnings information also play a role in the premium discounts around frauds (Koh et al. 2008)? How do we relate our findings to the research showing that investors shifted their focus from GAAP to non-GAAP earnings news after the 1990s (Bradshaw and Sloan 2002; Bartov et al. 2002)?

Figures 2 and 3 visualize the distributions of yearly incremental earnings multiples (tabulated in Panel C of Table 10) to firms with increasing earnings patterns. Figure 2 shows the premiums for increasing GAAP earnings patterns to firms with and without non-GAAP earnings history. The presence of non-GAAP earnings history may speak to the premium discounts in the years of fraud. When major scandals occurred in the early 2000s, the market offered lower rewards to UPEARN firms among firms with non-GAAP earnings history than it did to UPEARN firms among firms without such history. The thinner rewards found in the window with major accounting scandals from 2001 to 2005 corroborate Koh et al.’s (2008) pre-2007 evidence that investor skepticism when firms meet benchmarks is associated with lower market rewards in the post-scandal world. However, Fig. 2 shows that discounted premiums rebounded after 2007 (Edmonds et al. 2018). Based on Fig. 2, we conjecture that the discounts on premiums for increasing earnings patterns in the years of fraud occur mainly in firms with non-GAAP earnings history.

Fig. 2

Incremental Earnings Multiples Associated with the Patterns of Increasing GAAP Earnings for Firms with and without Non-GAAP Earnings History. Based on the subsamples from 1989 to 2020, this figure shows the yearly results of incremental earnings multiples estimated from Model (2) for UPEARN firms that have increasing GAAP earnings patterns. The dotted line presents the incremental earnings multiples (${a}_{1,up}$) for UPEARN firms among firms without non-GAAP earnings history. The dashed line presents the incremental earnings multiples (${a}_{1,up}$) for UPEARN firms among firms with non-GAAP earnings history. The premiums for increasing GAAP earnings patterns are stable over time for firms with and without non-GAAP earnings history. One notable exception is a temporary drop in value premiums for increasing GAAP earnings patterns to firms with non-GAAP earnings history between 2000 and 2005

Fig. 3

Incremental Earnings Multiples Associated with the Patterns of Increasing GAAP or Non-GAAP Earnings for Firms with Non-GAAP Earnings History. Based on the subsample of firms with non-GAAP earnings history from 1989 to 2020, this figure shows incremental earnings multiples yearly results for increasing earnings patterns. The solid line presents the incremental earnings multiples (${a}_{1,up}^{NG}$) for increasing non-GAAP earnings patterns after controlling for the increasing patterns of GAAP earnings. The dashed line presents the incremental earnings multiples (${a}_{1,up}$) for increasing GAAP earnings patterns after controlling for the increasing patterns of non-GAAP earnings. The market increased (reduced) the rewards for increasing non-GAAP (GAAP) earnings patterns since 1994. The incremental multiples for increasing GAAP earnings patterns declined after 2000 when several scandals occurred but rebounded in later years following the implementations of SOX and related regulations like Regulation G

Figure 3 focuses on firms with non-GAAP earnings history and shows the incremental premiums for increasing GAAP and for increasing non-GAAP earnings patterns. There are several observations. First, investors have increased (decreased) their rewards for increasing non-GAAP (GAAP) earnings patterns since 1994. Based on the sample of firms with non-GAAP earnings history, this observation confirms Bartov et al.’s (2002) and Bartov and Cohen’s (2009) views that investors shifted their focus from GAAP to non-GAAP earnings then. Second, the skepticism about GAAP earnings seems to be what mainly contributes to the discount for the premiums to firms that have non-GAAP earnings history and show increasing GAAP earnings patterns around the early 2000s when scandals arose (Koh et al. 2008). Third, the incremental premiums for increasing non-GAAP earnings patterns remain stable and increase modestly following the 2003 Regulation G under SOX. Relative to the premiums for increasing non-GAAP earnings patterns, the incremental premiums for increasing GAAP earnings patterns rebound but stay low after 2005. Based on Fig. 3, we conjecture that the increased premiums for increasing earnings patterns are associated with increasing non-GAAP earnings patterns, and the premiums strengthen after SOX.

Panel E of Table 10 shows modified price regression results after pooling the two subsamples to examine how market rewards differ between GAAP and non-GAAP earnings patterns across contexts of interest, including SOX ($SOX$) and frauds ($Scandal$). From the modified Model (2), the coefficients on ${D}_{up}^{NG}\times {NI}^{NG}\times Scandal$ are significantly negative (Coef. Est. =  − 4.07, robust t = − 2.58), consistent with our conjecture regarding discounts on premiums. On the other hand, the coefficient on ${D}_{up}\times NI\times SOX$ is significantly positive (Coef. Est. = 2.28, robust t = 2.09), partly supporting higher premiums for increasing GAAP earnings patterns after SOX among firms without non-GAAP earnings history. In both specifications, the coefficients on ${D}_{up}^{NG}\times {NI}^{NG}\times SOX$ are significantly positive, which supports that SOX is positively associated with increased investors’ pricing of increasing earnings patterns, and the higher premium is specific to firms that provide more non-GAAP earnings information.

A plausible case showing that SOX has a role in the premium changes is Regulation G. When GAAP and non-GAAP earnings information are mutually informed and verified, investors may find them more credible and informative, and be willing to offer higher premiums for increasing earnings patterns (Chen et al. 2021, 2022; Guggenmos et al. 2022). However, one should be cautious about using current evidence to infer the causal effect of SOX on earnings pricing. Overall, the increased (decreased) premiums for increasing earnings patterns after SOX (around frauds) are primarily driven by firms with non-GAAP earnings history.²⁰

5.3 The changing landscape of price-earnings multiples

This paper confirms the presence of value premiums for increasing earnings patterns before 1992, documents an increase in such premiums after 1993, and validates findings regarding pattern length. Higher (lower) premiums for increasing earnings patterns are associated with the introduction of SOX (the incidence of major frauds). Analyses of subsamples partitioned by the presence of non-GAAP earnings history further support the enhanced role of non-GAAP earnings in investors’ pricing of earnings patterns, especially after 1994.

These findings reflect changes in the reporting environment from different aspects. On the one hand, the evidence from firms with non-GAAP earnings history suggests that the premiums for increasing non-GAAP earnings patterns subsume those for increasing GAAP earnings patterns. It also supports investors’ shifted focus from GAAP to non-GAAP earnings since 1994 (Bartov et al. 2002; Lev and Gu 2016). The evidence from the non-GAAP subsample also supports valuation discounts for skepticism of earnings credibility around frauds (Koh et al. 2008). On the other hand, there are increased premiums for increasing earnings patterns after SOX in pooled samples. Firms with increasing non-GAAP earnings patterns especially drive the increased premiums. The pooled sample findings support that SOX is associated with investors’ increased pricing of earnings pattern information (Gipper et al. 2020). Thus, differences in samples based on whether firms have non-GAAP earnings history can reconcile seemly conflicting views about whether the market would price increasing earnings patterns higher or lower.

If SOX can partly explain the increased premiums for increasing earnings patterns, a further issue would be to examine the channels through which SOX affects the market pricing of earnings patterns: better information about firms’ prospects or increased management credibility (Koonce and Lipe 2010). Supporting the view that accruals are more subject to managerial opportunism than cash flows, Fig. 4 indicates higher (lower) market rewards associated with increasing earnings patterns with lower (higher) accruals. If management credibility increases after SOX and earnings pattern information reflects the credibility increase, one would expect increased rewards for increasing earnings patterns with higher accruals.²¹ However, untabulated Chow tests find that SOX has insignificant effects on the incremental rewards to UPEARN firms with either higher or lower accruals. Therefore, Fig. 4 offers neutral evidence regarding increased management credibility for UPEARN firms after SOX.

Fig. 4

Incremental Earnings Multiples Associated with the Patterns of Increasing GAAP Earnings with Different Levels of Accruals. Based on the sample from 1987 to 2020, this figure shows yearly results for incremental earnings multiples for increasing GAAP earnings patterns associated with different levels of accruals. The sample period starts in 1987 after the cash flow statement information became available. By year, UPEARN firms are sorted into quintiles by their signed accruals, scaled by total assets. UPEARN firms with higher (lower) accruals are from the first quintile (the other four quintiles) for each year. UPEARN firms in the first quintile for each year have positive accruals on average, and UPEARN firms in the other quintiles do not. The solid line presents the incremental earnings multiples (${a}_{1,up}$) for UPEARN firms with higher accruals. The dashed line presents the incremental earnings multiples (${a}_{1,up}$) for UPEARN firms with lower accruals. The market offers higher (lower) rewards for increasing GAAP earnings patterns associated with lower (higher) accruals. The premiums for increasing earnings patterns associated with different levels of accruals tend to be stable over time, except for the years when major scandals occur

By combining the findings that (1) the premiums for increasing GAAP earnings patterns increase after SOX, (2) increasing non-GAAP earnings patterns can account for the premiums for increasing GAAP earnings patterns, and (3) the premiums for increasing GAAP or non-GAAP earnings patterns increase after SOX, these results support the channel of better information on firms’ prospects. Increased disclosures under SOX, such as the reconciliation between GAAP and non-GAAP earnings under Regulation G, may boost the information value of increasing earnings patterns and help investors assess firms’ prospects, resulting in higher multiples associated with increasing earnings patterns. However, the two channels—prospect information and management credibility—may not be mutually exclusive (Koonce and Lipe 2010). How these two channels may interact and jointly affect investors’ pricing of earnings patterns can be a question for future research. This paper does not preclude the possibility that the increased predictive values of earnings patterns information arise from the increased reporting credibility after SOX (Altamuro and Beatty 2010; Gipper et al. 2020; Li 2019; Singer and You 2011).

To gauge the role of earnings and their patterns in the information mix in the changing environment over the past few decades, Table 11 tabulates key coefficients of earnings multiples and incremental premiums for increasing earnings patterns by decade. The results show that the magnitudes of market rewards for increasing earnings patterns are higher starting in the 1990s than before. Specifically, ${a}_{1, up}\hspace{0.17em}$= 4.16 in the 1990s, ${a}_{1, up}\hspace{0.17em}$= 2.22 in the 2000s, and ${a}_{1, up}\hspace{0.17em}$= 4.98 in the 2010s are higher than ${a}_{1, up}\hspace{0.17em}$= 1.82 in the 1980s. On an absolute basis, the market increases its rewards for increasing earnings patterns after the 1980s. Consistent with the increase in ${a}_{1, up}$, the percentage of premiums to UPEARN firms compared to multiples of OTHER firms in the same period (${a}_{1, up}/{a}_{1}$) is higher in 1990s (85%) compared to the 1980s (72%). However, it is lower in the 2000s (46%) and the 2010s (53%) suggesting that, on a relative basis, continuously profitable firms may play a less important role in the information mix after the 2000s because the general reactions to earnings have increased. This inconsistency between measuring market rewards on an absolute basis and on relative basis is a caveat of this paper. Future research is needed to reconcile this inconsistency.

Table 11 Price-earnings multiples by decade

BEF (1999) conjecture that the source of premiums can be growth path information, and Koonce and Lipe (2010) further suggest that prospect information and management credibility are contributing factors. While showing that the premiums for increasing earnings patterns are associated with notable changes in the reporting environment, like non-GAAP reporting, SOX, and accounting scandals, this paper notes that the sources of the premium changes are not fully understood and warrant more research.

Besides non-GAAP earnings, accounting scandals, or regulatory reforms, additional factors may explain the changing landscape of earnings multiples. For example, the downward trending interest rates since the 1990s may contribute to the upward trending pricing of equity stocks, earnings, and earnings patterns. The technological advances in firms’ accounting information production since the 2000s may enhance the credibility of earnings information. The increase in forward-looking and segment disclosures since the 2000s may enrich information about firms’ prospects conveyed through earnings and their patterns. How investors consume and use accounting information has been evolving with investors’ skills and sophistication, which may contribute to the observed differential pricing. Since the US economy has been increasing intangible investments since the late 1980s, another factor may be the accounting deficiencies associated with understated earnings and, thus, increased earnings multiples (Lev and Srivastava 2022). Another factor worthy of consideration for future research is the rising phenomenon of a winner-takes-all market where the top performers capture a disproportionately large share of the rewards.²² This paper focuses on how the variations in earnings multiples are associated with non-GAAP earnings, frauds, or regulatory reforms after controlling other time-variant factors with the yearly indicator design. Future research is needed to examine ways that other factors would affect investor pricing of earnings and earnings patterns.

6 Conclusion

The increase in annual earnings in a streak of more than 5 years is value-relevant. On average, the market rewards higher earnings multiples to firms exhibiting increasing earnings patterns. The reward in earnings multiples is almost twice as high as it was three decades ago. Additional analyses offer more insights into the roles of non-GAAP earnings, frauds, and SOX in investor pricing of earnings patterns. For firms with non-GAAP earnings history, investors have rewarded increasing non-GAAP (GAAP) earnings patterns more (less) since the 1990s. The market temporarily but sharply cuts its value premiums for increasing earnings patterns around major accounting frauds. SOX is associated with increased valuation premiums for increasing earnings patterns, especially for firms that demonstrate increasing non-GAAP earnings patterns. These results support the valuation relevance of earnings patterns and the effectiveness of regulatory efforts under SOX to restore investors’ trust in the accounting system.

A more general and high-level issue is what basis should be used to evaluate the value of accounting information altogether (Ball 2024b). Based on the magnitude of earnings multiples, this paper shows that the market boosts its rewards for increasing earnings patterns. Earnings multiples for continuously profitable firms increased on average, but the multiples for other firms increased at an even higher rate after the 2000s. SOX appears to leverage the value of earnings and earnings pattern information for both firm groups. Future research is needed to examine the effects of market-based incentives on firm activities when the incentives swing between short-term and long-term orientations on an absolute or relative basis. Puzzles remain in identifying other determinants of the premium changes, in explaining the relatively lesser role of earnings patterns in the information mix, and in better understanding the shifts in the mix and how the market prices elements within the mix (Barth et al. 2023).

The financial reporting environment has changed drastically since the late 1990s with shifting investor focus, accounting scandals, and impactful regulations. The market rewards associated with patterns of increasing earnings are maintained in the late 1990s, briefly gone with frauds, but stay high when SOX works effectively. Depicting the changing landscape of earnings multiples helps to understand factors associated with investors’ pricing of earnings and earnings patterns. Investors, managers, researchers, and regulators may find relevance in the contextual evidence about changing rewards for increasing earnings patterns over time.

Appendix: Variable definitions

Variable	Description
${PRICE}_{it}$	Price per share at fiscal year-end
${NI}_{it}$	Earnings per share before extraordinary items and discontinued operations
${D}_{yr}$	Indicator variable that equals one for the year $yr$; zero otherwise
${D}_{b\&u}$	Indicator variable that equals one for banks and utilities; zero otherwise. This paper assigns one to industries 8 (Utilities) and 11 (Finance) under Fama and French 12 industry codes
${D}_{up}$	Indicator variable that equals one for firm-years with at least five consecutive prior years of increasing earnings; zero otherwise
${D}_{up,N}$	Indicator variable that equals one for firms with $N$ consecutive years of increasing earnings, $N\in \left[0, 9 \text{or more}\right]$; zero otherwise
${D}_{dn, N}$	Indicator variable that equals one for firms with $N-5$ consecutive years of decreasing earnings following five or more consecutive years of increasing earnings for $N\in \left[6, 9\right]$; zero otherwise
${D}_{oth}$	Indicator variable that equals one for firms classified as UPEARN in one of the previous 4 years and having at least one but not consecutive years of decreasing earnings since that time; zero otherwise
${GROWTH}_{it}$	The 5-year compound annual growth rate of equity book value
${BVE}_{it}$	Book value of equity per share
${EVAR}_{it}$	The variance of the past 6 years’ percentage change in earnings
${DE}_{it}$	Debt-to-equity ratio
${AN\_GROW}_{it}$	The mean IBES long-term growth forecasts outstanding around the earnings announcement
${\Delta P}_{it}$	$({PRICE}_{it}-{PRICE}_{it-1})/{PRICE}_{it-1}$
${RET}_{it}$	The 12-month compound monthly market-adjusted return, beginning with the fiscal year’s first month. Specifically, it is the firm’s compound return minus the compound value-weighted market return obtained from CRSP
${\Delta NI}_{it}$	$({NI}_{it}-{NI}_{it-1})/{PRICE}_{it-1}$
${\Delta BVE}_{it}$	$({BVE}_{it}-{BVE}_{it-1})/{PRICE}_{it-1}$
${D}_{dn}$	Indicator variable that equals one for firms breaking a pattern of at least five consecutive years of increasing earnings through the prior year by reporting decreasing earnings in the current year; zero otherwise
${NI}_{it}^{NG}$	Non-GAAP earnings per share, measured as actual street earnings from IBES
${D}_{up}^{NG}$	Indicator variable that equals one for firm-years with at least five consecutive prior years of increasing non-GAAP earnings; zero otherwise