Earnings Comparability,Accounting Similarities,and Stock Returns: Evidence From Peer Firms’ Earnings Restatements

Abstract

Using a sample of earnings restatements, we provide evidence that an empirical measure of the comparability in two firms’ earnings (“earnings comparability”) captures the extent to which a firm’s accounting choices and estimates are similar to those of its restating peer firm. We then document that investors appear to underreact to the implications of this earnings comparability signal. Additional analyses reveal that large traders and short sellers react in a timely manner, and their trades trigger an immediate negative price reaction to earnings comparability. Small traders appear to behave inattentively, and their herding-driven delayed trades contribute to a negative drift in prices.

Keywords

earnings comparability restatements accounting choice stock returns

Introduction

Financial statement users frequently compare financial information across firms to judge firm performance. A growing body of academic research focuses on defining and understanding the benefits of financial statement comparability across firms (e.g., Chen, Collins, Kravet, & Mergenthaler, in press; De Franco, Kothari, & Verdi, 2011; S. Kim, Kraft, & Ryan, 2013; J.-B. Kim, Li, Lu, & Yu, 2016).¹ In this study, we examine whether the De Franco et al. (2011) earnings comparability measure captures the extent to which a firm’s accounting choices and estimates are similar to those of peer firms that have restated their earnings. We then examine whether the information conveyed by this signal improves investors’ ability to price the implications of a peer firm’s earnings restatement.

We are particularly interested in the consequences of financial statement comparability in the earnings restatements setting because prior work documents significant decreases in stock price for nonrestating peer firms (i.e., price contagion) around restatement announcements (Gleason, Jenkins, & Johnson, 2008). It is not entirely clear, however, whether the documented price contagion is consistent with market efficiency in the sense that the degree of contagion reflects an appropriate assessment that a firm has accounting problems similar to its restating peer firm. We thus address the following two specific research questions:

Research Question 1: Is earnings comparability between a firm and its restating peer firm a useful signal of the extent to which that firm’s accounting choices and estimates are similar to its restating peer firm?

Research Question 2: Do stock prices efficiently incorporate the implications of earnings comparability?

We first conduct two sets of tests to examine whether earnings comparability between a nonrestating firm and its restating peer firm is an incrementally useful signal of similarity in the two firms’ accounting choices and estimates beyond the signal used by previous literature (i.e., the level of the nonrestating firm’s accruals). First, we find that earnings comparability is positively associated with similarities in important accounting choices (i.e., capital leasing decisions and pension assumptions) between a nonrestating firm and its restating peer firm. Second, earnings comparability is positively associated with the likelihood that a nonrestating firm restates its own earnings in the future. At the same time, the accruals of the nonrestating firm are not significant in either of these tests. Taken together, these results suggest that earnings comparability with a restating peer firm is an incrementally useful signal that a firm’s financial statements are of a similarly poor quality.

Next, we examine the extent to which prices efficiently incorporate the implications of earnings comparability around peer firms’ restatement announcements. Under the efficient market hypothesis, investors are aware of the implications of earnings comparability and fully penalize firms for their potential financial misreporting around the peer firm’s earnings restatement announcement. For instance, evidence in Gleason et al. (2008) suggests no significant delayed price reactions on average (i.e., when earnings comparability is not considered). However, it is possible that investors do not accurately price the information conveyed by comparability. On one hand, comparability is not easy to discern.² A subset of investors may not be fully aware of the similarity in two firms’ accounting choices and assumptions due to limited attention, or may be aware of this similarity but are unable to trade due to their risk-bearing capacity or transaction costs. This would cause the market to underreact to the implications of comparability (Hirshleifer & Teoh, 2003). On the other hand, research suggests that investors sometimes overweight the extent to which one firm’s news maps into a second firm (Cheng & Eshleman, 2014; Thomas & Zhang, 2008). This would cause the market to overreact to the implications of comparability (D. Griffin & Tversky, 1992). Thus, whether prices efficiently react to the implications of earnings comparability is an empirical question.

Using a sample of earnings restatements published by the U.S. General Accounting Office (GAO), we find that a firm’s stock price reaction to a peer firm’s restatement announcement is negatively associated with earnings comparability between the firm and its restating peer firm. However, we also find a significant and negative drift in stock prices associated with comparability during the postrestatement announcement period. In multiple regression analyses, we find that as much as 80% of the price impact is delayed. These results are consistent with investor underreaction to the information conveyed by earnings comparability.

In additional analyses, we find that attentive and inattentive investors react differently to earnings comparability. Large traders and short sellers appear to trade on this signal immediately. Specifically, around the restatement announcements, large traders’ net buy (of nonrestating firms) is negatively associated with earnings comparability, and short-sell volume is positively associated with earnings comparability. In contrast, small (and less sophisticated) traders fail to react promptly to earnings comparability as we find that small traders’ net buy during the restatement announcement period is unassociated with earnings comparability.

During the postrestatement announcement (i.e., drift) period, we find that both large traders’ net buy and short-sell volume (of the nonrestating firms) are unassociated with earnings comparability. However, small traders’ net buy during the postrestatement announcement period is negatively associated with this signal. There are two alternative interpretations for these results: First, consistent with the limited attention model, sophisticated investors are bounded by their risk-bearing capacity and reluctant to trade beyond their initial positions, while less sophisticated investors’ delayed trading is triggered by herding (e.g., Barber & Odean, 2008; Barber, Odean, & Zhu, 2009) or the eventual revelation of poor earnings quality by the nonrestating firm. Alternatively, the observed small trades during the nonnews drift period reflect sophisticated investors’ trading in disguise by splitting up their large orders (e.g., Cready, Kumas, & Subasi, 2014). The totality of our evidence is consistent with the limited attention model.

Our results hold after controlling for economic similarities between the two firms that are unrelated to accounting choices and estimates. Similarly, additional analyses suggest that our results are unlikely to be due to a mechanical relation in the earnings comparability measure. Specifically, our results only hold for restatements that convey bad news (i.e., they do not hold for good news restatements), we do not find similar results around pseudo-restatement announcement dates, and we do not find similar results around the restating peer firm’s earnings announcement prior to the restatement announcement.

Our study makes several contributions to the academic literature. First, we contribute to the literature on financial statement comparability. Previous research shows that when firms exhibit greater comparability with their industry peer firms, they experience an improvement in their information environment (Brochet, Jagolinzer, & Riedl, 2013; De Franco et al., 2011; Horton, Serafeim, & Serafeim, 2013) and engage in more efficient acquisitions (Chen et al., 2015). We contribute to this literature by showing that comparability can also be associated with negative outcomes. That is to say, when there is evidence that a firm in an industry has restated its earnings, firms with financial statements that are more comparable with the restating peer firm also have financial statements of poor quality.

Second, our study contributes to the literature on market efficiency with respect to signals of financial statement comparability. Previous research begins with a maintained assumption that financial statement users (i.e., analysts, credit markets, equity markets, etc.) understand financial statement comparability, and then these studies examine the benefits of comparability. However, it is not easy for financial statement users to estimate comparability. Specifically, users need to (a) understand the firm’s accounting choices and estimates across multiple dimensions (i.e., rent/buy percentages, pension assumptions, fair value assumptions, accruals estimates), (b) understand the impact of each of these choices and estimates on the firm’s financial statements, and (c) assess the extent to which all of these same accounting choices and estimates for a second firm map into that second firm’s financial statements. Our findings support the view that investors underreact to the implications of comparability. This is consistent with investors having limited attention to complex accounting information (e.g., Bloomfield, 2002; Hirshleifer & Teoh, 2003), or being aware of the information but being unable to trade on it due to risk-bearing capacity or transaction costs. Our results are inconsistent with the moderated confidence hypothesis (D. Griffin & Tversky, 1992), which predicts that investors overweight the implications of a peer firm’s news and thus overreact to comparability.

Third, our study has implications for the literature investigating price formation around earnings restatements. P. Griffin (2003) finds that financial analysts largely react to (rather than anticipate) earnings restatement news. Similarly, Drake, Myers, Scholz, and Sharp (2015) find that sophisticated investors (i.e., short sellers) fail to anticipate restatement announcements but trade as if they understand the postannouncement stock price implications of restatement disclosures. We document that sophisticated users react promptly not only to the implications of earnings restatement announcements for the restating firm but also to the implications of the restatement announcement for restating firm’s peers with similar earnings quality.

Finally, our study may be of interest to auditors and regulators for two reasons: First, we find that after observing a restatement, the earnings comparability measure predicts accounting concerns and future restatements at nonrestating peer firms. Thus, this measure could potentially be used by auditors and regulators to identify accounting problems at peer firms in a more timely manner. Second, our results suggest that unsophisticated investors fail to appreciate the implications of earnings comparability. Instead, these investors exhibit herding behavior only after sophisticated investors trade on it. Thus, accounting standards that promote comparability should reduce the variation in comparability across firms and, thus, reduce the advantages that sophisticated investors have in processing differences in comparability across firms.

Prior Research and Hypothesis Development

Prior Literature on Financial Statement Comparability

A growing body of academic research focuses on defining and understanding the benefits of financial statement comparability across firms. De Franco et al. (2011) develop a measure of financial statement comparability, and find that this measure is associated with improvements in equity analyst forecast accuracy, coverage, and dispersion. Studies also find that comparability benefits investors by improving debt market assessments of credit risk (S. Kim et al., 2013), acquisition decisions (Chen et al., 2015), and by encouraging managers to make timely voluntary disclosures (Gong, Li, & Zhou, 2013; J.-B. Kim et al., 2016). Finally, studies examine comparability in the context of mandatory International Financial Reporting Standards (IFRS) adoption (e.g., Barth, Landsman, Lang, & Williams, 2012; Brochet et al., 2013; DeFond, Hu, Hung, & Li, 2011; DeFond, Hung, Li, & Li, 2013; Wang, 2014; Yip & Young, 2012). These studies generally find evidence supporting the proposition that investors benefit from comparability of financial statement information across firms.

A related literature documents that poor accounting practices tend to cluster across networks of firms. Chiu, Teoh, and Tian (2013) find that a firm is more likely to restate its earnings if one of its directors is also on the board of another firm that has restated its earnings in the past. That is, interlocking boards carry corporate practices between firms, and interlocking directors are conduits for the spread of behavior between firms. Two recent studies also show that firms’ geographic proximity acts as a network through which accounting choices and estimates are correlated. These studies document that firms within a given geographic location share similarities in religious and social norms (McGuire, Omer, & Sharp, 2012) and employee education levels (Call, Campbell, Dhaliwal, & Moon, 2017), both of which could influence their accounting choices and estimates. Finally, firms in the same industry engage in similar business transactions, face similar economic conditions, and use similar accounting practices, so any information about one firm is likely to apply to the others as well.

Using the restatement setting, Gleason et al. (2008) follow the earnings management literature and use the level of the nonrestating firms’ accruals as a signal of the extent to which accounting discretion across two firms is correlated. However, accruals lack power to assess comparability between firms for at least two reasons: First, accruals are influenced by underlying business shocks that are unrelated to accounting discretions. Second, focusing on the nonrestating firms’ accruals in isolation does not capture the similarity in financial reporting between the firm and its restating peer firm.

We argue that earnings comparability is a more powerful signal than accruals (in isolation) in detecting correlated accounting discretion between firms. To facilitate our discussion, we also refer to “relative accruals” and “accruals comparability” to bridge the gap between accruals and earnings comparability. “Relative accruals” is the difference in the level of accruals between the nonrestating firm and the restating firm. This measure at a minimum compares accruals across two firms but likely does not adequately control for common economic shocks across the industry. “Accruals comparability” would involve estimating the nonrestating firm’s accruals under the assumption that the firm experiences the same business shocks as the restating peer firm. In other words, conditional on the restating peer firm’s business shocks, a small difference between the estimated accruals of the firm and its restating peer firm’s actual accruals should be a strong indication of correlated earnings quality.

Earnings comparability is similar to “accruals comparability” in the sense that it compares accruals across firms and also controls for underlying economic events. However, we argue that similar accounting policies are more likely to be reflected through similar earnings rather than similar accruals because earnings includes cash flow information as well. Incorporating cash flows’ similarity is important for two reasons: First, cash flows are often influenced directly by accounting choices and estimates. For instance, operating leases generally require higher cash outflows than capital leases because the lessor assumes the residual value risk of the equipment. Second, to the extent that the restating peer firm is also managing real activities (e.g., overproduction to spread out costs), capturing similarity in cash flows increases the power of the signal.

On the contrary, it is possible that cash flow information incorporates randomness that is unassociated with accounting policies and estimates, and thus earnings comparability is less powerful than accruals comparability. We treat this as an empirical question, and conduct two sets of tests to examine whether earnings comparability is an incrementally useful signal for detecting correlated accounting discretions between firms than accruals, relative accruals, and accruals comparability. Our first hypothesis (stated in alternative form) is stated as follows:

Hypothesis 1A (H1A): Earnings comparability is positively associated with similarities in accounting choices and estimates between a nonrestating firm and its restating peer firm (after controlling for the level of accruals).

Hypothesis 1B (H1B): Earnings comparability is positively associated with the likelihood of a nonrestating firm’s future restatements (after controlling for the level of accruals).

Stock Returns Around Peer Firms’ Restatement Announcements

A large body of prior work focuses on stock price reactions to a firm’s own restatement announcement and documents large negative reactions to it.³ As an interesting departure from this work, Gleason et al. (2008) examine stock price reactions to a peer firm’s restatement announcement. The premise underlying this inquiry is that investors reassess the likelihood of poor earnings quality in nonrestating firms in the same industry after the restatement announcement. A negative stock price reaction is expected if investors revise the likelihood of poor earnings quality in the nonrestating firm. In a sample of restatements that convey bad news, they find an average −1.5% cumulative return among nonrestating firms around a peer firm’s restatement announcement. Furthermore, the degree of this price decline is related to the level of the nonrestating firm’s accruals. If earnings comparability is an incrementally useful signal (relative to accruals), we expect that, around peer firms’ restatement announcements, the price reaction is negatively associated with earnings comparability after controlling for accruals. We thus state the following hypothesis (in alternative form):

Hypothesis 2A (H2A): The nonrestating firm’s stock returns around the peer firm’s restatement announcement are negatively associated with earnings comparability (after controlling for the level of accruals).

It is less clear whether prices efficiently impound the information conveyed by earnings comparability. Gleason et al. (2008) find that price contagion is short lived, and there is no drift, on average, in prices after the restatement announcements. This finding suggests two possibilities: First, investors are sophisticated, and react swiftly and efficiently to signals of correlated accounting discretion (e.g., accruals of the nonrestating firm). A second possibility is that investors do not efficiently incorporate the information conveyed by earnings comparability, but without an empirical measure to detect comparability between two firms, researchers are unable to detect the underreaction or overreaction.⁴

It is possible that investors underreact to the information conveyed by earnings comparability. Hirshleifer and Teoh (2003) provide a limited attention model that examines firms’ reporting and disclosure decisions when investors have limited attention and processing power. In their model, there is a continuum of investors who have a probability of being attentive to an aspect of the economic environment (e.g., a particular accounting estimate made by the manager).⁵ Ex-post, attentive investors are those who react to the information while inattentive investors fail to do so. Because attentive and inattentive investors observe different signals or aspects of the economic environment, they disagree with each other, and this disagreement generates trading. In equilibrium, price is approximated by a weighted average of attentive and inattentive investors’ beliefs where the weights reflect their respective risk-bearing capacity.

It is also possible that investors overreact to the information conveyed by earnings comparability. D. Griffin and Tversky (1992) develop the moderated confidence hypothesis, which argues that Bayesian investors may systematically place too much weight on imprecise signals. Recent research examines the extent to which the market reaction to related firm news is efficient, and finds support for investor overreaction and, thus, the moderated confidence hypothesis (Cheng & Eshleman, 2014; Thomas & Zhang, 2008).

Given that prior theory and evidence support a prediction of efficient pricing, underreaction, or overreaction, we view investor reaction to the information conveyed by earnings comparability as an empirical question and state the following null hypothesis:

Hypothesis 2B (H2B): The nonrestating firm’s stock returns after the peer firm’s restatement announcement are not associated with earnings comparability (after controlling for the level of accruals).

Empirical Measures and Sample

Measure of Financial Statement Comparability

Accruals (Accruals_Ajt) of a nonrestating firm A in industry j for the most recent year prior to restatement announcement year t (by firm R) are defined as the firm’s net income less cash flow from operations, scaled by average total assets. Following Gleason et al. (2008), we adjust accruals for the industry average in our regression analyses.

Earnings comparability (Earnings Comparability_Ajt) involves estimating the nonrestating firm A’s earnings under the assumption that the firm experiences the same business shocks as its restating peer firm R. To do so, we hold constant any underlying business shocks for both firms. We use stock returns as a proxy for business shocks following De Franco et al. (2011), and estimate the mapping between earnings and business shocks using piece-wise linear regressions during the 4 years (i.e., 16 quarters) prior to the restatement announcement for firm A in Equation 1a and firm R in Equation 1b:

E a r n i n g s_{A j q} = a_{A} + b_{A} R e t u r n_{A j q} + c_{A} N e g_{A j q} + d_{A} R e t u r n_{A j q} \times N e g_{A j q} + e_{A j q},

(1a)

E a r n i n g s_{R j q} = a_{R} + b_{R} R e t u r n_{R j q} + c_{R} Ne g_{R j q} + d_{R} R e t u r n_{R j q} \times N e g_{R j q} + e_{R j q},

(1b)

where Return refers to the stock return during the quarter, and Neg is set to 1 if Return is negative and 0 otherwise.⁶ The impact of business shocks on firm A’s and firm R’s earnings is captured by the estimated coefficients vector (a_A, b_A, c_A, d_A) and (a_R, b_R, c_R, d_R).

Earnings comparability is measured as the “distance” between (a) the estimated quarterly earnings of firm A conditional on the restating peer firm R’s business shocks and (b) the estimated quarterly earnings of firm R conditional on its own business shocks. In other words, for each of the 16 prior quarters, we calculate the following conditional earnings:

C o n d i t i o n a l E a r n i n g s_{A j q} = a_{A} + b_{A} R e t u r n_{R j q} + c_{A} N e g_{R j q} + d_{A} R e t u r n_{R j q} \times N e g_{R j q} .

(1c)

C o n d i t i o n a l E a r n i n g s_{R j q} = a_{R} + b_{R} R e t u r n_{R j q} + c_{R} N e g_{R j q} + d_{R} R e t u r n_{R j q} \times Ne g_{R j q} .

(1d)

By using the restating peer firm R’s return (Return_Rjt) in both Equations 1c and 1d, the comparability measure explicitly holds constant economic events between the two firms (i.e., the measure conditions on firm R’s economic events).⁷

Earnings comparability is calculated as the negative value of the average absolute difference between quarterly conditional earnings:

E a r n i n g s Comparabilit y_{A j t} = - \frac{1}{16} \sum | C o n d i t i o n a l E a r n i n g s_{A j q} - C o n d i t i o n a l E a r n i n g s_{R j q} | .

(1e)

The average difference is then multiplied by −1, so that higher values capture higher similarity in the two firms’ conditional earnings. Finally, in our empirical analysis we decile-rank earnings comparability within each industry j and scale the ranks to fall within the [0, 1] range.

Sample

We obtain a sample of restatement announcements from the GAO publications (GAO, 2003, 2006), which include 2,705 restatements announced during the period between January 1997 and June 2006. We obtain stock price data from the Center for Research in Security Prices (CRSP). Following Gleason et al. (2008), we focus on restatements that convey bad news (i.e., the initial price drop based on CRSP is larger than −1%).⁸

We further require that each restatement event is the first restatement announcement from a firm within the previous 24 months, so that investors view the restatement announcement as news about the quality of the firm’s accounting processes. After deleting firms with prices less than US$5 per share or without requisite data to construct control variables, our sample of restating peer firms includes 442 restatement announcements. Panel A of Table 1 shows our sample selection process.

Table 1.

Sample Selection and Descriptive Statistics.

Panel A: Sample Selection.
	Number of restatements
Total GAO restatements for the period Jan 1997 through Jun 2006	2,705
Less:
Firms missing [−10, +10] returns from CRSP	(392)
Firms where [−10, +10] returns are greater than −1.0%	(937)
Firms for which no comparability measure could be calculated within 2 years from restatement date	(566)
Firms that restated within previous 2 years	(86)
Firms with stock price less than US$5 or missing requisite control variables	(282)
Final sample of restatement events	442

Panel B: Variable Distribution.
Variables	M	SD	25%	50%	75%
Accruals	0.30	0.75	−0.02	0.02	0.18
Earnings Comparability (continuous)	−0.05	0.12	−0.05	−0.02	−0.01
AbsDiff_Size	1.83	1.52	0.64	1.45	2.64
AbsDiff_Growth	0.34	0.43	0.04	0.12	0.52
AbsDiff_Earnings	0.08	0.21	0.01	0.03	0.07
AbsDiff_Leverage	0.10	0.13	0.01	0.06	0.13
CorrCFO	0.08	0.39	−0.16	0.06	0.34
CorrReturns	0.12	0.38	−0.11	0.11	0.38

Panel C: Correlation Between Earnings Comparability and Firm Characteristics.
	Accruals	AbsDiff_Size	AbsDiff_Growth	AbsDiff_Earnings	AbsDiff_Leverage	CorrCFO	CorrReturns
Expected sign	?	−	−	−	−	+	+
Earnings Comparability	−.226 (<.01)	.039 (<.01)	−.260 (<.01)	−.162 (<.01)	−.112 (<.01)	−.003 (.67)	.032 (<.01)

Note. Panel A shows our sample selection procedures. Panel B shows descriptive statistics of the variables in our regression analyses (n = 20,818 peer firms for the 442 restatement announcements). Panel C shows Pearson’s correlations between Earnings Comparability and various measures for similarity in two firms’ accounting choices and estimates. All variables are defined in the appendix, and continuous variables are winsorized at 1% and 99% to mitigate the influence of outliers. GAO = U.S. General Accounting Office; CRSP = Center for Research in Security Prices.

In untabulated results, we examine the distribution of restatements in our sample across time. We observe an increase in the number of restatements after 2004, potentially due to more rigorous internal control efforts after the Sarbanes–Oxley Act’s Section 404 became effective. We also find that the industry composition of our sample is similar to that of the Compustat population.

We obtain nonrestating firms that have the same historical four-digit Standard Industry Classification (SIC) code as the restating peer firm (Chiu et al., 2013; McGuire et al., 2012). Following Gleason et al. (2008), nonrestating firms are retained if (a) announcement period stock returns are available, (b) the preannouncement stock price is at least US$5, and (c) the nonrestating firm has not announced an accounting restatement within the preceding 24 months. This process produces a sample comprised of 20,818 firms or about 47 firms for each restatement event.⁹

Research Design and Empirical Results

Research Design for H1A: Similarity in Accounting Estimates and Choices

To test whether earnings comparability is a signal of correlated accounting discretion across firms, we first focus on the extent to which it captures similarities in two firms’ accounting choices and estimates. Specifically, we examine whether Earnings Comparability is associated with the similarity in a nonrestating firm and its restating peer firms’ accruals quality, capital leasing decisions, and assumptions for defined benefit pension plans.¹⁰ We use the following model to test our prediction:

D e p e n d e n t V a r i a b l e = α_{0} + α_{1} E a r n i n g s C o m p a r a b i l i t y + α_{2} A c c r u a l s + \sum α_{k} Com p_{k} + ε_{1},

(2)

where Dependent Variable is (a) the difference in accruals quality (AbsDiff_AQ) between the firm and its restating peer firm, (b) the difference in these two firms’ capital leasing decisions (AbsDiff_Lease), (c) the difference in these two firms’ assumptions regarding the rate of return for pension assets (AbsDiff_PensionRate), and (d) the difference in these two firms’ assumptions regarding the increases in employee future salaries for pension liabilities (AbsDiff_PensionSalary).

To hold economic similarities constant between the nonrestating and restating peer firms, we include in Equation 2 a vector of six firm comparability variables (Comp_k) as control variables, which include the absolute difference in firm size (AbsDiff_Size), growth opportunities (AbsDiff_Growth), leverage (AbsDiff_Leverage), and profitability (AbsDiff_Earnings) between the nonrestating and restating peer firms, and the historical correlations of cash flows (CorrCFO) and stock returns (CorrReturns) between the two firms. All variables are defined in the appendix.

As a validation test, we also follow the approach to provide evidence that the drivers of accounting discretion networks documented in prior studies are associated with earnings comparability. Based on prior research, we expect that Earnings Comparability should be positively associated with the presence of board interlocking with the restating peer firm as well as the geographic proximity with the restating peer firm in the following two regressions:

E a r n i n g s C o m p a r a b i l i t y = β_{0} + β_{1} B o a r d I n t e r l o c k + β_{2} A c c r u a l s + \sum β_{k} C o m p_{k} + ε_{2},

(3a)

E a r n i n g s C o m p a r a b i l i t y = γ_{0} + γ_{1} S a m e L o c a t i o n + γ_{2} A c c r u a l s + \sum γ_{k} Com p_{k} + ε_{3},

(3b)

where BoardInterlock is defined as 1 if the two firms share a director during the year of the restatement and 0 otherwise (Chiu et al., 2013), and SameLocation as 1 if the two firms’ headquarters are within 100 km of each other and 0 otherwise (Ayers, Ramalingegowda, & Yeung, 2011). We expect positive coefficients for BoardInterlock and SameLocation because prior work shows that poor earnings quality across firms is more likely when firms have interlocking directors and are geographically close to one another.

Research Design for H1B: Future Restatements

Second, we test whether earnings comparability is positively associated with nonrestating firms’ future restatement announcements. It is important to note that our construct of interest is the extent to which accounting choices and estimates are correlated across firms, and not the extent to which restatements are correlated across firms. We focus here on future restatement events because they are ex-post indicators of similar poor earnings quality in the (initially) nonrestating firm. We estimate the following Probit model:

\begin{matrix} P r o b i t (F u t u r e R e s t a t e = 1) = δ_{0} + δ_{1} E a r n i n g s C o m p a r a b i l i t y + δ_{2} A c c r u a l s \\ + \sum δ_{k} Com p_{k} + \sum δ_{m} F i r m_{m} + ε_{4}, \end{matrix}

(4)

where FutureRestate equals 1 if the nonrestating firm restates its earnings within 3 years after the peer firm’s restatement announcement, and Firm_m is a set of governance and other firm-specific control variables that prior literature associates with the likelihood of restatements (Beasley, 1996; Dechow, Sloan, & Sweeney, 1996; DeFond & Jiambalvo, 1991).¹¹

Empirical Results for H1

Panels B and C of Table 1 present descriptive statistics and correlations for our sample. In Panel B, Earnings Comparability has negative values because, as previously mentioned, we multiply the absolute difference by −1. The mean and median of CorrCFO and CorrReturns are positive, consistent with the expectation that firms in the same industry are subject to common business shocks. Panel C provides correlations between (ranked) Earnings Comparability and Accruals as well as firm comparability variables. Under the null that both Earnings Comparability and Accruals capture the extent to which earnings quality is correlated across firms, one would expect a positive association between them. However, the correlation between Earnings Comparability and Accruals is −.226 (p < .01), suggesting that the two measures capture different constructs. The correlations between Earnings Comparability and the six firm comparability variables are generally statistically significant, indicating that it is important to control for firm comparability in testing our hypotheses. On the contrary, the magnitude of these correlations is quite low.

Table 2 presents the baseline results of Regression Models 2, 3a, 3b, and 4. Specifically, Panel A reports the results regarding the associations between Earnings Comparability and absolute differences in accounting estimates and choices. We find that Earnings Comparability is negatively associated with the absolute difference in the two firms’ accruals quality (AbsDiff_AQ), capital leasing decisions (AbsDiff_Lease), and assumptions regarding the increases in employee future salaries for pension liabilities (AbsDiff_PensionSalary). The association between Earnings Comparability and the absolute difference in the assumptions regarding the rate of return for pension assets (AbsDiff_PensionRate) is negative but statistically insignificant.

Table 2.

Analyses of Accruals and Accounting Estimates and Choices, the Determinants of Poor Earnings Quality Across Firms, and Predicting Future Restatements.

Panel A: Regressions of Accounting Estimates and Choices on Level of Nonrestating Firm’s Accruals.
	(1)	(2)	(3)	(4)
	AbsDiff_AQ	AbsDiff_Lease	AbsDiff_Pension-Rate	AbsDiff_Pension-Salary
Intercept	0.120 (27.20)***	0.088 (9.38)***	0.057 (2.26)**	0.558 (12.74)***
Earnings Comparability	−0.103 (−20.48)***	−0.038 (−4.21)***	−0.044 (−1.51)	−0.239 (−4.69)***
Accruals	−0.004 (−8.06)***	−0.003** (−2.11)	0.028 (0.98)	0.056 (0.65)
AbsDiff_Size	−0.004 (−5.47)***	−0.001 (−0.12)	−0.007 (−2.08)**	0.002 (0.16)
AbsDiff_Growth	0.056 (14.74)***	0.032 (3.57)***	0.104 (2.08)**	0.056 (0.87)
AbsDiff_Earnings	0.038 (3.00)***	0.098 (4.12)***	−0.020 (−0.83)	−0.068 (−0.63)
AbsDiff_Leverage	−0.027 (−3.15)***	0.453 (14.05)***	−0.055 (−1.02)	−0.153 (−1.21)
CorrCFO	0.001 (0.46)	0.029 (4.61)***	−0.018** (−2.15)	−0.006 (−0.21)
CorrReturns	0.002 (0.50)	0.032 (4.51)***	0.008 (0.85)	−0.057 (−1.57)
Observations	14,012	20,470	820	1,866
Adj. R² (%)	9.65	5.40	2.51	1.89

Panel B: Regressions of Earnings Comparability on Board Interlocks and Distance.
	Dependent variable: Earnings Comparability
Intercept	0.128*** (6.97)	0.108** (2.54)
BoardInterlock	0.234*** (5.65)
SameLocation		0.042*** (4.43)
Accruals	0.023 (0.97)	−0.055*** (−5.12)
AbsDiff_Size	0.005 (1.89)	0.005*** (3.47)
AbsDiff_Growth	−0.037 (−1.49)	−0.028*** (−4.05)
AbsDiff_Earnings	−0.037** (−2.44)	−0.071*** (−6.37)
AbsDiff_Leverage	−0.041 (−1.34)	−0.139*** (−6.05)
CorrCFO	0.024** (2.02)	0.034*** (5.69)
CorrReturns	0.081*** (6.36)	0.024*** (3.25)
Observations	4,832	18,060
Adj. R² (%)	17.73	23.79

Panel C: Probit Model of Future Restatements on the Level of the Nonrestating Firm’s Accruals.
	FutureRestate = 1
Earnings Comparability	0.043** (2.23)
Accruals	0.005 (0.98)
AbsDiff_Size	−0.008 (−0.42)
AbsDiff_Growth	0.013 (0.31)
AbsDiff_Earnings	−0.070 (−0.36)
AbsDiff_Leverage	−0.227 (−1.32)
CorrCFO	0.133*** (2.80)
CorrReturns	−0.027 (−0.76)
Audit Committee Size	−0.012*** (−6.22)
Board of Directors Size	0.010*** (11.35)
CEO is Chairman Indicator	−0.005 (−1.68)
Leverage	0.025 (0.95)
Finance	−0.027 (−1.16)
Growth	0.004 (1.41)
Acquisitions	0.014** (2.31)
Observations	7,530
Adj. R² (%)	9.64

Note. Panel A presents regression results of proxies for accounting estimates and choices on accruals. All dependent and independent variables are defined in the appendix. Panel B presents the regression results of accruals on board interlocks and geographic distance between the nonrestating and restating peer firm pairs. All dependent and independent variables are defined in the appendix. Panel C presents the regression results of a Probit regression that predicts future restatements. Key variables are defined in the appendix. In addition, Audit Committee Size is the number of audit committee members. Board of Directors Size is the number of directors on the board. CEO is Chairman Indicator is equal to 1 if the CEO of the company is also the chairman of the board of directors. Leverage is equal to total debt divided by total assets. Finance is equal to the firm’s current year operating cash flow less the average capital expenditures of the past 3 years, all divided by current assets in the prior year. Growth is the average percentage change in total assets for the 2 years ending before the restatement year. Acquisitions is equal to the natural logarithm of assets acquired through acquisition during the year preceding the restatement year.

** and *** indicate statistical significance at two-tailed 5% and 1% levels, respectively. All statistical significances are based on heteroskedasticity-adjusted standard errors clustered by firm.

The results in Panel B indicate that the determinants of correlated accounting choices and estimates across networks of firms as identified by prior research (i.e., board interlocking and geographic proximity) are significantly associated with earnings comparability between the nonrestating firm and the restating peer firm. Panel C of Table 2 shows that Earnings Comparability predicts the nonrestating firm’s future restatements. Overall, the results in Table 2 indicate that Earnings Comparability is a useful financial-statement-based signal of the extent to which accounting choices and estimates are correlated across two firms.

Table 3 summarizes the estimated coefficients and statistical significance when we include either Earnings Comparability or Accruals in Regression Models 2, 3a, 3b, and 4 without controlling for the other. For brevity, coefficient estimates for other variables in the models are not reported. Consistent with the results in Table 2, we find that Earnings Comparability is associated with the measures of similarity in accruals quality and accounting choices (AbsDiff_AQ, AbsDiff_Lease, AbsDiff_PensionSalary), the determinants of accounting discretion across networks of firms (BoardInterlock, SameLocation), and the probability of future restatements (FutureRestate). However, Accruals is only associated with similarity in accruals quality (AbsDiff_AQ). Overall, the results are consistent with H1A and H1B, suggesting that Earnings Comparability is an incremental useful signal of the extent to which accounting choices and estimates are correlated across two firms, which in turn helps predict future restatements.

Table 3.

Summarized Associations Between Signals of Poor Earnings Quality Across Firms and Accounting Estimates and Choices, the Determinants of Poor Earnings Quality Across Firms and Future Restatements.

	(1)	(2)	(3)	(4)	(5)	(7)	(8)
	AbsDiff_AQ	AbsDiff_Lease	AbsDiff_PensionRate	AbsDiff_PensionSalary	BoardInterlock	SameLocation	Prob (FutureRestate = 1)
Prediction	−	−	−	−	+	+	+
Accruals only	−0.001*** (−3.39)	−0.001 (−1.59)	0.035 (1.31)	0.102 (1.26)	−0.043 (−0.54)	−0.044 (−0.86)	0.004 (1.09)
Earnings Comparability only	−0.097*** (−21.15)	−0.034*** (−4.61)	−0.046 (−1.63)	−0.242*** (−5.99)	0.197*** (5.31)	0.021** (2.21)	0.033*** (2.61)

Note. This table repeats the analyses from Table 2 but has either Accruals or Earnings Comparability in the regressions. For brevity, we only tabulate the coefficient estimates and t-statistic for the variable of interest.

** and *** indicate statistical significance at two-tailed 5% and 1% levels, respectively. All statistical significances are based on heteroskedasticity-adjusted standard errors clustered by firm.

Research Design and Results for H2: Stock Price Reaction and Drift

After demonstrating that Earnings Comparability is a signal of accounting discretion being correlated across two firms, we next examine whether investors detect and price the implications of this signal using the following regression to test H2:

\begin{matrix} D e p e n d e n t V a r i a b l e = δ_{0} + δ_{1} E a r n i n g s C o m p a r a b i l i t y + δ_{2} A c c r u a l s + \sum δ_{m} C o m p_{k} \\ + \sum δ_{k} F i r m_{l} + \sum δ_{l} R e s t a t e_{r} + ε_{5}, \end{matrix}

(5)

where Dependent Variable represents (a) the cumulative abnormal stock returns during the peer firms’ restatement announcement period (CAR-Event) for H2A, and (b) the cumulative abnormal stock returns during the postannouncement (i.e., drift) period (CAR-Post) for H2B. The appendix provides detailed variable definitions.

In Equation 5, we expect a negative coefficient for Earnings Comparability (i.e., δ₁ < 0) when CAR-Event is the dependent variable because we predict that investors attend to signals of correlated accounting discretion (H2A). We make no prediction for the coefficient for Earnings Comparability when CAR-Post (i.e., price drift) is the dependent variable (H2B).

Firm_l in Regression Equation 5 is a vector of firm-level control variables. We include industry-adjusted cash flows (CashFlows) to control for firm performance (Gleason et al., 2008). We also control for the market value of equity (Size) and the book-to-market ratio (BTM), as price declines are a function of firm size and distress risk, and because these factors are known to affect stock prices.¹²Restate_r is a vector controlling for investor attention effects related to restatement events. Specifically, a larger decline in the restating peer firm’s stock price at the announcement (Restating Firm Return) and a restating peer firm that is larger (Restating Firm Size) are both likely to catch more attention and trigger more pronounced negative price reactions.

Panel A of Table 4 reports the mean and median of Restating Peer Firm Return for the [−10, +10] (trading day) restatement announcement (i.e., event) period and for the [+11, +90] postrestatement announcement (i.e., drift) period. The average abnormal return over the event window is −12.6% (median = −8.8%), which is large in magnitude because we focus on restatements with a negative stock market reaction.¹³ The mean (median) cumulative abnormal return during the postrestatement announcement period is 0.6% (−0.9%). Thus, on average there is no significant drift in stock prices for the restating peer firms in our sample.

Table 4.

Univariate Analysis of Abnormal Returns.

Panel A: Peer Firms’ Abnormal Returns Around Their Restatement Announcements.
	[−10, +10] trading day		[+11, +90] trading day
	M	Median	M	Median
Restating Peer Firm Return	−0.126 (p < .01)	−0.088 (p < .01)	0.006 (p = .52)	−0.009 (p = .64)

Panel B: Price Response Around Peer Firms’ Restatement Announcements.
	CAR-Event		CAR-Post
	M	Median	M	Median
Response to Restatement	−0.011 (p < .01)	−0.016 (p < .01)	−0.013 (p < .01)	−0.015 (p < .01)

Panel C: Average CAR-Event for Tercile Portfolio Sorted on Earnings Comparability.
	Earnings Comparability tercile
	1 (least comparable)	2	3 (most comparable)	3 − 1 (diff.)
CAR-Event	−0.009	−0.010	−0.014	−0.005 (p = .02)

Panel D: Average CAR-Post for Tercile Portfolio Sorted on Earnings Comparability.
	Earnings Comparability tercile
	1 (least comparable)	2	3 (most comparable)	3 − 1 (diff.)
CAR-Post	−0.006	−0.008	−0.023	−0.017% (p < .01)

Note. Panel A shows the size-adjusted cumulative abnormal stock returns around and after the restatement announcement of restating peer firms. Day 0 is the restatement announcement date. Panel B shows the size-adjusted cumulative abnormal stock returns of nonrestating firms around and after the restatement announcements of restating peer firms. Panel C shows the average values of CAR-Event for tercile portfolios sorted on Earnings Comparability. Panel D shows the average values of CAR-Post for terciles portfolios sorted on Earnings Comparability. All variables are defined in the appendix.

Panel B reports the mean and median of CAR-Event (i.e., firms’ response to peer firms’ restatement news) and CAR-Post (i.e., the drift in firms’ prices following the peer firms’ restatement announcement). The mean of CAR-Event is −1.1% (p < .01), which confirms the documented price declines around earnings restatements by Gleason et al. (2008). In contrast to their evidence, we find a statistically significant drift during the [+11, +90] postrestatement announcement period. Average CAR-Post is −1.3% (p < .01), which is slightly larger than the price decline during the restatement announcement period, consistent with a market underreaction to peer firms’ restatement announcements.¹⁴

Panel C shows the average CAR-Event for tercile portfolios sorted on Earnings Comparability. Average CAR-Event decreases from the bottom tercile (= −0.9%) to the top tercile (= −1.4%). The spread between the top and bottom terciles is statistically significant (p = .02). Thus, the initial market reaction to the restatement announcement is more negative when the likelihood of poor earnings quality in the nonrestating firm is higher, which is consistent with H2a. Panel D shows the average CAR-Post for tercile portfolios sorted on Earnings Comparability. Average CAR-Post decreases from the bottom tercile (= −0.6%) to the top tercile (= −2.3%). The spread between the top and bottom terciles is relatively large (= −1.7%) and statistically significant (p < .01). Thus, the univariate evidence in Panel D suggests that the market underreacts to Earnings Comparability (and is inconsistent with efficient pricing or overreaction).

Table 5 presents multiple regression results. In column 1, the estimated coefficient for Earnings Comparability is −0.012 (t = −3.81) in the regression of CAR-Event, indicating that the nonrestating firms in the top Earnings Comparability decile experience an incremental −1.2% abnormal return relative to the nonrestating firms in the bottom Earnings Comparability decile. Consistent with H2A, this evidence suggests that investors attend and react to signals of correlated accounting discretion. Consistent with Gleason et al. (2008), we find a negative coefficient for Accruals in the regression of CAR-Event, indicating that investors also react negatively to the levels of accruals. The evidence that both Accruals and Earnings Comparability are significant suggests that investors with varying levels of financial sophistication react to different financial signals. As expected, we also find that price declines are less pronounced for firms with strong cash flows but more pronounced for firms with high book-to-market ratios. However, we find smaller price declines when the restating peer firm is larger.¹⁵

Table 5.

OLS Regressions of Abnormal Returns During the Restatement Announcement Period (CAR-Event) and During the Postrestatement Announcement Period (CAR-Post).

	(1) CAR-Event	(2) CAR-Post
Earnings Comparability	−0.012*** (−3.81)	−0.054*** (−8.35)
Accruals	−0.002*** (−2.64)	−0.003 (−1.89)
AbsDiff_Size	0.000 (0.34)	−0.008*** (−6.28)
AbsDiff_Growth	−0.001 (−0.28)	−0.015** (−2.14)
AbsDiff_Earnings	−0.007 (−0.80)	−0.072*** (−4.79)
AbsDiff_Leverage	−0.003 (−0.42)	−0.034 (−1.81)
CorrCFO	−0.002 (−0.97)	0.005 (1.22)
CorrReturns	−0.005** (−2.12)	−0.001 (−0.16)
CashFlows	0.018*** (3.02)	0.038*** (3.27)
Size	0.000 (0.40)	0.018*** (13.36)
BTM	−0.011*** (−3.62)	−0.059*** (−7.89)
Restating Firm Return	0.003 (0.29)	0.172*** (9.44)
Restating Firm Size	0.003*** (7.21)	−0.002*** (−3.07)
Observations	20,818	20,818
Adj. R² (%)	2.16	12.43

Note. This table presents the regression results of nonrestating firms’ size-adjusted stock returns around the restating peer firms’ restatement announcements. The dependent variable in column 1 is CAR-Event, and the dependent variable in column 2 is CAR-Post. All key variables are defined in the appendix. In addition, CashFlows is the industry-adjusted operating cash flows from the statement of cash flows, scaled by average total assets. To adjust for industry, the Compustat-wide average for the four-digit SIC is subtracted from the firm’s value. Size is the natural log of the firm’s market value of equity at the end of the year prior to peer firm’s restatement. BTM is the natural log of the firm’s book assets to market assets ratio at the end of the year prior to peer firm’s restatement. Restating Firm Return is the size-adjusted buy-and-hold return of the restating peer firm during the [−10, +10] trading day window centered on the restating peer firm’s restatement announcement day. The benchmark returns are the size-decile returns constructed from NYSE, AMEX, and NASAQ firms in CRSP. Restating Firm Size is the restating peer firm’s CRSP size-decile rank as of the beginning of the calendar year in which the restatement is announced. Estimated coefficients for the industry- and year-fixed effects are untabulated. OLS = ordinary least squares; SIC = Standard Industry Classification; NYSE = New York Stock Exchange; AMEX = American Stock Exchange; NASDAQ = National Association of Securities Dealers Automated Quotations; CRSP = Center for Research in Security Prices.

** and *** indicate statistical significance at two-tailed 5% and 1% levels, respectively. All statistical significances are based on heteroskedasticity-adjusted standard errors clustered by firm.

The coefficients in column 1 for the variables controlling for firm comparability between the two firms are mostly insignificant. The fact that we find a significant coefficient for Earnings Comparability but not for measures of firm comparability in the same regression provides assurance that the significant coefficient for Earnings Comparability is not driven by economic similarities between the two firms.

Column 2 reports regression results when CAR-Post is the dependent variable. We find a significantly negative coefficient for Earnings Comparability (−0.054, t = −8.35), which indicates that during the postrestatement announcement period, the nonrestating firms in the top Earnings Comparability decile experience an incremental −5.4% abnormal return in comparison with the nonrestating firms in the bottom Earnings Comparability decile. This evidence indicates that more than 80% of the price reaction associated with Earnings Comparability is delayed. This evidence is consistent with the market underreacting to this signal of correlated accounting discretion. Consistent with Gleason et al. (2008), we find that the estimated coefficient for Accruals is insignificant in the regression of CAR-Post. CAR-Post is positively associated with industry-adjusted cash flows (CashFlows), indicating that prices underreact to the information conveyed by cash flows (e.g., Sloan, 1996). Delayed reaction is also a function of size and book-to-market ratio.

Interestingly, we also find significantly negative coefficients for the variables that control for firm comparability in column 2. Given the result that the market underreacts to economic similarities between two firms, we would expect positive coefficients for these variables (that capture absolute differences). Thus, these results again do not support the idea that our results are a function of nonaccounting firm comparability between the two firms.¹⁶

Finally, untabulated results indicate that the documented drift cannot be explained by investor overreaction as we do not detect any price reversals at longer horizons. Specifically, we regress future long-term abnormal returns (i.e., during the [+91, +180] or [+91, +240] trading day window) on CAR-Post (i.e., the abnormal returns during the [+11, +90] window), an interaction of CAR-Post and Earnings Comparability, and all of the control variables from Equation 5. If the drift represents overreaction, we should find a negative coefficient for CAR-Post. Furthermore, if this overreaction is associated with the likelihood of earnings comparability, we should find a negative coefficient for the interaction terms of CAR-Post and Earnings Comparability. In untabulated results, we find that the estimated coefficients are insignificant in all cases.¹⁷

Additional Analyses

The Implications of Trade Size

Given that our primary analysis suggests that investors underreact to the information conveyed by earnings comparability, in additional analysis we now examine the types of investors that trade around a peer firm’s restatement announcement.

Restatement announcement trading

The observable trading activities of attentive investors (e.g., large traders and short sellers) and inattentive investors (e.g., small traders) are likely to differ with respect to earnings comparability. Evidence from investor trading generally supports the dichotomy of attentive and inattentive investors set forth in the limited attention model (e.g., Ayers, Li, & Yeung, 2011; Battalio, Lerman, Livnat, & Mendenhall, 2012; Mikhail, Walther, & Willis, 2007). Large (small) investors are more likely to be attentive (inattentive) to signals of correlated accounting discretion. This would suggest that large traders’ net buy (short-seller volumes) during the restatement announcement period is negatively (positively) associated with earnings comparability, but small traders’ net buy is unassociated with earnings comparability.¹⁸ To examine whether this is indeed the case, we use the following regression model to test the associations between Earnings Comparability and investor trading activity:

\begin{matrix} D e p e n d e n t V a r i a b l e = λ_{0} + λ_{1} E a r n i n g s C o m p a r a b i l i t y + λ_{2} A c c r u a l s + \sum λ_{m} C o m p_{k} \\ + \sum λ_{k} F i r m_{l} + \sum λ_{l} R e s t a t e_{r} + ε_{5}, \end{matrix}

(6)

where Dependent Variable represents (a) the cumulative abnormal net buy of small traders (SmallTrade-Event) and large traders (LargeTrade-Event) during the announcement period, and (b) the cumulative abnormal short-sell volumes during the announcement period (ShortVol-Event). The appendix provides detailed variable definitions. If large trades are more negatively associated with Earnings Comparability than small trades, we expect that λ₁ is more negative in the LargeTrade-Event regression than in the SmallTrade-Event regression. Furthermore, if that abnormal short-sell volume is positively associated with Earnings Comparability during the announcement period, then we expect λ₁ > 0.

We define LargeTrade-Event as the average daily abnormal net buy for large traders during the [−10, +10] trading window centered on the peer firm’s restatement announcement day. Following Ayers, Li, and Yeung (2011), daily net buy for large traders is defined as the daily dollar buy volume of large trades (i.e., above US$50,000 per trade) minus the daily dollar sell volume of large trades. We subtract average daily net buy for large traders during a 60-day control period (i.e., [−74, −15]) from the daily net-buy measure, and then scale it by the average nondirectional daily volume of large trades (i.e., buy volume plus sell volume) during the control period.¹⁹

Similarly, we define SmallTrade-Event as the average daily abnormal net buy for small traders during the [−10, +10] window centered on the peer firm’s restatement announcement date. Daily net buy for small traders is defined as the daily dollar buy volume of small trades (i.e., below US$5,000 per trade) minus the daily dollar sell volume of small trades. We then subtract average daily net buy for small traders during the 60-day control period from the daily net-buy measure and scale it by the average nondirectional daily volume of small trades during the control period.

Columns 1 and 2 of Table 6 present the multivariate results when abnormal small net buy (SmallTrade-Event) and large net buy (LargeTrade-Event) are the dependent variables. We find results consistent with the notion that around the peer firm’s restatement announcement, more sophisticated large traders react to the signal of correlated accounting discretion while less sophisticated small traders do not.

Table 6.

OLS Regressions of Trading Activities During the Restatement Announcement Period.

	Net-buy activity		Short-selling volume
	(1)	(2)	(3)
	SmallTrade	LargeTrade	ShortVol
Earnings Comparability	−0.006 (−0.61)	−0.240*** (−2.68)	0.011*** (2.67)
Accruals	−0.002 (−0.63)	−0.009 (−0.18)	0.004 (1.72)
AbsDiff_Size	−0.004 (−1.88)	0.015 (0.54)	0.001 (0.84)
AbsDiff_Growth	0.005 (0.80)	−0.107 (−0.83)	0.005 (1.06)
AbsDiff_Earnings	0.015 (0.58)	0.611 (1.18)	−0.032 (−1.55)
AbsDiff_Leverage	0.020 (1.20)	−0.178 (−0.80)	0.004 (0.43)
CorrCFO	−0.010 (−1.64)	−0.048 (−0.55)	0.003 (0.80)
CorrReturns	0.000 (0.05)	0.039 (0.39)	−0.002 (−0.73)
CashFlows	−0.016 (−1.53)	0.137 (0.85)	−0.003 (−0.31)
Size	0.000 (−0.15)	0.062*** (2.81)	−0.003*** (−3.02)
BTM	−0.009 (−1.65)	−0.058 (−0.60)	0.010** (2.35)
Restating Firm Return	−0.030 (−1.14)	0.070 (0.27)	−0.015 (−0.69)
Restating Firm Size	−0.001 (−0.96)	0.002 (0.09)	0.001 (1.30)
p value for large < small	p = .005
Observations	11,105	11,105	4,380
Adj. R² (%)	.31	.06	1.24

Note. This table presents regressions of SmallTrade, LargeTrade, and ShortVol. All key variables are defined in the appendix. In addition, CashFlows is the industry-adjusted operating cash flows from the statement of cash flows, scaled by average total assets. To adjust for industry, the Compustat-wide average for the four-digit SIC is subtracted from the firm’s value. Size is the natural log of the firm’s market value of equity at the end of the year prior to peer firm’s restatement. BTM is the natural log of the firm’s book assets to market assets ratio at the end of the year prior to peer firm’s restatement. Restating Firm Return is the size-adjusted buy-and-hold return of the restating peer firm during the [−10, +10] trading day window centered on the restating peer firm’s restatement announcement day. The benchmark returns are the size-decile returns constructed from NYSE, AMEX, and NASDAQ firms in CRSP. Restating Firm Size is the restating peer firm’s CRSP size-decile rank as of the beginning of the calendar year in which the restatement is announced. OLS = ordinary least squares; SIC = Standard Industry Classification; NYSE = New York Stock Exchange; AMEX = American Stock Exchange; NASDAQ = National Association of Securities Dealers Automated Quotations; CRSP = Center for Research in Security Prices.

** and *** indicate statistical significance at two-tailed 5% and 1% levels, respectively. All statistical significances are based on heteroskedasticity-adjusted standard errors clustered by firm.

Column 3 presents the results when abnormal short-sell volume (ShortVol) is the dependent variable. We define ShortVol as the average abnormal daily short volume during the [−10, +10] window centered on the peer firm’s restatement announcement date. Daily short volume is defined as daily number of shares for all intraday transactions that involve short selling scaled by total daily trading volume. If this ratio is greater than 1, the observation is deleted. To derive abnormal daily short volume, we subtract the average daily short volume during the 60-day control period (i.e., [−74, −15]) from each day’s short volume.²⁰ We find more intense selling when earnings comparability is higher. The results from the analysis on short volume also indicate that the price declines associated with Earnings Comparability are economically significant, because short sellers view the drift as being large enough to exploit after considering transaction costs and limits to arbitrage.

Postrestatement announcement (i.e., drift) period trading

During the postrestatement announcement (i.e., drift) period, however, it is less clear whether attentive and/or inattentive investors’ trading will cause a drift associated with correlated accounting discretion with the restating peer firm. Although the insufficient initial price adjustments provide attentive investors with arbitrage opportunities, the limited attention model argues that attentive investors are limited by their risk-bearing capacity and unwilling to trade. In other words, unconstrained sophisticated investors would have already traded away any mispricing if they are not subject to noise trader risk or liquidity constraints. It is also ambiguous whether inattentive investors exhibit delayed trading associated with earnings comparability. On one hand, because inattentive investors neither pay attention to nor learn about the quality of accounting systems, their trades should not be associated with comparability. On the other hand, unsophisticated investors tend to herd after observing conspicuous market-level activities such as large price movements and high volumes (Barber & Odean, 2008; Barber et al., 2009). Herding would lead to an association between inattentive investors’ postrestatement announcement period trades and earnings comparability to the extent that the initial price declines are also associated with earnings comparability.

Table 7 reports the multivariate results of abnormal small net buy (SmallTrade-Post), large net buy (LargeTrade-Post), and short-selling volume (ShortVol-Post) during the postrestatement announcement (i.e., drift) period. Following the same approach of measuring trading activities during the event window, we define LargeTrade-Post as the average daily abnormal net buy for large traders during the [+11, +90] trading day window. Similarly, SmallTrade-Post is the average daily abnormal net buy for small traders during the [+11, +90] window, and ShortVol-Post is the average daily abnormal short volume during the [+11, +90] window.

Table 7.

OLS Regressions of Trading Activities During the Postrestatement Announcement Period.

Panel A: Trading Activities During the Postrestatement Announcement Period [+11, +90].
	Net-buy activity		Short-selling volume
	(1)	(2)	(3)
	SmallTrade_Post	LargeTrade_Post	ShortVol_Post
Earnings Comparability	−0.017** (−2.57)	0.023 (0.66)	0.001 (0.32)
Accruals	−0.005 (−0.99)	0.018 (1.14)	0.002 (0.96)
CashFlows	0.004 (0.33)	−0.021 (−0.49)	0.013 (1.13)
Size	0.004** (2.38)	0.018** (2.31)	−0.003*** (−3.37)
BTM	0.007 (1.64)	−0.019 (−0.97)	0.014** (3.33)
Restating Firm Return	−0.027 (−1.65)	0.079 (0.90)	−0.009 (−0.43)
Restating Firm Size	0.001 (1.82)	0.000 (0.05)	0.001** (2.09)
AbsDiff_Size	0.000 (0.02)	−0.003 (−0.46)	0.002 (1.95)
AbsDiff_Growth	0.008 (1.50)	−0.018 (−0.73)	−0.001 (−0.42)
AbsDiff_Earnings	−0.020 (−0.97)	0.234** (2.14)	0.035 (1.73)
AbsDiff_Leverage	0.024 (1.65)	−0.011 (−0.18)	−0.005 (−0.57)
CorrCFO	−0.004 (−1.08)	0.015 (0.74)	0.001 (0.27)
CorrReturns	−0.005 (−1.34)	0.033 (1.42)	0.000 (0.04)
Observations	11,105	11,105	4,380
Adj. R² (%)	.41	.04	2.71

Panel B: Regressions of Net Buy for Small Trades After Controlling for Announcement Period Volume and Returns.
	(1)	(2)	(3)
Earnings Comparability	−0.010 (−1.59)	−0.009 (−1.48)	−0.009 (−1.49)
Announcement Volume	−0.001** (−2.15)		−0.001** (−2.11)
Announcement Returns		0.015*** (3.06)	0.015*** (2.98)
Controls and Fixed Effects	Yes	Yes	Yes
Observations	11,105	11,105	11,105
Adj. R² (%)	.44	.45	.53

Note. Panel A of this table presents regressions of SmallTrade-Post, LargeTrade-Post, and ShortVol-Post, and Panel B presents regressions of SmallTrade-Post. In Panel B, Announcement Volume is equal to the average trading volume for the firm’s stock as reported by CRSP over the [−10, +10] window centered on the peer firm’s restatement announcement day. Announcement Returns is equal to the firm’s size-adjusted buy-and-hold return during the [−10, +10] window centered on the peer firm’s restatement announcement day. The benchmark returns are the size-decile returns constructed from NYSE, AMEX, and NASDAQ firms in CRSP. All key variables are defined in the appendix. OLS = ordinary least squares; CRSP = Center for Research in Security Prices; NYSE = New York Stock Exchange; AMEX = American Stock Exchange; NASDAQ = National Association of Securities Dealers Automated Quotations.

*, **, and *** indicate statistical significance at two-tailed 10%, 5%, and 1% levels, respectively. All statistical significances are based on heteroskedasticity-adjusted standard errors clustered by firm.

Column 1 of Table 7 Panel A shows a significantly negative coefficient for Earnings Comparability in the regression of SmallTrade-Post. However, neither the estimated coefficient for Earnings Comparability in the regression of LargeTrade-Post in column 2 nor in the regression of ShortVol-Post in column 3 is significant. These results indicate that small trades (rather than large trades and short sellers) play a role in explaining the drift associated with earnings comparability.²¹ Theoretically, the insufficient initial price adjustments should provide attentive investors with arbitrage opportunities. The fact that we observe no evidence of postannouncement trades from attentive investors suggests that they are unwilling to trade, perhaps limited by their risk-bearing capacity. This evidence supports the assumptions set forth in the limited attention model.

At first glance, the delayed trading of small trades should not be associated with Earnings Comparability, if inattentive investors do not have the information processing power to learn about the quality of accounting systems from financial statements or from other traders. However, this association can be attributable to other established trading patterns of small traders. Barber and Odean (2008) suggest that individual investors may follow attention-grabbing stocks (i.e., high volumes). In addition, Barber et al. (2009) find that small trades exhibit a strong pattern of herding (i.e., they are correlated with prior returns). Thus, conspicuous market-level activities during the event window may attract investor attention and trigger selling from small investors. Herding would result in a postrestatement announcement (i.e., drift) period association between small trades and Earnings Comparability to the extent that the initial price declines are also associated with Earnings Comparability.

Panel B of Table 7 reports the results when we add two control variables to the model to capture the potential impact of attention-grabbing and herding, trading volume (Announcement Volume) and stock price reaction (Announcement Returns) during the announcement period. The results show that the estimated coefficient for Earnings Comparability becomes insignificant after controlling for either variable. Small traders’ net buy during the postrestatement announcement window is negatively associated with Announcement Volume and positively associated with Announcement Returns. Column 3 shows that the attention effect and the herding effect do not subsume each other when both are included in the regression.²² Overall, these results suggest that sophisticated investors react to Earnings Comparability, but their trading is limited by their risk-bearing capacity, while unsophisticated traders’ herding explains the drift.

Economic Comparability Between Firms

Despite our effort to control for the effects of firm comparability in our analyses, one may still be concerned that the observed price reactions associated with earnings comparability are driven by the similarity in firm characteristics between a nonrestating firm and its restating peer firm that are unrelated to accounting choices and estimates or financial reporting decisions.²³ To further rule out this explanation, we conduct a pseudo-event analysis by estimating regressions as in Equation 6 but using a sample where the restatement announcements convey good news to the market (e.g., abnormal return > 1% over the [−10, +10] window around the restatement announcement). If the observed effects associated with Earnings Comparability are largely driven by “mechanical” price comovements between firms, we expect a significant positive coefficient for Earnings Comparability in this subsample. During our sample period, we identify 314 restatements with positive returns > 1% over the [−10, +10] window around the restatement announcement, which leads to a sample of 16,735 firm i− firm j observations around those restatements. In this sample, the estimated coefficients for Earnings Comparability are insignificant. Thus, the observed price decline associated with Earnings Comparability appears unique to restatements conveying bad news (i.e., investors’ concern about accounting quality).

In a second pseudo-event analysis, we create pseudo-restatement announcement dates, and then estimate Regression Equation 6. A pseudo-event date is the first trading date that the restating peer firm experiences more than a 1% price drop within the 2-year window prior to the actual restatement announcement date. If the reported results in the article are mechanical, we should find similar results in this setting. However, we find no evidence of price declines related to Earnings Comparability around these pseudo-event dates.²⁴

Finally, we examine the association between firms’ stock returns around restating peer firms’ prior earnings announcements and Earnings Comparability. Each earnings announcement is the most recent earnings announcement prior to the restatement announcement. The most recent earnings announcement is used to minimize the possibility that either the restating peer firm or the nonrestating firm experiences significant changes in fundamentals. Similar to our main sample, we focus on events that convey negative news (i.e., announcement period return less than −1%). We find that firms’ stock returns around restating peer firms’ prior earnings announcements are unassociated with Earnings Comparability. Thus, it is unlikely that our results are explained by a mechanical relation, which would predict a significant negative association.²⁵

Conclusion

We examine whether earnings comparability between a firm and its restating peer firm is an incremental useful signal of the extent to which accounting discretion correlates across firms, and whether investors incorporate this signal into their pricing of the restatement event. We find that earnings comparability between a firm and its restating peer firm appears to have the power to detect similar accounting choices and estimates in the nonrestating firm, beyond the ability of accruals alone. Consistent with limited attention to similarities in firms’ accounting discretion, we find that investors underreact to the information conveyed by earnings comparability, and that approximately 80% of the price impact is delayed. In additional analyses, we find that large traders and short sellers attend to signals of earnings comparability and trade on it swiftly. In contrast, small traders fail to react to the information conveyed by earnings comparability with a restating peer firm. The entire body of our evidence is consistent with the limited attention model (i.e., sophisticated investors are bounded by their risk-bearing capacity, whereas less sophisticated investors’ delayed trading is triggered by herding). Our results cannot be attributed to economic similarities between the two firms that are unrelated to accounting choices and estimates.

Our study contributes mainly to the literature on financial statement comparability and on earnings restatements. We provide evidence that the earnings comparability measure is a useful signal to the market. However, our evidence suggests that the market on average underreacts to the implications of this signal for the underlying poor earnings quality of a nonrestating firm. This evidence may not be surprising, given the difficulty in evaluating the earnings quality of a nonrestating firm. Future research may consider developing a new earnings quality measure that utilizes earnings comparability to restating firms.

Footnotes

Appendix

Variable Definitions.

Signals of correlated accounting discretion
Accruals	Industry-adjusted accruals for the year prior to the restatement announcement year, defined as the firm’s net income less cash flow from operations, scaled by average total assets. To adjust for industry, the Compustat-wide average for the four-digit SIC is subtracted from the firm’s value.
Earnings Comparability	The estimated earnings comparability between a nonrestating firm A and a restating peer firm R in the same industry j, measured over the 4 years prior to the restatement year t, decile ranked, and scaled to be between [0,1]. Mathematically, it is calculated as follows: $- \frac{1}{16} \sum \| C o n d i t i o n a l E a r n i n g s_{A j q} - C o n d i t i o n a l E a r n i n g s_{R j q} \|,$ where Conditional Earnings is conditional earnings, calculated from the coefficients from a piece-wise regression of quarterly earnings on quarterly stock returns multiplied by the actual stock returns of the restating peer firm.
Main variables for testing the power of signals of poor earnings quality (H1)
AbsDiff_AQ	The absolute value of the difference in accruals quality between nonrestating and restating peer firms. Accrual quality is defined as the residual from regressing a firm’s accruals on prior year, current year, and future year cash flows, as defined in Dechow and Dichev (2002) and modified by Ball and Shivakumar (2006).
AbsDiff_Lease	The absolute value of the difference in lease accounting between nonrestating and restating peer firms. Lease accounting is coded 1 if the firm’s capital to operating lease ratio is below the two-digit SIC industry median in a given year and 0 otherwise. A firm’s capital to operating lease ratio is calculated as the firm’s capital lease obligation reported in Compustat divided by the present value of the firm’s operating lease payment obligations over the following year. For simplicity, we use a 10% discount rate for all firms.
AbsDiff_PensionRate	The absolute value of the difference in the pension rate of return assumption between nonrestating and restating peer firms. Pension rate of return assumption is coded 1 if the firm’s rate of return assumption for its pension assets is above the median in a given year and 0 otherwise.
AbsDiff_PensionSalary	The absolute value of the difference in pension salary increase assumption between nonrestating and restating peer firms. The pension salary increase assumption is coded 1 if the firm’s salary increase assumption for its pension liability is below the two-digit SIC industry median in a given year and 0 otherwise.
BoardInterlock	An indicator variable equal to 1 if the two firms share the same director (as in Chiu, Teoh, & Tian, 2013).
Distance	Equals the geographic distance between the two firms’ headquarter locations (as in Ayers, Ramalingegowda, & Yeung, 2011).
SameLocation	An indicator variable equal to 1 if the two firms’ headquarters locations are within 100 km of each other and 0 otherwise.
FutureRestate	Equal to 1 if a peer firm experienced a restatement within the years after the restating peer firm’s announcement.
Main variables for price reaction analyses (H2)
CAR-Event	The buy-and-hold return in excess of the buy-and-hold return from the benchmark size-decile portfolio during the [−10, +10] trading day window centered on the peer firm’s restatement announcement day. The benchmark returns are the size-decile returns constructed from NYSE, AMEX, and NASDAQ firms in CRSP. If a stock is delisted during the return accumulation period, we obtain delisting returns following Shumway (1997) and Shumway and Warther (1999), and assume the proceeds are reinvested to earn the average return of the matching size-decile portfolio.
CAR-Post	The size-adjusted buy-and-hold return during the [+11, +90] trading day window after the peer firm’s restatement announcement day.
Main variables for investor trading analyses (additional analyses)
LargeTrade-Event	Average daily abnormal net buy for large traders during the [−10, +10] trading day window centered on the peer firm’s restatement announcement day. Daily net buy for large traders is defined as daily dollar buy volume of large trades (i.e., above US$50,000 per trade) minus dollar sell volume of large trades. We subtract average daily net buy for large traders during a 60-day control period (i.e., [−74, −15]) from the daily net-buy measure, and then scale it by the average nondirectional daily volume of large trades (i.e., buy volume plus sell volume) during the control period.
LargeTrade-Post	Average daily abnormal net buy for large traders during the [+11, +90] trading day window.
SmallTrade-Event	Average daily abnormal net buy for small traders during the [−10, +10] trading day window centered on the peer firm’s restatement announcement day. Daily net buy for small traders is defined as daily dollar buy volume of small trades (i.e., below US$5,000 per trade) minus sell volume of small trades. We then subtract average daily net buy for small traders during the 60-day control period from the daily net-buy measure, and then scale it by the average nondirectional daily volume of small trades during the control period.
SmallTrade-Post	Average daily abnormal net buy for small traders during the [+11, +90] trading day window.
ShortVol-Event	Average daily abnormal short volume during the [−10, +10] trading day window centered on the peer firm’s restatement announcement day. Daily short volume is defined as daily number of shares for all intraday transactions that involve short selling scaled by total daily trading volume. To derive abnormal short volume, we subtract average daily short volume during the 60-day control period (i.e., [−74, −15]) from the daily short volume.
ShortVol-Post	Average daily abnormal short volume during the [+11, +90] trading day window.
Firm comparability control variables
AbsDiff_Size	The absolute value of the difference between the restating peer firm’s Size and the firm’s Size. Size is defined as the natural log of the firm’s market value of equity at the end of the year prior to peer firm’s restatement.
AbsDiff_Growth	The absolute value of the difference between the restating peer firm’s book-to-market ratio and the firm’s book-to-market ratio, which is the natural log of the firm’s book assets to market assets ratio at the end of the year prior to peer firm’s restatement.
AbsDiff_Earnings	The absolute value of the difference between the restating peer firm’s return on assets in the year prior to the restatement and the firm’s return on assets for the same time period.
AbsDiff_Leverage	The absolute value of the difference between the restating peer firm’s long-term debt scaled by total assets, both for the year prior to the restatement, and the firm’s long-term debt scaled by total assets, for the same time period.
CorrCFO	Pearson’s correlation between a firm’s quarterly cash flows from operations and the restating peer firm’s quarterly cash flows from operations during the 16 quarters prior to the restatement year.
CorrReturns	Pearson’s correlation between a firm’s quarterly stock returns and the restating peer firm’s quarterly stock returns during the 16 quarters prior to the restatement year.

Note. SIC = Standard Industry Classification; CRSP = Center for Research in Security Prices; NYSE = New York Stock Exchange; AMEX = American Stock Exchange; NASDAQ = National Association of Securities Dealers Automated Quotations.

Acknowledgements

We appreciate the helpful comments and suggestions from two anonymous reviewers, Andrew Acito, Ben Ayers, Steve Baginski, Linda Bamber, Larry Brown, Nerissa Brown, Brian Cadman, Jeffrey Callen, Agnes Cheng, Dan Dhaliwal, Jenna Feagin, Jennifer Francis, Lisa Hinson, Nicole Jenkins, John Jiang, Marilyn Johnson, Bin Ke, Melissa Lewis Western, Rick Mergenthaler, Linda Myers, Jeffrey Ng, Kathy Petroni, Marlene Plumlee, Bob Resutek, Bharat Sarath, Katherine Schipper, Casey Schwab, Jonathan Shipman, Roger Silvers, Steve Stubben, Jake Thornock, Isabel Wang, Dan Wangerin, workshop participants at the University of Georgia, Georgia State University, Lehigh University, Michigan State University, Rutgers University, Syracuse University, the State University of New York at Buffalo, the University of Utah, and the discussants and participants at the American Accounting Association (AAA) Annual Meeting, the Hong Kong University of Science and Technology (HKUST) Accounting Research Symposium, and the AAA Financial Accounting and Reporting Section (FARS) Mid-Year Meeting. We thank Julie Wu and Jake Thornock for providing short-interest data, and Anne Ehinger, Jenna Feagin, Lisa Hinson, and Jasmine Wang for research assistance.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article:John Campbell gratefully acknowledges financial support through a Terry Sanford Research Award from the Terry College of Business at the University of Georgia.

Notes

References

Ayers

Yeung

P. E.

(2011). Investor trading and the post earnings announcement drift. The Accounting Review, 86, 385-416.

Ayers

Ramalingegowda

Yeung

P. E.

(2011). Hometown advantage: The effects of monitoring institution location on financial reporting discretion. Journal of Accounting & Economics, 52, 41-61.

Ball

Kothari

S. P.

Robin

(2000). The effect of international institutional factors on properties of accounting earnings. Journal of Accounting & Economics, 29, 1-51.

Ball

Shivakumar

(2006). The role of accruals in asymmetrically timely gain and loss recognition. Journal of Accounting Research, 44, 207-242.

Barber

Odean

(2008). All that glitters: The effect of attention and news on the buying behavior of individual and institutional investors. Review of Financial Studies, 21, 785-818.

Barber

Odean

Zhu

(2009). Do retail trades move markets? Review of Financial Studies, 22, 151-186.

Barth

Landsman

Lang

Williams

(2012). Are IFRS-based and US GAAP-based accounting amounts comparable? Journal of Accounting & Economics, 54, 68-93.

Basu

(1997). The conservatism principle and the asymmetric timeliness of earnings. Journal of Accounting & Economics, 24, 3-37.

Battalio

Lerman

Livnat

Mendenhall

(2012). Who, if anyone, reacts to accrual information? Journal of Accounting & Economics, 53, 205-224.

10.

Beasley

(1996). An empirical analysis of the relation between the board of director composition and financial statement fraud. The Accounting Review, 71, 443-465.

11.

Beneish

(1999). Incentives and penalties related to earnings overstatements that violate GAAP. The Accounting Review, 74, 425-457.

12.

Bergstresser

Desai

Rauh

(2006). Earnings manipulation, pension assumptions, and managerial investment decisions. Quarterly Journal of Economics, 121, 157-195.

13.

Bloomfield

(2002). The “incomplete revelation hypothesis” and financial reporting. Accounting Horizons, 16, 233-243.

14.

Brochet

Jagolinzer

Riedl

(2013). Mandatory IFRS adoption and financial statement comparability. Contemporary Accounting Research, 30, 1373-1400.

15.

Call

Campbell

Dhaliwal

Moon

, Jr. (2017). The disciplining effect of rank-and-file employees on financial reporting quality (Working paper). Tucson: University of Arizona.

16.

Chen

C.-W.

Collins

Kravet

Mergenthaler

(in press). Financial statement comparability and the efficiency of acquisition decisions. Contemporary Accounting Research.

17.

Cheng

C. S. A.

Eshleman

(2014). Does the market overweight imprecise information? Evidence from customer earnings announcements. Review of Accounting Studies, 19, 1125-1151.

18.

Chiu

P.-C.

Teoh

Tian

(2013). Board interlocks and earnings management contagion. The Accounting Review, 88, 915-944.

19.

Cready

Kumas

Subasi

(2014). Are trade size-based inferences about traders reliable? Evidence from institutional earnings-related trading. Journal of Accounting Research, 52, 877-909.

20.

Dechow

Dichev

(2002). The quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review, 77, 35-59.

21.

Dechow

Sloan

Sweeney

(1996). Causes and consequences of earnings manipulation: An analysis of firms subject to enforcement actions by the SEC. Contemporary Accounting Research, 13, 1-36.

22.

DeFond

Hung

(2011). The impact of mandatory IFRS adoption on foreign mutual fund ownership: The role of comparability. Journal of Accounting & Economics, 51, 240-258.

23.

DeFond

Hung

(2013). Does mandatory IFRS adoption affect crash risk? The Accounting Review, 90, 265-299.

24.

DeFond

Jiambalvo

(1991). Incidence and circumstances of accounting errors. The Accounting Review, 66, 643-655.

25.

De Franco

Kothari

S. P.

Verdi

. (2011). The benefits of financial statement comparability. Journal of Accounting Research, 49, 895-931.

26.

Drake

Myers

Scholz

Sharp

(2015). Short selling around restatement announcements: When do bears pounce? Journal of Accounting, Auditing & Finance, 30, 218-245.

27.

Financial Accounting Standard Board. (2010, September). Conceptual framework for financial reporting (Statement of Financial Accounting Concepts No. 8). Retrieved from http://www.fasb.org/jsp/FASB/Document_C/DocumentPage?cid=1176157498129&acceptedDisclaimer=true

28.

Gleason

Jenkins

Johnson

(2008). The contagion effects of accounting restatements. The Accounting Review, 83, 83-110.

29.

Gong

L. Y.

Zhou

(2013). Earnings non-synchronicity and voluntary disclosure. Contemporary Accounting Research, 30, 1560-1589.

30.

Griffin

Tversky

(1992). The weighing of evidence and the determinants of confidence. Cognitive Psychology, 24, 411-435.

31.

Griffin

(2003). A league of their own? Financial analysts’ responses to restatements and corrective disclosures. Journal of Accounting, Auditing & Finance, 18, 479-517.

32.

Hirshleifer

Teoh

S. H.

(2003). Limited attention, information disclosure, and financial reporting. Journal of Accounting & Economics, 36, 337-386.

33.

Horton

Serafeim

(2013). Does mandatory IFRS adoption improve the information environment? Contemporary Accounting Research, 30, 388-423.

34.

Hribar

Jenkins

(2004). The effect of accounting restatements on earnings revisions and the estimated cost of capital. Review of Accounting Studies, 9, 337-356.

35.

Imhoff

Thomas

(1988). Economic consequences of accounting standards: The lease disclosure rule change. Journal of Accounting & Economics, 10, 277-310.

36.

Kahneman

Slovic

Tversky

(1982). Judgment under uncertainty: Heuristics and biases. Cambridge, MA: Cambridge University Press.

37.

Kim

J.-B.

L. Y.

(2016). Financial statement comparability and expected crash risk. Journal of Accounting & Economics, 61, 294-312.

38.

Kim

Kraft

Ryan

(2013). Financial statement comparability and credit risk. Review of Accounting Studies, 18, 783-823.

39.

Lee

(1992). Earnings news and small traders: An intraday analysis. Journal of Accounting & Economics, 15, 265-302.

40.

Lee

Ready

(1991). Inferring trade direction from intraday data. Journal of Finance, 46, 733-746.

41.

Lev

Ryan

(2008). Rewriting earnings history. Review of Accounting Studies, 13, 419-451.

42.

Libby

Bloomfield

Nelson

(2002). Experimental research in financial accounting. Accounting, Organizations and Society, 27, 775-810.

43.

Maines

L. A.

(1995). Judgment and decision-making research in financial accounting: A review and analysis. In Ashton

R. H.

Ashton

A. H.

(Eds.), Judgment and decision-making research in accounting and auditing (pp. 76-101). New York, NY: Cambridge University Press.

44.

McGuire

Omer

Sharp

(2012). The impact of religion on financial reporting irregularities. The Accounting Review, 87, 645-673.

45.

Mikhail

M. B.

Walther

Willis

R. H.

(2007). When security analysts talk, who listens? The Accounting Review, 82, 1227-1253.

46.

Palmrose

Richardson

Scholz

(2004). Determinants of market reactions to restatement announcements. Journal of Accounting & Economics, 37, 59-89.

47.

Palmrose

Scholz

(2004). The circumstances and legal consequences of non-GAAP reporting: Evidence from restatements. Contemporary Accounting Research, 21, 139-180.

48.

Shumway

(1997). The delisting bias in CRSP data. Journal of Finance, 52, 327-340.

49.

Shumway

Warther

(1999). The delisting bias in CRSP’s Nasdaq data and its implications for the size effect. Journal of Finance, 54, 2361-2379.

50.

Sloan

(1996). Do stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review, 71, 289-315.

51.

Thomas

Zhang

(2008). Overreaction to intra-industry information transfers? Journal of Accounting Research, 46, 909-940.

52.

U.S. General Accounting Office. (2003). Financial statement restatements: Trends, market impacts, regulatory responses, and remaining challenges (GAO-03-138). Washington, DC: Government Printing Office.

53.

U.S. General Accounting Office. (2006). Financial statement restatements: Update of public company trends, market impacts, and regulatory enforcement activities (GAO-06-1053R). Washington, DC: Government Printing Office.

54.

Wang

(2014). Accounting standards harmonization and financial statement comparability: Evidence from transnational information transfer. Journal of Accounting Research, 52, 955-992.

55.

Yip

Young

(2012). Does mandatory IFRS adoption improve information comparability? The Accounting Review, 87, 1767-1789.