Going Concern Risk and the Market’s Use of Earnings: Multi-Year Evidence Preceding First-Time Going Concern Opinions

Abstract

This study examines whether earnings become less informative as firms near a first-time going concern audit opinion. The question is important because the going concern assumption underlies the measurement of earnings, and growing doubt about that assumption may substantially alter the market’s use of earnings information before the auditor formally issues the opinion. Using firms that receive first-time going concern audit opinions, the analysis traces the intertemporal behavior of earnings informativeness for up to six years before the opinion through the opinion year, measuring informativeness through the relation between abnormal stock returns and unexpected earnings. The results show a substantial decline in earnings informativeness as the going concern opinion approaches, with the earnings response coefficient becoming indistinguishable from zero at least two years before, and in the year of, the opinion. The decline occurs earlier and is more pronounced for firms receiving consecutive going concern opinions. Additional tests indicate that known determinants of earnings informativeness explain part, but not all, of the decline, while evidence on abnormal earnings and accruals is more consistent with severe distress or conservative reporting than with systematic income-increasing earnings management. Overall, the findings suggest that the market revises downward the usefulness of earnings well before the issuance of a first-time going concern audit opinion.

1. Introduction

This study investigates the relation between going concern uncertainty and the informativeness of earnings for firms that receive first-time going concern audit opinions. The audit report, which represents the independent auditor’s opinion on whether the financial statements present fairly the firm’s financial position and results of operations, is the focus of considerable research. That research is motivated by, among other things, a desire to assess the information conveyed to the market through the auditor’s report. Evidence of this kind reflects on the timeliness and relevance of audit reports. One of the most important and debated features of an audit report is the going concern opinion, where auditors render a judgment on whether or not the firm is expected to continue as a going concern.1 Going concern risk is, at its core, a financial risk. It matters to lenders who price default, to investors who bear distress-related losses, and to auditors who must decide when that risk warrants formal disclosure. This study asks when the equity market begins to act on that risk.

Similarly, research on the informativeness of earnings constitutes another substantial focus of scholarly research. That research is motivated by, among other factors, a desire to assess the information conveyed through earnings, the determinants of earnings informativeness, and the residual implications of earnings reports (such as contracting and managerial remuneration). Evidence from this research bears on the relevance of earnings and related policy implications.2 A fundamental assumption underlying the computation of earnings is that of a firm’s going concern, with many key accounting measurements and disclosures predicated on the applicability of the going concern assumption.

This study links these two streams of research to provide evidence on the relation between going concern uncertainty as reflected in the auditor’s report and the informativeness of earnings. The emphasis is on the informativeness of earnings conditional on going concern uncertainty. A unique aspect of this study is the examination of earnings informativeness up to six years prior to a going concern audit opinion, extending through each year up to and including the going concern audit opinion. The evidence obtained from this investigation is applicable to the question of whether or not market participants differentially assess earnings informativeness preceding the going concern audit opinion.

The prospect of differential informativeness derives from several prior studies. Dong et al. (2015) examine the market’s response to earnings surprises after first-time going concern modifications and document a decline in earnings response coefficients after the opinion, especially when the modification is unexpected. Myers et al. (2018) show that measured market reactions to going concern modifications are often confounded by concurrent earnings announcements and that the incremental announcement effect of the modification itself is weaker than prior research suggests. Gutierrez et al. (2020) show that going concern opinions provide incremental information for predicting corporate defaults beyond conventional default predictors. Charitou et al. (2011) show that distress risk, growth, and earnings quality are related in important but non-monotonic ways. This study differs from these papers by examining the intertemporal behavior of earnings informativeness within firms that ultimately receive a first-time going concern audit opinion. It does so for the years preceding, rather than following or contemporaneous with, the opinion year and focuses on the within-firm temporal path of earnings informativeness rather than the cross-sectional relation between termination probability and informativeness. This design also complements Subramanyam and Wild (1996) by shifting the focus from the cross-sectional relation between termination probability and earnings informativeness to the within-firm temporal path of earnings informativeness as the going concern opinion approaches.

Evidence on these issues is important to both accounting and auditing.3 The design does not presume that investors know with certainty, in earlier periods, that a first-time going concern opinion will later be issued; rather, it examines whether the market’s use of earnings declines as the conditions associated with such an opinion accumulate. This design addresses a specific gap in the literature. Prior studies examine the information conveyed at or after the going concern opinion, or the cross-sectional link between distress and the returns–earnings relation. The within-firm temporal path of earnings informativeness in the years preceding a first-time opinion remains undocumented. Closing this gap contributes to theory on how the going concern assumption underpins the measurement and valuation relevance of earnings. It also speaks to a broader question—how capital markets process accounting information under mounting financial distress and risk—an issue of continuing importance to investors, auditors, and regulators.

There are several reasons to anticipate a decline in earnings informativeness in the years preceding a going concern audit opinion. Prior research shows that firms that receive a going concern audit opinion are in financial distress and are often subject to serious structural changes (e.g., Carson et al., 2013; Gutierrez et al., 2020). Prior research also shows that the going concern audit opinion is at least to a degree predictable using accounting and market variables and, in some settings, by the trading behavior of sophisticated investors (e.g., Blay & Geiger, 2001; Geiger & Kumas, 2018). Related evidence spans markets and settings. Averio (2021) finds that higher leverage, and weaker liquidity and profitability, increase the likelihood that a going concern opinion is issued. Models built on accounting indicators, such as the Altman Z-score for failure risk and the Beneish M-score for reporting irregularities, are widely used to identify these risks before serious consequences occur (Altman, 1968; Srebro et al., 2021). Recent work extends this line to auditors’ own risk assessments (Deis et al., 2026) and to the early-warning performance of distress-prediction models (Ashraf et al., 2019).

This study offers a complementary, market-based signal: the fading response of stock prices to earnings news. Changes in firm characteristics that aid the prediction of the going concern opinion can reasonably be expected to affect earnings informativeness. Further, firms receiving going concern audit opinions are subject to increased scrutiny and, as such, are likely to have increased predisclosure information. We also know that firms in distress are documented to alter reporting incentives and, in some cases, manage earnings (e.g., Charitou et al., 2011; Campa & Camacho-Miñano, 2015; Li et al., 2020). These factors predict a decline in earnings informativeness as the firm moves closer to the point at which the auditor issues a going concern opinion. Accordingly, this study’s main hypothesis predicts a decline in earnings informativeness during the years preceding and including the year of the going concern audit opinion. This hypothesis is distinct from prior research that has examined the information conveyed in the going concern opinion itself or in earnings contemporaneously disclosed with the going concern opinion (e.g., Dong et al., 2015; Myers et al., 2018). More specifically, this study looks at the temporal behavior in earnings informativeness for up to six years in advance of a going concern opinion. The six-year horizon is intended to capture the longer-run evolution of earnings informativeness while preserving sufficient observations for estimation in each pre-opinion year. This study also considers whether changes in firms’ economic characteristics and managers’ accounting choices help explain the hypothesized decline in earnings informativeness preceding going concern opinions.

As a preview, the evidence in this study reveals a marked and statistically significant decline in the informativeness of earnings at least two years prior to the actual receipt of the going concern audit opinion. Specifically, the earnings response coefficient, calculated as the slope coefficient from a regression of abnormal returns on unexpected earnings, declines by roughly 95 percent for the two years prior to, and the year of, the going concern audit opinion. None of the earnings response coefficients for those three years is statistically different from zero at conventional levels. In this sense, the decline in earnings informativeness is persistent in that it remains depressed from at least two years prior to the going concern opinion through the year of the going concern opinion.4

We extend the analysis to consider a corollary hypothesis that firms facing greater going concern uncertainty exhibit a greater, and accelerated, decline in earnings informativeness. We test this corollary hypothesis by conditioning the analysis on whether or not firms receive two consecutive going concern audit opinions, presumably indicative of greater going concern uncertainty. This study’s findings reveal that the decline in the informativeness of earnings occurs earlier and is more pronounced for firms with consecutive going concern audit opinions. This accelerated decline is evidenced by a significant drop in earnings informativeness three years prior to the going concern audit opinion. The pronounced decline in the earnings response coefficient persists as we observe a similarly significant drop in earnings informativeness for the year before and the year of the going concern audit opinion for those firms with two consecutive going concern opinions.

We extend the analysis to explore potential sources of the decline in earnings informativeness preceding the going concern audit opinion. We begin by looking at temporal variations in known determinants of earnings informativeness and investigate the possibility that changes in these determinants explain the significant drop in earnings informativeness leading up to the going concern opinion. The determinants consist of firm size, leverage, growth, systematic risk, earnings variability, and persistence. We find evidence of significant shifts in each of these determinants, for one or more years, leading up to the going concern opinion. Moreover, the direction of the shifts in these determinants is generally consistent with the decline in earnings informativeness. For example, both leverage and earnings variability significantly increase in each of the three years preceding the going concern audit opinion. Increases in both of these determinants are consistent with the observed deterioration in earnings informativeness for the same period. We also find that growth significantly decreases for the two years preceding the going concern opinion, which is similarly consistent with a decline in earnings informativeness. These results are consistent with the possibility that the market at least partially anticipates greater going concern uncertainty and its implication for earnings informativeness.

To explore further these determinants, we include them as explanatory variables in a regression of returns on earnings to assess whether or not they can account for the temporal behavior in earnings informativeness leading up to and including the going concern audit opinion. Results show that the inclusion of these determinants significantly adds to the explanatory power of earnings for returns as the multiple correlation coefficient (R2) increases by more than 50 percent. Yet, in spite of the fact that many of these determinants prove significant and, moreover, explain a significant portion of the temporal decline in earnings informativeness, a significant unexplained decline in earnings informativeness persists at least two years prior to, and continuing through the year of, the going concern audit opinion.

A final analysis of potential determinants considers the possibility that managers’ accounting choices, in light of greater going concern uncertainty, adversely impact the informativeness of earnings. That is, if managers’ accounting choices reflect attempts to portray the firm as less distressed, then earnings informativeness is predictably impaired. The evidence reveals large and significant negative abnormal earnings for the three years leading up to the going concern opinion, along with negative abnormal accruals in the year of the going concern audit opinion but less pronounced negative accruals for the two years prior to the going concern opinion. These findings suggest that managers’ accounting choices reflect either the firm’s severe distress or the managers’ inability to mask the firm’s dismal financial condition. Nonetheless, this exploratory analysis of managers’ accounting choices finds that they are insufficient to explain the pronounced and persistent decline in earnings informativeness two or three years in advance of a going concern audit opinion.

Sample selection and hypothesis development are described in Section 2. Section 3 investigates the temporal behavior of the informativeness of earnings leading up to, and including, the year of the going concern audit opinion. The first part of Section 3 focuses on results for the full sample, while the second part isolates firms on the basis of whether they received a subsequent-year going concern audit opinion or not. Section 4 and Section 5 explore explanations of the decline in earnings informativeness leading up to the going concern opinion. Specifically, Section 4 considers the temporal behavior in known determinants of earnings informativeness, and Section 5 examines managers’ accounting choices as a potential factor in the decrease in earnings informativeness. Section 6 summarizes the study.

2. Sample Selection and Hypothesis Development

2.1. Sample Selection and Data Sources

Our sample consists of firms that received first-time going concern audit opinions. We focus on first-time going concern opinions because market expectations and valuation implications are likely to differ for first-time and continuing going concern recipients (Blay & Geiger, 2001; Dong et al., 2015). To be included in the sample, the companies had to meet the following additional criteria: (1) their quarterly earnings-per-share, and their respective announcement dates, were available from the Compustat database and (2) their daily adjusted stock returns were available from the CRSP database. The initial list of companies that received going concern audit opinions was identified by searching the auditors’ reports filed with the Securities and Exchange Commission. The initial sample, with available Compustat and CRSP data, consisted of 135 companies that received going concern opinions. Next, we identified those companies that received a first-time going concern audit opinion. This criterion reduced the sample to 85 companies. Later tests partition this first-time opinion sample on the basis of whether the firms received a subsequent-year going concern audit opinion. The six-year pre-opinion window reflects a tradeoff between capturing the longer-run evolution of earnings informativeness and preserving sufficient observations for estimation in each pre-opinion year. The resulting quarterly sample comprised company-quarter observations with first-time going concern opinions drawn from eight consecutive years and spanned a broad cross-section of industries—including manufacturing, energy, financial services, health care, retailing, technology, transportation, and utilities.5 Tests in Section 3 report the number of observations available in each relative year.

2.2. Hypothesis Development

This study rests at the intersection of research examining going concern uncertainty and the informativeness of earnings. The connection is deep-rooted. The going concern assumption is fundamental to the measurement of earnings and many of its underlying components. When substantial doubt arises regarding the validity of that assumption, the persistence and valuation relevance of reported earnings are predictably called into question. Prior research further indicates that the relation between returns and earnings depends on the persistence of earnings innovations, the length of the revision horizon, the amount of predisclosure information, and the quality of the information released (Holthausen & Verrecchia, 1988; Subramanyam & Wild, 1996). Financial distress also has direct pricing consequences. Campbell et al. (2008) show that distressed stocks display distinctive, and on average poor, returns. As distress deepens, investors have strong incentives to look beyond reported earnings for valuation-relevant information. These considerations imply that the usefulness of earnings is likely to diminish as firms move closer to the point at which substantial doubt about the going concern assumption is reflected in the auditor’s report.

This study focuses on the relation between going concern uncertainty and the informativeness of earnings, emphasizing the conditionality of the latter on the former. The study is distinctive in examining earnings informativeness up to six years prior to the issuance of a going concern audit opinion, progressing each year up to and including the opinion year. Through this design, the study examines whether market participants assess earnings informativeness differentially in periods leading up to a going concern audit opinion. This approach differs from prior research that primarily examines the information conveyed in the going concern opinion itself, the announcement effect of that opinion, or the behavior of earnings informativeness after the opinion has already been issued. As reviewed above, prior studies document the market’s response at and after the opinion date (e.g., Dong et al., 2015; Myers et al., 2018). Even at the opinion date, prices do not adjust cleanly; Kausar et al. (2009) document investor underreaction to going concern opinions that persists well after their release. By contrast, the present study examines whether the market revises the usefulness of earnings before the opinion year. Accordingly, although expectations surrounding the opinion matter in announcement studies, the central design here is not intended to isolate the announcement effect of an expected versus unexpected opinion. Instead, it examines the ex post intertemporal path of earnings informativeness for firms that ultimately cross the going concern reporting threshold.

The central hypothesis of this study is that earnings informativeness declines in the years leading up to, and including, the year in which a first-time going concern audit opinion is issued. This prediction is consistent with the argument that a higher probability of termination shortens the expected revision horizon of earnings and thereby reduces earnings informativeness. It is also consistent with the prospect that going concern uncertainty increases predisclosure information and reduces the incremental role of earnings in valuation (Holthausen & Verrecchia, 1988).

H1. 

The informativeness of earnings is hypothesized to decline in the years leading up to, and including, the year of a first-time going concern audit opinion.

A corollary hypothesis emerges from this main research premise and focuses on firms facing greater going concern uncertainty. The logic is that the informativeness of earnings and other accounting measurements, which depend critically on the validity of the going concern assumption, should decline more sharply as the faithfulness of that assumption is more severely compromised. The primary hypothesis, H1, treats the first-time going concern opinion as a general indicator of uncertainty surrounding the going concern assumption. An extension of this perspective considers whether the firm does or does not receive a second consecutive going concern opinion in the subsequent year. A firm receiving a going concern opinion in the subsequent year is, all else equal, less likely to continue as a going concern. Put differently, the underlying factors driving the decline in earnings informativeness are expected to be more severe when a subsequent going concern opinion is issued. This extension is also consistent with prior research showing that market expectations matter for first-time going concern recipients (Blay & Geiger, 2001).

To test this corollary hypothesis, we categorize the initial first-time going concern opinion sample into two groups: firms that receive a going concern opinion in the subsequent year and firms that do not. We then examine the temporal evolution of earnings informativeness conditional on receipt of a subsequent-year going concern opinion.

Consistent with hypothesis H1, firms that receive a first-time going concern opinion but do not receive another one in the subsequent year are expected to exhibit a decline in earnings informativeness in the years leading up to that opinion. More important, the corollary hypothesis predicts that firms receiving a second going concern opinion in the immediate subsequent year will show a more marked decline in earnings informativeness in the years leading up to the opinion compared to firms that receive a going concern opinion for only one year.

H2. 

The informativeness of earnings is hypothesized to exhibit a more pronounced earlier decline for firms receiving two consecutive going concern audit opinions than for firms receiving a single-year going concern audit opinion.

Prior research indicates that earnings informativeness is influenced by several economic factors. Among these factors are firm size, which is often linked to the amount of predisclosure information; leverage, which reflects default risk and potential wealth transfers between debt and equity holders; growth, commonly measured by the market-to-book ratio; systematic risk, which affects the firm’s valuation discount factor; earnings variability; and the temporal persistence of earnings (Holthausen & Verrecchia, 1988; Subramanyam & Wild, 1996). In anticipation of significant changes in these factors in the years leading up to the issuance of a going concern audit opinion, this study investigates whether variations in these known determinants contribute to the observed decline in earnings informativeness. These determinants are expected to account, at least in part, for a portion of the decrease in earnings informativeness. At the same time, prior evidence suggests that the relation between going concern status and earnings informativeness can persist after controlling for related determinants. Accordingly, the purpose of this part of the analysis is to assess whether the decline in earnings informativeness remains after these determinants are included, rather than to treat the control specification as a separate formal hypothesis.

As a final exploratory analysis, we examine managerial accounting choices through the lens of discretionary accounting accruals and related components of reported earnings. Prior research indicates that financial distress can alter earnings quality and reporting incentives, but that the direction of such effects is not uniform across settings (Charitou et al., 2011; Campa & Camacho-Miñano, 2015; Li et al., 2020). We therefore investigate whether abnormal earnings, abnormal accruals, and abnormal cash flows in the years leading up to the going concern opinion help explain the observed decline in earnings informativeness. This final analysis is directed at possible explanatory channels and should not be interpreted as a direct test of earnings informativeness itself, which in this study is measured through the return–earnings relation.

The proposition that managers exploit discretion in accounting principles to influence reported earnings in periods preceding the going concern opinion is grounded in the possibility that distress changes both reporting incentives and the relative costs of alternative reporting choices. Managers may seek to reduce default-related costs, avoid covenant violations, or mitigate compensation and labor-market losses. At the same time, auditors and other stakeholders may constrain such behavior.

The incentives for managers to influence earnings in this context are not straightforward, and the direction of abnormal accruals is therefore an empirical question. We do not formulate a formal directional hypothesis for this final analysis. Instead, we offer two limited conjectures. First, abnormal earnings are expected to be negative, given the financial distress of the firms involved. Second, we do not predict the direction of abnormal accruals or abnormal cash flows because their observed sign will reflect the combined actions of managers, auditors, and other stakeholders in the period preceding the going concern opinion. In sum, the analysis proceeds in three linked steps that tie the central constructs together. Section 3 documents the temporal path of earnings informativeness leading up to the first-time going concern opinion (H1) and conditions that path on receipt of a subsequent-year opinion (H2). Section 4 asks whether known economic determinants of earnings informativeness account for the documented path. Section 5 examines managers’ accounting choices as a final candidate channel.

3. Going Concern Audit Opinion and Earnings Informativeness

This study focuses on the informativeness of earnings in the years preceding and including the issuance of an auditor’s going concern opinion. The tests in this section use the eventual receipt of a first-time going concern opinion as an ex post partitioning device for studying the temporal behavior of earnings informativeness. Tests do not assume that market participants know with certainty, in earlier periods, that such an opinion will later be issued. We measure the informativeness of earnings by the market response to unexpected earnings, as reflected in the earnings response coefficient from a regression of abnormal returns on unexpected earnings. Abnormal returns are computed as the difference between actual and expected stock returns cumulated over the day before and the day of the earnings announcement. The expected stock returns are obtained from the market model, where the model’s parameters are estimated using the 250 trading days prior to the earnings announcements. Unexpected earnings is measured as the difference between reported quarterly earnings and an expected value implied by a quarterly seasonal random walk model, deflated by stock price at the end of the quarter preceding the earnings announcement. To minimize the effects of extreme values, unexpected earnings observations that exceed three standard deviations from the mean are deleted; this affects roughly two percent of the observations. Robustness tests described later in this section examine whether the inferences are sensitive to alternative return windows, alternative truncation criteria, and alternative earnings expectation models.

3.1. Informativeness of Earnings Preceding a Going Concern Audit Opinion

Preliminary evidence shown in

Table 1. Correlations between unexpected earnings and abnormal returns for earnings preceding the going concern audit opinion.

Year Relative to Going Concern Opinion a	Number of Company- Quarter Observations	Correlation Between Earnings and Returns b
All Years	896	0.08
Year {−4}	147	0.31
Year {−3}	180	0.30
Year {−2}	193	0.29
Year {−1}	212	0.10
Year {0}	164	0.08

^a Sample comprises firms that received a going concern audit opinion. Year {0} is the period in which the going concern audit opinion is reported; years {−1} through {−4} represent the firm years immediately preceding the going concern opinion year. ^b This column reports the Spearman correlations between unexpected earnings and abnormal stock returns. The boldface correlations for years {−4} through {−2}, and for all years combined, are significant at the 0.01 level or better (two-tailed tests), whereas the correlations for years {−1} and {0} are not significant at conventional levels. Unexpected earnings is measured as the difference between reported quarterly earnings and an expected value implied by a quarterly seasonal random walk model and that quantity deflated by stock price taken from the end of the quarter preceding the earnings report. Abnormal stock returns are computed as the difference between actual and expected returns, cumulated over the day before and the day of the earnings report, where expected returns are obtained from the market model estimated using the 250 trading days preceding the returns cumulation period.

reveals a decline in the correlation between earnings and returns in the years preceding the issuance of a going concern audit opinion. Specifically, correlations for the full sample (see the first row of Table 1) show that earnings announcements are positively associated with returns, which is consistent with prior research on the average informativeness of earnings (0.08 is significantly greater than zero at the 0.01 level). The All Years row is the pooled correlation across all 896 company-quarter observations from years {−4} through {0}; it is not an average of the yearly correlations, and its magnitude reflects the mix of the larger correlations of the earlier years with the near-zero correlations of years {−1} and {0}. Unlike the year {0} correlation, the pooled correlation is reliably positive given the substantially larger number of observations.

Pertinent to this study is the pattern in the magnitude of the correlation between earnings and returns as firms near the going concern audit opinion. The correlations decline from a high of 0.31 for the earlier pre-opinion years to values that are insignificantly different from zero in the year immediately preceding, and the year of, the opinion. Table 1 also makes transparent the number of company-quarter observations in each relative year, which is relevant to assessing whether the six-year design induces any material loss of observations as the opinion year approaches. The pattern is consistent with a progressive weakening in the return–earnings relation as firms approach the going concern opinion year. The decline is such that the correlation between earnings and returns is insignificantly different from zero for the year before, and the year of, the going concern audit opinion. The economic magnitude of this decline is notable. The correlation falls by roughly two-thirds from 0.31 for year {−4} to 0.10 for year {−1}. By the time we reach the year before the opinion, earnings announcements carry little detectable news for returns.

We extend this analysis to consider the temporal behavior of the earnings response coefficient. We employ the following pooled time-series and cross-sectional regression model:

$$CA{R}_{i,q}={\alpha }_0+{\alpha }_1\cdot E_{i,q}/P_{i,q-1}+{\displaystyle \underset{\lambda =1}{\overset{5}{\Sigma }}}Y{R}_{\lambda -5}\cdot {\gamma }_{\lambda -5}\cdot E_{i,q}/P_{i,q-1}+{\epsilon }_{i,q}$$

(1)

CAR_i,q is the cumulative abnormal return for the day before and day of company i’s earnings announcement for quarter q. E is unexpected earnings-per-share, P is the price-per-common-share at the end of the quarter preceding the earnings announcement, and YR is an indicator variable denoting the year relative to the report (e.g., YR₋₄ equals one if the earnings announcement is from the fourth year preceding the going concern opinion). The γ₋₄ through γ₀ estimates measure the change in the earnings response coefficient for each of the years immediately preceding, and including, the going concern audit opinion. This specification is intended to describe the temporal path of the return–earnings relation as the opinion year approaches; it is not intended to attribute those yearly differences to investor certainty regarding the future issuance of the opinion.

The empirical evidence, reported in

Table 2. Regression of abnormal stock returns on unexpected earnings and the interaction of unexpected earnings with an indicator variable denoting the year preceding the going concern opinion. ^a

$\mathit{C}\mathit{A}{\mathit{R}}_{\mathit{i},\mathit{q}}={\mathit{\alpha }}_0+{\mathit{\alpha }}_1\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\displaystyle \underset{\mathit{\lambda }=1}{\overset{5}{\mathit{\Sigma }}}}\mathit{Y}{\mathit{R}}_{\mathit{\lambda }-5}\cdot {\mathit{\gamma }}_{\mathit{\lambda }-5}\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\mathit{\epsilon }}_{\mathit{i},\mathit{q}}$
Parameter Estimates							Sample		F-Value
α0	α1	γ−4	γ−3	γ−2	γ−1	γ0	Size	R2%	(Sig. Level)
−0.009 ***	0.190 ***	−0.013	−0.046	−0.191 ***	−0.184 **	−0.181 **	1142	2.42%	4.70
(−4.10)	(2.32)	(−0.11)	(−0.52)	(−2.34)	(−2.25)	(−2.18)			(0.001)

^a Cumulative abnormal returns (CAR) are computed as the difference between actual and expected returns, cumulated over the day before and the day of the earnings report, where expected returns are obtained from the market model estimated using the 250 trading days preceding the returns cumulation period. Unexpected earnings (E) is measured as the difference between reported quarterly earnings and an expected value implied by a quarterly seasonal random walk model and the quantity deflated by stock price taken from the end of the quarter preceding the earnings report. The indicator variable YR₋₄ [YR₋₃, YR₋₂, YR₋₁, YR₀] equals one if the company-quarter earnings report is from year {−4} [{−3}, {−2}, {−1}, {0}, respectively] immediately preceding the going concern audit opinion year, where year {0} is the period in which the going concern audit opinion is received. Parameter estimates and t-statistics (in parentheses) are shown for the regression in which the estimation period covers the current and six preceding years’ company-quarter observations for 85 firms which received first-time going concern audit opinions. A * (**/***) designates statistical significance at the 0.10 (0.05/0.01) level, one-tailed tests.

, shows a decline in the informativeness of earnings as the going concern opinion nears. Regression results show that the earnings response coefficient depends on proximity to the going concern opinion. Specifically, the earnings response coefficient is positive and significant (α₁ = 0.190) for the base period, defined as year {−6} and year {−5} prior to the going concern audit opinion year. The implied earnings response coefficient—computed by adding each year’s change estimate reported in Table 2 to the base-period coefficient—then declines to 0.177, 0.144, −0.001, 0.006, and 0.009 for years {−4} through {0}, respectively.6 Estimates of the change in the earnings response coefficient, or γ₋₂, γ₋₁, and γ₀, are each significantly less than zero at the 0.05 level for years {−2}, {−1}, and {0}. These results indicate that the earnings response coefficient significantly declines two years prior to the going concern opinion and, moreover, that this decline persists through the year of the going concern audit opinion. Stated differently, the evidence is consistent with a weakening in earnings informativeness as going concern uncertainty intensifies, not merely with a reaction to the formal report in the opinion year. The economic magnitude again bears emphasis. The implied earnings response coefficient in the opinion year is roughly one-twentieth of its base-period level. An earnings surprise that once moved prices in a meaningful way barely moves them once going concern risk takes hold.

Equation (1) and the results in Table 2 implicitly assume that the residual variance is independent of proximity to the going concern opinion. We therefore re-estimate the relation between abnormal returns and unexpected earnings separately for each of the years leading up to and including the going concern opinion. These separate regressions do not impose a constant residual variance assumption across years. The earnings response coefficients from these regressions are portrayed graphically in Figure 1. We again observe a substantial decline in the earnings response coefficient as the going concern opinion nears. This additional presentation is useful because it demonstrates that the decline in the earnings response coefficient from six years prior to the going concern opinion (0.216) to the year of the going concern opinion (0.009) is not an artifact of the pooled specification. Moreover, the decline in the earnings response coefficient occurs at least two years prior to the going concern opinion. This evidence is consistent with that in Table 1 and Table 2 and with hypothesis H1. Figure 1 also shows that the decline is gradual rather than abrupt. The earnings response coefficient erodes steadily from year {−4} onward and then flattens near zero. That path fits a market that updates as distress signals accumulate. It does not fit a market that waits for the auditor’s report.

Tests of Robustness. We conduct several tests to assess the robustness of the findings. First, we examine the effect of any extreme observations. Although prior tests excluded the most extreme two percent of observations, we now relax this criterion to include all observations. The regression results for Equation (1), using all observations, are reported in panel A of

Table 3. Sensitivity tests of the regression of abnormal stock returns on unexpected earnings and the interaction of unexpected earnings with an indicator variable denoting the year preceding the going concern opinion ^a.

$\mathit{C}\mathit{A}{\mathit{R}}_{\mathit{i},\mathit{q}}={\mathit{\alpha }}_0+{\mathit{\alpha }}_1\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\displaystyle \underset{\mathit{\lambda }=1}{\overset{5}{\mathit{\Sigma }}}}\mathit{Y}{\mathit{R}}_{\mathit{\lambda }-5}\cdot {\mathit{\gamma }}_{\mathit{\lambda }-5}\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\mathit{\epsilon }}_{\mathit{i},\mathit{q}}$
Parameter Estimates							Sample		F-Value
α0	α1	γ−4	γ−3	γ−2	γ−1	γ0	Size	R2%	(Sig. Level)
Panel A: Unexpected earnings not truncated for extreme observations
−0.010 ***	0.055 *	−0.038	−0.090 **	−0.056 *	−0.055 *	−0.066 *	1161	1.73%	3.39
(−4.46)	(1.33)	(−0.73)	(−2.12)	(−1.35)	(−1.31)	(−1.56)			(0.003)
Panel B: Abnormal stock returns cumulated over the quarterly reporting period
−0.032 ***	0.625 **	−0.444	−0.489 *	−0.605 **	−0.544 **	−0.521 **	1142	1.64%	3.15
(−3.85)	(2.05)	(−1.00)	(−1.48)	(−1.98)	(−1.78)	(−1.69)			(0.005)

^a Unless otherwise labeled in the panel’s heading, the variable definitions for the regression sensitivity tests are as follows: cumulative abnormal returns (CAR) are computed as the difference between actual and expected returns, cumulated over the day before and the day of the earnings report, where expected returns are obtained from the market model estimated using the 250 trading days preceding the returns cumulation period; unexpected earnings (E) is measured as the difference between reported quarterly earnings and an expected value implied by a quarterly seasonal random walk model and that quantity deflated by stock price taken from the end of the quarter preceding the earnings announcement; and the indicator variable YR₋₄ [YR₋₃, YR₋₂, YR₋₁, YR₀] equals one if the company-quarter earnings report is from year {−4} [{−3}, {−2}, {−1}, {0}, respectively] immediately preceding the going concern opinion year, where year {0} is the period in which the going concern opinion is received. Parameter estimates and t-statistics (in parentheses) are shown for the regression in which the estimation period covers the current and six preceding years’ company-quarter observations for 85 firms which received first-time going concern audit opinions. A * (**/***) designates statistical significance at the 0.10 (0.05/0.01) level, one-tailed tests.

. Although the statistical significance of the decline in the earnings response coefficient is not as striking as in Table 2, it is nevertheless significant at conventional levels (the 0.10 level). Moreover, none of the implied earnings response coefficients for years {−2}, {−1}, or {0} (namely, −0.001, 0.000, and −0.011, respectively, computed by adding the panel A change estimates, γ, to the base-period coefficient, α₁) differ significantly from zero, and each is markedly less than the base-period earnings response coefficient of 0.055. Further, experimentation with several different truncation criteria for the data did not alter any inferences. Hence, this evidence is consistent with the decreased informativeness of earnings as the going concern opinion nears.

Second, we consider an alternative measure of the dependent variable; specifically, we compute long-window cumulative abnormal returns that span the entire quarterly reporting period. This contrasts with the short-window cumulation period of previous tests.7 These long-window regression results, using the framework of Equation (1), are reported in panel B of Table 3. Consistent with prior results, the decline in the earnings response coefficient is evident as the going concern opinion nears. The implied earnings response coefficient for years {−2}, {−1}, and {0}—namely, 0.020, 0.081, and 0.104, respectively, computed by adding the panel B change estimates, γ, to the base-period coefficient, α₁—does not differ significantly from zero at any conventional level, and each is significantly lower than the base-period earnings response coefficient of 0.625.

Third, to examine the sensitivity of the findings to the earnings expectation model, we repeat all previous tests with several different sources of earnings expectations. As representative of these sensitivity tests and similar to the results in Table 2, the regression using a simple random walk model (results not shown for brevity) resulted in consistently negative γ₋₄ through γ₀ terms, reflecting the decline in the earnings response coefficient as the going concern opinion nears. For example, the implied earnings response coefficient equals 0.029, 0.046, and 0.055 for years {−2}, {−1}, and {0}, respectively, and each of these coefficients is significantly less than the base-period earnings response coefficient of 0.191. These results are again consistent with those in Table 2 and hypothesis H1, indicating that the informativeness of earnings declines with increasing proximity to the going concern audit opinion.

3.2. Informativeness of Earnings Conditional on a Subsequent-Year Going Concern Opinion

Tests in the prior section implicitly use the going concern opinion as a general indicator of doubtful going concern status. One extension is the consideration of the receipt or non-receipt of a subsequent-year going concern opinion. The idea is that a firm that receives a subsequent-year going concern opinion is, ceteris paribus, less likely to remain a going concern. Put differently, the underlying drivers of the decline in earnings informativeness are predictably more severe when a subsequent going concern opinion is received. To examine this conjecture, we stratify the first-time going concern opinion sample into firms that do (n = 38) and firms that do not (n = 47) receive a going concern opinion in the subsequent year because prior research shows that market expectations matter for first-time recipients. This partition should be read as capturing variation in the severity and persistence of going concern uncertainty. It is not a direct test of whether the original opinion was expected or unexpected.

We estimate the following regression to investigate differential behavior in the earnings response coefficient between these two groups. We define an indicator variable (SY, for Subsequent-Year) which equals one when a company is the recipient of a subsequent-year going concern opinion and zero otherwise. Next, we formulate a regression equation to test for a temporal change in the earnings response coefficient conditional on a subsequent-year going concern opinion.

$$CA{R}_{i,q}={\alpha }_0+{\alpha }_1\cdot E_{i,q}/P_{i,q-1}+{\displaystyle \underset{\lambda =1}{\overset{5}{\Sigma }}}{\gamma }_{\lambda -5}\cdot Y{R}_{\lambda -5}\cdot E_{i,q}/P_{i,q-1}+{\displaystyle \underset{\lambda =1}{\overset{5}{\Sigma }}}{\psi }_{\lambda -5}\cdot S{Y}_i\cdot Y{R}_{\lambda -5}\cdot E_{i,q}/P_{i,q-1}+{\epsilon }_{i,q}$$

(2)

The remaining regression variables are as defined in Equation (1).

Following H1, we predict that the earnings response coefficient will continue to exhibit a decline for companies that receive a first-time going concern opinion with no subsequent-year opinion (i.e., γ₋₄ through γ₀ are predictably less than zero, reflecting the change in the earnings coefficient from the base years {−5} and {−6} to the test period years of {−4}, {−3}, {−2}, {−1}, and {0}). Second, and more important, the corollary hypothesis H2 predicts that companies who receive an immediate subsequent-year going concern opinion will exhibit a more pronounced decline in the earnings response coefficient for years leading up to the going concern opinion in comparison to companies that receive a single-year going concern opinion. Alternatively stated, the parameters ψ₋₄ through ψ₀, for periods {−4}, {−3}, {−2}, {−1}, and {0}, respectively, are predictably less than zero, reflecting a smaller earnings response coefficient for companies that receive a subsequent-year going concern opinion in comparison to companies that do not receiving a subsequent-year going concern opinion.

The empirical evidence in

Table 4. Regression of abnormal stock returns on (i) unexpected earnings, (ii) unexpected earnings interacted with an indicator variable denoting the year preceding the going concern opinion, and (iii) unexpected earnings interacted with an indicator variable denoting the existence of a subsequent-year going concern opinion ^a.

$\mathit{C}\mathit{A}{\mathit{R}}_{\mathit{i},\mathit{q}}={\mathit{\alpha }}_0+{\mathit{\alpha }}_1\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\displaystyle \underset{\mathit{\lambda }=1}{\overset{5}{\mathit{\Sigma }}}}{\mathit{\gamma }}_{\mathit{\lambda }-5}\cdot \mathit{Y}{\mathit{R}}_{\mathit{\lambda }-5}\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\displaystyle \underset{\mathit{\lambda }=1}{\overset{5}{\mathit{\Sigma }}}}{\mathit{\psi }}_{\mathit{\lambda }-5}\cdot \mathit{S}{\mathit{Y}}_{\mathit{i}}\cdot \mathit{Y}{\mathit{R}}_{\mathit{\lambda }-5}\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\mathit{\epsilon }}_{\mathit{i},\mathit{q}}$
Parameter Estimates												Sample		F-Value
α0	α1	γ−4	γ−3	γ−2	γ−1	γ0	ψ−4	ψ−3	ψ−2	ψ−1	ψ0	Size	R2%	(Sig. Level)
−0.009 ***	0.190 ***	−0.126	0.086	−0.189 ***	−0.179 **	−0.145 **	0.202	−0.200 ***	−0.004	−0.075 ***	−0.053 **	1142	4.05%	4.33
(−3.98)	(2.34)	(−0.82)	(0.86)	(−2.32)	(−2.20)	(−1.71)	(1.15)	(−2.81)	(−0.51)	(−2.60)	(−1.69)			(0.001)

^a Cumulative abnormal returns (CAR) are computed as the difference between actual and expected returns, cumulated over the day before and the day of the earnings report, where expected returns are obtained from the market model estimated using the 250 trading days preceding the returns cumulation period. Unexpected earnings (E) is measured as the difference between reported quarterly earnings and an expected value implied by a quarterly seasonal random walk model and that quantity deflated by stock price taken from the end of the quarter preceding the earnings report. The indicator variable YR₋₄ [YR₋₃, YR₋₂, YR₋₁, YR₀] equals one if the company-quarter earnings report is from year {−4} [{−3}, {−2}, {−1}, {0}, respectively] immediately preceding the going concern audit opinion year, where year {0} is the period in which the going concern opinion is received, and the indicator variable SY_i equals one if the company receives a going concern opinion in the year immediately subsequent to the initial opinion year. Parameter estimates and t-statistics (in parentheses) are shown for the regression in which the estimation period covers the current and six preceding years’ company-quarter observations for 85 firms which received first-time going concern audit opinions. A * (**/***) designates statistical significance at the 0.10 (0.05/0.01) level, one-tailed tests.

is consistent with this hypothesis. First, and consistent with the main hypothesis H1, the significant (at the 0.05 level) and negative estimates of γ₋₂ (−0.189), γ₋₁ (−0.179), and γ₀ (−0.145) reflect a marked decline in the earnings response coefficient for companies that receive a one-time going concern opinion. Their earnings response coefficient declines from 0.190 for base years {−6} and {−5} to the implied values 0.001, 0.011, and 0.045 for years {−2}, {−1}, and {0}, respectively.8 Hence, the temporal behavior in the informativeness of earnings exhibits a marked decline for those companies which receive a one-time going concern opinion.

Second, the significant (at the 0.05 level) and negative estimates for ψ₋₃ (−0.200), ψ₋₁ (−0.075), and ψ₀ (−0.053) reveal that companies that receive a subsequent-year going concern opinion exhibit a more pronounced decline in their earnings response coefficient compared to companies that do not receive a subsequent-year going concern opinion. Their earnings response coefficient declines from 0.190 for base years {−6} and {−5} to the implied values 0.076, −0.064, and −0.008 for years {−3}, {−1}, and {0}, respectively, preceding the going concern opinion.9 This evidence is consistent with hypothesis H2, indicating less informative earnings from companies with consecutive going concern opinions, suggesting that the underlying factors that result in the low informativeness of earnings are tied to the greater uncertainty of the validity of the going concern assumption. Viewed through a risk lens, this pattern is a severity gradient. The market does not merely register the presence of going concern risk. It prices the intensity of that risk, with the earliest and deepest declines among firms whose uncertainty persists into a second opinion year.

The evidence in Table 4, based on regression Equation (2), implicitly assumes the residual variance to be stationary across years. Therefore, as a further check on the temporal behavior in the earnings response coefficient, we re-examine the data after categorizing firms into the four separate years preceding the going concern opinion—years {−3}, {−2}, {−1}, and {0}—and two distinct audit report samples (based on the receipt or non-receipt of a subsequent-year going concern opinion). We then regress abnormal stock returns on unexpected earnings for each of the eight portfolios (4 × 2). Like earlier, we expect a decline in the earnings response coefficient as the going concern opinion nears for both sub-samples, but more importantly we predict that this decline will be more pronounced for companies receiving a subsequent-year going concern opinion.

The evidence, portrayed graphically in Figure 2, is again consistent with the corollary hypothesis H2. Figure 2 reveals (i) an inverse relation between the magnitude of the earnings response coefficient and nearness to the going concern opinion for both samples and (ii) an earlier decline in the earnings response coefficient for companies that are less likely to be going concerns, as represented by a subsequent-year going concern opinion. Specifically, the earnings response coefficient for years {−3}, {−2}, {−1}, and {0} equals 0.330, 0.153, 0.038, and 0.015, respectively, for companies that did not receive a subsequent-year going concern opinion. This compares to an earnings response coefficient of 0.127, −0.003, 0.004, and −0.002 for years {−3}, {−2}, {−1}, and {0} for companies that did receive a subsequent-year going concern opinion. The contrast in Figure 2 is stark. For firms with a subsequent-year opinion, the earnings response is effectively zero by year {−2}. For the remaining firms, the decline arrives later and is less severe.

4. Firm-Specific Economic Factors Preceding the Going Concern Opinion

Results in the previous section show that the informativeness of earnings declines in the years leading up to the going concern opinion. There are several factors which can separately, or in combination, potentially explain a portion or all of this decline. The issue in this section is whether those factors merely move with the decline in earnings informativeness or whether the decline remains after they are taken into account.

4.1. Evidence on Determinants of Earnings Informativeness

The earnings response coefficient has been shown to be influenced by a number of economic factors.10 Among these factors are firm size, with a theory suggesting that larger firms exhibit a less pronounced reaction to earnings due to the availability of greater pre-disclosure information11; leverage, which reflects the riskiness of debt and is anticipated to be negatively correlated with informativeness; growth, typically measured by the market-to-book value ratio, which is expected to be positively related to returns; systematic risk, which affects the firm’s valuation discount factor and is inversely related to informativeness; earnings predictability, usually assessed by the time-series variability in earnings and expected to negatively correlate with informativeness; and finally, the temporal persistence of earnings, which is predicted to enhance informativeness. The purpose of this section is not simply to restate the relevance of these determinants, but to assess whether they change in ways that are consistent with the temporal decline in the earnings response coefficient documented in Section 3.

Table 5. Descriptive statistics on known determinants of the earnings response coefficient preceding the going concern audit opinion ^a.

	Year Relative to the Going Concern Audit Opinion
Determinant	−5	−4	−3	−2	−1	0
Firm size
Levels	294.49	287.01	327.01	329.99	291.50	264.99
Changes		18.37	81.06	38.66	−31.46	−34.71
t-statistic		(1.53) *	(4.76) ***	(2.27) **	(−1.87) **	(−2.34) **
Leverage
Levels	0.4148	0.4277	0.4204	0.4647	0.4906	0.5484
Changes		0.0130	−0.0052	0.0483	0.0472	0.0572
t-statistic		(1.25)	(−0.42)	(5.27) ***	(3.90) ***	(4.09) ***
Growth
Levels	1.1900	1.3211	2.2075	2.0969	1.5596	1.1400
Changes		−0.2030	0.7193	−0.3125	−0.5724	−0.6494
t-statistic		(−0.58)	(1.43)*	(−0.78)	(−1.60) *	(−1.73) **
Systematic risk
Levels	0.7931	0.8707	0.8112	0.8010	0.7983	0.6891
Changes		0.065	−0.0366	0.0116	−0.0327	−0.1315
t-statistic		(1.43) *	(−0.87)	(0.28)	(−0.95)	(−3.69) ***
Earnings variability
Levels	0.0637	0.0717	0.0713	0.1659	0.2657	0.3695
Changes		0.0111	0.0091	0.1089	0.1169	0.1129
t-statistic		(1.55) *	(1.10)	(2.82) ***	(3.89) ***	(5.11) ***
Earnings persistence
Levels	0.1590	0.1272	0.1541	0.1689	0.0456	0.0414
Changes		0.0146	0.0133	−0.0020	−0.0940	0.0117
t-statistic		(0.32)	(0.32)	(−0.05)	(−2.55) ***	(0.38)

^a Measurement of the determinants is defined as follows: persistence (PERST) is measured by the autocorrelation in the seasonal difference in earnings-per-share during the eight periods preceding the current quarter; earnings variability (EVAR) is measured as the variance in the seasonal difference in earnings-per-share during the eight periods preceding the current quarter; growth (GRWTH) is measured as the ratio of market value to book value of equity at the end of the current period; systematic risk (RISK) is the estimate of the firm’s beta from the market model; leverage (LEV) is computed as the ratio of liabilities to assets; and firm size (SIZE) is measured by the market value of equity at the end of the period. Levels are means of all available observations in each year; Changes are means of year-over-year changes for observations with data in both years (this means that reported changes need not equal differences between adjacent reported levels). The t-statistic is from a test of the null hypothesis that the year’s change in the determinant is equal to zero. The sample comprises firms that received a going concern audit opinion. Year {0} is the period in which the going concern audit opinion is received, and years {−1} through {−5} represent the firm years immediately preceding the going concern opinion year. A * (**/***) designates statistical significance at the 0.10 (0.05/0.01) level, one-tailed tests.

reports descriptive statistics for these six potential determinants along with changes in those determinants. Measurement of the determinants follows the prior literature: firm size (SIZE) is the market value of equity at the end of the period; leverage (LEV) is the ratio of liabilities to assets; growth (GRWTH) is the ratio of market value to book value of equity at the end of the current period; systematic risk (RISK) is the firm’s beta estimated from the market model; earnings variability (EVAR) is the variance of the seasonal difference in earnings-per-share over the eight quarters preceding the current quarter; and earnings persistence (PERST) is the autocorrelation of that seasonal difference over the same period. Each determinant, with the exception of persistence, experiences a significant change in the year of the going concern audit opinion. Further, many of the determinants experience significant changes in several of the years preceding the receipt of the going concern opinion. Specifically, each determinant exhibits at least one significant shift during the four-year period, {−5} to {−1}, preceding the year of the going concern opinion. These descriptive patterns do not, by themselves, establish that the determinants explain the decline in earnings informativeness; rather, they indicate whether the determinants move in directions consistent with the time-series pattern in the earnings response coefficient.

The determinant of firm size reveals a significant decline in years {−1} and {0}. Leverage exhibits a systematic and near-monotonic increase during the years preceding the going concern opinion, with changes in the final three years being significant at the 0.01 level. Growth shows a significant decline in years {−1} and {0}. Systematic risk exhibits a substantial decline in the year of the going concern opinion. Earnings variability shows a significant increase (at the 0.01 level) in the three years leading to the going concern opinion. That variability increases five-fold, from 0.0713 to 0.3695, over the period {−3} to {0}. Turning to earnings persistence, which is 0.1590 in year {−5}, declines to 0.0414 by year {0}. Although earnings persistence shows a significant decline of −0.0940 in year {−1}, no other years’ change in persistence is significant. Taken together, these shifts in determinants describe a risk profile deteriorating on several fronts. Leverage rises by roughly a third over the window. Earnings variability increases more than five-fold. Growth options fade. Each movement points in the same direction—toward less informative earnings.

This descriptive analysis indicates significant movement in several known determinants of the earnings response coefficient in both the year of and in the years preceding the issuance of the going concern opinion. This suggests that at least part of the decline in the earnings response coefficient is potentially attributable to variations in these fundamental determinants. Because this evidence is descriptive, however, the key question is whether the decline in earnings informativeness remains after these determinants are introduced jointly in the return–earnings specification.

4.2. Joint Tests of Known Determinants of Earnings Informativeness and the Going Concern Opinion

We jointly examine changes in known determinants of the temporal behavior of the earnings response coefficient together with the earnings information preceding and including the going concern audit opinion by estimating the following model. The objective is to determine whether the documented temporal decline in the earnings response coefficient is subsumed by these known determinants or whether an incremental decline remains after they are included.

$$\begin{array}{l}CA{R}_{i,q}={\alpha }_0+{\alpha }_1\cdot E_{i,q}/P_{i,q-1}+{\displaystyle \underset{\lambda =1}{\overset{5}{{\displaystyle \Sigma }}}}{\gamma }_{\lambda -5}\cdot Y{R}_{\lambda -5}\cdot E_{i,q}/P_{i,q-1}+{\delta }_1\cdot E_{i,q}\cdot SIZ{E}_{i,q}/P_{i,q-1}+{\delta }_2\cdot E_{i,q}\cdot LE{V}_{i,q}/P_{i,q-1}\hfill \\ +{\delta }_3\cdot E_{i,q}\cdot GRWT{H}_{i,q}/P_{i,q-1}+{\delta }_4\cdot E_{i,q}\cdot RIS{K}_{i,q}/P_{i,q-1}+{\delta }_5\cdot E_{i,q}\cdot EVA{R}_{i,q}/P_{i,q-1}+{\delta }_6\cdot E_{i,q}\cdot PERS{T}_{i,q}/P_{i,q-1}+{\epsilon }_{i,q}\hfill \end{array}$$

(3)

Measurement of the determinants is as defined in Section 4.1 (also see notes to Table 5).

If the decline in the earnings response coefficient is at least partly explained by variations in the determinants, then the coefficients γ₋₄ to γ₀ are predictably lower in magnitude (or insignificant) relative to the estimates from Equation (2). Because of the non-availability of certain data, the sample declines from 1142 to 702 observations, which constitute all available data for this regression test. It is therefore not possible to directly compare the results of this estimation with the results in Table 2. Consequently, to assess the effect of including these determinants, we separately estimate Equation (1) on the reduced sample so that the model with determinants can be compared against the appropriate reduced-sample benchmark.

Table 6. Regression of abnormal stock returns on (i) unexpected earnings, (ii) unexpected earnings interacted with an indicator variable denoting the year preceding the going concern opinion, and (iii) unexpected earnings interacted with known determinants of the earnings response coefficient ^a.

$\begin{array}{c}\mathit{C}\mathit{A}{\mathit{R}}_{\mathit{i},\mathit{q}}={\mathit{\alpha }}_0+{\mathit{\alpha }}_1\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\displaystyle \underset{\mathit{\lambda }=1}{\overset{5}{\mathit{\Sigma }}}}{\mathit{\gamma }}_{\mathit{\lambda }-5}\cdot \mathit{Y}{\mathit{R}}_{\mathit{\lambda }-5}\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\mathit{\delta }}_1\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}\cdot \mathit{S}\mathit{I}\mathit{Z}{\mathit{E}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\mathit{\delta }}_2\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}\cdot \mathit{L}\mathit{E}{\mathit{V}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\mathit{\delta }}_3\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}\cdot \mathit{G}\mathit{R}\mathit{W}\mathit{T}{\mathit{H}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}\\ +{\mathit{\delta }}_4\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}\cdot \mathit{R}\mathit{I}\mathit{S}{\mathit{K}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\mathit{\delta }}_5\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}\cdot \mathit{E}\mathit{V}\mathit{A}{\mathit{R}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\mathit{\delta }}_6\cdot {\mathit{E}}_{\mathit{i},\mathit{q}}\cdot \mathit{P}\mathit{E}\mathit{R}\mathit{S}{\mathit{T}}_{\mathit{i},\mathit{q}}/{\mathit{P}}_{\mathit{i},\mathit{q}-1}+{\mathit{\epsilon }}_{\mathit{i},\mathit{q}}.\end{array}$
Parameter Estimates													Sample
α0	α1	γ−4	γ−3	γ−2	γ−1	γ0	δ1	δ2	δ3	δ4	δ5	δ6	Size	R2%
Panel A: Regression without known determinants
−0.0125	0.2310	−0.1504	0.0823	−0.2184	−0.2130	−0.2403							702	5.28
(−4.60) ***	(2.21) **	(−1.05)	(0.69)	(−2.06) **	(−2.03) **	(−2.27) **
Panel B: Regression with known determinants
−0.0121	0.2089	−0.1700	0.1063	−0.1720	−0.1830	−0.1838	0.0003	−0.0670	0.0123	0.0066	−0.0084	0.0171	702	7.96
(−4.50) ***	(1.95) **	(−1.18)	(0.90)	(−1.62) *	(−1.70) **	(−1.70) **	(2.29) **	(−1.89) **	(3.64) ***	(0.19)	(−0.72)	(0.64)

^a Cumulative abnormal returns (CAR) are computed as the difference between actual and expected returns, cumulated over the day before, and day of, the earnings report, where expected returns are obtained from the market model estimated using the 250 trading days preceding the returns cumulation period. Unexpected earnings (E) is measured as the difference between reported quarterly earnings and an expected value implied by a quarterly seasonal random walk model and that quantity deflated by stock price taken from the end of the quarter preceding the earnings report. The indicator variable YR₋₄ [YR₋₃, YR₋₂, YR₋₁, YR₀] equals one if the company-quarter earnings announcement is from year {−4} [{−3}, {−2}, {−1}, {0}, respectively] immediately preceding the going concern audit opinion year, where year {0} is the period in which the going concern opinion is received. Parameter estimates and t-statistics (in parentheses) are shown for the regression in which the estimation period covers the current and six preceding years’ company-quarter observations for 85 firms which received first-time going concern audit opinions. The determinants are defined as follows: persistence (PERST) is measured by the autocorrelation in the seasonal difference in earnings-per-share during the eight periods preceding the current quarter; earnings variability (EVAR) is measured as the variance in the seasonal difference in earnings-per-share during the eight periods preceding the current quarter; growth (GRWTH) is measured as the ratio of market value to book value of equity at the end of the current period; systematic risk (RISK) is the estimate of the firm’s beta from the market model; leverage (LEV) is computed as the ratio of liabilities to assets; and firm size (SIZE) is measured by the market value of equity at the end of the period. A * (**/***) designates statistical significance at the 0.10 (0.05/0.01) level, one-tailed tests.

presents the results of the estimation of Equations (1) and (3). The regression results using Equation (1) on this subsample are first compared with the larger sample results from Table 2. The parameter estimates in panel A of Table 6 are similar to those in Table 2. Specifically, as with the full sample, this reduced sample evidences a significant decline in the earnings response coefficient for years {−2}, {−1}, and {0}; that is, γ₋₂, γ₋₁, and γ₀ are significantly less than zero at the 0.01 level (e.g., γ₋₂ = −0.2184, γ₋₁ = −0.2130, and γ₀ = −0.2403).

We next estimate Equation (3), which includes the additional known determinants. The results of this regression are displayed in panel B of Table 6. While size, leverage, and growth are significant in explaining the earnings response coefficient, the significant temporal decline in the earnings response coefficient remains. That is, the inclusion of the known determinants tempers the magnitude of the temporal decline in the earnings response coefficient; the γ coefficients for years {−2}, {−1}, and {0} are −0.1720, −0.1830, and −0.1838 in the model with the determinants, and this compares to −0.2184, −0.2130, and −0.2403 in the model without the determinants. However, each of these γ coefficients remain statistically significant (at the 0.10 level or better) for each of the years {−2}, {−1}, and {0}. Consequently, and consistent with the interpretation that going concern uncertainty conveys incremental information beyond these known determinants, the decline in the informativeness of earnings persists in the years leading up to the issuance of a going concern audit opinion after controlling for these known determinants.12 The evidence therefore indicates that the determinants explain a portion, but not all, of the temporal decline documented in Section 3. This residual decline matters for interpretation. Going concern uncertainty appears to carry pricing-relevant information beyond standard risk fundamentals. That reading parallels evidence that conventional characteristics capture distress risk only imperfectly (Campbell et al., 2008).

5. Managers’ Accounting Choices Preceding the Going Concern Opinion

This section extends our analysis of factors leading to the decline in the earnings response coefficient preceding a going concern opinion, with a specific focus on managers’ accounting choices. The selection and application of accounting principles in the determination of accruals is one area where managers (and auditors) can exercise discretion to affect reported earnings. We explore the evidence in light of the potential for accounting choices to affect the informativeness of earnings. The purpose of this section is not to test earnings informativeness directly, but to examine whether managers’ accounting choices provide a plausible explanatory channel for the decline documented earlier. However, the managerial incentives to manage earnings in this setting are not immediately evident.

Earnings management is typically predicated on blocked communication between interested parties, often attributed to proprietary costs of reporting, opaque disclosure rules, or other institutional/contractual constraints. Yet for companies nearing a going concern audit opinion, the scope for aggressive reporting may be more limited. Owing to their access to inside information, auditors can potentially constrain the ability of managers to adjust reported earnings and may have especially strong incentives to do so when client distress is severe. At the same time, managers may face unusually strong incentives to avoid default-related costs and other adverse consequences of financial distress. Still, the net effect of these competing incentives is ultimately an empirical question.13

To evaluate the influence of managers’ accounting choices in this setting, we look at several measurements. We employ the years {−6} and {−5} as base years to formulate expectations. We then compute abnormal amounts, defined as the difference between expected and current period amounts, for each year preceding the going concern opinion. For example, the level of abnormal accounting accruals serves as a proxy for the extent to which earnings are managed upwards or downwards, where accounting accruals (AC_t) for period t equals earnings (NI_t) less operating cash flows (CF_t). The expected amounts for accruals, earnings, and operating cash flows are defined as the average of their respective amounts for base years {−6} and {−5} preceding the going concern opinion. The abnormal amounts for earnings, accruals, and cash flow amounts are then computed by comparing actual amounts to their expected amounts and then scaled by total assets.14

Abnormal Earnings. Panel A of

Table 7. Abnormal earnings, abnormal accounting accruals, and abnormal cash flows preceding the going concern audit opinion ^a.

Year Relative to Going	Abnormal Changes		Statistical Tests
Concern Opinion	Mean	Median	t-Statistic	Wilcoxon Signed-Rank
Panel A: Abnormal earnings
Year {−4}	−0.011	−0.006	−0.59	−192.5
Year {−3}	−0.021	−0.018	−0.97	−367.0 ***
Year {−2}	−0.077	−0.026	−2.91 ***	−720.5 ***
Year {−1}	−0.211	−0.065	−2.73 ***	−633.5 ***
Year { 0}	−0.160	−0.114	−3.82 ***	−571.0 ***
Panel B: Abnormal accounting accruals
Year {−4}	−0.023	−0.028	−0.64	−57.5
Year {−3}	−0.004	−0.058	−0.12	−89.0
Year {−2}	−0.057	−0.038	−1.45	−120.0 *
Year {−1}	−0.093	−0.018	−1.55	−80.0
Year { 0}	−0.328	−0.178	−2.69 ***	−178.0 ***
Panel C: Abnormal cash flows
Year {−4}	0.020	0.015	0.78	55.0
Year {−3}	0.006	0.031	0.26	68.0
Year {−2}	−0.006	−0.014	−0.24	−39.0
Year {−1}	−0.003	−0.033	−0.11	−46.0
Year { 0}	0.164	−0.054	1.95 ***	97.0 *

^a The sample comprises firms that received going concern audit opinions. Year {0} is the period in which the going concern audit opinion is received; years {−1} through {−4} represent the firm years immediately preceding the going concern opinion year. The accrual for period t (AC_t) equals earnings (NI_t) less operating cash flows (CF_t), and the expected accrual, earnings, and operating cash flow figures are defined as the average of their respective figures for base years {−6} and {−5} preceding the going concern opinion. The abnormal accruals, earnings, and cash flow amounts are scaled by total assets. For the statistical tests of differences, a * (**/***) designates statistical significance at the 0.10 (0.05/0.01) level; two-tailed tests.

shows descriptive statistics and tests involving abnormal earnings for the years preceding and including the going concern opinion. As expected, the abnormal earnings are significantly negative for the two years preceding and the year of the going concern opinion. Specifically, the average level of abnormal earnings is −7.7%, −21.1%, and −16.0% of total assets for years {−2}, {−1}, and {0} preceding the going concern opinion—and each of those percentages are significantly less than zero at the 0.01 level. The median level of abnormal earnings is similarly negative, and the Wilcoxon test of medians shows significance at the 0.01 level for each of the years {−3} through {0} preceding the going concern opinion.

Abnormal Accounting Accruals. Panel B of Table 7 reports abnormal accounting accruals for the years preceding the going concern opinion. We see that only in the year of the going concern opinion do abnormal accounting accruals differ from zero. Specifically, in the going concern year, the mean and median levels of abnormal accounting accruals are −32.8% and −17.8% of total assets, respectively. Both measures are significantly less than zero at the 0.01 level. The negative direction of abnormal accounting accruals for year {0} is consistent with either more conservative reporting, an inability to mask the firm’s poor financial condition, or auditor pressure for more conservative reporting of accruals. There is some weak evidence of negative abnormal accounting accruals in the two years prior to the going concern opinion. The mean abnormal accounting accruals equal −5.7% and −9.3% of total assets for years {−2} and {−1}, respectively, although both means are only marginally significant.15

Abnormal Cash Flows. Panel C of Table 7 reports results for abnormal cash flows. The results show minimal evidence of any significant behavior in abnormal cash flows for years preceding the going concern opinion. Specifically, abnormal cash flows do not significantly differ from zero for any of the years {−4} through {−1}, although there is evidence of an increase in year {0}. This latter finding is unexpected given the distressed nature of these companies’ operations. Among the possible explanations is that managers might have substantially reduced discretionary expenditures (such as maintenance, advertising, and research and development) in the going concern opinion year. The cash-flow evidence should therefore be interpreted cautiously and primarily as supplementary to the abnormal earnings and abnormal accrual evidence.

Overall, the evidence on managers’ accounting choices is not consistent with our exploratory conjectures that managers use their discretion in the reporting of accruals to increase earnings in years preceding the going concern opinion. There is, instead, some evidence of negative earnings management in the two years leading up to the going concern opinion, including the year of the going concern opinion. This finding suggests that either managers’ accounting choices (or the economic setting) result in negative abnormal accruals, or that auditors can, and will, undo the accounting choices of managers when it is in their best interest to do so. Indeed, the behavior of abnormal accounting accruals is consistent with a more conservative depiction of the company’s performance in the going concern opinion year, at least in comparison to earlier years.16 These results suggest that managers’ accounting choices may contribute to the documented decline in earnings informativeness, but they do not provide a complete explanation for it. The year {0} accrual evidence deserves emphasis. Mean abnormal accruals of −32.8% of total assets point to sharply conservative reporting in the opinion year. That pattern is consistent with auditors constraining, or even reversing, optimistic accounting choices as going concern doubt crystallizes.

6. Summary and Conclusions

The findings of this study provide compelling evidence of a significant decline in the informativeness of earnings starting at least two years before the receipt of a first-time going concern audit opinion. The decline is substantiated by a decrease of approximately 95 percent in the earnings response coefficient, which represents the slope coefficient obtained from a regression linking abnormal returns to unexpected earnings, for the two years preceding and including the year of the going concern audit opinion. Notably, none of the earnings response coefficients for these three years exhibit statistical significance at conventional levels. These findings suggest that the market significantly reduces its reliance on earnings to the extent that there is no observable relation between earnings and returns at least two years prior to and including the year of the going concern audit opinion. The findings are consistent with the crucial applicability of the going concern assumption in determining the relevance of accounting earnings.

These findings connect with prior evidence in at least three ways. They parallel the lower earnings response coefficients reported for qualified-opinion firms (Choi & Jeter, 1992) and for firms with high termination probability (Subramanyam & Wild, 1996). They extend the post-opinion declines in Dong et al. (2015) to the years preceding the first-time opinion. They also complement evidence that losses and impaired reporting reduce the information content of earnings (Hayn, 1995; Wilson, 2008). The findings also help explain why announcement-date reactions to going concern opinions often appear muted or delayed (Kausar et al., 2009). If the market has largely revalued earnings in advance, the formal opinion confirms more than it reveals.

To further explore this phenomenon, we extend the analysis to investigate a corollary hypothesis proposing that firms facing greater uncertainty in continuing as a going concern would experience a more substantial and accelerated decline in earnings informativeness. To test this hypothesis, we condition the analysis based on whether firms receive two consecutive going concern audit opinions, which is assumed to be indicative of heightened going concern uncertainty. The results reveal an earlier and more pronounced decline in earnings informativeness for firms with consecutive going concern audit opinions. This accelerated decline is substantiated by a significant decrease in earnings informativeness three years before the going concern audit opinion. Moreover, the pronounced decline in the earnings response coefficient persists, as there is a similarly significant drop in earnings informativeness for the year preceding and the year of the going concern audit opinion for firms with two consecutive going concern opinions. Importantly, the decline in earnings informativeness for firms with consecutive going concern opinions exceeds the decline observed for firms that received a single-year going concern audit opinion.

We extend the analysis to examine potential factors contributing to the decline in earnings informativeness preceding the going concern audit opinion. Specifically, we investigate temporal variations in established determinants of earnings informativeness, including firm size, leverage, growth, systematic risk, earnings variability, and persistence. The findings indicate significant shifts in these determinants during one or more years leading up to the going concern opinion. Moreover, the direction of these shifts generally aligns with the observed decline in earnings informativeness. For instance, both leverage and earnings variability significantly increase in each of the three years preceding the going concern audit opinion, which corresponds to the observed deterioration in earnings informativeness during the same period. Additionally, the results reveal a significant decrease in growth for the two years preceding the going concern opinion, further supporting the decline in earnings informativeness. Overall, these results are consistent with the possibility that the market partially, but not fully, anticipates greater going concern uncertainty and its implications for earnings informativeness.

To delve deeper into these determinants, we include them as explanatory variables in a regression analysis that examines the relation between returns and earnings to assess their capacity to explain the temporal patterns in earnings informativeness leading up to and including the going concern audit opinion. The results indicate that the inclusion of these determinants significantly improves the explanatory power of earnings for returns. However, despite the statistical significance and ability of many of these determinants to explain a significant portion of the temporal decline in earnings informativeness, there remains a significant unexplained decline in earnings informativeness at least two years prior to and throughout the year of the going concern audit opinion. Thus, while several of these determinants prove significant, the inferences regarding earnings informativeness and going concern uncertainty after controlling for these determinants remain similar to those without such controls. These results are consistent with going concern uncertainty being associated with a decline in earnings informativeness that is not fully subsumed by these known determinants.

Lastly, the analysis explores the possibility that managers’ accounting choices, influenced by increased going concern uncertainty, adversely impact the informativeness of earnings. Specifically, if managers’ accounting choices reflect attempts to portray the firm as less distressed, this would predictably impair earnings informativeness. The evidence from this study reveals substantial and significant negative abnormal earnings for the three years leading up to the going concern opinion, along with negative abnormal accruals in the year of the going concern audit opinion, and less pronounced negative accruals for the two years before the going concern opinion. These findings suggest that managers’ accounting choices either reflect the severe distress of the firm or their inability to conceal the firm’s dire financial condition. Nevertheless, this exploratory analysis was insufficient to explain fully the decline in earnings informativeness for the years leading up to and including the going concern audit opinion. More broadly, the Section 5 evidence is more consistent with conservative reporting or an inability to mask distress than with a systematic pattern of income-increasing accrual choices.

The findings carry practical implications. For investors and analysts, earnings lose valuation content well before a first-time going concern opinion is issued, so the analysis of firms with deteriorating fundamentals warrants greater weight on cash flows, leverage, and other distress indicators, including accounting-based measures of failure risk. In risk-management terms, a fading earnings response serves as a market-based early-warning indicator. It can complement model-based tools built on accounting ratios (Altman, 1968; Ashraf et al., 2019). For auditors, the evidence that the market revalues earnings at least two years before the formal opinion bears on the timeliness of going concern reporting and supports earlier, graduated communication of going concern risk. For regulators and standard setters, the results inform disclosure policy for firms under going concern uncertainty by identifying when reported earnings convey less valuation-relevant information to the market.

This study suggests several avenues for future research. Informativeness here is measured primarily from short-window announcement returns. Such designs cannot fully separate a decline in earnings informativeness from growth in competing predisclosure information. Future research could ask whether the temporal path documented here holds under more recent reporting regimes, including expanded auditor disclosures such as critical audit matters (Vandamas & Myring, 2026). It could also test that path in other markets and enforcement environments. Research could also link the pre-opinion decline in earnings informativeness to the drivers of auditors’ opinion decisions and to accounting-based indicators of distress and reporting irregularities (Averio, 2021; Srebro et al., 2021). In short, going concern risk leaves a market footprint well before it reaches the audit report. Reading that footprint through the earnings response offers researchers and practitioners an early window into financial distress.