1 Introduction

Restatement of financial statements and its consequences are becoming an important issue among the investors, corporate managers, regulators, and auditing firms, particularly in the aftermath of the Sarbanes–Oxley Act (SOX) of 2002. Investors and regulators are concerned over restatements to correct non-GAAP accounting in previously issued financial statements. A Governmental Accounting Office (GAO) report published in 2003, documents a marked increase in the total number of restatements over the period from 1997 to 2002 and a detrimental impact on financial markets from such restatements as evident by an estimated total loss of $100 billion in market capitalization in the period immediately following the restatement announcement.

When firms restate their financial statements, investors reassess their perceptions pertaining to the quality of financial information about the firms. Investor reaction to such restatements may be mixed. On one hand, they may view a restatement announcement translating into improved financial reporting quality as it corrects the past errors. On the other hand, it is also reasonable to expect that the restatement increases information asymmetry that may arise from the questionable integrity of the management, from the lack of reliability of financial reporting, and from uncertain viability of the firm. Thus restatements may be viewed as a signal for poor quality of information because the investors now believe that not only the past accounting information was of low quality, but future information may be equally suspect.

In this paper, we study the impact of financial restatement announcements on information asymmetry and market microstructure activities in the stock market. We use a sample of restatement announcements from 1997 to 2005 collected from GAO Financial Statement Restatement Database.Footnote 1 We investigate information asymmetry around the announcement dates by examining various trading activities, such as, stock returns, trading volume, number of transactions, volatility, order size, and spreads in both NYSE-AMEX and NASDAQ markets. We document economically and statistically negative mean cumulative abnormal returns around the announcement dates. The restatements attributed to auditors are associated with more negative returns than those attributed to management and the SEC. However, we do not find any significant difference between the abnormal returns resulting from restatements associated with core and non-core accounts. This implies that the market reacts to the core restatement announcements in a manner similar to the non-core restatement announcements. The abnormal return results are consistent with the notion that restatements tend to diminish company prospects and increase uncertainty.

Next, we examine the time series of changes in volume, number of transactions, average order size, volatility, and various measures of spreads around the restatement announcements. Prior to the announcement, we find little evidence of increased spreads and of changes in other trading metrics. However, post-announcement, we find a significant increase in the trading metrics. The increase in volume or number of transactions could be an increase in adverse trades because informed investors exploit their informational advantage. Easley and O’Hara (1987) suggest that informed traders choose to trade larger amounts. Since order size (average trading volume per transaction) serves as a proxy for informed trading, an increase in order size should be usually observed around the event date. We document a significant increase in the abnormal order size after the restatement announcement, which is consistent with Easley and O’Hara (1987)‘s model.

Further, we find a significant increase in both abnormal relative quoted spread and abnormal relative effective spread after the announcement. A restatement is an indication to investors that managers are not credible, and that the public financial information provided by managers is less than reliable. Therefore investors may be more skeptical about the reliability of the firm’s past financial statements as well as the firm’s future financial information yet to be released. The investor skepticism tends to heighten uncertainty and information asymmetry environment in the market. As a result, bid-ask spreads widen to compensate for adverse selection risk that the dealers or specialists may face. We also document a significant increase in abnormal daily return volatility measured as standard deviation of the trade-to-trade returns within the day after the announcements. Assuming that the return volatility measures market participants’ perceived uncertainty over the firm value, a larger increase in the return volatility is consistent with greater firm value uncertainty created by the restatement events.

In order to account for other events unrelated to restatements that may potentially affect the trading activities, we compare the results of the restatement firms with those of the matching firms based on SIC code and size. Unlike the restatement firms, we do not find any significant increase in any of the measures of the trading metrics for the matching firms around the announcement date. The abnormal returns of the matching firms are stable over time with no significant changes around the event dates. Thus we conclude that the increases in uncertainty, in adverse selection, and in negative market reactions following the restatement announcements should be associated with the effect of restatement announcements only and not to other events that might have affected the restatement firms during the event periods.

We also conduct a cross sectional regression analysis to further examine the impact of trading activities on the spread during both nonevent and event trading periods. We observe a positive overall relation between spread and order size during periods of normal trading. This finding is consistent with Easley and O’Hara (1987), which views large trades arising most likely from access to private information, an evidence that supports adverse selection. The relation between spread and order size during the restatement event period is also significantly positive implying an increase in adverse selection during the event window. This is also consistent with the univariate analysis.

Our paper investigates the trading activities and adverse selection around the restatement announcements of NASDAQ restatement firms. Many studies have examined the difference in market microstructures between the NYSE and NASDAQ and find that NASDAQ has significantly higher transaction costs than those in NYSE (e.g., Christie and Schultz 1994; Huang and Stoll 1996; Bessembinder and Kaufman 1997; Weston 2000) and adverse selection costs are significantly higher on NYSE than on NASDAQ (Affleck-Graves et al. 1994; Lin et al. 1995; and Huang and Stoll 1996). The results with NASDAQ restatement firms are generally similar to those with NYSE-AMEX restatement firms. The univariate results show that there is a significant increase in abnormal trading activities and adverse selection after the restatement announcements in NASDAQ. However, it appears that the level of adverse selection in NASDAQ around the event window is less severe than that in NYSE-AMEX. Additionally, we do not detect any evidence of increased adverse selection around the announcement in the cross sectional regression analysis.

Our paper contributes to the literature in several ways. Prior research shows that the accounting restatements are salient negative firm-specific events, associated with a decrease in firm value, future earnings prospects and creditability of management. Unlike earning announcements and other popular corporate events, restatement events are little known to investors before they are made public. Investors seem to know about the restatement event only ex post. Thus, the restatement announcements provide an ideal setting to study information asymmetry and microstructure activities. Our paper is the first study that examines the trading activities and information asymmetries related to the restatement announcement using intraday data and the matching firm approach. An important advantage of using high frequency data is the added precision with which microstructure based empirical measures of information asymmetry may be estimated. Additionally, we provide empirical evidence that validates important theoretical microstructure models of Easley and O’Hara (2004) and Kim and Verrecchia (1994, 1997), which support the view that poor quality information may give rise to information asymmetry among the market participants. Our paper is also the first study that directly examines the difference in trading activities and information asymmetry around the restatement announcement event in NYSE-AMEX and NASDAQ.

The rest of the paper is organized as follows. Section 2 discusses background of restatement and related literature. Section 3 describes the data and method employed in the paper. Section 4 discusses the results of both univariate and cross sectional analyses and NASDAQ restatements. Section 5 concludes the paper.

2 Background and related literature

A restatement is an indication to investors that managers are not credible and that the public financial information provided by managers is less than reliable. There has been a recent upsurge in interest in issues concerning restatement. For example, Richardson et al. (2002) show that the primary motivation to manipulate earnings is to attract external financing at a lower cost. Restating firms have high market expectations for future earnings growth and have higher levels of outstanding debt. They also document that information in accruals, specifically operating and investing accruals, are key indicators of the earnings manipulations that lead to restatements. Palmrose et al. (2004) examine the market reaction to restatement announcements and find an overall significant negative abnormal return (about 9 %) over a 2-day event window. Palmrose et al. (2004) also find a significant association between the dispersion of earnings forecasts by analysts and restatement announcements. Wu (2002) examine a 3-day price response around the restatement announcements and find that restatements are regarded bad news in the market and the market reaction is stronger when the amount restated is larger. Anderson and Yohn (2002) examine the market returns and the bid-ask spread effects at the announcement of the accounting problem that leads to restatement using a 7-day window and negative market returns for accounting problem announcements and the negative reaction is most pronounced for firms with revenue recognition issues. They also find an increase in spreads surrounding the announcement of revenue recognition problems.

Using conditional total accruals measure as their proxy for reporting strategies, Moore and Pleiffer (2009) find that restatement firms do not change their reporting strategies after a restatement even though investors have learned that the strategy has resulted in misstated financial statements. Their findings suggest investors’ increased skepticism of management and decreases in the quality of earnings following a restatement. Desai et al. (2006) investigate the impact on adverse managerial reputations and penalties imposed by both the labor market and regulators. Srinivasan (2005) also showed that directors of companies that have restatements incur significant labor market penalties. The impact on litigation is investigated by Palmrose and Scholz (2004) and the tax consequences are examined by Erickson et al. (2004). Richardson (2000) suggests a systematic relationship between the magnitude of information asymmetry and the level of earnings management. Richardson and Waegelein (2002) find that firms that compensate managers with long-term performance plans are associated with lower levels of managed earnings than firms that have only short-term bonus plans.

The relationship between companies and auditors after restatements has also been examined. Auditors face the trade-off between risks and benefits in their decisions on whether to resign or not. The benefits for the auditors come from the audit fees. The risks mainly come from auditors’ litigation costs, which are the primary factor driving auditor resignations. Prior literature argues that auditors strategically manage clients’ audit risk by selectively refusing to accept new, risky engagements or by declining to continue with engagements where audit risk has increased (Pratt and Stice 1994). Shu (2000) and Krishnan and Krishnan (1997) also find empirical evidence that auditor resignation is positively associated with the likelihood of litigation. The auditors are more likely to resign if they perceive that the probability of hidden audit risk is not sufficiently compensated by the audit fess (Bockus and Gigler 1998). Recently, Huang and Scholz (2012) find that auditors are more likely to resign from the clients with severe restatements (e.g., fraudulent accounting) that indicate weak corporate governance and internal control. They also find the auditors’ risk-based incentives to resign are balanced by their economic incentives to retain clients. Companies that experience greater increases in audit fees from the year before to the year of restatement announcements are less likely to have audit resignations. Their results suggest that auditors are more likely to stay with the current clients or accept new clients if they perceive increased fees are sufficient to compensate for the increased audit risk.

Prior research suggests that public financial disclosure should reduce information asymmetry in the stock market (Verrecchia 1982; Diamond 1985). However, other research contends that not only the existence but also the quality of the information may affect the information asymmetry in the market, hence firm’s cost of capital. For example, Lambert et al. (2007) show that the quality of accounting information can influence the cost of capital directly because higher quality disclosures reduce the firm’s assessed covariances with other firms’ cash flows. Easley and O’Hara (2004) propose that poor accounting information quality can induce information asymmetry and therefore, increase non diversifiable information risk because less informed investors are always at a disadvantage relative to informed investors in adjusting their portfolio weights. As a result, investors expect higher returns from a firm with high information asymmetry or low quality accounting information. Kim and Verrecchia (1994, 1997) demonstrate that low quality accounting information can increase information advantage of some sophisticated investors who have superior ability to process the information. This information advantage, in turn, may increase the information asymmetry among the market participants. Restatement can be viewed as low quality information because the investors now believe that not only the past accounting information was of low quality, but also the future information may not be reliable, therefore, information asymmetry is expected to increase around the restatement announcements.

Theoretical model by Kyle (1985), Copeland (1976), Glosten and Milgrom (1985) predict that information asymmetry can increase adverse selection risk for liquidity providers or market makers who in turn, demand a larger compensation for providing liquidity. As a result, these market makers widen bid-ask spread to compensate for the adverse selection risk.

There are two principal theories that explain the bid-ask spread: (1) asymmetric information model and (2) inventory control model. In asymmetric information model, dealers (market makers) trade with liquidity traders and informed traders. The latter groups have information which is superior to the dealers, so bid and ask prices are set in order to compensate dealers for the perceived adverse selection risk. Kyle (1985), Copeland (1976), Glosten and Milgrom (1985) all argue that if market conditions are such that dealers become concerned that there is a higher proportion of informed traders in the market or that the informed traders have better information, they will widen bid-ask spread to compensate for the adverse selection risk.

The prediction of the adverse selection models is that spread should widen before an announcement as there is increased probability that trades are initiated by investors with superior information, while spreads should fall after an announcement, once the information has become public. However, it is possible that within context of these models, spreads may not fall immediately after the announcement, as there is still some advantage to be gained by market participants who did not have superior information but have superior-information processing abilities. For example, Kim and Verrecchia (1994) argue that the directors or corporate insiders may have superior information but they are prohibited from trading before the announcement dates, so they are able to make use of it only after the announcements. Therefore, Kim and Verrecchia (1994) suggest that disclosure of information would cause increased information asymmetry risk, so that spread should widen after the announcement rather than before it. However, in either case, one would expect spreads to return to normal levels within a few days of the announcement.

These studies suggest a positive relationship between spreads and order size (proxy for adverse selection cost), since dealers interpret an increase in order size as a sign of an increased number of informed traders and widen their spreads accordingly. Furthermore, Kim and Verrecchia (1991a, b) argue that heterogeneous beliefs around the corporate announcements induce market participants to trade. Therefore, they suggest that increased information asymmetry at announcement dates should result in higher trading volumes as well as increased spreads.

According to the inventory control models, risk-averse market makers have a desired (optimal) inventory position. To maintain this optimal inventory level, the market makers face two types of risk: (1) the risk of being unable to trade the stock and (2) the risk that prices will change while stocks are being held. Amihud and Mendelson (1980) and Ho and Stoll (1980) argue that the higher the first risk, the more difficult for the market makers to return to their optimal inventory level. In a liquid market characterized by high trading volumes, the dealer (market maker) will only set a narrow inventory spread, since he/she is assured of being able to quickly restore an out-of-equilibrium position. The inventory model, therefore, predict that as the liquidity of stock increases (i.e., trading volume increases), the market maker will reduce the spread since the compensation during this period is lower, resulting in a negative relationship between trading volumes and spreads.

The second feature of inventory risk is related to the underlying variability of the stock return. Garber and Silber (1979) and Ho and Stoll (1981) show that the more volatile the stock price is, the more the market maker is exposed to the risk of adverse price movements, and consequently, the spread will be widened to compensate the market maker, leading to a positive relation between return variability and the spread.

Finally, besides adverse selection and inventory components as discussed above, Roll (1984) and Stoll (1989) identify another component of bid-ask spread which is the order processing cost. According to the order processing cost model, the dealers need to recover fixed transaction costs through the bid-ask spread. The fixed cost will be lower if the dealers make a large volume of trades. Therefore, the model will imply the negative relationship between number of transactions and spread.

3 Data and method

3.1 Data

We use a statement announcement sample from United States General Accounting Office (GAO) from 1997 to 2005. This is the most comprehensive public sample we could obtain. We separate the firms in NYSE-AMEX and NASDAQ to investigate the difference of the effect of restatement announcements on the two markets. In order to compute bid-ask spread and other microstructure variables (volume, number of transactions, volatility, order size), we use intraday quotes and trades from TAQ database. Following the microstructure literature, before computing the variables, we screen the data as given below.

For the trades file, we retain the following:

  • Trades inside regular trading hours (9:30–16:00)

  • Good trades (corr = 0, 1) under regular sale conditions (cond = blank or *)

  • Trades with positive trade price (price > 0) and positive trade size (siz > 0)

For the quotes file, we retain the following:

  • Quotes inside regular trading hours (9:30–16:00)

  • Regular quotes (mode = 12)

  • Quotes with positive bid price (bid > 0), positive ask price (ofr > 0), bid price greater than ask price (ofr > bid), positive bid size (bidsiz > 0) or positive ask size (ofrsiz > 0)

Following Lee and Ready (1991), we match each trade with the latest available quote at least 5 s earlier. To classify the order flow of each trade as a buyer-initiated or seller-initiated, we use the standard Lee-Ready algorithm, which involves a “quote test” and a “tick test”. For the “quote test”, any trade that takes place above (below) the midpoint of the current quoted spread is classified as a buy (sell) order because trades originating from buyers (sellers) will most likely be executed at or near the bid (ask). For trades taking place at the midpoint, we use a “tick test” to classify a trade as a buy (sell) order if the trade price is above (below) the previous price. In the event there is no change in the trade price, the order flow is regarded as indeterminable and this trade is not used in computations that require an order flow indicator.

We use intraday data to estimate daily variables, spread, trading volume, number of transactions, volatility, and order size. For each stock, we compute daily trading volume and number of transactions as the total trading volume and total number of transactions within the day. Daily volatility is the standard deviation of the stock return within the day. Daily order size is the average number of shares traded per transaction on that day. For measures of spread, we use the two relative spread measures commonly used in market microstructure literature, relative quoted spread and relative effective spread. These relative spreads are used to estimate market makers’ revenue and conversely, investors’ cost for a round-trip transaction. The underlying assumptions of the quoted spread are that market makers set the prevailing quotes and stand on the other side of the customer trades, and that the investors cannot trade within the quoted spread. For each quote at time s for firm-quarter i, we compute the intraday relative quoted spread, IntraQSpread i,s , as follows:

$$ IntraQspread_{i,s} = \frac{{ask_{i,s} - bid_{i,s} }}{{(ask_{i,s} + bid_{i,s} )/2}} $$
(1)

We compute daily relative quoted spread, QSpread, by equal-weighting IntraQSpread within each day. The effective spread is based on the notion that trade is only costly to the investor to the extent that the trade price deviates from the true price, approximated by the bid-ask midpoint. For each trade-matched quote at time s for firm-quarter i, we compute the intraday relative effective spread, IntraESpread i,s , as follows:

$$ IntraEspread_{i,s} = \frac{{2\left| {{\text{trade}}\,{\text{price}}_{i,s} - (bid_{i,s} + ask_{i,s} )/2} \right|}}{{{\text{trade}}\,{\text{price}}_{i,s} }} $$
(2)

where trade price i,s is the trade price at which the trade executed at time s for firm-quarter i. We compute daily relative effective spread, ESpread, by equal-weighting IntraESpread within each day. As a check for robustness, we also use daily quoted spread and daily effective spread computed in a similar manner.

3.2 Univariate test

An important part of this study is to examine the impact of restatements on abnormal changes in spreads, volume, number of transactions, volume per transactions (order size), volatility, and return using univariate analysis. In order to reduce the effect of outliers, our analysis employs natural-log transforms of all variables except return. The following method helps to determine the abnormal changes.

First we compute average daily values of spreads, trading volumes, number of transactions, order size, and volatility over a ‘normal trading period’ that spans from day −250 to day −30 relative to the restatement announcement date. Next, we compute a time series of abnormal daily spreads and other variables by taking the daily value on a given day less normal value calculated from the normal trading period as defined above. We then obtain the abnormal return from the market model in the event study approach with the estimation period that also ranges from day −250 to day −30 relative to the announcement date.

The above firm specific abnormal trading activities do not take into account other events unrelated to restatement announcements, which may potentially affect the trading activities. It is quite possible that abnormal trading patterns may emerge from events other than restatements. In order to account for this possibility, the matching firm approach based on SIC code and size is often very useful. For a restatement firm, we select the matching firm closest in size (market capitalization at least 30 days before the announcement) within the same four-digit SIC code. If we are unable to find a match, we relax the four-digit to three-digit level and find the matching firm closest in size to the restatement firm. Then we calculate the abnormal trading activities for the matching firm and compare them with those of the restatement firm. The difference between the two samples should be attributed to the effect of restatement announcements.

3.3 Cross sectional tests

A cross sectional regression analysis is useful to conduct an in-depth examination of the impact of trading activities on the spread during both non-event and event trading. Dummy variables facilitate testing for event-related shifts in the intercept and the slope coefficients on the variables studied. The event window is split relative to the announcement date (day 0) into three sub periods: PRE (day −10 through day −1), DURING (day 0 and day 1), and POST (day +2 through day +10).

Many microstructure artifacts considered here proxy for the sources of bid-ask spread documented in the literature. For instance, number of transactions proxy for order processing costs, order size for adverse selection cost, and price and return variability for inventory cost. Similar control variables are used in Conrad and Niden (1992), Lee et al. (1993), Welker (1995), Yohn (1998), Ertimur (2004), among others. To the extent that these control variables are adequate, one can interpret the coefficients on order size variable as the effect of adverse selection on the spread during non-event trading period, and the coefficient on the order size variable during the restatement announcement period as effect of adverse selection on the spread during the event period. We use first differences to control for serial correlation in the variables.

The model is as follows:

$$ \begin{aligned} \Updelta \ln (SPREAD)_{it} & = a_{0} + a_{1} \Updelta \ln (PRICE)_{it} + a_{2} \Updelta \ln (VOLATILITY)_{it} + a_{3} \Updelta \ln (NTRANS)_{it} \\ & \quad + a_{4} \Updelta \ln (ORDERSIZE)_{it} + a_{5} PRE + a_{6} DURING + a_{7} POST \\ & \quad + a_{8} PRE \times \Updelta \ln (ORDERSIZE)_{it} + a_{9} DURING \times \Updelta \ln (ORDERSIZE)_{it} \\ & \quad + a_{10} POST \times \Updelta \ln (ORDERSIZE)_{it} + a_{11} COMPANY + a_{12} SEC_{it} + a_{13} AUDITOR + \varepsilon_{it} \\ \end{aligned} $$
(3)

where \( \Updelta \ln (SPREAD)_{it} \): the change in natural log of spread (quoted spread, effective spread, relative quoted spread, and relative effective spread); \( \Updelta \ln (PRICE)_{it} \): the change in natural log of mean price of day t of stock i; \( \Updelta \ln (VOLATILiTY)_{it} \); the change in natural log of volatility of day t of stock i; \( \Updelta \ln (NTRANS)_{it} \): the change in natural log of number of transactions; \( \Updelta \ln (ORDERSIZE)_{it} \): the change in natural log of order size; \( PRE \), \( DURING \), \( POST \) are the event period dummies corresponding to day −10 through day −1, day 0 through day 1, day 2 through day 10 (relative to the announcement date), respectively; \( COMPANY, SEC, AUDITOR \) are the dummies for restatement initiators, company, SEC, and auditor, respectively. The data are stacked across firms and days from day −250 to day 250 (relative to the announcement date). We employ White (1980)’s method to correct for heteroscedasticity.

4 Results and analysis

4.1 Descriptive statistics

Table 1 presents descriptive statistics for NYSE-AMEX and NASDAQ restatements. The number of restatements increases from 23 (2.5 %) in 1997 to 240 (26.1 %) in 2005 for NYSE-AMEX, and from 57 (5.8 %) in 1997 to 207 (20.9 %) in 2005 for NASDAQ. Manufacturing industry accounts for the most number of restatements in both NYSE-AMEX and NASDAQ (29.1 and 28.2 %, respectively). About half of the restatements pertains to revenue recognition (22.5 % in NYSE-AMEX and 28.7 % in NASDAQ) and to cost or expense (29.5 % in NYSE-AMEX and 27.7 % in NASDAQ). The company initiates most of the restatements (55 %) in both NYSE-AMEX and NASDAQ, while Auditor and SEC initiate lesser number of restatements (10.7 and 10.1 % in NYSE-AMEX and 11.4 and 9.5 % in NASDAQ, respectively). Figure 1 graphically shows the distribution of restatements by year, industry, reason, and initiators for both NYSE-AMEX and NASDAQ. It is noteworthy that the distribution in each category is very similar for both markets regardless of the differences in their market micro-structure.

Table 1 Descriptive statistics by year, industry, reasons and initiators
Fig. 1
figure 1

Distributions of restatement firms in NYSE-AMEX and NASDAQ by year, industry, reasons, and initiators. The figures show the distribution of the restatement firms in NYSE-AMEX and NASDAQ by year, industry, reasons and initiators as reported in GAO (2002, 2006) database. Industry classification are defined using the following SIC codes as in Palmrose et al. (2004): agriculture, mining and construction = 0–1999, manufacturing = 2000–3999 (except codes assigned to technology), technology = 3570–3579 plus 7370–7379, transportation = 4000–4799, communications = 4800–4899, utilities = 4900–4999, wholesale/retail = 5000–5999, financial services = 6000–6999, services = 7000–8999 (except codes assigned for technology). The total of restatement reasons and initiators are not equal to the total sample size due to some restatements having multiple reasons or initiators

Table 2 presents descriptive statistics for both the restatement firms and matching firms in NYSE-AMEX and NASDAQ. Our matches show fairly close fit along most dimensions. The mean and median share price (pre-announcement) for restatement firms are 21.23 and 17.25, respectively in NYSE-AMEX while those of the matching firms are 24.49 and 19.84, respectively. Restatement firms and matching firms also have similar number of shares outstanding, market cap and book-to-market ratio. We also include a comparison of the beta, a measure of systematic risk of the firms. As can be seen from Table 2, the betas are similar for both restatement firms and matching firms. This implies that any difference between the restatement and matching firms might not be explained by beta. The matching firms closely fit to the restatement firms in NASDAQ as well. The restatement firms and matching firms in NASDAQ are much smaller in size as measured by the product of outstanding shares and their price than those in NYSE-AMEX.

Table 2 Descriptive statistics—restatement and matching firms

4.2 Univariate test

Table 3 presents mean cumulative abnormal return (CAR) for restatements by core and non-core reasons, and by initiators. Following Palmrose et al. (2004), we define core restatements as the ones in which the restatement is due to cost or expense and revenue recognition. For all other reasons, the restatements are considered non-core restatements. We document negative abnormal returns around the announcement. From 1997 to 2006, the CARs on the announcement date, day 0, and on day +1 are −1.86 and −1.66 %, respectively. For the 2-day window [0, 1], the CAR is −3.5 %. Our estimates of CAR are lower than those in some early studies (Palmrose et al. 2004; GAO 2002), but are consistent with those in the recent studies (Scholz 2008; Hranaiova and Byers 2007; Wang and Yu 2008), which attribute the reduced market reactions to restatement to the SOX effect. These studies suggest that the SOX increases the probability of detecting errors in the previously issued financial statements thereby reducing the market reactions to the announcements. Table 4 shows the average daily abnormal trading activities around the announcement day for all the restatement companies in NYSE-AMEX. All variables are log-transformed for further computation. We report both mean and median for each variable for stronger results. The t statistics in the parenthesis are reported for the mean while the p values of the non-parametric Wilcoxon test in the squared bracket are reported for the median to examine whether the mean and median are statistically different from zero. As can be seen from the table, there is no significant abnormal activity before the announcement. It implies that, unlike earning announcements and other popular corporate events, restatement event is little known to investors before it is made public. Investors seem to know about the restatement event only after the public announcements. However, when the event is made public, the market reacts quite strongly to the news. We observe that significant positive abnormal changes begin on the announcement day and continue for all 5 days after the announcement in all trading activities such as volume, number of transactions, order size, volatility, and the two measures of spreads (relative quoted spread and relative effective spread). The evidence clearly supports the existence of informed trading around the announcement day. The increase in volume or number of transactions could be an increase in adverse trades because informed investors exploit their informational advantage (Kyle 1985). However, in the literature, there is some ambiguity on whether trading volume increases in adverse selection during corporate event announcements. For example, if liquidity traders have discretion over timing of the trades, then there could be a decrease in volume (Admati and Pfleiderer 1988; Tinic and West 1972). It may be due to trades arising from the lack of consensus among market participants (Verrecchia 1981). Some other studies suggest alternative measures to proxy for adverse selection. For example, Easley and O’Hara (1987) contend that informed traders choose to trade larger amounts. Hence the order size (average trading volume per transaction) may proxy for informed trading. If informed trading is associated with large orders, an increase in the order size and a stronger positive relation between the order size and spread should be observed around the announcement. This hypothesis has been supported by Conrad and Niden (1992). Consistent with Easley and O’Hara (1987)‘s prediction our results also document a significant increase in the abnormal order size after the restatement announcement.

Table 3 Mean cumulative abnormal return around the announcements by core/non-core reasons and by prompters for NYSE-AMEX firms
Table 4 Average daily trading activities around the restatement announcement for NYSE-AMEX firms

Table 4 also presents the response of spreads around the restatement announcements. We document a significant increase in both abnormal relative quoted spread and abnormal relative effective spread.Footnote 2 A restatement is an indication to investors that managers are not credible and that the public financial information provided by managers is less than reliable. Therefore investors are more likely to be skeptical about the reliability of the firm’s past financial statements as well as the firm’s future financial information to be released. Such a skepticism inevitably creates information asymmetry in the market and a high degree of uncertainty. As a result, bid-ask spreads should increase to compensate for adverse selection risk that the dealers or specialists may face. The result is consistent with adverse selection literature (Kyle 1985; Easley and O’Hara 1987; Glosten and Milgrom 1985), which suggests that if market conditions are such that dealers or specialists become concerned that there is a higher proportion of informed traders in the market or that the informed traders have better information, they will widen bid-ask spread to compensate for the adverse selection risk.

Our results on bid-ask spreads are also consistent with the theoretical model of Kim and Verrecchia (1994, 1997). They suggest that restatement announcements can increase information advantage of some sophisticated investors who have superior ability to process the information. Once the information is released, these sophisticated investors are able to analyze and advantageously act upon it better than other participants. This information advantage, in turn, increases the information asymmetry among the market participants. As a result, the dealers or specialists widen bid-ask spread to compensate for the adverse selection risk as reported in this paper.

We also observe a significant increase in abnormal daily return volatility measured as standard deviation of the trade-to-trade returns within the day after the announcements. Assuming that return volatility measures market participants’ uncertainty over firm value, a larger increase in return volatility is consistent with greater uncertainty in the firm value created by the restatement events. Interestingly, the return volatility pattern is similar to other patterns pertaining to spread and order size. These results provide support for the positive association between uncertainty about the firm value and adverse selection as discussed earlier.

In order to account for non-events that may affect the trading activities, we use matching firm approach based on SIC code and size as discussed earlier. The matching firms and the restatement firms have very similar risk characteristics. Thus any difference in their abnormal market trading activities should be attributed to the effect of the restatement announcements. Figure 2 shows the abnormal trading activities around the announcement date for both matching and restatement firms. The results are quite remarkable. We do not find any significant increase in volume, number of transactions, order size, volatility, and spreads for the matching firms around the announcement date. However, we do find significant increases in the trading activities for the restatement firms. Similarly, the abnormal returns of the matching firms are stable over time with no significant changes around the event dates, whereas the abnormal returns of the restatement firms decrease significantly after the announcements implying a negative market reaction to the restatement event. Thus, one must conclude that the significant increase in the abnormal trading activities and the significant decrease in the abnormal returns are mainly due to the effect of restatement announcements only. Other variables unrelated to restatement may have no impact on the restatement firms during the event period.

Fig. 2
figure 2

Abnormal trading activities around the restatement announcement for NYSE-AMEX firms. These figures report the average daily abnormal volume, abnormal transactions, abnormal order size, abnormal volatility, abnormal relative quoted spread, abnormal relative effective spread, and abnormal return for a sample of NYSE-AMEX restatement and matching firms from 1997 to 2005 collected from GAO (2002, 2006) database. All variables are log-transformed before further computation. The normal trading period value is the average daily values over an estimation period, day −250 to day −30 relative to the restatement announcement date. Abnormal daily spreads (volume, number of transactions, order size, volatility) are computed by taking the daily value on a given day less the normal trading period value. Abnormal returns are based on a single-factor market model estimated from day −250 to day −30 for each sample firm, using the CRSP value-weighted index. a Abnormal volume, b abnormal number of transactions, c abnormal order size, d abnormal volatility, e abnormal relative quoted spread, f abnormal relative effective spread, g abnormal return

The next important result deals with the market reaction and abnormal trading activities of core restatements and non-core restatements as defined earlier. It is reasonable to believe that investors perceive mis-statements of the core accounts as severe breaches that diminish the reliability of earnings and directly impact their ability to forecast earnings. Hence the market may react more negatively with the core restatements than they do with the non-core restatements. In this regard, there are mixed results in the previous literature about whether there is an association between restatement effects and the accounts that are restated. Palmrose and Scholz (2004) find that core restatements are positively associated with shareholder litigation, while non-core are not, which suggests that investors regard restatements of core accounts as more serious. Palmrose et al. (2004) and Anderson and Yohn (2002) find that restatements related to revenue recognition are associated with more negative market response than other restatements, but Hribar and Jenkins (2004) do not find significant evidence that core account restatements are associated with a greater implied cost of capital effect.

Table 3 and Fig. 3 exhibit our results for core and non-core restatements. The two-day window CAR [0, 1] for core restatements (−3.74 %) is somewhat smaller than that of non-core restatements (−4.14 %). However, the difference between them is not statistically significant. We also do not find any significant difference in spreads and other trading metrics between the core and non-core restatements. This implies that the market reacts negatively in a similar manner between the two types of restatements.

Fig. 3
figure 3

Abnormal trading activities around the restatement announcement by core/non-core reasons for NYSE-AMEX firms. These figures report the average daily abnormal volume, abnormal transactions, abnormal order size, abnormal volatility, abnormal relative quoted spread, abnormal relative effective spread, and abnormal return for a sample of NYSE-AMEX core and non-core restatements from 1997 to 2005 collected from GAO (2002, 2006) database. All variables are log-transformed before further computation. The normal trading period value is the average daily values over an estimation period, day −250 to day −30 relative to the restatement announcement date. Abnormal daily spreads (volume, number of transactions, order size, volatility) are computed by taking the daily value on a given day less the normal trading period value. Abnormal returns are based on a single-factor market model estimated from day −250 to day −30 for each sample firm, using the CRSP value-weighted index. a Abnormal volume, b abnormal number of transactions, c abnormal order size, d abnormal volatility, e abnormal relative quoted spread, f abnormal relative effective spread, g abnormal return

A further distinction between revenue and cost restatements in core accounts reveals that revenue restatements occur when companies prematurely recognize revenues. Even worse so, when they recognize fictitious revenues. Misstating costs or expenses is one of the most common ways of twisting current-year profits. Techniques used include overstating inventory, overstating long-term assets, underestimating or overestimating reserves, and shifting expenses from one period to another. While both cost and revenue restatements reflect serious issues with a company’s accounting framework and reporting practices, revenue restatements may indicate a more serious problem since sales are the core of a firm’s ability to grow and prosper (Stickney et al. 2006). A firm’s restatement of overstated revenue implies not only that its historical performance is exaggerated, but also that its future growth may be crippled. Investors’ concern over the potential for growth leads to a downward revision of the firm’s value. Hence, revenue recognition restatements may have a greater negative market reaction than cost restatements. In a result whose details are not included in the paper in the interest of brevity but available upon requests, we find that the two-day window CAR [0, 1] and 3-day window CAR [−1, 1] of revenue restatements (−4.36 and −4.74 %, respectively) are higher than those of cost restatements (−3.24 and −3.54 %, respectively). However, the differences between the two restatements are not statistically significant. In addition to abnormal returns results, we also document significant increases in trading volume, number of transactions, order size, volatility and spreads for both types of restatements. Again, the increases are stronger for revenue restatements than for those of cost restatements. Our results are in line with Palmrose and Scholz (2004) in which they find core restatements are associated with a higher likelihood of litigation and this association is driven primarily by revenue restatements.

Moving further, our next focus is on estimating the differences in market reaction to restatements initiated by SEC, auditors, and management of the companies. Restatements attributed to auditors signal that the internal monitoring functions of companies have failed not only to prevent a significant mis-statement, but also to identify and correct them. Similarly, restatements initiated by company management may provide mixed signal to market. On one hand, detection and revelation by the company management indicate relatively stronger internal controls and oversight by management, boards and audit committees, thereby reducing the likelihood of top management involvement in creating mis-statements and mitigating some of the uncertainty surrounding the management’s creditability. On the other hand, these restatements still give negative signals to the market that the management may not be reliable since they do not prevent the errors in the first place. Thus, we expect that the market reaction is more negative for auditor initiated restatements than for company initiated restatements.

Just as the mixed signals generated by a company, the restatements by SEC may also send mixed signals to the market. While the market may view these restatements negatively because of the uncertainty related to management creditability, the market reaction to these restatements may be attenuated if market participants perceive the issues as technical matters or judgment disagreements between the SEC and companies and/or auditors. In fact, Palmrose et al. (2004) find that the number of core restatements attributed to auditors is more than twice the core restatements attributed to the SEC (25 versus 12 %). Thus, investors may view SEC intervention based on reviews of company filings as technical compliance issues, rather than fundamental financial reporting violations.

Peterson (2012) develops a model to predict which restatement firms will be sanctioned by the SEC. He documents that firms with complex revenue recognition processes are more likely to restate revenue and less likely to receive an Accounting and Auditing Enforcement Release (AAER) from the SEC. Likewise, Files (2012) argues that the SEC lacks sufficient resources to investigate every case of potential fraud. Accordingly, only a small percentage of restatement firms (generally less than 20 %) are officially sanctioned by the SEC.

The SEC, faced with time and monetary constraints, may choose to target firms with more transparent disclosures to limit their information generating costs. Also, the SEC employees often rely on external cues, including news reports and company press releases, to determine which restatement merits additional scrutiny (DeFond et al. 2008). This practice suggests that transparent disclosures are more likely to come to the attention of the SEC and are, therefore, more likely to result in a sanction. The studies cited here suggest that restatements initiated by the SEC tend to associated with companies that have less complex revenue recognition processes more corporate transparency. These companies are less likely to restate revenue, which is an important item in the core account. Thus, we expect that market reaction to restatements initiated by the SEC is less negative than auditor-initiated statements.

Our results indicate that the restatements initiated by the auditors have the most negative abnormal returns (−4.79 %) while the SEC initiated restatements have the least negative abnormal returns (−2.21 %). The CAR for the restatements initiated by the companies themselves is −3.42 %. Palmrose et al. (2004) and Hribar and Jenkins (2004) have also documented a more negative market response associated with auditor-initiated restatements than those associated with the SEC. Restatements initiated by the auditors may ensue a negative market reaction and possibly a greater uncertainty and information asymmetry. Restatements initiated by company management and SEC may have less negative reaction since they indicate that management may be forthright and trustworthy and that the companies tend to have less complex revenue recognition process and more corporate transparency. This notion is supported by our results presented in Fig. 4. We observe that the auditor-initiated restatements generate far greater abnormal volatility and spreads, and more negative abnormal returns than the management and SEC-initiated restatements.

Fig. 4
figure 4

Abnormal trading activities around the restatement announcement by initiators for NYSE-AMEX firms. These figures report the average daily abnormal volume, abnormal transactions, abnormal order size, abnormal volatility, abnormal relative quoted spread, abnormal relative effective spread, and abnormal return for a sample of NYSE-AMEX restatements by initiators from 1997 to 2005 collected from GAO (2002, 2006) database. All variables are log-transformed before further computation. The normal trading period value is the average daily values over an estimation period, day −250 to day −30 relative to the restatement announcement date. Abnormal daily spreads (volume, number of transactions, order size, volatility) are computed by taking the daily value on a given day less the normal trading period value. Abnormal returns are based on a single-factor market model estimated from day −250 to day −30 for each sample firm, using the CRSP value-weighted index. a Abnormal volume, b abnormal number of transactions, c abnormal order size, d abnormal volatility, e abnormal relative quoted spread, f abnormal relative effective spread, g abnormal return

We also examine whether the companies tend to release all of the restatement bad news out at one time or they tend to make the restatement announcement in several times. In other words, we look at whether there is a higher probability that a company will make other restatements once the first restatement is announced. In our sample, there are 902 restatements from period 1995–2005. However, these restatements are from only 692 companies since some companies have multiple restatements over time. In a result not tabulated but available upon requests, we majority of the restatement firms (76 %) have only one statement in the sample period. The number of firms that restated their financial statement twice is about 18 % of the sample. Only about 6 % of the firms restated 3 times or more during the period. Our result is consistent with the notion that restatement event is viewed as negative news that could impact the company’s value negatively, therefore, companies tend to make the restatement announcement only once.

We also investigate how the market reacts to the second, the third or higher restatements compare with the first restatement. One would expect that the restatements after the first announcement would contain less news and evoke less reaction from investors than the first since the investors already anticipate problems with the companies’ information quality. In our sample, there are 166 companies that have more than one restatement. We estimate the market reaction to the first restatements and compare with the market reaction to the restatements after the first restatements of these companies. We find that the 2 day window CAR (0, 1) and the 3 day window CAR (−1, 1) of the first restatements are significantly more negative than those of after the first restatements. The result indicates the market reacts more negatively to the first restatements than to the ones after the first. We also look at the graph of abnormal volume, number of transaction, order size, volatility, and spreads. In all cases, these variables are significantly higher for the first restatements than for the ones after the first.Footnote 3 Therefore, our results suggest that the first restatement announcement would contain more news and evoke more negative reactions from investors than those restatements after the first.

In summary, the univariate tests provide evidence of increased uncertainty (measured by abnormal volatility), increased adverse selection (measured by abnormal volume, number of transactions, order size, and spreads), and increased negative market reactions (measured by negative abnormal returns) immediately following the restatement announcements. A caveat with the univariate tests is that we have not controlled for the changes in the inventory holding and order processing cost components of the spreads as suggested in the literature (Stoll 1989). In the next section, we present a cross-sectional analysis that includes appropriate control variables in the regression tests so that we can make stronger inference on the extent of adverse selection associated with restatement announcements.

4.3 Cross-sectional tests

The purpose of the cross sectional regression is to examine further the impact of adverse selection, order processing, and inventory holding costs on the spreads during both nonevent and event trading. The model is discussed earlier in the Sect. 3. Table 5 presents results on all the four measures of spread, namely proportional effective spread, relative quoted spread, effective spread, and quoted spread. As can be seen from the table, for core restatements, volatility significantly and positively affects all the four measures of spread. However, volatility’s influence on spread is neither significant for the non-core restatements nor for the overall sample. This suggests that the core restatements may create more uncertainty in the market than the non-core restatements causing a positive relation between spreads and volatility. The positive relation between price and effective and quoted spreads is consistent with the inventory carrying cost model and with the extant results documented in prior studies. The positive coefficient on the order size variable for the overall sample as well as for core and non-core restatements is consistent with the model of Easley and O’Hara (1987). If the informed traders do tend to trade in larger quantities as Easley and O’Hara (1987) suggest, then this result implies that the specialist tends to increase the spreads for large orders in order to compensate for a higher probability that these traders represent private information.

Table 5 Cross-sectional determinants of changes in spreads during non-event trading and around the restatement announcement for NYSE-AMEX firms

The significantly positive coefficient on the changes in the number of transactions may be a result of estimating the model in the first difference rather than levels. As such we also estimate the model using the levels of the variables and find the significantly negative cross sectional relation between spread and number of transactions, which is consistent with order processing cost literature.

During the event window, which is partitioned into pre-announcement, during and post-announcement periods, we document positively significant coefficient of order size during the event period [0, 1]. In particular, most of the coefficients of the variable \( DURING \times \Updelta \ln (ORDERSIZE) \) are significantly positive at 5 % level or below for all four measures of spreads in the core and non-core restatements as well as in the overall sample. This result implies an increase in adverse selection during the event window which is consistent with the univariate analysis.

Our result provides support to the theoretical model of Kim and Verrecchia (1994, 1997). They suggest that after the announcement, there is still some advantage to be gained by market participants, who possess no superior information but have superior-information processing ability. Corporate insiders, who are in possession of superior information but are prohibited from trading before the announcement dates, can make use of the information after the announcements. Our findings on wider spreads post announcements are consistent with the above cited studies in that the disclosure of information generates increased information asymmetry and adverse selection risk.

A similar cross sectional regression as above for the matching firms estimates whether there is any increase in adverse selection around the restatement announcement dates after controlling for other components of bid-ask spread. Table 6 reports the results. During non-event trading periods, the signs of variables are consistent with those of the restatement firms. However, we do not find any significant increase in order size, which serves as a proxy for adverse selection cost of the bid-ask spread during the event window. The coefficients of the variable \( DURING \times \Updelta \ln (ORDERSIZE) \) are not significant for all the four measures of spread. The result with regard to the matching firms suggests that the significant increase in adverse selection during the event window is unique only to the restatement firms and not to the non-restatement firms with similar risk characteristics.

Table 6 Cross-sectional determinants of changes in spreads during non-event trading and around the restatement announcement for NYSE-AMEX matching firms

4.4 NASDAQ restatements

Existing microstructure studies on bid-ask spread and information asymmetry are primarily developed within the framework of an auction market such as NYSE or AMEX as against the dealer market such as NASDAQ. The NYSE and AMEX, as auction markets, rely on a single specialist to make a market in a stock.

Specialists facilitate continuous trading by posting quotes for their own account or by reflecting the best quotes on their limit order book, which represent a centralized depository for limit orders to buy or sell stocks at specified prices or better. The NASDAQ market, on the other hand, being a dealer based market, typically has several dealers making a market in a given stock. Thus order flow is broken up across the dealers. This structure provides greater opportunity for an informed trader to distribute his/her trades across a number of dealers. Also, the NYSE-AMEX has the advantage of having direct face-to-face interaction between the market participants and the specialist, while no such interaction occurs on the NASDAQ. This direct interaction provides additional information to the market maker about parties with whom he/she is trading, leading to reduced adverse selection (Stoll and Whaley 1990; Biais 1993). In addition, due to his/her limited monopoly power, the specialist is better able to spread adverse selection risk across trades as suggested by Glosten (1989). All these factors are likely to impact the amount of adverse selection cost that is impounded in the bid-ask spread.

Many studies have examined the difference in market microstructures between the NYSE and NASDAQ. Christie and Schultz (1994), Huang and Stoll (1996), Bessembinder and Kaufman (1997), Weston (2000), and Chung et al. (2003) find that the spreads are higher on NASDAQ than on NYSE. Christie and Huang (1994), Christie and Schultz (1994), and Jain and Kim (2006) provide evidence that spreads become lower for stocks, which are relocated from NASDAQ to NYSE. These studies suggest that NSADAQ has significantly higher transaction costs than NYSE. Affleck-Graves et al. (1994), Lin et al. (1995), and Huang and Stoll (1996) study components of bid-ask spread on markets of different structures and suggest that adverse selection costs are significantly higher on NYSE than on NASDAQ. However, Van Ness et al. (1999) suggest that information asymmetry is less severe on the NYSE than on the NASDAQ.

In this section, we examine the trading activities and adverse selection around the restatement announcements of NASDAQ restatement firms. Table 7 and Fig. 5 present the univariate results of abnormal trading activities for both restatement and matching firms. The results show a very similar pattern as observed with NYSE-AMEX. There is a significant increase in abnormal trading volume, number of transactions, volatility, order size, and spreads. The abnormal returns fall substantially after the announcements. Our results also indicate increased uncertainty, increased adverse selection, and increased negative market reactions after the restatement announcements.

Table 7 Average daily abnormal trading activities around the restatement announcement for NASDAQ firms
Fig. 5
figure 5

Abnormal trading activities around the restatement announcement for NASDAQ firms. These figures report the average daily abnormal volume, abnormal transactions, abnormal order size, abnormal volatility, abnormal relative quoted spread, abnormal relative effective spread, and abnormal return for a sample of NASDAQ restatement and matching firms from 1997 to 2005 collected from GAO (2002, 2006) database. All variables are log-transformed before further computation. The normal trading period value is the average daily values over an estimation period, day −250 to day −30 relative to the restatement announcement date. Abnormal daily spreads (volume, number of transactions, order size, volatility) are computed by taking the daily value on a given day less the normal trading period value. Abnormal returns are based on a single-factor market model estimated from day −250 to day −30 for each sample firm, using the CRSP value-weighted index. a Abnormal volume, b abnormal number of transactions, c abnormal order size, d abnormal volatility, e abnormal relative quoted spread, f abnormal relative effective spread. g abnormal return

Further, we find the market reacts more negatively with NASDAQ restatements than with NYSE-AMEX restatements. The cumulative abnormal returns of the restatement firms in NASDAQ on day 0 and +1 are −2.55 and −3.14 %, respectively, while those in NYSE-AMEX are −1.86 and −1.66 %, respectively. However, the level of adverse selection around the restatement announcements in NASDAQ is lower than that in NYSE-AMEX. The significance levels of t statistics of all the abnormal trading activities after the announcements in NASDAQ are lower than those in NYSE-AMEX, especially in abnormal relative quoted spread and abnormal relative effective spread. The abnormal spreads increase significantly in day 0 and day +1, but beyond these days, they are rendered insignificant. In the NYSE-AMEX, the significant increase in abnormal spreads lasts longer (more than 5 days). Table 8 reports the cross-sectional regression results showing the impact of adverse selection, order processing, and inventory holding costs on the spreads during both nonevent and event trading. Unlike the NYSE-AMEX, we do not report any significant increase in adverse selection around the event window. None of the coefficients on the intercept, and shift dummy variables (PRE, DURING, POST) are statistically significant. The impact of order size on the spreads is not evident during the event window. The coefficient of the variable \( DURING \times \Updelta \ln (ORDERSIZE) \) is not significant for all the four measures of spread.

Table 8 Cross-sectional determinants of changes in spreads during non-event trading and around the restatement announcement for NASDAQ restatement firms

In summary, the univariate results indicate a significant increase in abnormal trading activities and adverse selection after the restatement announcements in the NASDAQ, which is similar to the evidence found in the NYSE. However, it appears that the level of adverse selection in NASDAQ around the event window is less severe than that in NYSE-AMEX. Moreover the significant increase in adverse selection around the event date is not documented in the cross sectional regression analysis, which controls for other components of spread.

5 Conclusions

In this study, we investigate how accounting restatement announcements affect the trading activities and information asymmetry in both quote-driven NYSE and AMEX markets and order-driven NASDAQ market using a sample of restatement announcements from 1997 to 2005 collected from GAO Financial Statement Restatement Database. We conduct a univariate analysis of daily return, volume, number of transactions, average order size, volatility, and various measures of spreads for the restatement firms. We also conduct a cross sectional analysis of the spread relation with control variables that proxy for adverse selection cost, inventory holding cost, and order processing cost to further examine the impact of trading activities on spreads during both non-event and event trading. In addition, we compare both univarite and cross sectional results of the restatement firms those of the matching firms based on size and SIC code to account for other events unrelated to the restatement announcement, which may affect the trading activities. We also perform similar analyses for NASDAQ restatements to examine the extent of the order-driven market reactions to the restatement announcement event compared with the quote-driven market.

Our univariate tests demonstrate economically and significantly negative mean cumulative abnormal returns around the announcement dates. The restatements attributed to auditors are associated with more negative returns than the restatements attributed to management and the SEC. However, we do not find any significant difference between the abnormal returns of restatements related to revenues and expenses (core accounts) and restatements related to non-core accounts. In addition to abnormal returns, we document a significant increase in volume, number of transaction, average order size, volatility, and various measures of spreads after the restatement announcement indicating increased uncertainty and adverse selection following the announcement.

The abnormal return and trading metrics results suggest that investors believe that the restatements may diminish company prospects, and that the financial information provided by managers is less than reliable. Investors, therefore, are more likely to be skeptical about the reliability of the firm’s past financial statements as well as the firm’s future financial statements. As a result, investors experience increased uncertainty, information asymmetry, and adverse selection, which are reflected in wider volume, volatility and bid-ask spreads.

In the cross sectional analysis, we observe a positive overall relation between spread and order size during periods of normal trading, which is consistent with the view expressed in Easley and O’Hara (1987) that large trades occur primarily due to private information and provide evidence on the presence of adverse selection. The relation between spread and order size during the restatement event period is also significantly positive implying an increase in adverse selection during the event window, which is consistent with the univariate analysis. In addition, in both univariate and cross sectional analyses, we do not document any significant increase in all trading metrics or increased negative abnormal returns for the matching firms around the announcement date. There is no evidence of increased order size during the event window. Therefore, the increased uncertainty, increased adverse selection, and increased negative market reactions after the restatement announcements should be attributed to the effect of restatement announcements only, not to other events that might have affected the restatement firms during the event periods.

We provide evidence supporting the theoretical model of Kim and Verrecchia (1994, 1997) that information asymmetry is not just due to inside information but also due to the superior analysis of publically available data by skilled information processors. Disclosure of information causes increased information asymmetry risk because some investors are able to make use of their superior information processing ability. Accordingly, spreads widen after the announcement rather than before it, which is consistent with our results.

Finally, we also document significant increase in market negative reaction, abnormal trading activities and adverse selection after the restatement announcements for NASDAQ, which is similar to the evidence documented for NYSE. However, the level of adverse selection in NASDAQ around the event date seems to be lower than that in NYSE-AMEX.