Earnings Management Around Employee Stock Option Reissues ∗
Jeffrey L. Coles Department of Finance, W. P. Carey School of Business, Arizona State University,
[email protected] Tel: (480) 965-4475 Michael Hertzel Department of Finance, W. P. Carey School of Business, Arizona State University,
[email protected] Tel: (480) 965-6869 Swaminathan Kalpathy Department of Finance, W. P. Carey School of Business, Arizona State University,
[email protected] Tel: (480) 965-0804
First Draft: May 17, 2002 Revised: August 13, 2003 Comments and suggestions are welcome
∗
We thank session participants at the 2002 Financial Management Association meeting, seminar participants at Arizona State University, Jim Boatsman, Joe Comprix, Christopher Dussold, Sanjay Gupta, Lynn Rees, Ross Watts (the editor) and an anonymous referee for helpful comments. Any remaining errors are our own.
Abstract We investigate analyst and investor behavior and market pricing in a setting where the managerial incentive to manipulate earnings and market price should be apparent ex ante to market participants.
We find evidence of abnormally low accruals following
announcements of cancellations of executive stock options up to the time that the options are subsequently reissued. Nevertheless, analysts and investors are not misled. Over the nine months prior and six months following the reissue date, abnormal accruals have little power to explain stock price performance. Moreover, the estimated abnormal accruals response coefficient is insignificantly different from zero. Finally, abnormal accruals have no power to explain subsequent analyst forecast errors. Thus, our findings suggest that, in this transparent setting, analysts and investors are not misled by managerial attempts to manipulate stock price by managing discretionary accruals.
Earnings Management Around Employee Stock Option Reissues
1. Introduction and Motivation The extent to which earnings management by firms affects investor perceptions continues to attract interest. Academic research focuses largely on accounting adjustments to the firm’s cash flows from operations. Since the timing of cash transactions does not necessarily match the timing of economic transactions, financial reporting regulations allow firms discretion over accruals.1 The intention of standard setters and regulators in allowing some degree of reporting flexibility is to provide enough latitude so that financial statements can be made more informative. Nevertheless, in a world of asymmetric information and agency problems, the discretionary nature of accrual accounting may lead to earnings manipulation. As Healy and Wahlen (1999, p. 368) suggests, “… managers (may) use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers.” Although there is some evidence that accruals management yields more informative financial statements,2 there also is convincing evidence that investors do not necessarily understand precisely what is communicated by accruals. Sloan (1996) finds that future abnormal stock returns are negative (positive) for firms whose earnings include positive (negative) current
1
For example, managers must estimate lives and salvage values of long-term assets, pension benefit obligations, losses from bad debts, asset impairment, and deferred taxes. Moreover, managers choose among different methods for reporting depreciation and inventory, for managing working capital, and for structuring corporate transactions such as a business combination. See the survey article by Healy and Wahlen (1999) for a more complete discussion. 2 For example, Dechow (1994) finds that current earnings are better than current cash flows in predicting future cash flow and Subramanyam (1996) finds that discretionary accruals are value-relevant.
1
accruals. Collins and Hribar (2000), using quarterly earnings data, reports similar evidence that the market overprices accruals. Xie (2001), uses a modification of the Jones (1991) model to estimate normal (benchmark) accruals and finds that Sloan’s results are due primarily to the “abnormal” or “discretionary” component of accruals.
Such overpricing could reflect an
inability of the market to recognize the mean reversion in earnings due to the accruals component. While evidence of mispricing raises concerns about the usefulness of discretionary accrual accounting, it does not necessarily imply opportunistic manipulation by managers. To get at this issue, a number of studies have focused on firm-specific events that potentially provide strong incentives for management to manipulate market price by managing accruals. There is considerable evidence that manipulation occurs around such events3 and growing evidence that the market does not see through this type of manipulation. Of particular interest is recent evidence in Teoh et al. (1998a, 1998b) and Rangan (1998) that managers aggressively manage accruals prior to initial public offerings and seasoned equity offerings, that the market overprices these accruals, and, thus, that the market overvalues the new issues. A follow-up study by Teoh and Wong (2002) examines the role of financial analyst credulity in this process. The results are consistent with the hypothesis that analysts are more overoptimistic about firms with larger pre-issue accruals and investors are not aware of the full extent of this bias. In other words, around equity issues investors rely on analysts who, in turn, are misled by accruals manipulation. It is particularly striking that abnormal negative post-announcement stock price
3
Healy (1985) documents earnings manipulation in connection with managerial bonus compensation. There is strong evidence of earnings management by banks of loan loss provisions (Beaver et al. (1989), Moyer (1990), Scholes et al. (1990), and successors) and by insurers of property and casualty claim loss reserves (among others, see Petroni (1992), Beaver and McNichols (1998)). Perry and Williams (1994) finds that unexpected accruals are income-decreasing (negative) prior to an MBO. There is also evidence that managers tend to report positive abnormal accruals prior to stock-financed acquisitions (Erickson and Wang (1999)).
2
performance persists for up to five years and predictable (from prior accruals) analyst forecast errors have significant power to explain the long-run underperformance.4 Thus, the evidence in these studies suggests that analysts and investors are boundedly-rational in that they are slow to recognize and unravel these accounting manipulations once the equity issue has been announced and the incentive to manage accruals prior has been revealed. In this paper, we investigate the extent of analyst and investor rationality in a unique experimental setting in which the incentive to manage earnings in the future should be apparent ex ante to market participants. In particular, we examine earnings management around the cancellation and subsequent reissue of executive (and other employee) stock options. A firm that alters the structure of managerial compensation using this method first cancels the outstanding options and then, after a clearly- and publicly-specified period of time, reissues new options.5 In the typical case, the firm announces the plan to reissue and then allows a month or so for employees to decide how many options to tender. This offer period is then closed, and six months and one day later the options are reissued with the strike price set at the then-current (reissue day) market price. Since managers of such “6-and-1” firms benefit from a lower strike price for the reissued options, investors and analysts should be able to anticipate managerial attempts to manage accruals downward prior to the reissue date.6 7
4
These findings suggest that either investors make the same sort of mistakes as analysts or analysts’ errors are impounded in market prices by investors who rely on analysts for guidance. 5 Recent accounting changes [FASB FIN 44 (March 31, 2000) and the December 4, 1998 FASB announcement foreshadowing FIN 44] have given rise to this method for dealing with underwater, out-of-the-money employee stock options. Alternatives are simply to issue new options and/or reprice existing options. 6 When the event itself may not be anticipated (in contrast to a cancellation and reissue), as in the case of equity issues, standard executive option repricing, or control contests, market participants will not know prior to the announcement of the event that managers had the incentive to manage accruals. Only at the announcement of the issue or control contest, which will be after the manipulation of earnings and market price occurred, will market participants have confirmation of the incentive to manipulate. 7 The New York Times article (March 15, 2002), “Option absurdity: Hoping for lower prices,” expresses the commonly held suspicion that a cancellation and reissue provides the perverse incentive for employees to drive down the stock price by the time of reissue.
3
Based on a sample of 128 reissues, we find strong evidence of abnormally low accruals leading up to the option reissue date.
Reissuing firm abnormal accruals are significantly
negative and significantly lower than abnormal accruals for a control sample of firms with similar characteristics, specifically, firms that reprice their employee options in the traditional manner. Thus, even in a setting where investors are aware of the incentives to manipulate stock price, it appears that managers still attempt to manage accruals to their own benefit. Our analysis of contemporaneous and future stock returns, however, suggests that analysts and investors are not misled. Abnormal accruals between the announcement date and the reissue date have little power to explain stock price performance over the nine months prior or the six months following the reissue date. The same is true for the collection of five-day windows surrounding quarterly earnings announcements prior to the reissue date – the estimated abnormal accruals response coefficient is insignificantly different from zero.
Moreover,
abnormal accruals have no power to explain subsequent analyst forecast errors. Thus, in contrast to evidence from earlier studies where the incentives to manipulate are not necessarily foreseeable, in our setting, in which earnings management can be anticipated, neither analysts nor investors are fooled by manipulation of accruals. One interpretation of our results is that they are consistent with the hypothesis that the simplicity of our setting, based on cancellation and reissue of executive stock options, is well within the bounds of rationality of investors and analysts. That is, market participants possess cognitive capacity that exceeds the demands of our setting, in which it is relatively easy to evaluate incentives, accruals, and market price. In contrast, in other studies, such as Teoh, Welch and Wong (1998a,b) and Teoh and Wong (2002), investors and analysts appear to be unable to recognize the accounting deception
4
ex ante and, apparently, are slow to unravel the deception ex post. One possibility is that investors and analysts have reasonable cognitive capabilities but incentives and accounting manipulation around these particular corporate events are not particularly transparent. Another possibility is that investors and analysts, limited by behavioral biases and lack of ability to penetrate and comprehend the situation, can be misled by simple accounting deceptions. Our evidence provides more support for the first of these possibilities, that investors possess at least some sophistication, and less support for the second. Section 2 describes the details of resetting the strike price of employee stock options by cancellation and reissue. Section 3 describes how we select the sample of cancellations and reissues, collect the control sample of standard employee stock option repricings, and construct measures of earnings management. Section 4 reports the results on the extent of earnings management and Section 5 presents results on the effects of such management on analyst forecasts and market prices. Section 6 checks for the robustness of our results using a control sample of firms that reprice stock options prior to the December 1998 FASB announcement. Section 7 concludes.
2. Earnings management and cancellation/reissue of employee stock options 2.1. Institutional background on cancellation and reissue of stock options The Financial Accounting Standards Board (FASB) announced on December 4, 1998 that it was planning to implement detailed standards on how companies must account for their employee (including executive) stock options. On March 31, 2000, the FASB officially issued Interpretation No. 44 (FIN 44), under which repricing is considered a modification of an option. Thus, under the new rules, any option repriced after July 1, 2000 must be marked to the market
5
every accounting period from the repricing date through the date of exercise (or expiration, if left unexercised).
The rules, while applicable to all options granted after July 1, 2000, are
retroactively applicable to options that were repriced after December 15, 1998.8 Prior to the rule change, firms that repriced employee stock options were not required to record compensation charges when the stock price moved above the reset (lowered) exercise price. Under the new rules, this is no longer the case. The repricing can imply significant compensation charges over the option life if the stock price moves above the strike price at which the options are reset. Companies that cancel options and then wait for at least six months and one day before reissuing new options can avoid recording compensation charges resulting from FIN 44.9 This is done by way of an exchange agreement between the company and the option holder. The Securities and Exchange Commission (SEC) Exchange Act of 1934 considers option exchanges to be issuer tender offers. A company must comply with Rule 13e-4 under the Securities Act, and provide information required by Schedule TO with the SEC. In this exchange offer, firms provide the option holders the choice to tender their options for cancellation and subsequent reissue. Figure 1 shows the timeline associated with this procedure. In the typical case, the firm closes the offer a month or so after the initial announcement/filing. The cancellation date is usually the day following the closing of the offer. The firm typically sets the reissue date at six months and one day following the cancellation date. Firms commit to reissue new “at-themoney” options on the reissue date. The date of the reissue and the percentage of employees who tendered their options are communicated by way of a Schedule 14(d)(1) filing. For ease of
8
The period from December 15, 1998 to July 1, 2000 is considered a transition period. Any repricing during this period was subject to prospective marking-to-the-market relative to the stock price as on July 1, 2000. Carter and Lynch (2002) documents an abnormally-high rate of repricings during the 12-day window between the announcement date (December 4, 1998) and the proposed effective date (December 15, 1998). 9 Under FIN 44, repricing is deemed to have occurred if the grant of new options and cancellation of existing options occur within six months of each other.
6
exposition, we often refer to the date of reissue of stock options, for firms that cancel and reissue stock options, and the repricing date for firms that reprice existing options, as the strike price reset date. 2.2. Prior literature on timing of stock option grants and stock option repricing We are not aware of any literature on earnings management around the cancellation and reissue or the repricing of stock options. Nevertheless, there is some evidence on timing of executive stock option grants (Aboody and Kaznik (2000), Yermack (1997)) and repricings (Callaghan, Saly, and Subramaniam (2002)). Callaghan et al. (2002), using data over the period 1992 through 1997, tests whether firms time executive stock option repricing around favorable stock price movements. In particular, Callaghan et al. finds that the repricing dates are set such that they precede favorable earnings announcements or follow unfavorable earnings releases. The authors interpret the run-up in stock price immediately following the repricing as evidence of opportunistic managerial behavior rather than positive incentive effects.10
3. Sample selection and measurement of earnings management 3.1. Sample selection and sample characteristics Our sample consists of 128 firms that cancelled and reissued employee stock options. To obtain this sample of 6-and-1 firms, we use Primark’s Global Access (Disclosure) to identify 10
Several papers examine the cross-sectional determinants of stock option repricing (see Carter and Lynch (2001), Chance, Kumar and Todd (2000), Chidambaran and Prabhala (2002)). Brenner, Sundaram and Yermack (2000) derives a model that values stock options with a repricing feature, and find that the ex post gain in option value arising from repricing is substantial. Johnson and Tian (2000) examines value and incentive effects of nontraditional options, and observes that stock options with repriceable features have lower deltas and higher vegas than traditional options. Acharya, John and Sundaram (2000) derives conditions under which some amount of repricing is beneficial. Jin and Meulbrook (2001) suggests that the sensitivity of executive stock option value to share price does not fall much with a decline in stock price. Driving this stickiness is the longer maturity of the options and higher stock price volatility. Subramanian, Chakraborty, and Sheikh (2003) finds that managerial retention is a motive for repricing. Rogers (2002), Coles, Daniel, and Naveen (2003), and Kalpathy (2003) find that realignment of incentives, particularly managerial risk-taking incentives, is a motivation for resetting the option strike price.
7
firms that undertook an option exchange program in 2001 or the first quarter of 2002.11 Of these 128 firms, 42 firms also made a public announcement of the exchange program that appeared on Dow Jones Newswire. Cancellation and reissue announcements almost always precede the exchange offer filing and always precede the actual cancellation.12 We collect a control sample of firms that reasonably could have canceled and reissued options, but chose instead simply to reprice existing stock options in the traditional way. The primary reason for using repricing firms as a control sample, a reason which we discuss in more detail below, is to compare our 6-and-1 firms to firms with similar characteristics, including industry, prior performance, and the incentive to reset employee stock option strike prices. To obtain the control sample of firms that repriced employee stock options, we search Primark's Global Access.13 For the period 1999-2001, 70 firms repriced employee stock options, of which 64 also have CRSP and Compustat data available. In each case, we obtain the repricing date from the relevant proxy statement. We obtain stock price data through the end of 2001 from the Center for Research in Security Prices (CRSP) database, accounting data from Standard & Poor’s COMPUSTAT database, and analyst forecasts of quarterly earnings per share from the Institutional Brokers Estimate Systems (IBES) database. We obtain returns data for 2002 from Datastream. Panel A of Table 1 shows the distribution of cancellation and reissue dates by calendar year and quarter. Panel A also contains the distribution of repricing dates for the control sample. 11
We conduct a text search in Primark’s Global Access (Disclosure) using the search string (“six month!” w/3 “one day”). We specify “Filing Type = SC 14D1”. This yields a total of 138 firms. Of these, CUSIP numbers are available for 129 firms. Kana Software Inc. canceled their exchange offer because of its merger with Broadbase Software Inc. The resulting sample size is 128. We find no events for the year 2000 and skip the last three quarters 2002 because some data are not yet available. 12 All announcements either preceded or coincided with the date of exchange offer filing, with the exceptions of Freemarkets Inc. and Meta Group Inc. where the announcement took place four days and one day following the filing, respectively. 13 We conduct a text search in Primark’s Global Access (Disclosure) using the search string (“year w/5 (option! repric!)”).
8
During 1999 and 2000 there were a total of 52 repricings and no cancellations and reissues. Starting in 2001, there is a significant shift from repricings to cancellation and reissues. Entering 2002, cancellation and reissue appears to have completely displaced traditional repricings, with a quarterly volume that exceeds (matches) the quarterly rate of repricings during 1999-2000 (1998 and earlier).14 The switch from repricing to cancellation and reissue appears to have been relatively slow. Although the formal FASB announcement came in March 2000, the proposed change was announced in December 1998. Yet the move to cancellation and reissue gained impetus only in the second quarter of 2001. Finally, Panel A reveals that the time pattern of reissue is consistent with a gap of at least six months after the cancellation of the original options. Panel B of Table 1 reports the industry distribution of the reissuers and repricers. For the reissuers, we observe that 50% of the firms are clustered in the “business services” industry segment. These are firms primarily engaged in software development and networking services. Carter and Lynch (2001) documents that over 54% of their sample consists of firms that belong to the high-technology industry, while the number is about 38% in Chidambaran and Prabhala (2002).15
Other than in business services (SIC 73), reissuers and repricers are distributed
similarly across industries (p(chi-squared) > 0.05). Table 2 reports statistics on various characteristics of the reissuing (column 1) and repricing (column 2) firms over the fiscal year ending most recently prior to the strike price reset date. In addition, column 4 of Table 2 reports the same statistics for reissuing firms for the fiscal 14
Carter and Lynch (2001) obtains a final sample of 22 companies that reprice executive stock options in 1999 (after the December 1998 FASB announcement); Chidambaran and Prabhala (2002), and Brenner, Sundaram and Yermack (2000) analyze samples containing an average of 36 and 34 (respectively) companies per year that reprice. 15 The difference in percentage of firms in high-technology industry in these two papers may be due to differences in sample selection. In Chidambaran and Prabhala (2002), the sample is limited to firms that are covered by EXECUCOMP, while in Carter and Lynch (2001) non-EXECUCOMP firms are also included in the sample. See Carter and Lynch (2001) for explanation.
9
year ending most recently prior to the cancellation/reissue announcement date, which typically precedes the reissue date by seven months.
Basing the calculation for reissuers on the
announcement date, because there is no prior announcement in a traditional repricing, allows a comparison of reissuers and repricers oriented around the effective announcement date when the market learns that the option strike price will be reset. Property, plant & equipment (PPE), change in sales, change in receivables, and operating cash flow all are deflated by beginning-of-period total assets. Return on assets (ROA) is defined as income before extraordinary items divided by total assets, market-to-book is (total assets common equity + market value of equity) divided by total assets, and leverage is (long-term debt + short-term debt) divided by total assets. Analyst forecast error is the average of the last four quarters [(actual EPS - consensus EPS estimate)/prior period stock price], and analyst following is the average over the last four quarters of the number of analysts following the company. Stock return is the compounded total stock return over the fiscal year ending most recently prior to the relevant date, while market-adjusted-return is the buy-and-hold market-adjusted return using the CRSP value-weighted index. Column 3 of Table 2 contains the results of statistical tests comparing 6-and-1 firms with repricers for data defined in reference to the strike price reset date. As measured by ROA, earnings per share, operating cash flow, stock return, and market-adjusted stock return, both repricers and reissuers, as expected, have poor prior performance. Reissuers have higher prior market-adjusted stock returns than repricers, but for no other measure of performance, including ROA and sales growth, is there a significant difference. In univariate comparisons, reissuing firms tend to have higher analyst following and lower leverage, PPE, and receivables growth than repricers. Otherwise, reissuers and repricers are quite similar. Based on column 5 of Table
10
2, which contains the results of statistical tests comparing 6-and-1 firms with repricers for data defined in reference to the reset announcement date, the results generally are similar.16 The exceptions are, based on the announcement date, that for reissuers prior sales growth is higher and prior raw stock return is lower. 3.2. Measuring earnings management There is considerable debate surrounding the measurement of abnormal accruals. Dechow, Sloan and Sweeney (1995) suggests that the modified Jones (1991) model provides the most power in detecting earnings management.
We follow this approach in measuring
nondiscretionary accruals. We model accruals as:
ACCRi ,t / TAi ,t −1 = α 1 [1 / TAi ,t −1 ] + α 2 [∆REVi ,t / TAi ,t −1 ] + α 3 [ PPEi ,t / TAi ,t −1 ] + ε i ,t where ACCRi,t is total accruals17 for firm i in quarter t, TAi,t-1 refers to beginning of quarter total assets, ∆REVi,t refers to change in net revenue, and PPEi,t refers to property, plant, and equipment. We estimate the above cross-sectional regression separately for each quarter and twodigit SIC code. Based on the above model, predicted or nondiscretionary accruals is given by: ∧
∧
∧
PREDACCRi ,t = α 1 [1 / TAi ,t −1 ] + α 2 [(∆REVi ,t − ∆REC i ,t ) / TAi ,t −1 ] + α 3 [ PPEi ,t / TAi ,t −1 ]
where ∆RECi,t refers to change in net receivables, and a hat indicates that the parameter has been estimated. This version of the Jones model implicitly assumes that all changes in receivables in the event period result from earnings management. Therefore, the change in receivables is
16
Reissuers are more likely to be business services (SIC 73) firms than are repricers (Table 1, Panel B). We compare reissuers in SIC 73 versus reissuers in all other industries on the dimensions in Table 2. We find no significant differences. 17 Total accruals is measured as the change in current assets (Compustat quarterly data item #40) minus change in current liabilities (#49) minus change in cash and cash equivalents (#36) plus change in debt included in current liabilities (#45) minus depreciation and amortization expense (#5).
11
removed from total sales in computing the nondiscretionary accruals.
Discretionary (or
abnormal) accruals is defined as the residual: ∧
ABNACCRi ,t = ACCRi ,t / TAi ,t −1 − PREDACCR i ,t
The (original) Jones model is the same as above except there is no adjustment for change in receivables when calculating nondiscretionary accruals (PREDACCR). While the use of the modified Jones and the original Jones models to estimate discretionary accruals is widespread (see McNichols (2000)), there are concerns about specification, accuracy, and power (e.g., Guay, Kothari, and Watts (1996) and Fields, Lys, and Vincent (2001)), particularly when firm performance is extremely high or low (Dechow, Sloan, and Sweeney (1995)). Accordingly, Kothari, Leone, and Wasley (2002) examines the properties of various measures of discretionary accruals when controlling for performance (return on assets) and industry (2-digit SIC) with firms matched on those characteristics.
They find that
discretionary accruals estimated from both the Jones and modified Jones models, adjusted for discretionary accruals from a performance-industry-matched firm, are well-specified. Our use of a control sample represents a similar approach. In particular, the control sample of repricing firms is chosen to match to important characteristics of the sample of reissuers. Tables 1 and 2 indicate that the control sample of repricers is reasonably similar to the sample of reissuers in ROA and industry representation (except for SIC 73), as well as in other respects. Table 2, however, is based on univariate comparisons. In order to isolate differences between the sample of reissuers and repricers while controlling for other characteristics, we estimate several logit models. The implicit indicator variable takes the value of one if the firm
12
reissues and zero if the firm reprices in the traditional way. The right-hand-side variables are as defined in Table 2. Table 3 reports the results. Aligning the data on the option reset date, models 1 and 2 show that control firms and reissuers have similar performance using a broad set of measures, including ROA, operating cash flow, and sales growth. Reissuers and repricers also are roughly similar in terms of analyst forecast error, market value of equity, and market to book. Reissuers, however, have significantly (at the 5% level) higher prior raw stock return, higher marketadjusted stock return, lower PPE, higher analyst following, and lower leverage relative to repricers. Models 3 and 4, which align the repricing date for repricers with the announcement date for reissuers, yield roughly similar results. Repricers and reissuers have similar ROA, operating cash flow, and sales growth, as well as similar market value of equity and market-tobook (of assets). Unlike the results in models 1 and 2, raw and market-adjusted stock return are both similar across 6-and-1 firms and repricers. Analyst forecast error, leverage, PPE, and unadjusted stock return are lower for reissuers than for repricers, but analyst following is higher. The results in Table 3 are essentially the same when we include an indicator variable for SIC 73 (business services). Moreover, these results broadly are consistent with those of Zheng (2003), who, based on a more parsimonious specification, finds that reissuers have higher analyst following and lower book-to-market (of equity).18 In summary, using repricers as a comparison sample appears to tackle several concerns. Repricers and reissuers tend to come from the same industries and have the same prior ROA, in which case, per Kothari, Leone, and Wasley (2002), the modified and original Jones models should be reasonably well-specified. In addition, firm characteristics likely to be associated with the benefits and costs of resetting the option strike price, such as the extent of incentive 18
Also see Kalpathy (2003) for a comparison of repricers and 6-and-1 firms using qualitative response models.
13
realignment, employee retention, and wealth transfer to employees, are similar for reissuers and repricers. For example, recent prior performance and market-to–book, which are likely to be correlated with all three of these motives for resetting strike price, are alike for reissuers and repricers. The results from Table 3 suggest it is not particularly likely that firm characteristics drive both the decision to reissue (versus reprice) and the use of discretionary accruals.
4. Earnings management around reissues and repricings
4.1. Modified Jones discretionary accruals prior to reissues and repricings We investigate discretionary accruals for the two samples prior to the strike price reset date. Again, the strike price reset date is the date of reissue of stock options for firms that cancel and reissue stock options and the repricing date for firms that reprice existing options. We focus on the four quarters prior to the reset date to examine the time-series pattern of accruals management for these firms. We scale discretionary, nondiscretionary and total accruals by total assets at the beginning of the quarter and, following Rangan (1998), we report return on assets (ROA) over the same period in order to provide a sense for the magnitude of these accrual measures. Panel A of Table 4 reports asset-scaled discretionary accruals.
For the sample of
reissuers, mean and median discretionary accruals are significantly negative over the year prior to the reissue date. The t-test comparing discretionary accruals to zero is significant in quarters -3 (p = 0.011), -2 (p = 0.000), and -1 (p = 0.006) relative to the reissue date, as well as for all four prior quarters pooled (t = -6.03, p = 0.000). Mean discretionary accruals accumulated over all four quarters prior to the reissue date equals -0.075 with a t-statistic of -5.93 (p = 0.000). Based on the Wilcoxon signed-rank test, median discretionary accruals are significantly less than
14
zero at the one percent level in quarters -4 (p = 0.000), -3 (p = 0.000, -2 (p = 0.000), and -1 (p = 0.000), pooled (p = 0.000), and cumulated (p = 0.000) relative to the reissue date. The negative discretionary accruals prior to the reissue date are large in economic terms. Based on Panel F of Table 4, the mean abnormal accruals constitute about 23%, 47%, 22% and 17% of ROA for quarters -1, -2, -3 and -4. Quarterly abnormal accruals pooled over the entire year average 26% of quarterly ROA.
The magnitude of discretionary accruals appears to be
considerable for the sample of reissuers over the year prior to the reissue date. In sum, reissuing firms appear to manage discretionary accruals downward in advance of the option strike price reset date. In contrast, it appears that repricing firms do not manage discretionary accruals downward prior to the strike price reset date. For repricing firms, mean abnormal accruals are close to zero prior to the repricing date. Median abnormal accruals tend to be negative but insignificant. As a percent of ROA, abnormal accruals are not as large. 4.2. Discretionary accruals for reissuers compared to repricers While the results for repricers are interesting on their own, the primary motivation for collecting data on these firms is to construct a natural control sample for our experiment on earnings management in a transparent environment. The key feature in this regard is that traditional repricings differ from reissues in that, for a cancellation and reissue, investors have been informed of the strike price reset in advance by way of an SEC exchange offer filing and/or a public announcement. Equally important, the use of repricers as a control allows us to address specification problems as described in Kothari, Leone, and Wasley (2002); we evaluate discretionary accruals management of reissuers relative to repricers, thereby controlling for industry, prior performance (ROA), and other firm characteristics
15
Panel A (Table 4) reports tests for differences in discretionary accruals between the two samples. The Wilcoxon Z-statistic from the rank-sum test is significant at the 1% level for cumulative scaled abnormal accruals and abnormal accruals averaged across all quarters. The ttest for the difference in means yields the same inferences at the 5% level. Of the four prior quarters, the statistical significance of the difference is most prominent in quarter -2 (p < 0.01 for both tests).
Relative to repricers, reissuing firms appear to manage discretionary accruals
downward over the period prior to the option strike price reset date. Overall, the results in Panel A indicate that abnormal accruals are consistently negative and economically significant for reissuers for the four quarters leading up to the reissue of stock options. 19 It is reasonable to suppose that repricers have at least as strong an incentive to manage accruals downward as reissuers. Repricers do not announce the option strike price reset in advance, so it could be more likely that investors and analysts can be misled and accruals management would have the impact on stock price desired by management.
The results,
however, show that reissuers manage discretionary accruals downward more aggressively than repricers.
Perhaps repricers, because of flexibility in timing, have less need to manage
discretionary accruals to reset strike price at a low level. Or, because investors don’t see it coming, repricers get a larger market response out of the same abnormal accruals and, thus, manage accruals downward but so slightly that our empirical methods fail to detect it. Finally, if abnormally low accruals serve as a signal to investors and analysts that a repricing is forthcoming, investors then will trade on this expectation, thereby moving market price, which would interfere with the ability to time the strike price reset at the bottom. Perhaps firms that
19
We find no strong evidence that the results on accruals are driven by business services firms. Abnormal accruals for resissuers tend to be lower for business services (SIC 73) reissuers than for reissuers in other industries, but the results of statistical tests are mixed. Comparing abnormal accruals for reissuers and repricers, when we exclude SIC 73 firms our tests again indicate that reissuers have significantly lower discretionary accruals than repricers.
16
expect to reprice existing options avoid signaling that intention by exercising restraint in managing discretionary accruals. 4.3. Additional discussion and robustness checks Having presented results based on what we believe to be the most appropriate method (modified Jones model with a control sample), we return to examine some of the issues associated with use of the modified Jones model in this empirical context. 4.3.1. Are nondiscretionary accruals low because performance is poor? An alternative to the downward manipulation explanation is that the negative discretionary accruals reflect poor firm performance. These firms are performing poorly (as suggested by the decline in their stock price and hence their selection in the sample as a reissue firm) and the discretionary accruals could, because of measurement error, capture some performance related nondiscretionary accruals.
To partially address this concern, for both
samples we report total accruals in Panel B and nondiscretionary accruals in Panel C. Panel B indicates that mean and median total accruals are negative for both reissuers and repricers, though some of the tests do not reject the null that the measure of central tendency is zero. Total accruals for reissuers are statistically indistinguishable from those of repricers.
Because
discretionary accruals are lower for reissuers, this suggests that nondiscretionary accruals are larger for reissuers. Panel C confirms this supposition. Nondiscretionary accruals for repricers and reissuers generally are significantly negative but tend to be significantly higher (not lower) for reissuers than for repricers. These results suggest that, while poor performance plays a role, the negative discretionary accruals we document in Panel A are unlikely to be driven in large part by prior poor performance of the reissuing firms. We now explore more directly the question of poor prior performance and the usefulness of the modified Jones model.
17
4.3.2. The unmodified Jones model The modified Jones model involves regressing accruals on revenue changes and then predicting nondiscretionary accruals using (i) revenue changes minus the change in accounts receivable and (ii) the coefficient estimate from the original regression. This causes all of the effect of a change in accounts receivable to be placed in discretionary accruals despite the fact that it is highly unlikely that all of the receivables change is due to manipulation (i.e., is discretionary).
Indeed, it is extremely likely that most of the receivables change is
nondiscretionary. To see the full potential effect of this bias assume the receivables change is non-discretionary and positive.
Then nondiscretionary accruals are understated by the
receivables change times the positive estimated α2 coefficient and discretionary accruals are overstated by the same amount. This suggests several possibilities. First, if indeed the reissue firms are good performers (in the sense of increases in sales and hence accounts receivable), then the discretionary accruals are even more negative than reported. Moreover, if reissue firms are better performers than repricing firms, their abnormal accruals would be more overstated than those of the repricing firms going against the observed negative difference in Panel A, i.e., suggesting that abnormal accruals of reissuers relative to repricers are even more negative than reported.
Second, if reissuing firms are worse performers than the repricing firms, their
abnormal accruals would be more understated causing a bias in favor of the results reported in Panel A. To explore these possibilities, in Panel D of Table 4 we report asset-scaled revenue changes for the four quarters prior to the reset date for both samples. Pooled across all four quarters the reissuing firms have a negative and significant change in revenues. Given the defect in the modified Jones model, this finding suggests that the negative discretionary accruals we
18
document for the reissuing firms (Table 4, Panel A) may, in part, be driven by the poor performance of the reissuing firms. The reissuing firms also appear, in a univariate comparison, to be worse performers than the repricing firms. The pooled difference in scaled revenue changes is –0.017, which is significant at the 1% level. This result suggests that our finding that reissuers report significantly more negative accruals than repricers over the year prior to the reset date may similarly reflect the bias in the modified Jones model. To address concerns about potential bias, we generate results using the original Jones model. The results, which are reported in Panel E of Table 4, are qualitatively similar to those reported in Panel A. Discretionary accruals for reissuers tend to be significantly negative, while there is little evidence of negative abnormal accruals for repricers. Comparing the two samples, reissuing firms tend to report discretionary accruals that are significantly lower than repricing firms. The Jones and modified Jones models yield similar inferences. Thus, throughout the remainder of the paper we rely on the modified Jones model and the control sample of repricers. To check robustness of our results, however, we repeat all empirical analysis using the original Jones model and the control sample. Except where noted, results from these two models are qualitatively the same.
5. Are analysts or investors misled?
The evidence above suggests that managers manipulate discretionary accruals prior to the reissue of stock options. We now consider whether investors and analysts are fooled by these manipulations. In particular, we test whether managerial attempts to manage earnings downward are successful in driving down analyst forecasts and stock price prior to the reset date.
19
5.1. Performance patterns around the option price reset date The raw and market-adjusted stock price performance of firms both before and after the reset date is displayed in Figures 2a (cancellation and reissues) and 2b (repricings). For 6-and-1 firms there is a run-up in stock price starting 50 trading days prior to the reset date and a decline after the reset date. This pattern is not consistent with a successful manipulation of stock price through accruals management. If anything, the pattern of returns is just the opposite of the pattern that would benefit option holders in terms of resetting the exercise price. While similar incentives would seem to apply to the repricing of existing stock options, we find little evidence of earnings management by these firms. Nevertheless, Figure 2b shows that stock price declines prior to the reset date and increases thereafter. These results are similar to the findings in Callaghan, Saly and Subramaniam (2002) based on an earlier sample (19921997, prior to the new FASB rule) of repricings. There are several possible explanations. One is that accruals management by repricers, though undetected by our empirical methods, helps management reset the option strike price at a low point. Also, the pattern of returns around the reset date is consistent with either outstanding timing by managers, without having to resort to accruals management, or the capitalization of positive net incentive effects associated with the reset of the exercise price.20 Based on this last possibility, for firms that cancel and reissue, the positive stock performance between trading days –175 and –120 (Figure 2a) potentially is due to the anticipated positive net incentive effects associated with the option price reset. It is during this time window, six to eight months prior to the reset date, or equivalently 120 to 175 trading days prior to the reset, that announcement (and/or filing) of the plan to cancel and reissue occurs.
20
See Acharya, John, and Sundaram (2000) for a discussion of the tension between the ex ante and ex post incentive effects of repricing executive stock options.
20
5.2. Does the market price pre-reset date abnormal accruals? We now test whether accruals management over the three quarters prior to the reset date has any impact on stock price over the same period. The left-hand-side variable in the regression analysis is CARi, which is the compounded stock return for firm i net of the value-weighted index (TOTMKUS from Datastream), both adjusted for dividends and splits. We rely on Datastream because CRSP returns are not available for some of our sample period. We calculate the cumulative abnormal stock return over the nine months prior to the option strike price reset date. We also aggregate each explanatory variable over three quarters. The reason, for both the dependent and independent variables, is that, for all firms in our sample that cancel and reissue, the period from the cancellation announcement date to the reissue date contains exactly three quarterly earnings announcements. Elapsed calendar time from the first to the last quarterly announcement date tends to be slightly more than six months.
Thus, we cumulate each
accounting-based explanatory variable, as well as the dependent return variable, across the three earnings announcements between the announcement of the cancellation and reissue and the reset date. For traditional option repricings, we follow the same procedure relative to the option reset date. Thus, again, all of the explanatory variables are cumulative over the three fiscal quarter announcements preceding the reset date. Because Subramanyam (1996), in comparing various estimated specifications, finds that scaled operating cash flow has power to explain cumulative annual stock returns, we control for scaled operating cash flow, aggregated over the three quarters prior to the reset date. In order to measure earnings surprises, we also include the cumulative change in earnings over the three quarters prior to the reset date (CUMCHEPS) as an independent variable. This measure is based
21
on a simple random walk model of earnings and is calculated as the accumulation across the three quarters of the change in EPS scaled by prior period stock price. In some specifications we use the “analyst forecast error” (AFE) as a measure of earnings surprise. We obtain data on analyst EPS forecasts from the IBES detailed estimates tape. CUMAFE is the aggregation across three quarters of quarterly actual EPS minus consensus estimate of EPS scaled by priorquarter stock price.21 Cumulative abnormal discretionary and nondiscretionary accruals (CUMABNACCR and CUMPREDACCR respectively) are cumulative versions of abnormal and predicted accruals (as defined in section 3.2). Teoh and Wong (2002) argues that discretionary accruals most likely are contaminated by nondiscretionary components. We address the possibility of bias in regression coefficients by including nondiscretionary accruals in the model. To control for industry effects, we include dummy variables for industries, DIND, in some regression specifications. Whenever we do so, we use the 17 industries in the classification scheme of Teoh and Wong (2002). Table 5a reports results from regressing the cumulative abnormal returns over the period leading up to the reissue date on the above set of explanatory variables for the sample of reissuers. The estimated coefficient on abnormal accruals is negative and insignificant. If management can move market price up or down with higher or lower, respectively, abnormal accruals, we would observe a significantly positive coefficient.
Instead, the insignificant
coefficient suggests that the market is well aware of the incentives of managers to report lower accruals during this time period. Consistent with Subramanyam (1996), we find that operating cash flow has significant explanatory power; the estimated coefficient on cumulative operating 21
Though Teoh and Wong (2002) and DeFond and Park (2001) note that this definition of analyst forecast error is standard in the prior literature, AFE and CUMAFE are best thought of measures of EPS surprise relative to the benchmark of the consensus analyst forecast. Thus, these measures are the negative of forecast error, as would be conventionally defined, and negative (positive) values of AFE and CUMAFE represent analyst optimism (pessimism).
22
cash flow is positive and significant at the 1% level. Unlike Subramanyam (1996), however, we do not find evidence that the market attaches value to nondiscretionary accruals; the coefficient on cumulative normal accruals is not statistically significant at any conventional levels. Finally, we find that the cumulative earnings surprise variables have little explanatory power. Table 5b reports results obtained by pooling the reissue and repricing samples. To capture and test for differences between the two methods of resetting strike price, we include the variables from Table 5a as well as the interaction of those variables with a dummy variable indicating that the option reset was executed by cancellation and reissue (Dum = 1; 0 if repricing).
The estimated coefficient on abnormal accruals is now positive but, again, is
insignificant. Thus, as is the case for cancellation and reissues, the market is not influenced significantly by abnormal accruals of firms in advance of the repricing of existing options. The interaction of the cancellation and reissue dummy with the abnormal accruals measure is negative and, generally, statistically significant.
This suggests that in comparing the two
methods, when the reset date is known in advance, as it is in the case of cancellation and reissue, the market is significantly less responsive to manipulation of abnormal accruals than when the reset date is not explicitly revealed in advance, as with repricings. Still, regardless of the method used to reset the option price, the market does not appear to be misled by abnormal accruals. As in the previous table, normal accruals and earnings surprise variables have little explanatory power.22 5.3. Market reaction to earnings and accruals announcements prior to the reset date 22
Prior research such as Bhattacharya, Black, Christensen, and Larson (2002) indicates that pro forma earnings are frequently reported for business service firms and that the investors focus on pro forma earnings (even to their detriment (e.g., Doyle, Lundholm, Soliman (2003)) rather than bottom line earnings. If the discretionary component of accruals contains unusual items that are excluded from pro forma earnings, then this would be consistent with the lack of association we document between contemporaneous returns and unexpected earnings/discretionary accruals. Thus, the market and analysts may be ignoring the abnormally low accruals because management is encouraging them to do so. We currently are investigating this possibility.
23
In the tests presented in Tables 5a and 5b, we use a window containing three fiscal quarters prior to the reset date to test whether the market contemporaneously prices the abnormal accruals. For earnings and accruals surprises, however, the market is likely to respond more quickly. Using a longer window for returns, that contains other days in addition to earnings announcement days, could introduce noise in the returns measure and reduce explanatory power. Thus, we narrow the window over which we measure returns in order to estimate response coefficients for earnings, normal accruals, and abnormal accruals. In particular, we estimate the same specifications as in Tables 5a and 5b but use the five-day return from the window centered on the earnings/accruals announcement day (days –2 through +2 relative to the announcement day 0). Pooling all three quarterly announcement dates, we regress the five-day abnormal returns on measures of earnings surprise (change in earnings or analyst forecast error), operating cash flow, normal accruals, and discretionary accruals. All explanatory variables are as defined in the previous subsection but are measured separately for each quarter rather than cumulated across the three quarters. We also include a fiscal fourth quarter dummy to control for differential market reaction for fiscal fourth quarter earnings announcements. Table 6 presents results for three quarters preceding the reset date. Models 4-6 include both the reissuers and repricers, while Models 1-3 include the reissuers only. Again, the results show that abnormal accruals have little power in explaining announcement period returns. Specifically, the coefficient on abnormal accruals, the abnormal accruals response coefficient, is slightly negative and insignificant in all six models.
This finding suggests that market
participants are aware of the incentives of the managers to mark down accruals prior to the reissue of options.
24
An alternative explanation that applies both here (in Tables 5a, 5b, and 6) and in some of what follows is that our insignificant findings on accruals are due to measurement error. As Guay, Kothari, and Watts (1996) and others note, estimates of abnormal accruals using the modified and original Jones models can suffer from considerable imprecision. In reference to Table 6, we note that the significantly negative coefficient for predicted accruals is something of a puzzle. Perhaps the market attaches a negative interpretation to the components of the modified Jones model that are associated with a high predicted value of accruals.23 Predicted accruals loses significance when we use the original Jones model. 5.4. The relation between pre-reset abnormal accruals and post-reset stock returns Figures 2a and 2b show that reissuers (repricers) experience negative (positive) returns over the six months following the reset date. The empirical question is whether the post-reset returns are explained by pre-reset accruals in our cross-sectional setting. Teoh, Welch and Wong (1998a, 1998b) and Rangan (1998) address the analogous question for firms conducting seasoned equity offerings (SEO) and initial public offerings (IPO), find a strong negative relation between pre-issue accruals and post-issue long-run stock performance, and conclude that the market overprices the positive abnormal accruals of the issuing firms. We measure post-reset stock return as the compounded return for a firm net of the Datastream total market index. We compound daily returns for six months following the reset date. Pre-reset abnormal accruals (CUMABNACCR) are calculated exactly the same way as in section 5.2. Similarly, the other explanatory variables are calculated as in Section 5.2 and, until
23
These results are broadly consistent with the findings in DeFond and Park (2001). DeFond and Park (2001) partitions the earnings announcements into good and bad earnings news associated with income increasing accruals. They find that the coefficient on the forecast error for the income-increasing good news is negative and significant, suggesting that positive earnings surprises associated with income increasing abnormal accruals are, on average, associated with lower abnormal returns.
25
contemporaneous, post-reset-date data become available, those variables are defined over the three quarters prior to the reset date. Table 7a (7b) reports regression results for reissuers only (both repricers and reissuers). For reissuers, the coefficient on abnormal accruals, the variable of interest, is positive and significant at the five percent level. In manipulating stock price by using accruals, managers would hope for a rebound after the reset date, in which case stock return after the reset data would be negatively related to abnormal accruals prior to the reset date. We find the opposite relation. When we control for past analyst forecast errors, however, the relation is weaker. It appears that the stock price performance does not respond to past abnormal accruals in the direction that may help option holders recover value from their reset options. The coefficient on lagged abnormal accruals is insignificant at conventional levels. Table 7b shows that abnormal accruals prior to a repricing do not affect post-reset abnormal returns. The estimated coefficient on the interaction of abnormal accruals and the indicator for cancellation and reissue is positive and generally significant at the ten percent level. As in Table 7a (reissuers only), after controlling for past analyst forecast errors, we do not find significance at any conventional levels. Until we have additional data, the results in this section are preliminary. Thus, while some sort of explanation related to a delayed reaction to abnormal accruals or mean reversion in earnings due to discretionary accruals might be reasonable, it is inappropriate at this point to attach any strong interpretation. The results of Teoh, Welch and Wong (1998a and 1998b) and Rangan (1998) are based on post-issue performance over long horizons up to four years. When
26
the data become available, we expect to use contemporaneous (rather than lagged) explanatory variables to examine returns over longer horizons following the stock option reissue. 5.5. Analyst forecast errors and abnormal accruals Evidence in Teoh and Wong (2002) indicates that analysts are credulous to discretionary accruals reported by firms and do not incorporate properly the mean reversion in earnings on account of accruals when they issue their EPS forecasts. This credulity extends not only to firms that issue new equity, but also to non-issuer firms with high accruals. Teoh and Wong observe that the credulity of analysts is reflected in their forecasts for a full four years subsequent to the equity issues, and that the forecast errors that follow from discretionary accruals significantly explain long-run underperformance following the equity issue. Their evidence suggests that investors overvalue firms that issue equity on account of their reliance on the overoptimistic forecasts of analysts and that both analysts and investors are surprised when earnings of high accrual firms do not persist as they originally thought. Given the process by which reissuers conduct option exchange offers, financial analysts should be aware of the incentives of these firms to record lower accruals and to manage earnings downwards. Consequently, analyst EPS forecasts should not be overly pessimistic based on low abnormal accruals from earnings releases. Of course, tests based on forecast errors over a longer horizon (e.g., up to four years) following the option strike price reset will need to wait for more data, but some tests are possible now using available data. In particular, we examine the analyst EPS consensus forecasts that are issued for the first quarter subsequent to the option strike price reset date. Since most reissues are executed in 2002, in most cases we have data for only one quarter of reported earnings following the reissue.
27
Following Teoh and Wong (2002), we test whether analysts incorporate the mean reversion in accruals when they issue the forecasts. In particular, we regress analyst forecast errors on abnormal accruals, predicted accruals, and controls for size, industry, and fiscal quarter. The accruals variables, ABNACCR and PREDACCR are computed as defined in section 3.2. To maintain comparability with prior literature, we include controls for size (natural logarithm of market value, log(MV)), industry (industry dummies, DIND), time (year dummies, DYR), and differential forecast error for fiscal fourth quarter (fiscal fourth quarter dummy, Q4). As before, we use Dum = 1 to indicate that the firm cancels and reissues stock options (Dum = 0 if the firm reprices), and all explanatory variables are defined over the three quarters prior to the option strike price reset date. Recall that analysts' forecast error (AFE) is the actual EPS minus the consensus forecast of EPS, scaled by the prior period (end of quarter) stock price. Thus, AFE measures EPS surprise relative to analysts’ forecasts. We follow the literature and winsorize AFE at +/- 10% to make sure outliers do not affect the estimated regression coefficients. We use the detailed estimates tape from IBES to construct the consensus forecasts. Table 8 reports regression results. Models 1 and 2 use the sample of reissuers only, while models 3 and 4 pool repricers with reissuers. Consistent with prior evidence, we find that AFE is positively and significantly related to size. Analysts tend to be more pessimistic when the firm is larger. The coefficient on abnormal accruals is close to zero and not statistically significant. When we interact abnormal accruals with a dummy for reissuers, the sign is positive but, once again, not statistically significant. This is entirely consistent with the notion that analysts are not fooled by abnormally low accruals in this simple setting. A negative and significant coefficient for the abnormal accruals variable (so higher abnormal accruals would be associated with more
28
optimism in analysts’ forecasts) would indicate that analysts are credulous over an extended period to accruals manipulation. We do not observe this in the data. Abnormal accruals lack explanatory power and analysts are not misled by earnings manipulation around employee stock option reissues. Of course, measurement error also is a potential explanation for not detecting a relation.
6. Repricing prior to the December 1998 FASB announcement
Our control sample consists of firms that have the alternative to cancel and reissue employee stock options, but instead, choose simply to reprice the options.
One potential
problem with this design is that we might be picking up firms in our control sample that differ in some important ways from firms that chose to reprice prior to FASB’s announcement of FIN 44 in December 1998. For example, such firms prior to December 1998 did not have as strong an incentive to reset the executive stock option strike price by way of a cancellation and reissue. Moreover, because of the increase in stock prices over that period, the incentive to reset option strike price using any method was weaker. We address this concern by collecting a sample of 128 firms that repriced their employee stock options in the years 1996, 1997 and 1998.24 We investigate whether accruals (both total and abnormal) for this sample of “old” repricers differs from that for the reissuers and the “new” repricers in our original control sample (repriced after the FASB announcement in December 1998). Descriptive statistics for total accruals (Table 9, Panel A) indicate that the difference between the reissuers and the old repricers is statistically significant at the 1% level when we 24
We conduct a text search in Primark’s Global Access (Disclosure) using the search string (“year w/5 (option! repric!)”). This yields a total of 657 firms for the years 1996 through 1998. We randomly choose 128 firms from this list. Following Carter and Lynch (2001), we exclude firms that reprice stock options from December 4, 1998 to December 15, 1998 because of unusual repricing activity during this window in response to FASB’s announcement.
29
pool total accruals over the four quarters prior to the strike price reset date. We get similar results when we combine the two sets of repricers and compare them with the reissuers. This finding differs from the result, reported in panel A of Table 4, that the difference in total accruals between the reissuers and the new repricers is not statistically different from zero. This suggests that the old repricers seem to mark down accruals to a lesser degree when compared to the new repricers, though the difference in total accruals between new and old repricers is not statistically significant (Table 9, Panel A). We now turn to a comparison of abnormal accruals. Panel B of Table 9 indicates that the difference in abnormal accruals between the reissuers and repricers gets larger with the inclusion of the old repricers. As in Table 4, we find that the difference in pooled and aggregate abnormal accruals for the two groups is statistically different from zero at the 1% level. In contrast, we do not find any difference in accruals management between the old and new repricers. Nonparametric results for the difference between the reissuers and repricers confirm the results we obtain in the t-tests. We also check whether our cross-sectional regression results on stock returns and analyst forecast errors still hold when we include the old repricers. In unreported regressions, we estimate the specifications in Tables 5 through 8 for the expanded sample pooling both old and new repricers. The results using the expanded control sample, which includes firms that reprice employee stock options prior to the December 1998 FASB announcement, are very similar to those reported in Tables 5 through 8. 7. Conclusion
In this paper, we investigate the extent of analyst and market rationality around the cancellation and subsequent reissue of employee stock options. This event provides a unique
30
experimental setting in which the incentives to manipulate earnings should be absolutely transparent to market participants. There is an obvious incentive to manage the stock price downward over a well-defined period in advance of the reissue date because managers prefer options with lower strike prices. Moreover, investors and analysts, because the cancellation and reissue is announced at least six months ahead of the option strike price reset date, should be able to anticipate such managerial attempts to manage earnings and price. Using methods employed in the prior literature, we investigate whether managers still attempt to manage earnings in this simple environment, whether these attempts at manipulation have the desired effect on stock price, and the role of analyst credulity in the process. Based on a sample of 128 reissues, we find strong evidence of abnormally low accruals leading up to the option reissue date. Thus, even in a setting where investors are aware of the incentives to manipulate stock price, it appears that managers still attempt to manage accruals to their own benefit. Our analysis of current and near-term stock returns, however, suggests that this strategy is futile. Abnormal accruals between the announcement date and reissue date have little power to explain stock price performance over the nine months prior to the reset date and over the six months following the reset date. Moreover, abnormal accruals have little effect on stock price around the earnings announcement date. Prior to the reissue date, for the five-day window surrounding earnings announcements, the estimated abnormal accruals response coefficient is small and insignificantly different from zero. Moreover, abnormal accruals have no power to explain subsequent analyst forecast errors. Thus, in contrast to the evidence from earlier studies, where the incentives to manipulate are not necessarily foreseeable, in our setting, in which
31
earnings management can be anticipated, neither analysts nor investors are misled by manipulation of accruals. Our findings stand in sharp contrast to evidence in other studies, such as Teoh, Welch and Wong (1998a,b) and Teoh and Wong (2002), suggesting that investors and analysts are “naive” in that they are unable to recognize and very slow to unravel accounting deceptions. One possibility suggested by our findings is that investors and analysts are reasonably sophisticated but incentives and accounting manipulation around these particular corporate events are not particularly transparent. In this sense, the more transparent cancellation and reissue setting we investigate is well within the bounds of rationality of investors and provides evidence on the extent of investor and analyst gullibility. A more aggressive interpretation of our findings is that it shows that investors and analysts do see through accounting deceptions and, thus, something else is driving the results in the earlier studies. We leave a further investigation of these alternative interpretations to future study.
32
References Aboody, D., Kaznik, R., 2000. CEO stock option awards and the timing of corporate voluntary disclosures. Journal of Accounting and Economics 29, 73–100. Acharya, V., John, K., Sundaram, R., 2000. On the optimality of resetting executive stock options. Journal of Financial Economics 57, 65–101. Balachandran, S., Carter, M., Lynch, L., 2002. Sink or swim: Firm’s responses to underwater options before and after the accounting change for stock option repricing. Columbia University working paper. Beaver, W., Eger, C., Ryan, S., Wolfson, M., 1989. Financial reporting, supplemental disclosures and bank share prices. Journal of Accounting Research , 157–178. Beaver, W., McNichols, M., 1998. The characteristics and valuation of loss reserves of property casualty insurers. Review of Accounting Studies 3 (1-2), 73–95 Bernard, V., Skinner, D., 1996. What motivates managers’ choice of discretionary accruals? Journal of Accounting and Economics 22, 313–325. Bhattacharya, N., Black, E., Christensen, T., Larson, C., 2002. Assessing the relative informativeness and permanence of pro forma earnings and GAAP operating earnings. Journal of Accounting and Economics 36, XXX-XXX. Brenner, M., Sundaram, R., Yermack, D., 2000. Altering the terms of executive stock options. Journal of Financial Economics 57, 103–128. Callaghan, S., Saly, P., Subramaniam, C., 2002. The timing of option repricing. Texas Christian University working paper. Carter, M., Lynch, L., 2001. An examination of executive stock option repricing. Journal of Financial Economics 61, 207–225. Carter, M., Lynch, L., 2002. The consequences of the FASB’s 1998 proposal on accounting for stock option repricing. Journal of Accounting and Economics (forthcoming). Chance, D., Kumar, R., Todd, R., 2000. The repricing of executive stock options. Journal of Financial Economics 57, 129–154. Chidambaran, N., Prabhala, N., 2002. Executive stock option repricing, internal governance mechanisms, and management turnover. Journal of Financial Economics (forthcoming). Coles, J., Daniel, N., Naveen, L., 2003. Option repricing and incentive alignment. Working paper, Georgia State University.
33
Collins, D. and Hribar, P., 2000. Earnings-based and accrual-based market anomalies: one effect or two? Journal of Accounting and Economics 29, 101–123. Dechow, P., 1994. Accounting earnings and cash flows as measures of firm performance: The role of accounting accruals. Journal of Accounting and Economics 18, 3–40. Dechow, P., Sloan, R., Sweeney, A., 1995. Detecting earnings management. The Accounting Review 70, 193–225. DeFond, M., Park, C., 2001. The reversal of abnormal accruals and the market valuation of earnings surprises. The Accounting Review 76, 375–404. Doyle, J., Lundholm, R., Soliman, M., 2003. The predictive value of expenses excluded from “pro forma” earnings. Review of Accounting Studies 8, 145-174. Erickson, M., Wang, S., 1999. Earnings management by acquiring firms in stock for stock mergers. Journal of Accounting and Economics 27, 149–176. Fields, T., Lys, T., Vincent, L., 2001. Empirical research on accounting choice. Journal of Accounting and Economics 31, 255-307. Financial Accounting Standards Board. Financial Interpretation No. 44: Accounting for certain transactions involving stock compensation: an interpretation of APB Opinion No. 25. March 2000. Guay, W., Kothari, S. P., Watts, R. 1996. A market-based evaluation of discretionary accrual models. Journal of Accounting Research 34, 83–115. Healy, P., 1985. The effect of bonus schemes on accounting decisions. Journal of Accounting and Economics 7, 85–107. Healy, P., Wahlen, J., 1999. A review of the earnings management literature and its implications for standard setting. Accounting Horizons 13, 365–383. Jin, L., Meulbrook, L., 2002. Do underwater executive stock options still align incentives? The effect of stock price movements on managerial incentive alignment. Harvard Business School working paper. Johnson, S., Tian, Y., 2000. The value and incentive effects of nontraditional executive stock option plans. Journal of Financial Economics 57, 3–34. Jones, J., 1991. Earnings management during import relief investigations. Journal of Accounting Research 29, 193–228. Kalpathy, S., 2003. Incentive realignment and resetting the strike price of employee stock options. Working paper, Arizona State University.
34
Kothari, S., Leone, A., Wasley, C., 2002. Performance matched discretionary accruals. Working paper, MIT and University of Rochester. McNichols, M., 2000. Research design issues in earnings management studies. Journal of Accounting and Public Policy 19, 313–345. Moyer, S., 1990. Capital adequacy ratio regulations and accounting choices in commercial banks. Journal of Accounting and Economics 13, 123–154. Perry, S., Williams, T., 1994. Earnings management preceding management buyout offers. Journal of Accounting and Economics 18, 157–179. Petroni, K., 1992. Optimistic reporting in the property-casualty insurance industry. Journal of Accounting and Economics 15, 485–508. Rangan, S., 1998. Earnings management and the performance of seasoned equity offerings. Journal of Financial Economics 50, 101–122. Rogers, D. A., 2002. Managerial incentives and executive stock option repricing: A study of US casino incentives, Working Paper, Portland State University Scholes, M., Wilson, G., Wolfson, M., 1990. Tax planning, regulatory capital planning, and financial reporting strategy for commercial banks. Review of Financial Studies 3, 625–650. Shivakumar, L., 2000. Do firms mislead investors by overstating earnings before seasoned equity offerings? Journal of Accounting and Economics 29, 339–371. Sloan, R., 1996. Do stock prices fully reflect information in accruals and cash flow about future earnings? The Accounting Review 71, 289–315. Subramanyam, K., 1996. The pricing of discretionary accruals. Journal of Accounting and Economics 22, 249–281. Subramanian, N., A. Chakraborty, and S. Sheikh, 2003. Repricing is motivated by turnover, Working Paper, Brandeis University. Teoh, S., Welch, I., Wong, T., 1998a. Earnings management and the underperformance of seasoned equity offerings. Journal of Financial Economics 50, 63–99. Teoh, S., Welch, I., Wong, T., 1998b. Earnings management and the long-run market performance of initial public offerings. Journal of Finance 53, 1935–1974. Teoh, S., Wong, T., 2002. Why new issues and high-accrual firms underperform: The role of analysts’ credulity. Review of Financial Studies 15, 869–900. Xie, H., 2001. The mispricing of abnormal accruals. The Accounting Review 76, 357-373.
35
Yermack, D., 1997. Good timing: CEO stock option awards and company news announcement, Journal of Finance 52, 449–476. Zheng, L., 2003. Six-month-one-day employee stock option repricing: An examination of accounting considerations and incentive implications. Working paper, University of Southern California.
36
Figure 1 Timeline of cancellation and reissue of employee stock options This figure displays the timeline associated with the cancellation and reissue of employee stock options for 128 firms that announce the program in 2001.
Announcement
Filing
Cancellation
Reissue
Mean = 5 days Median = 0 days
Mean = 36 days Median = 31 days
Mean = 184 days Median = 184 days
37
Figure 2a Plot of buy-and-hold returns around reissue date for firms that cancel and reissue employee stock options This figure displays the mean cumulative buy-and-hold returns for the sample of firms that cancel and reissue employee stock options. Day 0 is the date these firms reissue the canceled stock options.
0.3
0.25 0.2
0.15 0.1
0.05 0 -250
-200
-150
-100
-50
0
50
100
150
-0.05 -0.1
Trading days relative to the date of reissue of options
38
Market-adjusted returns Raw returns
Figure 2b Plot of buy-and-hold returns around repricing date for firms that reprice employee stock options This figure displays mean cumulative buy-and-hold returns for the sample of firms that reprice employee stock options. Day 0 is the date these firms reprice the options.
0.2 0.15 0.1 0.05 0 -250
-200
-150
-100
-50
0
50
100
150
-0.05 -0.1 -0.15 -0.2 -0.25
Trading days relative to the date of repricing of options
39
Market-adjusted returns Raw returns
Table 1 Time and industry distribution for a sample of 128 firms that cancel and reissue executive stock options and 64 firms that reprice stock options The cancellation and reissue dates are obtained from Primark’s Global Access (Disclosure). The executive stock option reset dates are obtained from proxy statements of firms that cancel and reissue or reprice.
Panel A: Time distribution Cancellation (For Reissuers)
Reissue (For Reissuers)
Repricing (For Repricers)
1999 Q1 Q2 Q3 Q4
5 5 6 8
2000 Q1 Q2 Q3 Q4
7 4 6 11
2001 Q1 Q2 Q3 Q4 2002 Q1 Q2 Q3 Q4 Total
5 34 39 38
1 3 3 5
5 34
12
39 36 14
128
128
64
40
Table 1 (continued) Time and industry distribution for a sample of 128 firms that cancel and reissue the employee stock options and 64 firms that reprice the stock options Industry groupings are based on two-digit SIC codes.
Panel B: Industry distribution SIC code Industry 10 13 15 23 25 27 28 30 34 35 36 37 38 39 44 48 49 50 51 56 58 59 60 62 63 64 67 73 78 79 80 83 87
Reissuers Frequency %
Metal mining Oil & gas extraction General building contractors Apparel & other textile products Furniture & fixtures Printing & publishing Chemical & allied product Rubber & misc. plastics products Fabricated metal products Industrial machinery & equipment Electronic & other electric eqpmt. Transportation equipment Instruments & related products Misc. manufacturing industries Water transportation Communications Electric, gas, & sanitary services Wholesale trade- Durable goods Wholesale trade- Nondurable goods Apparel & accessory stores Eating & drinking places Miscellaneous retail Depository institutions Security & commodity brokers Insurance carriers Insurance agents, brokers, & service Holding & other investment offices Business services Motion pictures Amusement & recreation services Health services Social services Engineering & management services
Total
1 1 2 1 1 11 18 1 3
0.8 0.8 1.6 0.8 0.8 8.6 14.1 0.8 2.3
7 1 3 1
5.5 0.8 2.3 0.8
2
1.6
1
0.8
1 64
0.8 50.0
1
0.8
1 7
0.8 5.5
128
100.0
Pearson chi-square statistic for difference in proportion is reported in the last column. ***, **, * denote significance at the 1%, 5% and 10% respectively.
41
Repricers Frequency % 1 3 1 1
1.6 4.7 1.6 1.6
3 1 1 5 6
4.7 1.6 1.6 7.8 9.4
4 2 1 5 1 1
6.3 3.1 1.6 7.8 1.6 1.6
1 1 3 2
1.6 1.6 4.7 3.1
2 1 4 12 1
3.1 1.6 6.3 18.8 1.6
1
1.6
64
100.0
Chi-square
17.42***
Table 2 Descriptive statistics for reissuers and repricers The following table provides descriptive statistics for 128 firms that cancel and reissue employee stock options, and 64 firms that reprice stock options. Median figures are reported in paranthesis. Accounting data are measured at the fiscal year-end immediately prior to the earlier of announcement and filing dates, and reissue date for the reissuers, and repricing date for the repricers. Figures in columns (3) and (5) denote p-values for difference in means test/Wilcoxon rank-sum test of equality of medians.
Total Assets ($ mil) Change in sales Return on Assets Prop., plant & eqpmt. Change in receivables Market-to-book ratio Sales ($ mil) MV of equity ($ mil) Leverage Earnings per share ($) Operating cashflow Analyst forecast error Analyst following (#) Stock return Market-adjusted return
128 Reissuers (before reissue date)
64 Repricers (before repricing date)
Reissuers (before reissue date) versus Repricers
128 Reissuers (before ann. or filing date)
Reissuers (before ann. or fil. date) versus Repricers
(1) 1468.07 (170.77) 0.10 (0.03) -0.35 (-0.15) 0.23 (0.17) -0.01 (-0.01) 1.81 (1.29) 656.96 (125.56) 1209.09 (211.90) 0.12 (0.01) -1.67 (-1.06) -0.05 (-0.01) -0.01 (0) 8.77 (6) -0.28 (-0.57) -0.14 (-0.43)
(2) 586.16 (119.39) 0.12 (0.04) -0.39 (-0.12) 0.50 (0.33) 0.04 (0.01) 2.02 (1.33) 223.07 (56.99) 449.97 (95.54) 0.26 (0.22) -1.83 (-0.89) -0.13 (-0.03) 0 (0) 4.48 (2) -0.41 (-0.52) -0.50 (-0.66)
(3) 0.331/0.112
(4) 1713.71 (182.99) 0.27 (0.17) -0.21 (-0.10) 0.29 (0.23) 0.09 (0.04) 2.59 (1.43) 729.56 (120.35) 3292.53 (308.42) 0.11 (0.01) -1.00 (-0.85) -0.07 (-0.02) -0.01 (0) 8.20 (6) -0.62 (-0.71) -0.45 (-0.53)
(5) 0.259/0.015
0.792/0.245 0.697/0.255 0.000/0.000 0.005/0.004 0.426/0.372 0.196/0.006 0.116/0.030 0.000/0.000 0.818/0.720 0.055/0.358 0.861/0.759 0.000/0.000 0.207/0.362 0.000/0.000
0.051/0.015 0.111/0.415 0.000/0.023 0.094/0.018 0.187/0.478 0.228/0.003 0.069/0.000 0.000/0.000 0.209/0.430 0.236/0.574 0.944/0.906 0.000/0.000 0.001/0.020 0.375/0.007
Change in sales, prop., plant & eqpmt., change in receivables, and oper. cashflow are deflated by beginning-of-year assets. Return on assets is defined as the income before extraordinary items divided by total assets. Market-to-book ratio is defined as (total assets-common equity+market value of equity) divided by total assets. Leverage is defined as (long-term debt+short-term debt) divided by total assets. Analyst forecast error is the average of the last four quarters [(actual eps-consensus eps estimate)/prior period stock price]. Analyst following is the average of the last four quarters number of analysts following the company. Stock return is the compounded total stock returns one year prior to the relevant dates. Market-adjusted-return is the buy-and-hold market-adjusted return using CRSP value-weighted index.
42
Table 3 Logit regressions of the determinants of cancel and reissue, repricing The following table displays the results of logit regressions of the determinants of cancel and reissue, and repricing of stock options. The accounting data, described in Table 2, are measured at the fiscal year-end immediately prior to the earlier of announcement and filing dates, and the reissue date for the reissuers, and the repricing date for the repricers. The indicator variable is set to 1 if the firm cancels and reissues stock options, and 0 if the firm reprices the stock options
Intercept Change in sales Return on assets Prop., plant & eqpmt. Market-to-book ratio Log(MV of equity) Leverage Earnings per share Operating cashflow Analyst forecast error Analyst following Market-adjusted return
Before reissue date for reissuers, and repricing date for repricers
Before earlier of ann. or filing dates for reissuers, and repricing date for repricers
Model 1 3.55*** (8.97) 0.30 (0.26) 0.69 (2.08) -2.13** (5.62) -0.31 (2.44) -0.35 (2.05) -3.05** (5.45) -0.08 (0.52) -0.42 (0.11) -21.69* (2.93) 0.23*** (10.07) 1.56*** (9.08)
Model 3 2.35** (6.18) 0.94 (1.50) 0.41 (1.00) -1.70** (4.89) -0.04 (0.23) -0.18 (0.91) -3.03*** (7.71) 0.10 (0.90) -0.58 (0.50) -27.83** (5.72) 0.19*** (7.25) 0.45 (0.63)
Stock return
Pseudo R2 Wald chi-square
Model 2 3.37*** (9.13) 0.22 (0.19) 0.70 (2.29) -2.08** (6.23) -0.25 (2.20) -0.36 (2.47) -3.07** (6.08) -0.08 (0.54) -0.16 (0.01) -22.19* (3.53) 0.23*** (10.19)
0.87** (4.13) 34.4% 29.17***
Model 4 1.72* (3.31) 0.74 (0.94) 0.39 (0.93) -1.54** (4.20) -0.02 (0.04) -0.15 (0.62) -3.08*** (8.06) 0.13 (1.74) -0.33 (0.15) -27.56** (5.53) 0.16** (5.34)
-0.78 (2.04)
30.4% 25.95***
Wald chi-square statistic is reported in parentheses. ***, **, * denote significance at the 1%, 5% and 10% respectively.
43
24.8% 27.80***
25.5% 28.64***
Table 4 Descriptive statistics of accruals, return on assets, and change in revenue of firms that cancel and reissue (or reprice) their stock options for four quarters prior to the date of option strike price reset Accruals is computed for the quarters prior to the date the strike price is reset for the stock options. The relevant date is the repricing date for the sample of repricers and the reissue date for the sample of firms that cancel their stock options and reissue them after six months and one day. Total accruals is measured as the change in current assets (Compustat quarterly data item 40) minus change in current liabilities (49) minus change in cash and cash equivalents (36) plus change in debt included in current liabilities (45) minus depreciation and amortization expense (5). Total accruals is scaled by the beginning of quarter total assets. Abnormal accruals is the asset-scaled excess accruals estimated using (i) the modified Jones model used in Dechow, Sloan and Sweeney (1995), and (ii) the Jones (1991) model. Nondiscretionary accruals is estimated as: NDAi,t = α1[1/TAi,t-1] + α2[(∆REVi,t–∆RECi,t )/TAi,t-1] + α3[PPEi,t/TAi,t-1] for the modified Jones model, and NDAi,t = α1[1/TAi,t-1] + α2[∆REVi,t/TAi,t-1] + α3[PPEi,t/TAi,t-1] for the Jones model. TAi,t-1 refers to beginning of quarter total assets, ∆REVi,t refers to change in net revenue, ∆RECi,t refers to change in net receivables and PPEi,t refers to property, plant, and equipment. Abnormal accruals are measured as Total accruals (described earlier) minus the estimated nondiscretionary accruals. Return on assets is measured as the income before extraordinary items (Compustat quarterly item number 8) divided by beginning of quarter total assets. Pooled and cumulative numbers denote the average and summation (respectively) of relevant variables across the four quarters. Numbers in parentheses denote t-statistics. ***, **, and * denote significance at less than the 1%, 5% and 10% levels, two-tailed tests, respectively using the t-test for means and signed-rank test for medians. Wilcoxon Z denotes the rank sum test of equality of medians for the reissuers and repricers
Panel A: Asset-scaled abnormal accruals of reissuers and repricers relative to the option reset date estimated using modified Jones model Qtr –4 Qtr –3 Qtr –2 Qtr –1 Pooled Cumulative Mean Reissuers -0.011* -0.018** -0.033*** -0.017*** -0.020*** -0.075*** (-1.75) (-2.56) (-5.08) (-2.76) (-6.03) (-5.93) Repricers 0.007 -0.023 -0.003 0.003 -0.004 -0.015 (0.44) (-1.32) (-0.23) (0.30) (-0.59) (-0.69) Difference -0.018 0.005 -0.030*** -0.020* -0.016** -0.059** Reiss-Repr (-1.27) (0.32) (-2.41) (-1.78) (-2.30) (-2.44) Median Reissuers -0.009*** -0.015*** -0.024*** -0.015*** -0.015*** -0.052*** ** * Repricers -0.003 -0.016 0.002 -0.003 -0.004 -0.032 Wilcoxon Z -1.126 0.172 -2.642*** -1.480 -2.729*** -2.356*** Observations Reissuers 121 120 119 108 468 108 Repricers 48 51 50 50 199 48
Panel B: Asset-scaled total accruals of reissuers and repricers relative to the option reset date Qtr –4 Qtr –3 Qtr –2 Qtr –1 Pooled Mean Reissuers -0.025*** -0.027*** -0.033*** -0.021*** -0.027*** ** * Repricers -0.016 -0.038 -0.016 -0.006 -0.019*** Difference -0.009 0.011 -0.017 -0.015 -0.008 Reiss-Repr (-0.81) (0.73) (-1.58) (-1.51) (-1.32) Median Reissuers -0.024*** -0.023*** -0.032 -0.022*** -0.024*** ** *** * Repricers -0.012 -0.029 -0.013 -0.009 -0.015*** * Wilcoxon Z -0.662 0.613 -1.945 -1.552 -1.880*
44
Table 4 (continued) Descriptive statistics of accruals, return on assets, and change in revenue of firms that cancel and reissue (or reprice) their stock options for four quarters prior to the date of option strike price reset Panel C: Asset-scaled non-discretionary accruals of reissuers and repricers relative to the option reset date Qtr –4 Qtr –3 Qtr –2 Qtr –1 Pooled Mean Reissuers -0.013*** -0.009*** -0.000 -0.004* -0.006*** ** *** ** *** Repricers -0.021 -0.014 -0.015 -0.012 -0.015*** ** ** Difference -0.008 0.005 0.015 0.008 0.009*** Reiss-Repr (-0.97) (1.05) (2.30) (2.05) (2.92) Median Reissuers -0.003*** -0.002*** -0.000 -0.001*** -0.001*** *** *** ** *** Repricers -0.003 -0.003 -0.002 -0.003 -0.003*** *** Wilcoxon Z 0.560 1.254 1.582 2.809 2.993*** Panel D: Asset-scaled change in revenues of reissuers and repricers Qtr –4 Qtr –3 Qtr –2 Mean Reissuers 0.002 -0.016*** -0.006* Repricers 0.010 0.013 0.009 Difference -0.008 -0.029*** -0.015** Reiss-Repr (-0.99) (-3.63) (-2.37) Median Reissuers 0.003 -0.005*** -0.001** ** Repricers 0.002 0.004 0.001 Wilcoxon Z -0.040 -3.611*** -2.130**
Qtr –1
Pooled
-0.011* 0.007 -0.018* (-1.88)
-0.008*** 0.009*** -0.017*** (-4.29)
-0.002*** 0.001 -2.266**
-0.001*** 0.001*** -4.119***
Panel E: Asset-scaled abnormal accruals of reissuers and repricers relative to the option reset date estimated using Jones model Qtr –4 Qtr –3 Qtr –2 Qtr –1 Pooled Cumulative Mean Reissuers -0.008 -0.016** -0.029*** -0.017*** -0.017*** -0.068*** Repricers Difference Reiss-Repr Median Reissuers Repricers Wilcoxon Z
(-1.17)
(-2.41)
(-4.67)
(-2.99)
(-5.48)
(-5.37)
0.004 (0.34) -0.012 (-0.92)
-0.024 (-1.45) 0.008 (0.56)
-0.001 (-0.12) -0.028** (-2.28)
-0.005 (-0.59) -0.012 (-1.23)
-0.007 (-1.01) -0.010* (-1.68)
-0.025 (-0.69) -0.043* (-1.75)
-0.010*** -0.002 -0.916
-0.014*** -0.020** 0.008
-0.025*** 0.004 -2.819***
-0.017*** -0.004 -1.240
-0.015*** -0.003* -2.473**
-0.069*** -0.044 -2.100**
Qtr –2
Qtr –1
Pooled
-0.070 -0.066
-0.075 -0.084
-0.073 -0.082
-0.044 -0.026
-0.033 -0.023
-0.039 -0.024
Panel F: Return on assets of reissuers and repricers Qtr –4 Qtr –3 Mean Reissuers -0.066 -0.081 Repricers -0.082 -0.097 Median Reissuers -0.034 -0.049 Repricers -0.023 -0.025
45
Table 5a Cross-sectional regression of 9-month buy-and-hold returns (prior to strike price reset date) on cumulative abnormal accruals for firms that cancel and reissue employee stock options This table displays estimated coefficients from an OLS model of 9-month buy-and-hold returns of firms that cancel and reissue employee stock options CARi = α0 + α1CUMABNACCRi + α2CUMPREDACCRi + α3CUM_EARNINGS_SURPRISEi + α4CUMOPCFi + α DIND j + εi
∑
5, j
j
The dependent variable, CAR, is the compounded market-adjusted return computed using the value weighted Datastream total market index (Datastream mnemonic TOTMKUS) for 9 months prior to the date the option price is reset. CUMABNACCR is the cumulative abnormal accruals for quarters 1, 2 and 3 prior to the date of reissue of stock options and following the announcement of cancellation and reissue of stock options for the reissuers, and quarters 1, 2 and 3 prior to the date of repricing of stock options for the repricers. CUMPREDACCR is the cumulative estimated nondiscretionary accruals for the above quarters. The abnormal accruals and the estimated nondiscretionary accruals are calculated as described in Table 4. CUM_EARNINGS_SURPRISE is measured using both CUMAFE and CUMCHGEPS. CUMAFE is the cumulative analyst forecast error. Analyst forecast error is computed as: [(actual EPS – consensus EPS estimate)] and is scaled by prior period stock price. CUMCHGEPS is the cumulative change in EPS for the above quarters. Change in EPS is computed using the random walk model and is measured as this period’s EPS minus previous period’s EPS, and is scaled by prior period stock price. CUMOPCF is the cumulative operating cash flow for the above quarters. Operating cash flow is the asset-scaled operating cash flow (Compustat quarterly item # 108). DIND are industry dummy variables. Estimates for these variables are not reported for brevity, and are available upon request. ***, **, and * denote significance at less than the 1%, 5% and 10%
Model 1 0.072 (0.68)
Model 2 -0.139 (-0.35)
Model 3 -0.006 (-0.05)
Model 4 -0.144 (-0.35)
Model 5 0.069 (0.66)
Model 6 0.261 (1.00)
CUMABNACCR
-0.835 (-0.73)
-0.539 (-0.49)
-1.611 (-1.29)
-1.123 (-0.96)
-0.871 (-0.78)
-0.704 (-0.65)
CUMPREDACCR
2.103 (0.82)
0.830 (0.31)
0.529 (0.18)
-0.144 (-0.05)
2.028 (0.80)
1.419 (0.56)
CUMCHGEPS
0.018 (0.17)
0.016 (0.16)
CUMOPCF
3.477*** (4.02)
2.973*** (3.37)
2.627*** (2.84)
2.407*** (2.65)
3.459*** (4.05)
2.901*** (3.34)
4.929 (1.57)
1.197 (0.34)
Intercept
CUMAFE IND CONTROL
No
Yes
No
Yes
No
Yes
Adj. R2 F-Statistic No. of obs.
15.74% 4.78*** 82
25.81% 2.66*** 82
14.34% 3.76*** 67
28.73% 2.90*** 67
16.79% 6.45*** 82
26.43% 3.24*** 82
47
Table 5b Cross-sectional regression of 9-month buy-and-hold returns (prior to strike price reset date) on abnormal accruals for reissuers and repricers This table displays estimated coefficients from an OLS model of 9-month buy-and-hold returns of firms that cancel and reissue employee stock options (reissuers) and firms that reprice the stock options (repricers)
CARi = α0 + α1CUMABNACCRi + α2(Cancel dummy*CUMABNACCRi) + α3CUMPREDACCRi + α4(Cancel dummy*CUMPREDACCRi) + α5CUM_EARNINGS_SURPRISEi +α6(Cancel dummy* CUM_EARNINGS_SURPRISEi) + α7CUMOPCF+α8(Cancel dummy* CUMOPCFi) + α 9 , j DIND j + εi
∑ j
The dependent variable, CAR, is the compounded market-adjusted return computed using the value weighted Datastream total market index (Datastream mnemonic TOTMKUS) for 9 months prior to the date the option price is reset. CUMABNACCR, CUMPREDACCR, CUM_EARNINGS_SURPRISE and CUMOPCF are as described in Table 5a. Cancel dummy is set to 1 if the firm cancels and reissues the employee stock option, and 0 if the firm reprices the stock option. DIND are industry dummy variables. Estimates for these variables are not reported for brevity, and are available upon request. ***, **, and * denote significance at less than the 1%, 5% and 10%
Model 1 -0.123 (-1.35)
Model 2 -0.492 (-1.45)
Model 3 -0.176** (-2.03)
Model 4 -0.317 (-1.08)
Model 5 -0.123 (-1.36)
Model 6 -0.507 (-1.46)
CUMABNACCR
0.434 (0.52)
0.522 (0.53)
0.456 (0.64)
0.354 (0.41)
0.571 (0.74)
0.933 (1.07)
Dum*CUMABNACCR
-2.587* (-1.88)
-2.469 (-1.65)
-3.205** (-2.53)
-2.626* (-1.94)
-2.707** (-2.06)
-2.895** (-2.09)
CUMPREDACCR
-0.416 (-0.28)
-1.036 (-0.65)
-0.363 (-0.27)
-0.554 (-0.38)
-0.439 (-0.30)
-0.897 (-0.57)
Dum*CUMPREDACCR 0.512 (0.18)
-0.024 (-0.01)
-0.268 (-0.09)
-0.783 (-0.25)
0.574 (0.20)
-0.283 (-0.09)
CUMCHGEPS
0.111 (0.43)
0.226 (0.78)
Dum*CUMCHGEPS
-0.123 (-0.44)
-0.244 (-0.79)
CUMAFE
-0.299 (-0.14)
0.118 (0.05)
Dum*CUMAFE
5.083 (1.40)
2.437 (0.63)
Intercept
CUMOPCF
0.037 (0.08)
0.336 (0.67)
0.635 (1.56)
0.921* (1.91)
0.030 (0.07)
0.305 (0.61)
Dum*CUMOPCF
2.851*** (2.89)
2.756** (2.55)
1.502* (1.66)
1.212 (1.24)
2.868*** (2.94)
2.794*** (2.61)
IND CONTROL Adj. R2 F-Statistic No. of obs.
No 6.50% 2.09** 126
Yes 5.15% 1.31 126
No 12.46% 2.74*** 99
Yes 14.25% 1.78** 99
No 7.92% 2.79*** 126
Yes 6.30% 1.42 126
48
Table 6 Market reaction to earnings surprise and accruals around earnings announcements: Cross-sectional regression of cumulative abnormal returns on abnormal accruals and earnings surprise for reissuers and repricers This table displays estimated coefficients from an OLS model of cumulative abnormal returns of firms that cancel and reissue employee stock options (reissuers) and firms that reprice the stock options (repricers)
CARi = α0 + α1ABNACCRi + α2(Cancel dummy*ABNACCRi) + α3PREDACCRi + α4(Cancel dummy*PREDACCRi) + α5EARNINGS_SURPRISEi +α6(Cancel dummy*EARNINGS_SURPRISEi) + α7OPCFi + α8(Cancel dummy*OPCFi) + α9Q4 + εi The dependent variable, CAR, is the cumulative abnormal return computed using the CRSP equally weighted index for the window (-2,+2) around three earnings announcements prior to strike price reset. ABNACCR, the abnormal accruals and PREDACCR, the estimated nondiscretionary accruals are calculated as described in Table 4. Q4 dummy is set to 1 for fiscal fourth quarter, and estimates are not reported for brevity. EARNINGS_SURPRISE is measured using both AFE and CHGEPS. OPCF, AFE and CHGEPS are as described in Table 5a. AFE and CHGEPS, the earnings surprise variables, are scaled by prior period stock price. The reference date (t=0) is the repricing date for the sample of repricers, and the reissue date for the sample of firms that cancel their stock options and reissue them after six months and one day. Consensus EPS estimates are obtained from IBES’ detailed estimates tape and uses one-quarter-ahead EPS forecast. Cancel dummy is set to 1 if the firm cancels and reissues the employee stock option, and 0 if the firm reprices the stock option. ***, **, and * denote significance at less than the 1%, 5% and 10%
Intercept ABNACCR
Reissuers only (Models 1-3) Model 1 Model 2 Model 3 -0.010 -0.010 -0.010 (-0.68) (-0.59) (-0.65)
Reissuers and Repricers (Models 4-6) Model 4 Model 5 Model 6 -0.001 -0.004 -0.002 (-0.15) (-0.34) (-0.18)
-0.159 (-0.61)
0.053 (0.30)
-0.098 (-0.44)
0.038 (0.22)
-0.076 (-0.26)
-0.043 (-0.12)
-0.064 (-0.22)
-0.370 (-1.07)
-0.944** (-1.97)
-0.382 (-1.11)
-0.650 (-1.12)
-0.310 (-0.43)
-0.637 (-1.10)
-0.185 (-0.64)
-0.155 (-0.60)
Dum*ABNACCR PREDACCR
-1.201** (-2.52)
-1.288** (-2.40)
-1.191** (-2.50)
Dum*PREDACCR CHGEPS
0.029 (0.59)
-0.026 (-0.50)
Dum*CHGEPS
0.055 (0.72) 1.353** (2.24)
AFE
0.552 (1.02)
Dum*AFE OPCF
0.813 (0.93) -0.028 (-0.12)
-0.054 (-0.19)
-0.024 (-0.10)
0.140 (1.10)
0.209 (1.49)
0.124 (1.02)
2.98% 2.28* 167
-0.101 (-0.37) 0.03% 1.01 309
-0.230 (-0.70) 2.79% 1.72* 228
-0.088 (-0.32) 0.52% 1.23 309
Dum*OPCF Adj. R2 F-Statistic No. of obs.
2.59% 1.88* 167
4.36% 2.24* 137
49
Table 7a Cross-sectional regression of 6-month buy-and-hold returns (following strike price reset date) on past abnormal accruals for firms that cancel and reissue employee stock options This table displays estimated coefficients from an OLS model of 6-month buy-and-hold returns of firms that cancel and reissue employee stock options
CARi = α0 + α1CUMABNACCRi + α2CUMPREDACCRi + α3CUM_EARNINGS_SURPRISEi + α4CUMOPCFi +
∑α
5, j
DIND j + εi
j
The dependent variable, CAR, is the compounded market-adjusted return computed using the value weighted Datastream total market index (Datastream mnemonic TOTMKUS) for 6 months after the date the option price is reset. CUMABNACCR, CUMPREDACCR, CUM_EARNINGS_SURPRISE and CUMOPCF are lagged variables, and are computed as in Table 5a. DIND are industry dummy variables. Estimates for these variables are not reported for brevity, and are available upon request. ***, **, and * denote significance at less than the 1%, 5% and 10%
Model 1 -0.025 (-0.41)
Model 2 0.178 (0.68)
Model 3 -0.125** (-2.19)
Model 4 0.079 (0.36)
Model 5 -0.034 (-0.56)
Model 6 0.190 (0.72)
CUMABNACCR
1.709** (2.55)
1.759** (2.59)
0.910 (1.48)
1.110* (1.99)
1.640** (2.46)
1.703** (2.53)
CUMPREDACCR
2.528 (1.59)
1.540 (0.89)
1.718 (1.12)
1.157 (0.78)
2.329 (1.47)
1.358 (0.80)
CUMCHGEPS
0.085 (1.02)
0.061 (0.73)
CUMOPCF
0.553 (1.15)
0.143 (0.27)
0.502 (1.05)
0.132 (0.29)
0.512 (1.06)
0.103 (0.20)
-0.102 (-0.06)
-2.700 (-1.31)
Intercept
CUMAFE IND CONTROL
No
Yes
No
Yes
No
Yes
Adj. R2 F-Statistic No. of obs.
3.82% 1.87 89
5.26% 1.29 89
0.01% 0.83 70
18.41% 2.11** 70
3.77% 2.15* 89
5.87% 1.34 89
50
Table 7b Cross-sectional regression of 6-month buy-and-hold returns (following strike price reset date) on past abnormal accruals for reissuers and repricers This table displays estimated coefficients from an OLS model of 6-month buy-and-hold returns of firms that cancel and reissue employee stock options (reissuers) and firms that reprice the stock options (repricers)
CARi = α0 + α1CUMABNACCRi + α2(Cancel dummy*CUMABNACCRi) + α3CUMPREDACCRi + α4(Cancel dummy*CUMPREDACCRi) + α5CUM_EARNINGS_SURPRISEi +α6(Cancel dummy* CUM_EARNINGS_SURPRISEi) + α7CUMOPCFi+α8(Cancel dummy* CUMOPCFi) +
∑α
9, j
DIND j + εi
j
The dependent variable, CAR, is the compounded market-adjusted return computed using the value weighted Datastream total market index (Datastream mnemonic TOTMKUS) for 6 months after the date the option price is reset. CUMABNACCR, CUMPREDACCR, CUM_EARNINGS_SURPRISE and CUMOPCF are lagged variables, and are computed as in Table 5a. Cancel dummy is set to 1 if the firm cancels and reissues the employee stock option, and 0 if the firm reprices the stock option. DIND are industry dummy variables. Estimates for these variables are not reported for brevity, and are available upon request. ***, **, and * denote significance at less than the 1%, 5% and 10%
Model 1 0.059 (0.89)
Model 2 0.294 (1.15)
Model 3 0.015 (0.21)
Model 4 0.621** (2.56)
Model 5 0.052 (0.77)
Model 6 0.169 (0.58)
CUMABNACCR
-0.122 (-0.19)
-0.008 (-0.01)
0.702 (1.09)
0.728 (1.03)
0.301 (0.50)
0.536 (0.79)
Dum*CUMABNACCR
2.365** (2.32)
2.416** (2.18)
1.101 (0.99)
1.398 (1.30)
1.877* (1.89)
1.733* (1.66)
CUMPREDACCR
0.368 (0.32)
0.339 (0.28)
0.190 (0.16)
1.355 (1.15)
0.227 (0.20)
0.546 (0.45)
Dum*CUMPREDACCR 3.020 (1.37)
1.898 (0.82)
2.494 (0.94)
0.428 (0.17)
2.953 (1.33)
1.131 (0.49)
CUMCHGEPS
0.354* (1.80)
0.301 (1.38)
Dum*CUMCHGEPS
-0.253 (-1.12)
-0.219 (-0.91)
CUMAFE
-2.703 (-1.42)
-0.685 (-0.38)
Dum*CUMAFE
2.946 (0.85)
-0.030 (-0.01)
Intercept
CUMOPCF
0.414 (1.20)
0.550 (1.44)
0.307 (0.84)
0.860** (2.20)
0.376 (1.08)
0.493 (1.30)
Dum*CUMOPCF
0.435 (0.65)
0.041 (0.06)
0.624 (0.76)
-0.370 (-0.46)
0.433 (0.64)
0.085 (0.12)
IND CONTROL Adj. R2 F-Statistic No. of obs.
No 3.59% 1.61 133
Yes 3.10% 1.19 133
No 0.64% 1.08 102
Yes 20.50% 2.24*** 102
No 1.99% 1.45 133
Yes 3.64% 1.24 133
51
Table 8 Cross-sectional regression of analyst forecast error on expected accruals and abnormal accruals for reissuers and repricers This table displays estimated coefficients from an OLS model of analyst forecast errors of firms that cancel and reissue employee stock options (reissuers) and firms that reprice the stock options (repricers)
AFEi = α0 + α1ABNACCRi + α2(Cancel dummy*ABNACCRi) + α3PREDACCRi + α4(Cancel dummy*PREDACCRi) + α5log(MVi) +
∑α j
6, j
DIND j + ∑ α 7,k DYRk + α8Q4 + εi k
The dependent variable, AFE, is the analyst forecast error and is computed as: [(actual EPS – consensus EPS estimate)/prior period stock price]. Consensus EPS estimates are computed for the quarter immediately following the date the strike price is reset for both the samples and are obtained from IBES’ detailed estimates tape. We require that analysts have information on that quarter's reported accruals in constructing the consensus EPS forecast for the quarter immediately following the date the strike price is reset. AFE is winsorized at +/- 10%. ABNACCR, the abnormal accruals, and PREDACCR, the estimated nondiscretionary accruals are calculated as described in Table 4. DIND, DYR and Q4 are industry, year and fiscal fourth quarter dummy variables respectively. The reference date (t=0) is the repricing date for the sample of repricers, and the reissue date for the sample of firms that cancel their stock options and reissue them after six months and one day. Cancel dummy is set to 1 if the firm cancels and reissues the employee stock option, and 0 if the firm reprices the stock option. Estimates for these variables are not reported and are available upon request. log(MV) is the natural logarithm of the firm's market value. ***, **, and * denote significance at less than the 1%, 5% and 10% levels using White T, two-tailed tests
Intercept ABNACCR
Reissuers only (Models 1 and 2) Model 1 Model 2 -0.042*** -0.034** (-3.54) (-2.12)
Reissuers and Repricers (Models 3 and 4) Model 3 Model 4 -0.036*** -0.057*** (-3.52) (-3.90)
0.064 (1.04)
0.022 (0.31)
0.004 (0.06)
0.043 (0.49)
0.059 (0.74)
0.017 (0.19)
0.068 (0.80)
0.145 (0.70)
-0.136 (-0.71)
0.049 (0.93)
Dum*ABNACCR PREDACCR
0.160 (0.83)
-0.136 (-0.76)
Dum*PREDACCR log(MV)
0.006*** (3.31)
0.004** (2.41)
0.005*** (3.23)
0.005*** (3.46)
IND CONTROL
No
Yes
No
Yes
Adj. R2 F-Statistic No. of obs.
6.65% 3.21** 125
37.27% 6.67*** 125
2.50% 1.48 171
26.67% 3.94*** 171
52
Table 9 Descriptive statistics of accruals and return on assets of firms that cancel and reissue (or reprice) their stock options for four quarters prior to the date of reissue (or repricing) The table provides descriptive statistics of total accruals for firms that cancel and reissue employee stock options, and firms that reprice the options. Old repricers reprice their employee stock options prior to FASB’s announcement of FIN 44 in December 1998. New repricers reprice their employee stock options after the announcement. The relevant date is the repricing date for the sample of repricers, and the reissue date for the sample of firms that cancel their stock options and reissue them after six months and one day. Total accruals are computed as in Table 4. Numbers in parentheses denote t-statistics. ***, **, and * denote significance at less than the 1%, 5% and 10% levels, two-tailed tests, respectively. Wilcoxon Z denotes the rank sum test of equality of medians
Panel A: Asset-scaled total accruals of reissuers, new repricers, and old repricers Qtr –4 Qtr –3 Qtr –2 Qtr –1 Mean Reissuers -0.025 -0.027 -0.033 -0.021 New Repricers -0.016 -0.038 -0.016 -0.006 Old Repricers 0.004 -0.010 -0.001 -0.011 All Repricers -0.001 -0.018 -0.006 -0.009
Pooled -0.027 -0.019 -0.004 -0.009
Difference (Reissuers-All Rep)
-0.023*** (-2.92)
-0.008 (-0.75)
-0.026*** (-3.46)
- 0.011 (-1.43)
-0.017*** (-3.95)
Difference (Reissuers-Old Rep)
-0.030*** (-3.51)
- 0.017 (-1.32)
-0.031*** (-3.76)
-0.009 (-1.04)
-0.022*** (-4.41)
Difference (Old Rep-New Rep)
0.021 (1.42)
0.027 (1.35)
0.015 (1.31)
-0.005 (-0.45)
0.014 (1.90)
121 49 106 155
123 51 114 165
123 52 112 164
114 52 114 164
481 204 446 650
No. of observations Reissuers New Repricers Old Repricers All Repricers
53
Table 9 (continued) Descriptive statistics of accruals and return on assets of firms that cancel and reissue (or reprice) their stock options for four quarters prior to the date of reissue (or repricing) The table provides descriptive statistics of abnormal accruals for firms that cancel and reissue employee stock options, and firms that reprice the options. Old repricers reprice their employee stock options prior to FASB’s announcement of FIN 44 in December 1998. New repricers reprice their employee stock options after the announcement. The relevant date is the repricing date for the sample of repricers, and the reissue date for the sample of firms that cancel their stock options and reissue them after six months and one day. Abnormal accruals are computed as in Table 4. Pooled and cumulative numbers denote the average and summation (respectively) of relevant variables across the four quarters. Numbers in parentheses denote t-statistics. ***, **, and * denote significance at less than the 1%, 5% and 10% levels, two-tailed tests, respectively. Wilcoxon Z denotes the rank sum test of equality of medians
Panel B: Asset-scaled abnormal accruals of reissuers, new repricers, and old repricers Qtr –4 Qtr –3 Qtr –2 Qtr –1 Mean Reissuers -0.011* -0.018** -0.033*** -0.017*** (-1.75) (-2.56) (-5.08) (-2.76)
Pooled
Cumulative
-0.020*** (-6.03)
-0.075*** (-5.93)
New Repricers
0.007 (0.44)
-0.023 (-1.32)
-0.003 (-0.23)
0.003 (0.30)
-0.004 (-0.59)
-0.015 (-0.69)
Old Repricers
0.009 (1.31)
-0.008 (-0.76)
0.002 (0.32)
-0.002 (-0.29)
0.000 (-0.01)
0.000 (-0.01)
All Repricers
0.008 (1.31)
-0.013 (-0.76)
0.000 (0.32)
0.000 (-0.29)
-0.001 (-0.01)
-0.005 (-0.01)
Difference (Reissuers-All Rep)
-0.019** (-2.09)
-0.004 (-0.39)
-0.033*** (-3.78)
-0.016* (-1.93)
-0.018*** (-3.78)
-0.070*** (-3.88)
Difference (Reissuers-Old Rep)
-0.020** (-2.15)
-0.009 (-0.68)
-0.035*** (-3.71)
-0.015 (-1.61)
-0.020*** (-3.70)
-0.075*** (-3.72)
Difference (Old Rep-New Rep)
0.002 (0.89)
0.014 (0.68)
-0.005 (-0.38)
-0.005 (-0.41)
0.004 (0.52)
0.016 (0.58)
Wilcoxon Z (Reissuers, All Rep)
-3.171***
-2.232**
-4.552***
-1.948*
-5.898***
-4.431***
Wilcoxon Z (Reissuers, Old Rep)
-3.553***
-2.867***
-4.525***
-1.756**
-6.272***
-4.544***
Wilcoxon Z (Old Rep, New Rep)
1.433
1.794*
0.937
0.163
2.074**
1.257
121 48 105 153
120 51 113 164
119 50 111 161
108 50 113 163
468 199 442 641
124 50 116 166
No. of observations Reissuers New Repricers Old Repricers All Repricers
54
