An Analysis of the Sources of Value Loss Following Financial Restatements

Abstract

We decompose the total value loss around firms’ announcements of financial restatements into components arising from investors’ revisions in cash flows and discount rates. First, relative to population benchmarks, restatements represent circumstances in which the cash flow component becomes more important in explaining valuations. While we find significant contributions from both sources, with the cash flow component explaining more than 33% of the variation in stock returns surrounding restatement announcements, this component explains only 13% to 22% in comparable non-restating firms. When restatements are caused by underlying financial fraud, the discount rate impact becomes more important, explaining about 88% of return variation. On the contrary, the cash flow impact is relatively larger for firms with higher earnings persistence or restatements associated with errors. Our decomposition of the value loss helps explain returns in the post-announcement period. Firms with a higher relative discount rate impact experience a significant downward stock price drift after the initial announcement-related price decline. For firms with a higher relative cash flow impact, the evidence suggests the initial impact of the restatement announcement is more complete with no subsequent drift pattern. Our findings close gaps in the evidence on financial restatements and extend the literature on the drivers of stock price movements.

Keywords

financial restatement variance decomposition equity valuation cost of equity post-announcement drift

Introduction

This study examines the underlying sources of declines in firm value following financial restatements, with the aim of estimating the relative contributions of value loss from such sources and understanding whether and how post-restatement stock returns are affected by investors’ initial reactions to restatements.¹ The sharp rise in financial restatements in the new millennium has shaken investor confidence in the corporate financial reporting system (GAO, 2002, 2006). The annual frequency of restatements by U.S. companies filing with the Securities and Exchange Commission (SEC) peaked at nearly 1,850 in 2006, declining thereafter to around 900 in 2012 (Murphy, 2014). Since the first listing of restatement firms published by the Government Accountability Office (GAO) in 2002 (GAO, 2002), academic research has provided consistent evidence that restatements result in a significant negative stock price reaction, on average. For instance, early work finds that short-window cumulative average abnormal returns surrounding the restatement announcement range from −6% (Dechow et al., 1996) to −10% (GAO, 2003; Palmrose et al., 2004). In subsequent studies, researchers have examined changes in specific components of the firm valuation model, that is, projections of the firm’s future cash flows or the discount rate used to evaluate the impact of restatements on firm value (Hribar & Jenkins, 2004). However, empirical evidence on the linkages from these valuation components to the announcement-related value change is scant.

The various factors that contribute to the change in firm value around the announcement of financial restatements can be grouped into two main sources. First, restatements affect investors’ priors about the firm’s ability to generate or sustain future cash flows. The resulting impact on value here could be positive or negative depending on the direction of the restatement. Second, restatements create incremental uncertainty about the quality of the firm’s earnings, which raises the discount rate.² Here, it is more likely that the impact on firm value is negative. Prior studies examine each of these causes on a standalone basis and provide some evidence that suggests they are both significant contributors to the decline in firm value. While it is intuitive that restatements influence each of the cash flow and discount rate components of the firm valuation function, what is missing in the literature is a joint examination of these components. Specifically, no prior study has examined the relative importance of the two sources of stock price declines at restatement announcements. In this study, we undertake such an examination by decomposing the total restatement-related value change that the firm experiences into a cash flow–related component and a discount rate–related component.

Based on the assumption that stock valuations reflect market efficiency, we decompose stock price changes around the restatement announcement into components attributed to revisions in expected cash flows and the discount rate. Cash flow changes are estimated as the changes in the analyst consensus estimate of earnings per share (EPS) across the restatement announcement. Using these estimates of cash flow changes, we estimate the implied cost of equity from the Claus and Thomas (2001) model, both before and after the restatement announcement.³ Following Chen et al. (2013) and Da et al. (2016), we use the estimated revisions in cash flows and discount rates in a variance decomposition framework that decomposes the total change in the stock price into components attributable to each of these factors. By construction, the variance decomposition constrains the total variation in stock prices to be apportioned to one of the two components. This variance decomposition is, in principle, similar to the return decomposition of Campbell and Shiller (1988a, 1988b) and Vuolteenaho (2002).⁴

We use a sample of over 1,800 restatements from 1995 to 2013 drawn from the Audit Analytics database for our analysis. We find that approximately 33.5% of the variation in stock price changes across restatements is attributable to revisions in the future earnings of the firm, with the remaining attributable to revisions in its implied cost of equity. The importance of cash flow news in explaining changes in valuation is much higher in the restatement sample compared with only 12.9% for the entire population around analyst’s monthly earnings forecast revisions. In addition, we also find that when the restatement involves a positive adjustment to cumulative earnings of prior periods (i.e., positive impact restatement subsample), revisions in cash flow explain 44.8% of the variation in the announcement return, but only 35.7% for the negative impact restatement subsample. We interpret this finding as being consistent with negative cash flow impacts having greater ramifications for firm uncertainty than positive cash flow impacts.

To better understand the relative importance of cash flow revisions versus discount rate revisions, we partition the restatement sample based on the underlying cause of the restatement: (a) accounting rule application failures, (b) financial fraud, (c) errors, and (d) other.⁵ Because accounting rule application failures occur in more than 90% of the cases, the relative importance of cash flow and discount rate components for this type of restatement is similar to that of the full sample. However, when the restatement is a result of an error, we find that cash flow revisions explain a relatively higher proportion of variation in price changes, at about 61%. This is consistent with the interpretation that investors indeed view such restatements as mainly arising out of errors and impose a smaller increase in the cost of capital of the firm.

By contrast, for restatements due to financial fraud, changes in the discount rate explain a significantly larger portion of the change in value (88%). This is consistent with the interpretation that investors view such restatements as a reflection of the poor governance of the firm and that the consequent revisions they make in firm uncertainty become the overwhelming factor in explaining the variation in price changes. As investors’ weighting of cash flow revisions (relative to discount rate revisions) is highly dependent on the persistence of earnings (Brav & Heaton, 2002; Francis et al., 2007), we also examine subsamples of firms with high and low earnings persistence, and find that for subsample of firms whose earnings persistence is lower, the discount rate–related component of the value loss is significantly larger (70.9%). Thus, our cross-sectional tests help to (a) quantify the overall relative importance of revisions in the cost of equity made by investors and (b) highlight the kinds of restatements where cash flow revisions have an above-average importance in explaining the variation in firm value changes.

We next examine how our decomposition of the value loss informs post-restatement announcement stock returns. On average, we do not find any evidence that stock prices for our sample firms drift systematically in the 1 year after the restatement when measured by risk-adjusted buy-and-hold returns (Daniel et al., 1997). However, we find a downward drift among the firms whose stock price decline around the restatement announcement is driven relatively more by discount rate revisions. These firms have a mean negative buy-and-hold abnormal return of approximately −4.7% in the subsequent 252-trading day period. By contrast, firms with above-median cash flow–related components of value loss remain relatively stable over the year, with a mean buy-and-hold abnormal return of about 0.9% at the end of the 252-day period. These results suggest that value losses from cash flow revisions are relatively immediate and complete compared with value losses from cost of equity revisions, which seem to be realized relatively more in the long term. Additional tests rule out the possibility that this pattern in the post-announcement returns of restating firms is driven by firm risk, book-to-market ratios, leverage, beta, momentum, or accruals.

The contribution of this article to the literature is twofold. First, the findings of this article extend the research that examines financial restatements. While prior work has examined earnings and discount rate impacts around restatements, the analysis of these factors has been performed separately with the result that evidence on their linkages to the decline in firm value is limited. Using a unified valuation framework that decomposes the change in value into discount rate and cash flow effects, we are able to go beyond the intuitive and well-documented result that the earnings response coefficient (ERC) will decrease (reflecting an increase in the discount rate), to quantifying the contribution of each component of the valuation function—cash flows and discount rates.

Second, our findings also shed incremental light on a fundamental question in the finance and accounting literature—whether stock prices are driven primarily by cash flow news or discount rate news (Campbell & Shiller, 1988a).⁶ In a summary of prior work, Cochrane (2011) concludes that much of the variation in stock prices is driven by discount rate news. In recent work, Chen et al. (2013) document the existence of a significant cash flow component in the drivers of stock price movements. We extend this literature by examining significant event-related movements in stock prices—financial restatements are widely acknowledged to affect both cash flow and discount rate aspects of valuation and are thus ideally suited for the purpose.

Our results should also be of interest to practitioners because the challenge to managers of affecting the cash flows of the firm is significantly different from the challenge of affecting the discount rate of the firm. Whereas the former is in large part a result of managerial effort, the latter is heavily influenced not only by investor perception of the business risk of the firm but also by investor perception of the quality of firm management and the risks thereof. Finally, our findings are of potential interest to investors as our variance decomposition helps explain differences in the post-announcement performance of restating firms. Our return results, suggesting a slower reaction to discount rate revisions than to cash flow revisions, help provide insight into the process by which investors revise their stock valuations around restatements. Combined with the proportion of announcement return variation explained by discount rate revisions, these results help quantify the importance of investors’ risk perceptions of firms surrounding earnings restatements.

Review of Related Work and Development of Hypotheses

The Short-Term Market Response to Restatements and Changes in Expected Cash Flows and Discount Rates

Prior studies provide consistent evidence of a significant negative stock market reaction to financial restatement announcements (e.g., Dechow et al., 1996; GAO 2003; Palmrose et al., 2004). The market’s reaction is attributable to its revisions of two main components of the firm valuation model, that is, projections of the firm’s future cash flows and the discount rate used to value these cash flows (Hribar & Jenkins, 2004). Because investors rely on accounting information to predict future cash flows, they are likely to adjust the future cash flow potential of the firm downward/upward when it restates its earnings and/or revises other related statements. Palmrose et al. (2004) and Akhigbe et al. (2005) document a significant downward revision by analysts of their earnings forecasts following restatement announcements. Although not the main focus of their study, Palmrose et al. (2004) document a positive univariate correlation between the revision in EPS forecasts and the announcement return.

Financial restatements are also likely to decrease the credibility of reported earnings. Hribar and Jenkins (2004) argue that this can happen due to “uncertainty regarding managerial competence and integrity, and perceptions about overall earnings quality.” Consistent with the reduced credibility of reported earnings, Anderson and Yohn (2002) and Wu (2003) show that the stock return response to earnings news is dampened, at least temporarily. More directly, Hribar and Jenkins (2004) find an increase of between 7% and 19% in the cost of capital of firms following restatements. However, Hribar and Jenkins (2004) do not examine the extent to which the change in cost of capital explains the stock price decline around restatements by econometrically relating the two. Using a discretionary information risk factor in a Fama and French (1993) model, Kravet and Shevlin (2010) also document that the cost of capital increases by 86 basis points. They go a step further and demonstrate a positive link between the increase in the cost of capital and the value change around the restatement announcement. However, they do not examine cash flow changes.

Although the above studies document changes in the cash flow and discount rate components of the firm valuation function individually, they do not examine the relative importance of these two components as sources of the stock price decline at restatement announcement. Therefore, it is difficult to gauge the extent to which the changes in stock value associated with restatements are driven by cash flow revisions versus discount flow revisions. Prior work on the drivers of stock price movements in general suggests that discount rates are overall a more significant determinant than cash flows (e.g., Cochrane, 2008). However, in the context of financial restatements, investors and analysts could make significant revisions to cash flow expectations. At the same time, restatements are also likely to lead to significant revisions in the cost of capital of the company. Therefore, it is not clear whether discount rates will remain a dominant factor in explaining stock price changes surrounding the restatement announcement. Therefore, we state our first testable hypothesis non-directionally, as follows:

Hypothesis 1: Changes in cash flow expectations and changes in the discount rate equally explain the variation in stock prices changes around restatement announcements.

Prior studies document that the reasons for the restatement also affect investors’ and analysts’ assessment of the scope and the impact of restatements on firm valuation (Palmrose et al., 2004; Wu, 2003). For example, Wu (2003) documents a larger negative stock price reaction to restatement news that involves admitted fraud (−23.0%) compared with restatements that do not involve fraud (−9.9%). To examine the impact of the underlying reason for the restatement, we use the rubric of “managerial competence and integrity” as suggested by Hribar and Jenkins (2004, p. 337). We expect that restatements arising from errors and lapses in technical knowledge fall on the competence end of the spectrum and are likely to elicit smaller revisions from investors in informational and disclosure-related uncertainty. On the contrary, restatements arising from deliberate wrongdoing by managers fall on the integrity end of the spectrum and are likely to cause investors to react more sharply, leading to larger revisions in the discount rate. Below, we discuss the specific categories of restatements with reference to this rubric.

Financial Fraud as the reason for the restatement refers to intentional misinterpretation of generally accepted accounting principles (GAAP) and represents the most egregious, integrity-related reasons for the restatement. Therefore, for this category, holding the extent of the cash flow revision constant, it is likely that there will be larger revisions in the firm’s cost of capital than cash flow revisions. On the contrary, where the reason for the restatement is Errors, we project that discount rate changes are going to be less drastic. This is because errors are clerical in nature or technical mistakes in material amount. Therefore, we expect ripple effects of such errors in the calculation of future cash flows but a relatively small impact on the underlying risks of the firm.

For Accounting Rule Application Failures as the reason for the restatement, we project that the magnitude of value loss due to cash flow revisions or discount rate changes will be somewhere between those of the Financial Fraud–related and Errors-related restatement types. This is because unlike errors, which are pure clerical or technical wrongdoings, Accounting Rule Application Failures–related restatements involve the management’s judgment and could reflect not only managers’ competence but also, in some cases, their intentions.⁷ Therefore, we conjecture that the magnitude of value loss due to cash flow or discount rate revisions will fall between those of the fraud-related and error-related restatements. Accordingly, we state our hypotheses as follows:

Hypothesis 2a: The importance of changes in cash flow expectation [changes in discount rate] decreases [increases] in explaining stock price changes surrounding restatement announcements when the restatement is financial fraud–related.

Hypothesis 2b: The importance of changes in cash flow expectation [changes in discount rate] increases [decreases] in explaining stock price changes surrounding restatement announcements when the restatement is associated with errors.

Hypothesis 2c: The importance of changes in cash flow expectation [changes in discount rate] in explaining stock price changes surrounding restatement announcements falls between those of fraud-related and error-related restatements when the restatement is initiated because of accounting rule application failures.

In addition to examining how the relative importance of cash flow and discount rate revisions varies with the underlying reason for the restatement, as a corollary to the above hypothesis, we examine this variation across subgroups formed on the basis of earnings persistence. Investors’ weighting of cash flow news (relative to news related to the firm’s discount rate) is highly dependent on the persistence of earnings (see, for example, Brav & Heaton, 2002; Francis et al., 2007). This characteristic of earnings is likely to be of particular salience at the time of financial restatements which introduce incremental uncertainty. In such circumstances, we expect that investors are likely to weight cash flow news more when cash flows (earnings) are more persistent.

Changes in Expected Cash Flows/Discount Rates and the Long-Term Market Response to Restatements

Prior studies that examine the market’s response to financial restatements, generally find that the short-term stock market response is negative (e.g., Chakravarthy et al., 2014; Drake et al., 2015; Files et al., 2009; Gordon et al., 2013; Lev et al., 2008; Myers et al., 2013; Palmrose et al., 2004). Under the efficient market hypothesis, the short-term market reaction to the announcement should be complete and efficient in the usual way, which would imply that variation in the drivers of restatement news (cash flow revisions vs. discount rate revisions) would constitute past information and should not affect the firm’s long-term stock returns. However, prior studies do not provide consistent evidence that the post-announcement returns of restating firms are non-anomalous. Badertscher et al. (2011) document that stock prices of restatement firms remain stable in the 180-day period after the initial negative reaction to the restatement announcement. On the contrary, Gleason et al. (2008) document a significant downward price drift of 10.3% in the 60 trading days after restatement. Similarly, Desai et al. (2006) document a downward price drift up to 6 months after the restatement announcement and find that this negative post-restatement return is associated with the level of short interest in the firm.⁸

We attempt to shed new light on the impact of financial restatements on long-term post-announcement returns by examining differences between subgroups formed on the basis of whether the initial change in stock valuation is driven relatively more by changes in future cash flow expectations or by changes in the firm’s discount rate. In addition to extending the prior studies discussed above, our post-restatement return tests are also related to prior work examining changes in the ERC of restating firms. Here, our studies are most similar in spirit to those in Chen et al. (2014) and Wilson (2008). Chen et al. (2014) find that firms with material restatement amounts exhibit a significant long-term decrease in ERC and document credibility concerns as a main factor contributing to this decline. In their setting, although risk is controlled for in the ERC equation, neither the value loss from the change in risk nor the value loss from cash flow revisions can be directly measured. On the contrary, Wilson (2008) documents that the drop in the information content of earnings is short-lived after restatements, including severe restatements. Her results suggest that firm ERCs exhibit a U-shaped pattern, suggesting that declines in ERC are only temporary.⁹

To the extent that there is significant variation in the relative impact of cash flow and discount rate factors in determining the announcement-related change in value, it is important to examine whether this variation has implications for the long-term returns of restating firms. Specifically, we examine post-restatement returns to better understand whether investors’ initial response to restatement announcements has been rational and to see whether it has any investment value. Because our variance decomposition framework isolates the impact of the discount rate component, it helps us directly address the issue of incremental uncertainty faced by investors at restatements and naturally links to the firm’s subsequent, long-term returns. Specifically, we examine whether there is any difference in post-restatement returns between subgroups formed on the basis of whether the initial stock price drop is driven relatively more by changes in future cash flow expectations or by changes in the discount rate. To the extent that investors take longer to ascertain and incorporate discount rate effects into stock valuations, we expect that for restatements where revisions in the discount rate are more salient in explaining the announcement-related change in value, post-announcement returns are more likely to be anomalous. Furthermore, the weight of prior evidence on post-restatement stock returns suggests that the post-announcement returns of discount rate–driven restatements will be anomalously negative. Based on this discussion, we state our final hypothesis as follows:

Hypothesis 3: Post-restatement stock returns for the subgroup where the initial stock price change is driven relatively more by changes in expectations of discount rate will be lower than for the subgroup where the initial stock price change is driven relatively more by changes in cash flow expectations.

Method

Following Chen et al. (2013) and Da et al. (2016), we adopt a variance decomposition framework to quantify the impact of the main drivers of the stock price movement immediately after a firm’s accounting restatement announcement: cash flow news and discount rate news.¹⁰ While analysts’ earnings and growth forecasts can be used to develop a proxy for investors’ cash flow expectations, the firm’s discount rate (or cost of equity capital) is not directly observable and is therefore model-dependent. To estimate the firm’s cost of equity capital, we first model its equity value as the present value of residual income and a terminal value as follows:

P_{t} = bvp s_{t} + \sum_{i = 1}^{T} \frac{E_{t} (R I_{t + i})}{{(1 + r)}^{i}} + \frac{E_{t} (TV)}{{(1 + r)}^{T}} .

(1)

In Equation 1, $P_{t}$ is the stock price per share at time t, $bvp s_{t}$ is the book value of equity per share at time t, $R I_{t + i}$ is the residual income for period t+i computed as $ep s_{t + i} - r * bvp s_{t + i - 1}$ (where $ep s_{t + i}$ denotes the EPS forecast for fiscal period t+i, and r denotes the discount rate), $TV$ is the terminal value, and E[.] is the expectations operator.¹¹ This model implicitly assumes that markets are efficient in that cash flows are rationally forecast into the future and the stock price is estimated purely as the sum of these cash flows discounted at a market-wide rate. Following Claus and Thomas (2001), we explicitly forecast future EPS up to t + 5 and compute the terminal value that captures all residual income beyond year t + 5, assuming that residual income growth occurs at the 10-year government bond yield less than 3%. We use analyst consensus forecasts of EPS from the I/B/E/S database for the current fiscal year (eps_t) and the next fiscal year (eps_t+₁), and earnings forecasts beyond year t + 2 are computed as follows:

ep s_{t + i} = ep s_{t + i - 1} \times (1 + lt g_{t}) for i = 3, 4, and 5,

(2)

where $(lt g_{t})$ is the mean long-term growth forecast, also drawn from I/B/E/S. At time t, all other factors except the cost of equity (r) are known, and r can be backed out from Equation 1.

We next model the decomposition of the price change around earnings restatements. From Equation 1, the stock price (P_t) is a function of cash flow (cf_t) and discount rate (dr_t):

P_{t} = f (c f_{t}, d r_{t}),

(3)

where cash flow (cf_t) is proxied by earnings forecasts ( $ep s_{t}$ ), and growth rate forecasts are given by ( $lt g_{t}$ ). By a simple algebraic transformation of Equation 2, we decompose the return ( $D P_{t, t + j})$ between t and t + j into a component related to cash flow news ( $CF N_{t, t + j}$ ) and a component related to discount rate news ( $DR N_{t, t + j}$ ). Specifically, $CFN$ is the change in value caused only by changes in future cash flow expectations, and DRN is the change in value caused only by changes in discount rates. Each component is measured as follows:

D P_{t, t + j} = \frac{P_{t + j -} P_{t}}{P_{t}}

(4)

= \frac{f (c f_{t + j}, d r_{t + j}) - f (c f_{t}, d r_{t})}{P_{t}}

(5)

= \frac{f (c f_{t + j}, d r_{t + j})}{P_{t}} + \frac{[f (c f_{t + j}, d r_{t}) - f (c f_{t + j}, d r_{t}) + f (c f_{t}, d r_{t + j}) - f (c f_{t}, d r_{t + j})]}{2 P_{t}} - \frac{f (c f_{t}, d r_{t})}{P_{t}}

(6)

= CF N_{t, t + j} + DR N_{t, t + j},

(7)

where

CF N_{t, t + j} = \frac{(f (c f_{t + j}, d r_{t + j}) - f (c f_{t}, d r_{t + j}) / P_{t}) + (f (c f_{t + j}, d r_{t}) - f (c f_{t}, d r_{t}) / P_{t})}{2},

(8)

DR N_{t, t + j} = \frac{(f (c f_{t}, d r_{t + j}) - f (c f_{t}, d r_{t}) / P_{t}) + (f (c f_{t + j}, d r_{t + j}) - f (c f_{t + j}, d r_{t}) / P_{t})}{2} .

(9)

This decomposition approach enables us to estimate the changes in stock prices due to the revision in one component by allowing the component to vary over time while holding the other component fixed. Thus, cash flow news ( $CF N_{t, t + j}$ ) is computed by holding the discount rate constant, and it captures the price change driven solely by changing cash flow expectations from t to t + j. Similarly, discount rate news ( $DR N_{t, t + j}$ ) is computed by holding the cash flow expectation constant, and it captures the price change driven solely by changing discount rate expectations from t to t + j. The decomposition equation (7) provides a convenient way to express the variance of stock price changes (or stock returns) as the sum of the two covariances:

Var (DP) = Cov [DP, CFN] + Cov [DP, DRN]

(10)

Dividing both sides of Equation 10 by $Var (DP)$ , we obtain the following equation:

1 = \frac{Cov [DP, CFN]}{Var (DP)} + \frac{Cov [DP, DRN]}{Var (DP)} .

(11)

Each term on the right-hand side of Equation 11 can be estimated by regressing $CFN$ and $DRN$ , respectively, on $DP$ . The slope coefficient of each regression is interpreted as the proportion of the total variation in price changes that is explained by variation in revisions of each component. By construction, the coefficients from each regression sum to 1. Vuolteenaho (2002) and Campbell and Shiller (1988a, 1988b) provide similar decompositions of the stock return as opposed to the stock price. More recently, He et al. (2019) use option market reactions to estimate the change in the implied volatility of the firm across the announcement of financial restatements associated with disclosure of a material weakness.

Data and Summary Statistics

We obtain restatement data from Audit Analytics, earnings and long-term growth rate forecasts from I/B/E/S, stock price data from CRSP, and accounting data from Compustat. We use the most recent consensus forecasts of annual EPS (eps_t) and long-term earnings growth rate ( $lt g_{t}$ ) available from I/B/E/S before the announcement date to estimate the pre-announcement implied cost of equity. Similarly, the corresponding forecasts that are available most immediately after the announcement are used to estimate the post-announcement implied cost of equity. We choose these forecast dates to ensure that we reliably attribute the change in the earnings forecast to the restatement. Accounting data are taken from the most recent quarterly statement in Compustat before the accounting restatement date t. Stock prices (P_t, P_t+1) from CRSP are taken 2 days after the date of the computation of the consensus earnings forecast (eps_t, eps_t+1) to ensure that information in the forecast is impounded in the price.

We identify 5,824 restatement announcements from Audit Analytics with non-missing stock price data in CRSP between 1995 and 2013. After dropping 655 observations with missing firm characteristics in Compustat and missing analyst forecast data from I/B/E/S (3,320 observations), our final restatement data set consists of 1,849 observations. Table 1 reports summary statistics for the full sample and subsamples based on the restatement’s cumulative impact on net income (NI). Cumulative Impact on NI > 0 [= 0, or < 0] restatement subsample includes any restatement whose impact on cumulative net income is positive [zero, or negative].

Table 1.

Descriptive Statistics.

Variables	Full sample			Cumulative Impact on NI > 0			Cumulative Impact on NI = 0			Cumulative Impact on NI < 0
	N	M	Median	N	M	Median	N	M	Median	N	M	Median
BP	1,849	0.557	0.485	237	0.590	0.514	627	0.573	0.497	985	0.539	0.472
MktCap	1,849	4,954	1,070	237	6,816	1,031	627	4,656	1,191	985	4,695	980
MOM	1,849	0.195	0.078	237	0.203	0.143	627	0.221	0.116	985	0.177	0.048
ROA	1,849	0.079	0.067	235	0.071	0.057	606	0.087	0.074	970	0.077	0.066
LTG	1,849	0.157	0.150	237	0.160	0.150	627	0.149	0.130	985	0.161	0.150
RestateSize	1,849	0.100	0.007	237	0.225	0.011	627	0.000	0.000	985	0.133	0.022
DFEPS	1,849	0.016	0.000	237	0.076	0.000	627	0.021	0.000	985	–0.002	0.000
DCOE	1,849	0.002	0.001	237	0.002	0.001	627	0.000	0.000	985	0.003	0.001
DP	1,849	–1.3%	–0.6%	237	–0.8%	–0.6%	627	0.9%	1.0%	985	–2.8%	–1.3%
DPCFN	1,849	1.1%	0.4%	237	3.3%	0.3%	627	1.1%	0.5%	985	0.7%	0.2%
DPDRN	1,849	–2.5%	–1.3%	237	–4.1%	–1.1%	627	–0.2%	0.5%	985	–3.5%	–2.5%

Note. The table reports summary statistics for the firms with restatement announcements from Audit Analytics database during 1995–2013. Cumulative Impact on NI > 0 [NI = 0, or NI < 0] restatement sample includes any restatement whose impact on cumulative NI is positive [negative or zero]. Variables are defined as follows. Book-to-market (BP) is the ratio of the book value of equity to market capitalization at each quarter-end. Market capitalization (MktCap) is the market capitalization in millions. The past return or momentum (MOM) is the 12-month cumulative return from month t – 12 to month t – 1. Return on assets (ROA) is quarterly earnings before extraordinary items divided by the total assets at the beginning of each quarter. Long-term growth rate (LTG) is analysts’ monthly consensus long-term earnings forecasts from I/B/E/S. RestateSize is total cumulative changes in net income divided by the book value of equity in absolute terms. Revisions in earnings forecasts (DFEPS) is measured between t − 1 and t + 1 dates, where t − 1 is the most recent earnings forecast consensus date right before the restatement announcement date, and t + 1 is the most immediate earnings forecast consensus date right after the restatement announcement date. Changes in cost of equity capital estimates (DCOE) is measured between the same t − 1 and t + 1 dates surrounding restatement announcements used in measuring DFEPS. Cost of equity capital measure is based on the method used in Claus and Thomas (2001). Stock price change (DP) is a log returns between the same t − 1 and t + 1 dates surrounding restatement announcements used in measuring DFEPS. Cash flow news (DP_CFN) is the stock return driven solely by changes in cash flow expectations from t − 1 to t + 1 surrounding restatement announcements. Discount rate news (DP_DRN) is the stock return driven solely by changes in cost of equity capital from t − 1 to t + 1 surrounding restatement announcements. NI = net income.

Statistics for the full sample show that restating firms are on average quite large with a mean (median) market capitalization of US$4.95 (US$1.07) billion. Firms are reasonably well performing with an average return on assets (ROA) of 7.9%, book-to-market ratio of 0.557 (BP), and long-term growth forecasts of 15.7%. Firm characteristics of the three restatement subsamples, Cumulative Impact on NI > 0 [=0, or < 0], are not very different when measured at the median (i.e., book-to-market ratios, market capitalization, ROA, and long-term growth expectations).¹²

Table 1 also reports restatement announcement stock returns, the revision in EPS forecasts, and the model-derived implied cost of equity capital. On average, financial analysts revise earnings up by about 1.6 cents (DFEPS), whereas the cost of equity capital increases by 20 basis points (DCOE). The combined effect of the increase in earnings and an increase in the cost of equity capital is that restating firms experience an average price decline of 1.3% (DP) in the 1-month window (t− 1 to t+ 1 month) surrounding the restatement announcement.

This negative return is mainly attributable to changes in the cost of equity capital, holding revisions in earnings and earnings growth constant; the average price change due to changes in the cost of equity capital (i.e., DP_DRN) is −2.5%.While analysts revise earnings downward on average, earnings revisions seem to have a slightly positive average impact on price changes (DP_CFN = 1.1%). Expectedly, the three subsamples based on the cumulative impact on net income show significantly different earnings revisions—the positive subsample has a mean revision of 7.6 cents upward compared with a mean downward revision of 0.2 cents in the negative subsample or upward revision of 2.1 cents in the zero subsample. The mean revisions in the positive and negative subsamples are statistically different at the 1% level, although at the median they are statistically indistinguishable. However, the changes in the cost of equity capital are more comparable across the two subsamples (positive vs. negative subsamples)—an average (median) upward move of 30 (10) basis points for the negative restatement subsample and 20 (10) basis points for the positive restatement subsample. The net impact of restatements for the negative subsample is a decline in stock price (DP) of 2.8% compared with a decrease in stock price for the positive subsample of 0.8%. The total stock price change attributable to discount rate revisions (DP_DRN = -3.5%) is larger than that attributable to future earnings expectations (DP_CFN = 0.7%) for the negative subsample. On the contrary, for the positive subsample, the stock price change surrounding the restatement is driven both by changes in future earnings expectations (DP_CFN = 3.3%) and by discount rate revisions (DP_DRN = −4.1%), but in opposite directions, resulting in a small net change in stock price. The firm characteristics for the subsample with zero cumulative impact on NI are mainly between those of the positive and negative cumulative impact subsamples.

Overall, these descriptive statistics show that the decrease in stock price following accounting restatement announcements is on average caused more by an increase in the cost of equity capital than by revisions in expectations of future cash flows. However, for the positive restatement subsample, the stock price increase is mainly driven by the revisions in expectations of future cash flows. In the following section, to understand the drivers of stock price changes surrounding the restatement announcement in a relative sense, we adopt the variance decomposition framework.

Results

Variance Decomposition—Full Sample Analysis

To examine the two main drivers of stock price movements around accounting restatements, we adopt the variance decomposition framework as detailed in section “Method.” The coefficient from a regression of cash flow news CFN_i,t (discount rate news DRN_i,t) on the stock price change across the restatement announcement (DP_i,t) measures the proportion of returns driven by changes in investors’ cash flow expectations (cost of equity capital).

We measure changes in stock price, earnings and earnings growth forecasts, and cost of equity capital between the closest I/B/E/S consensus forecast date right before (t − 1) and after (t + 1) the accounting restatement announcement date (t). As I/B/E/S consensus forecasts are available on a monthly basis, the time interval between t − 1and t + 1 is, on average, about 30 days, and the accounting restatement falls somewhere between the t − 1 and t + 1 consensus forecast dates.¹³ Table 2 reports the proportion of total price variation, Var(DP_i,t), that is explained by the covariance between DP_i,t and CFN_i,t and DRN_i,t, which are estimated as $Cov [DP, CFN] / Var (DP)$ and $Cov [DP, DRN] / Var (DP)$ , respectively. As a baseline case, we first conduct the variance decomposition exercise with the intersection of the entire universe of CRSP, I/B/E/S, and Compustat firms during the same sample period surrounding the analyst’s monthly earnings forecast revisions. For the entire population, using a 1-month revision interval, cash flow news explains 12.9% of stock price movements, and discount rate news explains the remaining 87.1% (Table 2, Panel A). The relatively small proportion of cash flow news is consistent with prior studies and the notion that short-term price movements can be driven by the investors’ sentiment unrelated to cash flow expectations (Cochrane, 2008, 2011). Compared with the entire population, the proportion of cash flow news explaining stock price variation is much higher for the restatement sample (33.5%).¹⁴ This difference captures the significance of the immediate impact of a firm’s restatement on investors’ expectations regarding future cash flows relative to investors’ assessment of the firm’s risk reflected in its cost of equity capital.

Table 2.

The Drivers of Stock Price Changes Following Accounting Restatement: Cash Flow News Versus Discount Rate News.

Sample	Observations	Coefficients
		CFN	DRN
Panel A: Full Sample
Population	821,395	0.129	0.871
		[2.72]	[18.36]
Matching Sample	1,849	0.218	0.782
		[7.44]	[26.74]
Restatement—Full Sample	1,849	0.335	0.665
		[8.78]	[17.44]
Panel B: Subsample based on restatement impact on net income
Cumulative Impact on NI > 0	237	0.448	0.552
		[3.99]	[4.92]
Cumulative Impact on NI = 0	627	0.240	0.760
		[3.51]	[11.10]
Cumulative Impact on NI < 0	985	0.357	0.643
		[6.95]	[12.52]

Note. The table reports the proportion of stock return variation that is explained by cash flow news (CFN) and discount rate news (DRN) following accounting restatements. Population includes all the firms from the intersection of CRSP, I/B/E/S, and Compustat universe from year 1995 to 2014. The Matching Sample includes non-restatement firms that have similar firm characteristics to those of restatement firms in the same industry. The matching sample was constructed based on the propensity score measured by the logistic prediction model of restatement versus non-restatement firms using determinant variables of the book-to-market, market capitalization, profitability (return on assets), and momentum (past 12-month returns) measures. Restatement—Full Sample includes all the firms with restatement announcements from Audit Analytics database with non-missing variables from CRSP, Compustat, and I/B/E/S during 1995–2013. Cumulative Impact on NI > 0 [NI = 0, or NI < 0] restatement sample includes any restatement whose impact on cumulative NI is positive [negative or zero]. Obs is the total number of observations. CFN is the stock return driven solely by changes in cash flow expectations from t − 1 to t + 1 surrounding restatement announcements. DRN is the stock return driven solely by changes in cost of equity capital from t − 1 to t + 1 surrounding restatement announcements. The CFN coefficient is estimated as $Cov [DP, CFN] / Var (DP)$ , where DP is the log return from the same t − 1 to t + 1 period surrounding the restatement announcements. Similarly, the DRN coefficient is estimated as $Cov [DP, DRN] / Var (DP)$ . t-statistics are reported in brackets. NI = net income.

We also conduct the variance decomposition exercise with a matching sample surrounding pseudo-restatement dates. Using the intersection of all non-restatement firms in Compustat, CRSP, and I/B/E/S, we first measure each firm’s propensity score by running a logistic regression by industry with market capitalization, market-to-book ratio, ROA, and past 12-month stock returns all measured in year t − 1. For each restatement firm, we find a matched firm with the closest propensity score in the same industry and in the same year. The matched sample is created without replacement.¹⁵ By this matching process, we create a pseudo-announcement date for the matched firm that mirrors the restatement date of a restatement firm. Using the same decomposition framework, for the matching sample, we find that cash flow news explains 21.8% of stock price movements, significantly less than for the restatement sample.¹⁶

We next partition our sample on the basis of the cumulative impact of the restatement on the net income of the firm (Table 2, Panel B).¹⁷ By doing this, we are able to control for the cash flow impact of the restatement on an ex ante basis and then quantify the extent to which changes in discount rate expectations explain the change in value due to the restatement. Consistent with prior work, we find that the proportion of income-increasing restatements is quite small (less than 13%). More than 53% of the restatements are income-decreasing, and the rest have no cumulative impact on net income. We find that CFN explains a larger part of the restatement-related stock price change for the firms where the restatement has some impact on net income (our proxy for cash flows). Specifically, CFN explains 44.8% [35.7%] of the variation for the subsample where the restatement is income-increasing [income-decreasing] compared with only 24.0% for the subsample with the restatement that has no impact on net income.¹⁸ Overall, these results provide greater insight into the relative power of these two factors in explaining restatement-related value changes.

Variance Decomposition and Types of Restatements

Audit Analytics documents more than 60 different reasons for firms to restate their financial statements. These can be grouped into four broad categories: Restatements due to (a) accounting rule application failures (ARAF), (b) Errors, (c) Financial Fraud (Fraud), and (d) Others.¹⁹ In Panel A of Table 3, we report how restatement type affects the variance decomposition. ARAF occur in more than 93% (1,725) of the cases, Errors in about 6.9% (128), Financial Fraud in 2.5% (47), and Others in 10.9% (202) of our sample.²⁰ Because ARAF includes more than 90% of total restatements, the relative importance of CFN (31.4%) and DRN (68.6%) in explaining stock return variation is very similar to that of the full sample, and consistent with our hypothesis, the magnitude of the relative importance of CFN for the ARAF subsample is between those of the Fraud and Errors restatement subsamples. We also find that restatements that fall into the Others category show a similar pattern to that of the overall restatement sample. However, when errors (Errors) lead to financial restatements, changes in stock value are driven to a larger extent by future cash flow expectations (CFN coefficient = 0.606) than changes in discount rate (DRN coefficient = 0.394). We find that this subgroup has the highest proportion of cash flow news–related variation across the various subsamples examined in our analysis. This is consistent with investors viewing such restatements as being less indicative of management weakness than other types of restatements. On the contrary, when restatements are associated with financial fraud, stock price changes are mainly driven by changes in the discount rate—for this subsample, discount rate revisions explain 88.0% of the total stock return variation. The difference in the coefficient for DRN (or CFN) between the Fraud and Errors subsample is statistically significant at least at the 10% level. This suggests that when investors’ confidence in a firm’s financial reporting system is shaken, the cost of equity (or discount rate) explains more of the return variation than revisions in the firm’s expected cash flows. The results of Table 3 support our second set of hypotheses that for restatements that are associated with fraud, the change in value is driven to a greater extent by revisions in the discount rate of the firm than the restatements associated with errors.

Table 3.

The Drivers of Stock Price Changes Following Accounting Restatement by Restatement Types.

		Coefficients
Sample	Observation	CFN	DRN
Panel A: Audit Analytics–based reasons for the restatement
Restatement—Full Sample	1,849	0.335	0.665
		[8.78]	[17.44]
Accounting Rule Application Failure	1,725	0.314	0.686
		[7.83]	[17.12]
Errors	128	0.606	0.394
		[3.62]	[2.36]
Financial Fraud	47	0.120	0.880
		[0.27]	[1.96]
Others	202	0.378	0.622
		[3.25]	[5.35]
Panel B: Restatements based on firms’ earnings persistence
Restatement—Full Sample	1,849	0.335	0.665
		[8.78]	[17.44]
Low Persistence	548	0.291	0.709
		[3.45]	[8.42]
High Persistence	547	0.432	0.568
		[6.63]	[8.72]

Note. The table reports the proportion of stock return variation that is explained by cash flow news (CFN) and discount rate news (DRN) following accounting restatement by restatement types (Panel A) and by earnings persistence (Panel B). Restatement—Full Sample includes all the firms with restatement announcements from Audit Analytics database with non-missing variables from CRSP, Compustat, and I/B/E/S during 1995–2013. Accounting Rule Application Failures is a subsample of firms whose restatement is due to misapplication of accounting rules. Financial Fraud is a subsample of firms whose restatement is caused by financial fraud. Errors is a subsample of firms whose restatement is due to a simple error. Others include all other restatements. Obs is the total number of observations. CFN is the stock return driven solely by changes in cash flow expectations from t − 1 to t + 1 surrounding restatement announcements. DRN is the stock return driven solely by changes in cost of equity capital from t − 1 to t + 1 surrounding restatement announcements. The CFN coefficient is estimated as $Cov [DP, CFN] / Var (DP)$ , where DP is the log return from the same t − 1 to t + 1 period surrounding the restatement announcements. Similarly, the DRN coefficient is estimated as $Cov [DP, DRN] / Var (DP)$ . t-statistics are reported in brackets. Earnings persistence is measured as the coefficient b₁ from a time-series regression of ROA(t) = b₀+b₁*ROA(t− 1) at the firm level using the past 6 years of ROA with a minimum of 4 years of data where ROA is measure net income divided by total assets.

In Panel B of Table 3, we examine how our decomposition varies across subgroups formed on the basis of earnings persistence. Brav and Heaton (2002) and Francis et al. (2007) argue that Bayesian investors will place more weight on less uncertain earnings information. Therefore, we conjecture that investors will place more weight on earnings information for firms with higher earnings persistence. Our aim here is to extend the evidence of the rational variation in the decomposition results of Panel A above. For this test, we drop restatements with no earnings impact and split the sample on the basis of subsample median value of earnings persistence. Earnings persistence is measured as the coefficient b₁ from the time-series regression ROA(t) = b₀+b₁*ROA(t− 1) +ϵ at the firm level using the past 6 years of ROA with a minimum of 4 years of data, where ROA is measured as the net income divided by total assets. As the results in Panel B of Table 3 show, firms with low earnings persistence have 29.1% of the change in value explained by cash flow news, whereas for firms with high earnings persistence this is 43.2%. However, these are not statistically different (p = .169).

Post-Restatement Announcement Stock Returns

We next turn to an examination of how the variance decomposition of stock price changes around the restatement announcement informs post-announcement stock returns. Table 4 reports post-announcement abnormal returns for the full sample and by restatement type: High versus Low CFN/DP restatements based on the median breakpoint. As a baseline case, we report raw returns in Panel A. We find that post-announcement returns are significantly positive for the whole sample as well as for each of the Low and High subgroups. Furthermore, we find that the High subgroup has significantly higher returns than the Low subgroup in the 126- and 252-day trading periods (p < .10).

Table 4.

Post-Restatement Buy and Hold Return on Full Sample.

Return periods	Full sample	Low CFN/DP restatement	High CFN/DP restatement	High–Low
Panel A: Raw returns
Raw Ret (−1,1)	−0.015***	−0.017***	−0.014***	0.002
Raw Ret (2,63)	0.033***	0.023**	0.044***	0.021*
Raw Ret (2,126)	0.053***	0.036***	0.069***	0.034*
Raw Ret (2,252)	0.123***	0.092***	0.154***	0.063**
Panel B: Abnormal returns (Raw returns minus DGTW(1997) characteristics portfolio returns)
BHAR (−1,1)	−0.018***	−0.018***	−0.017***	0.001
BHAR (2,63)	−0.001	−0.009	0.007	0.016
BHAR (2,126)	−0.014*	−0.028**	−0.001	0.027*
BHAR (2,252)	−0.019	−0.047***	0.009	0.055**

Note. The table reports raw returns in Panel A, and buy-and-hold abnormal returns (BHAR) in Panel B from 1 day before the restatement date to 252 trading days after the restatement by the magnitude of cash flow news (CFN) relative to stock price changes (High vs. Low) in quarterly intervals. The High [Low] CFN/DP sample includes any restatement firms with high [below] median of CFN/DP between t − 1 to t + 1 dates, where t − 1 is the most recent earnings forecast consensus date right before the restatement announcement date and t + 1 is the most immediate earnings forecast consensus date right after the restatement announcement date. BHARs are calculated during each trading period we choose by subtracting the corresponding 125 characteristics-based benchmark portfolio returns based on the size, book-to-market ratio, and momentum portfolio following Daniel et al. (1997) from the raw returns. High—Low column reports the difference in BHAR between the High and Low CFN/DP samples. *, **, and *** represent the significance levels at 5%, 1%, and 0.1%, respectively.

To compute abnormal returns, we subtract characteristics-based benchmark portfolio returns from raw returns following Daniel et al. (1997). After adjusting for risk and other pricing factors, the mean buy-and-hold abnormal return after a year for the full sample is −1.9% (Panel B, Table 4). However, while both the High and Low CFN/DP restatement subsamples have an initial negative return similar to that of the full sample, post-announcement returns of the two subsamples show marked differences. The Low CFN/DP restatement firms continue to drift downward, and their mean buy-and-hold abnormal return is −4.7% 252 days after the restatement announcement date. On the contrary, the average buy-and-hold return for the High CFN/DP restatement subgroup remains relatively stable at 0.9% (insignificant) with no significant peaks or dips in the intervening period. The return difference between the High and Low CFN/DP restatement subsamples is 5.5% and is statistically significant (p<.05), which supports our third hypothesis.²¹

The result that post-announcement returns are systematically linked to the variation in cash flow revisions versus discount rate revisions provides several insights. At a descriptive level, we do not find evidence that post-announcement returns are consistent with an overall underreaction to restatement announcements, which is similar to findings in Badertscher et al. (2011). However, the extent of underreaction is significantly more pronounced when the initial reaction is driven to a greater extent by changes in investors’ cost of equity expectations. For restatements characterized by a larger content of cash flow news, post-restatement returns suggest the immediate market reaction is relatively complete and therefore efficient. These regularities are consistent with the argument that investors are able to estimate the value impact of shocks to earnings relatively quicker than the value impact of changes in the firm’s discount rate. Thus, by tying long-term returns to the drivers of short-term stock price movements, our results provide a more comprehensive analysis of the investor reaction to restatement announcements.

Robustness Tests

We conduct additional tests to verify whether our results documented above are sensitive to key aspects of our test design. We consider two reasons that could lead to a mis-estimation of the portion of stock price variation explained by discount rate news. First, the discount rate is unobservable and is estimated on the basis of a catch-all measure. Specifically, because the discount rate is derived from the stock price and earnings and earnings growth forecasts, noise in the stock price or biases in earnings and earnings growth forecasts are also likely to be captured in the discount rate and innovations thereof. This is unlike innovations in cash flows which are estimated more directly from analyst consensus forecasts. Second, our valuation model (as shown in Equation 1) and variance decomposition (as shown in Equation 7) are based on the assumption that analysts’ earnings forecasts accurately reflect investors’ expectations of earnings in a timely fashion. However, analysts’ delayed revisions of earnings forecasts can lead to an underestimation of the impact of cash flow news.

To reduce the likelihood of stale forecasts in the post-announcement period, we first extend the measurement of cash flow expectations after the restatement announcement by another month (so to t + 2). We report the results based on this alternative measurement in Panel A of Table 5. Second, we delete any observations that have earnings forecast revisions of zero value and report the results in Table 5, Panel B. Finally, recognizing that our results are model-dependent, in Panel C of Table 5, we change the valuation model that serves as the basis for the variance decomposition.

Table 5.

The Drivers of Stock Price Changes Following Accounting Restatement: Cash Flow News Versus Discount Rate News—Additional Tests.

		Coefficients
Sample	Observation	CFN	DRN
Panel A: The revision horizon t− 1 to t+ 2
Restatement—Full Sample	1,786	0.280	0.720
		[8.30]	[21.35]
Cumulative Impact on NI > 0	227	0.485	0.515
		[3.53]	[3.75]
Cumulative Impact on NI = 0	617	0.186	0.814
		[2.89]	[12.69]
Cumulative Impact on NI < 0	942	0.293	0.707
		[7.37]	[17.81]
Panel B: Non-zero forecast revisions
Restatement—Full Sample	1,350	0.385	0.615
		[7.97]	[12.74]
Cumulative Impact on NI > 0	175	0.530	0.470
		[3.85]	[3.41]
Cumulative Impact on NI = 0	458	0.295	0.705
		[3.30]	[7.89]
Cumulative Impact on NI < 0	717	0.399	0.601
		[6.19]	[9.34]
Panel C: Alternative valuation model (Pástor et al., 2008)
Restatement—Full Sample	1,828	0.311	0.689
		[6.10]	[13.51]
Cumulative Impact on NI > 0	231	0.479	0.522
		[4.04]	[4.41]
Cumulative Impact on NI = 0	617	0.054	0.946
		[0.42]	[7.33]
Cumulative Impact on NI < 0	980	0.389	0.611
		[7.09]	[11.13]
Panel D: Restatement announcements within (t− 1, t+ 1) days of earnings announcement date
Restatements Without Earnings Announcements	1,716	0.329	0.671
		[8.32]	[16.98]
Restatements With Earnings Announcements	133	0.445	0.555
		[3.03]	[3.78]
Panel E: The use of GAO restatement data
GAO Full Sample	854	0.305	0.695
		[5.07]	[11.55]
Restatement—Revenue/Expense	543	0.419	0.581
		[4.90]	[6.80]
Restatement—Other	311	0.102	0.898
		[1.48]	[13.03]

Note. The table reports the proportion of stock return variation that is explained by cash flow news (CFN) and discount rate news (DRN) following accounting restatement. Panel A reports the results with extended revision horizon from t− 1 to t+ 2. Panel B reports the results after deleting firms whose earnings forecasts that financial analysts do not revise following accounting restatements. Panel C reports the results based on the valuation model used in Pástor et al. (2008). Panel D reports results of subsample analysis portioned by the restatements that are made with and without earnings announcements. Panel E reports results based on the GAO restatement data. When the restatement involves prior period adjustment related to revenue or expenses, the restatement is classified as Restatement—Revenue/Expense subsample, otherwise Restatement—Other subsample. Population includes all the firms from the intersection of CRSP, I/B/E/S, and Compustat universe from year 1995 to 2013. Restatement—Full Sample includes all the firms with restatement announcements from Audit Analytics database during 1995–2013. Cumulative Impact on NI > 0 [NI = 0, or NI < 0] restatement sample includes any restatement whose impact on cumulative net income is positive [negative or zero]. Obs is the total number of observations. CFN is the stock return driven solely by changes in cash flow expectations from t − 1 to t + 1 [or t+ 2] surrounding restatement announcements. Discount rate news (DRN) is the stock return driven solely by changes in cost of equity capital from t − 1 to t + 1 [or t + 2] surrounding restatement announcements. The CFN coefficient is estimated as $Cov [DP, CFN] / Var (DP)$ , where DP is the log return from the same t − 1 to t + 1 period surrounding the restatement announcements. Similarly, the DRN coefficient is estimated as $Cov [DP, DRN] / Var (DP)$ . t-statistics are reported in brackets. GAO = Government Accountability Office.

Extension of the Consensus Forecast Revision Period

When we extend the revision period to 2 months to mitigate the stale forecast issue, for the overall population (baseline case) the importance of CFN in explaining stock price variation increases from 12.9% to 22.4% (not tabulated). This is consistent with the notion that while in the short term stock price changes can be driven relatively more by sentiment or other factors unrelated to firm fundamentals, over the long term stock returns will eventually be more closely tied to expected changes in future cash flows (Chen et al., 2013; Da et al., 2016; Easton et al., 1992). However, in the restatement sample, we do not observe an increase in the coefficient for CFN; rather, it decreases slightly. For the whole restatement sample, cash flow news explains only 28.0% of return variation compared with 33.5% with the 1-month forecast revision. The pattern of results for the negative, positive, and no-impact on net income subsamples is qualitatively similar to that reported in Table 2. These results suggest that our proxy of cash flow news (earnings forecast revisions) does not suffer from a significant timing issue.

Removal of Potentially Stale Forecasts

There are 499 observations for which the revision in the consensus analyst forecast from t − 1 to t + 1 is zero. All else equal, these are observations that are more likely susceptible to the stale forecast issue. In Panel B of Table 5, we report our main variance decomposition results after dropping these observations from the analysis. Expectedly, when we drop the stale forecasts, the proportion of CFN explaining stock price variation is somewhat higher. For example, in the full sample, the coefficient for CFN is 38.5% compared with 33.5% in Table 2. The pattern of results for the negative, positive, and no-impact on net income subsamples is also qualitatively similar to that reported in Table 2. These results suggest that while stale forecasts have the effect of inflating the impact of DRN in explaining stock price variation, our conclusions are not driven by this effect.

Alternative Equity Valuation Model

In estimating the discount rate and conducting variance decomposition, we rely on the valuation model proposed by Claus and Thomas (2001). To test whether our results are sensitive to our choice of the valuation model, we adopt an alternative valuation model used in Pástor et al. (2008) as follows:

P_{t} = \sum_{i = 1}^{T} \frac{E_{t} (R I_{t + i})}{{(1 + r)}^{i}} + \frac{E_{t} (TV)}{{(1 + r)}^{T}},

(12)

where all variables are defined as in Equation 1 except that Pástor et al. (2008) use a 15-year horizon (T = 15). Results in Panel C of Table 5 are qualitatively very similar to our main findings reported in Table 2. For example, the coefficient for CFN for the restatement sample is 31.1% with the model used in Pástor et al. (2008) compared with 33.5% with our main model.²²

Exclusion of Restatement Announcements Bundled With Earnings Announcements

There are 133 restatement announcements made within the 3-day window around an earnings announcement. Because this may affect our variance decomposition results, in Panel D of Table 5 we report our main variance decomposition results after dropping these restatements. The exclusion of these observations does not yield a major difference in the proportion of CFN explaining stock price variation. The coefficient for CFN in this subsample is 32.9% compared with 33.5% in Table 2. Interestingly, for firms that have restatements bundled with earnings announcements, the fraction of value changes explained by cash flow news is much higher at 44.5%.

Using Alternative Data Sources of Restatements—GAO

Our analyses are based on the Audit Analytics database which provides more recent restatement information. As a robustness test, we replicate our main variance decomposition analyses using the restatements obtained from the GAO database, which covers years from 1997 to 2006. Compared with Audit Analytics which records 1,849 restatements, there are only 854 restatements in the GAO database (about 46% of Audit Analytics data) with more of the observations from years prior to 2002. About 407 firms show up in both databases. We find that the results are consistent across the two samples. Results reported in Table 5, Panel E show that the impact of CFN [DRN] for the GAO full sample is 30.5% [69.5%], comparable to the findings in Table 2. In addition, GAO classifies any restatements that require adjustments to prior period revenues or expenses as “Revenue/Expense-related restatement.” Out of 854 restatements, GAO reports 543 (64%) restatements as revenue/expense-related restatements. As expected, for the revenue/expense-related restatement subsample, the importance of CFN is greater than that of DRN, with CFN explaining 41.9% of stock price changes versus a mere 10.2% for the remaining firms. The difference is statistically significant at less than 5%. These results suggest that when a firm makes a restatement announcement that affects its revenue or expenses, the market’s confidence in the firm’s ability to generate future cash flows deteriorates, and consequently, the stock price variation is driven more by expectation changes in future cash flow.

Conclusion

This article examines the fundamental drivers of stock price changes around financial restatements—investors’ revisions in the firm’s cash flow expectations and implied cost of equity capital. Despite extensive research on financial restatements, the evidence on the links between these drivers of price movements and the price decline around restatements is relatively sparse. Our findings address this gap in the literature by undertaking a joint analysis of these drivers within a single valuation model. Using a variance decomposition framework, we find that revisions in the discount rate are generally a bigger driver of the short-term price reaction to financial restatements. However, the proportion of stock price variation explained by cash flow news is higher for restating firms than for the population or in a propensity score-matched sample. Consistent with a financial reporting risk argument, we find that discount rate news is more important when the underlying reason for the restatement is fraud, whereas cash flow news is more important when errors are the reason for the restatement. These results make contributions to the literature on financial restatements as well as to the broader literature on the drivers of stock price movements.

We find that our decomposition also informs post-announcement stock returns. For restatements where the importance of discount rate changes is relatively higher, we find a significant negative drift in stock prices. No drift pattern is obtained for restatements whose announcement price changes are mainly explained by revision in future earnings. These findings help shed additional light on the link between investors’ revaluation of stocks around restatements and suggest that revisions in the financial reporting risk of the company are relatively long-lived.

Two important caveats are in order. First, our examination is a joint test of the relative importance of cash flows versus discount rates as value drivers and the valuation model that is employed. While we attempt to mitigate this issue by employing different valuation models, commonalities in the features of these models naturally limit the extent to which our examination can be model-free. To that extent, our results are to be interpreted with caution. In this regard, it is useful to note that our tests of post-announcement returns do not depend parametrically on the decomposition. Second, because our study requires the market’s earnings expectation for restatement firms, a large number of observations with missing I/B/E/S variables are dropped out of the sample. Therefore, even though we have a large sample of restatement firms, the inferences we draw from our study may not be generalizable to all firms covered by I/B/E/S.

Footnotes

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: Philip K. Hong is grateful for the financial support from the Childress-Klein Faculty Scholars Program through the Belk College of Business, University of North Carolina at Charlotte. Sukesh Patro is grateful for financial support from the Summer Research Grant Program at the College of Business, Northern Illinois University.

ORCID iD

Sang-Hyun (Sam) Park

Notes

References

Akhigbe

Kudla

R. J.

Madura

(2005). Why are some corporate earnings restatements more damaging? Applied Financial Economics, 15(5), 327-336.

Anderson

Yohn

(2002). The effect of 10-K restatements on firm value, information asymmetries, and investors' reliance on earnings [Working paper]. Georgetown University.

Baber

W. R.

Gore

A. K.

Rich

K. T.

Zhang

J. X.

(2013). Accounting restatements, governance and municipal debt financing. Journal of Accounting and Economics, 56(2), 212-227.

Badertscher

B. A.

Hribar

S. P.

Jenkins

N. T.

(2011). Informed trading and the market reaction to accounting restatements. The Accounting Review, 86(5), 1519-1547.

Brav

Heaton

J. B.

(2002). Competing theories of financial anomalies. Review of Financial Studies, 15(2), 575-606.

Campbell

Shiller

(1988a). The dividend yield and expectations of future dividends and discount factors. The Review of Financial Studies, 1(3), 195-228.

Campbell

Shiller

(1988b). Stock Prices, earnings, and expected dividends. The Journal of Finance, 43(3), 661-676.

Chakravarthy

DeHaan

Rajgopal

(2014). Reputation repair after a serious restatement. The Accounting Review, 89(4), 1329-1363.

Chen

Zhao

(2013). What drives stock price movement? The Review of Financial Studies, 26, 841-876.

10.

Chen

Zhao

(2009). Return decomposition. The Review of Financial Studies, 22(12), 5213-5249.

11.

Chen

Cheng

A. K.

(2014). Is the decline in the information content of earnings following restatements short-lived? The Accounting Review, 89, 177-207.

12.

Claus

Thomas

(2001). Equity premia as low as three percent? Evidence from analysts' earnings forecasts for domestic and international stock markets. The Journal of Finance, 56(5), 1629-1666.

13.

Cochrane

J. H.

(2008). The dog that did not bark: A defense of return predictability. Review of Financial Studies, 21(4), 1533-1575.

14.

Cochrane

J. H.

(2011). Presidential address: Discount rates. The Journal of Finance, 66(4), 1047-1108.

15.

Hong

Lee

(2016). What drives target price forecasts and their investment value? Journal of Business Finance & Accounting, 43(3–4), 487-510.

16.

Daniel

Grinblatt

Titman

Wermers

(1997). Measuring mutual fund performance with characteristic-based benchmarks. The Journal of Finance, 52(3), 1035-1058.

17.

Dechow

P. M.

Sloan

R. G.

Sweeney

A. P.

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

18.

Desai

Krishnamurthy

Venkataraman

(2006). Do short sellers target firms with poor earnings quality? Evidence from earnings restatements. Review of Accounting Studies, 11(1), 71-90.

19.

Drake

M. S.

Myers

L. A.

Scholz

Sharp

N. Y.

(2015). Short selling around restatement announcements when do bears pounce? Journal of Accounting, Auditing & Finance, 30(2), 218-245.

20.

Easton

Harris

Ohlson

(1992). Aggregate accounting earnings can explain most of security returns: The case of long return intervals. Journal of Accounting and Economics, 15(2–3), 119-142.

21.

Fama

E. F.

French

K. R.

(1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33(1), 3-56.

22.

Files

Swanson

E. P.

Tse

(2009). Stealth disclosure of accounting restatements. The Accounting Review, 84(5), 1495-1520.

23.

Francis

LaFond

Olsson

Schipper

(2007). Information uncertainty and post-earnings-announcement-drift. Journal of Business Finance & Accounting, 34(3–4), 403-433.

24.

Gleason

C. A.

Jenkins

N. T.

Johnson

W. B.

(2008). The contagion effects of accounting restatements. The Accounting Review, 83(1), 83-110.

25.

Gordon

E. A.

Henry

Peytcheva

Sun

(2013). Discretionary disclosure and the market reaction to restatements. Review of Quantitative Finance and Accounting, 41(1), 75-110.

26.

Government Accountability Office. (2002). Financial statement restatements: Trends, market impacts, regulatory responses, and remaining challenges (GAO-03-138). https://www.gao.gov/products/GAO-03-138

27.

Government Accountability Office. (2003). Financial restatement database (GAO-03- 395R). https://www.govinfo.gov/content/pkg/GAOREPORTS-GAO-06-1053R/html/GAOREPORTS-GAO-06-1053R.htm

28.

Government Accountability Office. (2006). Financial restatements: Update of public company trends, market impacts, and regulatory enforcement activities (GAO06-678). https://digital.library.unt.edu/ark:/67531/metadc301103/

29.

Bharath

Wans

(2019). Material weakness disclosures and restatement announcements: The joint and order effects. Journal of Business Finance & Accounting, 46(1–2), 68-104.

30.

Hribar

Jenkins

N. T.

(2004). The effect of accounting restatements on earnings revisions and the estimated cost of capital. Review of Accounting Studies, 9(2–3), 337-356.

31.

Kravet

Shevlin

(2010). Accounting restatements and information risk. Review of Accounting Studies, 15(20), 264-294.

32.

Lev

Ryan

S. G.

(2008). Rewriting earnings history. Review of Accounting Studies, 13(4), 419-451.

33.

Murphy

(2014, November 4). Restatements fall on improved internal controls. The Wall Street Journal. https://huntrisksolutions.com/restatements-fall-on-improved-internal-controls/.

34.

Myers

L. A.

Scholz

Sharp

N. Y.

(2013). Restating under the radar? Determinants of restatement disclosure choices and the related market reactions [Working paper]. University of Arkansas.

35.

Ohlson

(1995). Earnings, book values, and dividends in equity valuation. Contemporary Accounting Research, 11(2), 661-687.

36.

Palmrose

Z. V.

Richardson

V. J.

Scholz

(2004). Determinants of market reactions to restatement announcements. Journal of Accounting and Economics, 37(1), 59-89.

37.

Pástor

L'.

Sinha

Swaminathan

. (2008). Estimating the intertemporal risk–return tradeoff using the implied cost of capital. The Journal of Finance, 63(6), 2859-2897.

38.

Vuolteenaho

(2002). What drives firm-level stock returns? The Journal of Finance, 57(1), 233-264.

39.

Wilson

W. M.

(2008). An empirical analysis of the decline in the information content of earnings following restatements. The Accounting Review, 83(2), 519-548.

40.

(2003). Earnings restatements: A capital markets perspective [Working paper]. Hong Kong University of Science and Technology.