The Effects of Expensing Employee Stock Options and A New Approach to the Valuation Problem

Transcription

1 The Effects of Expensing Employee Stock Options and A New Approach to the Valuation Problem May 2004 Marika Semerdzhian Advisor: Professor John Shoven Stanford University, Stanford CA USA m.semerdzhian@stanford.edu ABSTRACT Although the Statement of Financial Accounting Standards (SFAS) No. 123 requires firms to report stock-based compensation expense based on the fair value of options granted, it allows firms to disclose this information in footnotes, allowing them to avoid expensing in the body of their financial statements. After the wave of recent corporate scandals, the controversy over employee stock options has intensified, and arguments are being made in favor of expensing in the actual income statement. The objective of this paper is to find whether the announcement to expense employee stock options under SFAS 123 affects the stock return of the expensing firm. The findings show that for all but the handful of early announcers, companies experience neither negative nor positive returns from the announcement. This invalidates some arguments against expensing, however the issue of a working valuation method for employee stock options continues to be a problem. I will propose new valuation methods and make recommendations for other incentive instruments. Acknowledgements: I would like to thank my advisor Professor John Shoven for his invaluable insight, guidance, and support. I would also like to acknowledge Dr. Geoffrey Rothwell, whose dedication and advice made this thesis possible.

2 I. INTRODUCTION There is intense debate is being held nationally and internationally as to whether firms should be required to deduct stock-based compensation as an expense when reporting their income. By the late 1990s outstanding employee stock options (also referred to as ESOs) averaged 7% of total outstanding shares at large corporations; top executives held one third of the total ESOs (Guay, Kothari and Sloan 1). Since executives and investors disagree in the debate (in addition to standard setters, academics, and even politicians) it is ironic that employee stock options were developed to align the interests of shareholders and firm employees. This paper investigates the announcement effect of recognizing employee stock options as an expense. Although the Statement of Financial Accounting Standards (SFAS) No. 123 requires firms to report stock-based compensation expense based on the fair value of options granted, it allows firms to disclose this information in the footnotes of the financial statement. Therefore investors only see the impact of employee stock options on net income if they read the footnotes. Those against recognition in the financial statement itself argue that this would result in a reduction in firms earnings per share because weaker financial results will be presented to investors. This will in turn raise the firms cost of financing, and thereby negatively effect investment and innovation. If required to expense employee stock options, large technology companies have alleged that they would give employees fewer stock options as compensation; this in turn they claim will make attracting and retaining top talent difficult. Prominent business leaders such as Harvey Golub (former CEO of American Express) and Andrew Grove (CEO of Intel) argue that the economic impact of ESOs manifests itself through shareholder dilution, and that the denominator in the computation of earnings per share already adequately captures the dilution cost of ESOs (Guay, Kothari, and Sloan 4). The 2

3 denominator in earnings per share computation is the average number of outstanding common shares and an adjustment for outstanding employee stock options, the conversion of outstanding ESOs into an equivalent number of common shares. Hence expensing employee stock options will double-count their costs. Golub and Grove however are wrong in their argument, which makes the mistake of confusing the financing implications of raising equity capital with the operating costs of paying for operating resources (Guay, Kothari, and Sloan 4). The true reason behind the objection against expensing employee stock options, as summarized by a comment letter to the Financial Accounting Standards Board (FASB), at whom the voiced concerns were directed, is that the recognition of compensation costs for fixed stock options will result in lower stock prices and higher costs of capital and will therefore cause many companies to eliminate or significantly curtail their stock-price compensation programs (fasb.org). It is important to realize that employee stock options are a cost associated with employee labor, and therefore should be treated like other labor costs. Many investors fail to understand or even skip over the supplemental information in the footnotes, where the true economic costs of employee stock options are disclosed. This inevitably results in the mispricing of stocks because investors simply look at the reported earnings instead of taking into account the employee stock option expense. The absence of employee stock option expense also increases the likelihood of ineffective corporate governance, because it allows the benefactors of ESOs to receive excessive compensation without impacting the firm s earnings. By expensing employee stock options, transparency is increased and investors are better able to see the actions of the managers. Also, since ESO expense lowers net income in future periods, employee stock option recognition can send a favorable signal to investors because it shows favorable future prospects for the firm. In this case, management benefits from expensing and therefore would be more willing to do so. 3

4 Companies that do not have favorable future prospects would be unwilling to expense employee stock options, because net income is often used in contracts such as debt agreements. The objective of this paper is to understand whether the announcement to expense employee stock options affects the stock return of the expensing firm. In other words, the results should indicate whether expensing is looked upon favorably. If daily stock returns increase after the announcement, then investors clearly view expensing as a positive signal. Between the summer of 2002 and early 2003, 156 firms announced their intention to recognize ESO expense. If the findings are significant, and there are positive returns associated with expensing, then this would provide strong support in favor of expensing. The overall results for all the expensing firms is insignificant, however it was hypothesized that this was due to the fact that the market anticipates firms announcements as gradually more firms show interest in announcing and begin to announce. Hence the firms that announced at the beginning of the summer of 2002 should have experienced the greatest increase in their stock return the day after the announcement because the market is less likely to have expected this, and would therefore react to it the day after instead of having anticipated the action and thereby acted previous to the announcement. Dividing up the firms between early and late announcers, the results show that in fact earlier announcers experience positive announcement returns that are significant, whereas the late announcers do not. This shows that a firm s decision to recognize employee stock option expense in the income statement does indeed create a positive signal for investors. This is easily observed for the early announcers, because investors begin to expect the announcements (especially from certain companies that have debate the possibility of doing so). 4

5 I then try to explain why companies should expense, and discuss the problems with ESO valuation. Solutions are presented, including other incentive instruments. II. BACKGROUND Accounting for employee stock options has received renewed attention in the past few years after a succession of corporate scandals relating to improper or creative accounting. Yet employee stock option expense recognition was an issue long before the recent events. The Accounting Principles Board Opinion (APB) No. 25 was issued in 1973 by the Financial Accounting Standards Board (FASB), the designate organization in the private sector for establishing standards of financial accounting and reporting (fasb.org). Under APB 25 stock based compensation is measured as intrinsic value. In other words, stock based compensation is measured as the difference between the stock price and option exercise price at the date when the exercise price and the number of options are known for each option. For fixed option grants the measurement date is the grant date, and for performance-based grants the measurement date is the date the performance criteria is met. Compensation expense under APB 25 is made to equal zero because for most fixed option grants exercise price is set equal to the stock price at the grant date. Hence companies were able to issue employee stock options without incurring extra costs that would be reported on their financial statement. The accounting treatment for employee stock options was (and is) inconsistent with other types of equity transactions. When stock is issued for equipment, the fair market value of the stock is ultimately expensed as depreciation (Rees and Stott 5) whereas employee stock options result in zero expense. In addition, different types of stock options were treated differently. The number of options granted and the exercise price of the options was not known on the grant date for performance based stock options (those linked to performance goals) since they depend on the attainment of set goals. Hence the variable 5

6 necessary to measure the intrinsic value of performance based options are available only after a certain time period, so the accounting treatment under APB 25 usually results in a significant expense deduction from earnings. For almost a decade after releasing APB Opinion No. 25, the FASB continued to debate the issue without much action. The development of the option pricing model did not solve the measurement problem. After 1973, the Black-Scholes or binomial option pricing model was used to value publicly traded call options, which differ from employee stock options. ESOs usually contain trading restrictions, have vesting requirements before they are exercisable, and have much longer exercise dates (Rees and Stott 6). However Black and Scholes did contribute to the ESO debate because they stated that options have a value other than the difference between the stock price and exercise price. In 1992 the option compensation expense issue once again made headlines when President George Bush visited Japan with 12 American executives. During the trip the Japanese focused attention on the disparity in executive compensation packages between America and Japan with the insinuation that this disparity contributed significantly to higher costs of American products (Rees and Stott 6). Democratic Senator Carl Levin soon after introduced a bill to Congress that would force recognition of the fair value of employee stock options as compensation expense by companies. The FASB was prompted to act, which was Senator Levin s main objective. In June 1993 the FASB released Exposure Draft: Accounting for Stock Based Compensation. The Exposure Draft required employers to recognize as an intangible asset the fair value of stock options at the grant date, to amortize this asset, and to record the asset s amortization as employee compensation expense (Bell, Landsman, Miller, and Yeh 1). 6

7 In other words, the Exposure Draft proposed that companies recognize employee stock option compensation as an expense using the fair value of ESOs. The Exposure Draft received strong opposition and criticism; over 1700 letters of comment in opposition to the Exposure Draft were received by the FASB. Business executives and company managers argued that recognizing the employee stock options as an expense using the fair value method to price them would adversely affect their stock price. This in return they argued would cut stock plans. Company managers even argued that the US economy would be seriously harmed due to: 1. the stifling of new business, 2. lost jobs from companies going out of business, and 3. a competitive disadvantage with foreign companies that do not follow the fair value method (Rees and Stott 6). The reliability of measuring the fair value of employee stock options was also brought into question. Because of risk-averse employees who are likely to exercise their employee stock options early and because of the long vesting requirements, option pricing models overstate the cost of employee stock options to companies (Huddart and Lang 9). The American Stock Exchange surveyed 200 chief executive officers listed on the exchange of which 84% opposed the Exposure Draft s recognition requirement. FASB s proposal was met with disagreement even by the Accounting Standards and Executive Committee (AcSEC) which voted 9 to 6 against the Exposure Draft. AcSEC claimed the measurement methods used to find the fair value of employee stock options were unreliable. The controversy surrounding the Exposure Draft continued becoming even more of a political issue as lobbyists went to Capitol Hill. Democratic Senator Joseph Lieberman introduced the Equity Expansion Act of The act, if passed, would have the power to overrule the Exposure Draft, or any decision by the FASB that required expensing the fair value 7

8 of granted ESOs (Rees and Stott 8). Although it never passed, Treasury Secretary Lloyd Bentsen and President Clinton indirectly pressured the FASB to discard the Exposure Draft and maintain the status quo. President Clinton went as far as to warn against the proposal outlined by the Exposure Draft, claiming that it would weaken the competitiveness of America s high-tech companies. President Clinton s comment was also in reference to the fact that high-tech and emerging companies rely on stock-based compensation more than others, and therefore would be impacted more by the proposal. The Financial Accounting Standards Board revisited the proposal outlined in the Exposure Draft due to the strong opposition it received, and deliberated other proposals. One suggestion proved to be a compromise: recognizing compensation expense would only be required if employee stock options were exercised quickly after the grant date. Other proposals included eliminating stock-price volatility from the regular option pricing formula, and recording the valuation of employee stock options at the vesting date. Yet there was no ideal solution. In October of 1995, the FASB issued the Statement of Financial Accounting Standards No. 123: Accounting for Stock-Based Compensation (SFAS 123). The new proposal was voted 5 to 2 in favor of accepting the pro forma disclosures of earnings that indicate what earnings would be if the fair value of stock-based compensation were recognized (Rees and Stott 8). Under SFAS 123 firms are required to disclose the pro forma effect of employee compensation expense on earnings credited to the amortization of the fair value of employee stock options on the grant date. The disclosure usually occurs in the footnotes to the financial statements because SFAS 123 does not require employee compensation expense to be recognized in the income statement, although this is encouraged. SFAS 123 instead allows firms to use APB 25 (in the 8

9 income statement), which does not require firms to recognize employee stock compensation expense if the grant meets two requirements at the grant date: the exercise price and the number of options are fixed, and the exercise price equals or exceeds the stock price (Bell, Landsman, Miller and Yeh 1). However SFAS 123 makes it clear that the fair value method is preferable to the APB 25 method for purposes of justifying a change in accounting principle under APB Opinion No. 20, Accounting Changes (fasb.org). Despite this, only five publicly traded firms recognized stock-based compensation expense based on the measurement provisions in SFAS 123 before the summer of Although the FASB compromised, it is important to recognize that the FASB was unanimous in its belief that recognizing stock-based compensation expense would lead to better financial reporting. Others favored the compromise, claiming it to be fair, while about 400 companies sent a letter to the FASB arguing against the new requirement of footnote disclosures that provide a specific value for granted employee stock options. They did not want to acknowledge the cost of issuing employee stock options anywhere in the financial statement. After the recent corporate scandals and the consequent decline in consumer confidence, issues surrounding transparency have once again become a hot topic. In October of 2002 the FASB issued another Exposure Draft, Accounting for Stock-Based Compensation Transition and Exposure, which would amend SFAS 123. The proposed changes, stated in the draft, would provide three methods of transition for companies that voluntarily adopt the fair value method in addition to propos[ing] clearer and more prominent disclosures about the cost of stock-based employee compensation and an increase in the frequency of those disclosures to include publication in quarterly financial statements (fasb.org). However the FASB has not amended SFAS 123 since soliciting opinions on its website for the amendment. 9

10 The topic of stock-based compensation has drawn much criticism from both sides. Many argue in favor of doing away with the APB 25 method option, especially as public scrutiny over stock-based compensation has increased over the last few years. On the other hand, managers and executives are once again opposing change in regulation due to their long held belief that the reduction in net income caused by recognition would decrease capital flow to the company. With lobbying power, they have brought the topic to a standstill as the FASB has not yet made a decision. III. RELATED RESEARCH Accounting for employee stock options is controversial as it has many economic consequences. There is extensive literature on the topic describing the consequences of expensing. The research tries to understand the impact of expensing on equity market values, future cash flows, and market returns. Dechow, Hutton, and Sloan (1996) try to examine stock price reactions to announcements concerning the new financial reporting rules (Dechow, Hutton, and Sloan 1). Firms fear that expensing employee stock options could lead to lower reported earnings, causing investors to view future cash flows negatively. This in turn would lead to lower stock prices, and a greater difficulty in raising equity capital. Therefore DHS, in addition to predicting that cash constrained firms will be more likely to lobby against mandatory expensing, and compensate employees with stock options under the old reporting rules, predict that such firms will experience negative stock price reactions to events increasing the probability of mandatory expensing of employee stock options. DHS test their hypothesis by examining stock price changes during certain events following the release of the Exposure Draft. The three observed events include: 1. Announcement of the FASB vote to move forward with an expensing 10

11 requirement; 2. Date of the 1993 Exposure Draft; and 3. Date the FASB announced its vote to drop the proposal for mandatory expensing of stock options. Events 1 and 2 will increase the probability of mandatory expensing, while event 3 reduces the probability. The data spans from 1994 till The following model is estimated for 3 portfolios of firms Ø jt = a j + b j Ø mt + Ó K k=1 ã jk D kt + jt where Ø jt is the return on portfolio j on day t; Ø mt is the return on the CRSP value-weighted index on day t; D kt is a dummy variable equaling one during the three-day period beginning the day before and ending the day after the announcement of event k and zero otherwise. DHS did not find a stock price reaction. The results provide no evidence of systematic evidence of a stock price reaction to announcements related to the mandatory expensing of stock options. Investors did not react to the news requiring mandatory expensing of employee stock options. More importantly, this means investors did not react negatively to the expensing requirement. Aboody (1996) finds a significant negative correlation between share price of a firm s stock and value of outstanding employee stock options (calculated using an option pricing model). Aboody s research shows that the value of stock options can be estimated well enough to be reflected in the share price of the firm s stock. However, Aboody s article predates SFAS 123 disclosures. In their 2001 article Aboody, Barth and Kasznik research the effects of expensing stock options. Mainly they try to understand the effects of disclosing stock-based compensation (but not recognizing it in regards to calculating the net income) and equity market values. They find that stock-based compensation expense has a negative relation with share price, because firms 11

12 view it as an expense for the firm. ABK also predict and find that future earnings are positively affected by the firm s decision to expense stock-based compensation. Their tests examine the relationship between share price and stock-based compensation while controlling for net income, book value equity, and expected future earnings growth (so that their incentive effects are not reflected in the results). Data was collected between 1996 and 1998 in the Standard and Poors (S&P) mid-capitalization and small-capitalization indices (these firms were chosen because they have higher stock-based compensation expense as a percentage of net income compared to larger firms). Initially all the firms were ranked according to a ratio of the number of shares reserved for stock option plans to the number shares outstanding. The 600 firms with the largest ratios were selected, however only 534 disclosed all the data required for the tests in at least one year. Data relating to stock-based compensation was collected from firms 1998 financial statement footnotes. For each year, 1996, 1997, and 1998 pro forma net income, fair value of granted options, option vesting period, and inputs for the option pricing model were collected. The first model uses the data to try to understand the implications of expensing stock-based compensation P = á 0 + á 1 BV it + á 2 NI it + á 3 LTG it + á 4 COMPX it + å it P is share price, BV is book value of equity, NI is net income, LTG is mean year-end analyst three-to-five year I/B/E/S earnings growth forecast, and COMPX is net income minus pro forma net income disclosed under SFAS 123. Each was measured at the end of the fiscal year. BV, NI, and COMPX are deflated by number of shares outstanding and i and t denote firms and years. ABK make a prediction that the coefficients on BV, NI, LTG (á 1, á 2, á 3 ) will be positive. Since ABK hypothesize that investors view stock-based compensation as an expense, they predict that 12

13 á 4 will be negative. However, because option-pricing theory suggests that stock options are a positive function of the price of the underlying stock, an inherent relationship exists between price and stock-based compensation. COMPX is therefore replaced by a predicted value from a regression of COMPX on instruments that we expect to be highly correlated with stock-based compensation expense, including expected stock price volatility, option life from grant to exercise, discount rate, dividend yield, number of options granted in the fiscal year, and the percentage of options vesting each year COMPX *. The results are consistent with the predictions made by ABK: the coefficients on BV, NI, and LTG are significantly positive for all three years. However, the coefficient on COMPX is also positive for all three years, and significantly in As ABK mention, it is important to remember that these results are probably due to the positive mechanical correlation between option prices and share price (Aboody, Barth, and Kasznik 18). The same regression is then performed using COMPX *. The coefficients on BV, NI, and LTG are significantly positive in these results as well; the coefficients on COMPX * are negative in all three years, in addition to being significant in 1997 and The results indicate that investors view stock-based compensation as an expense. This provides a basis for the research in this paper. If shareholders view stock options as a cost, this creates an argument in favor of expensing them. Although SFAS 123 requires footnote disclosure, the next step would be to include employee stock options in the balance sheet as a cost, since research shows that it is considered an expense. However, firms still argue against expensing due to the potential effects doing so will have on their expected future earnings. Therefore ABK in addition test their hypothesis that future earnings are positively affected by the firm s decision to expense stock-based compensation. They alter the original model slightly, and look to see whether there is a positive 13

14 interaction between stock-based compensation expense and expected future earnings. Specifically, look to see whether the coefficient on COMPX * is less negative for firms with higher LTG. The prediction states that it will be negative P = á 0 + á 1 BV it + á 2 NI it + á 3 LTG it + á 4 COMPX it + á 5 COMPX * it x LTG it + å it The coefficient on COMPX * it x LTG it is positive in all three years, and significant in 1997 and Next ABK remove LTG to test whether COMPX * will be more negative. Their prediction is based on positing a positive relation between stock-based compensation expense and the valuation effects of expected future earnings. If the costs of stock-based compensation reflected in COMPX * and the positive valuation effects of expected future earnings reflected in price and LTG essentially offset one another, the coefficient on COMPX * will not significantly differ from zero; if the positive valuation effects exceed costs, the coefficient on COMPX * will be positive (Aboody, Barth, and Kasznik 20-21) The results show that after omitting LTG, the coefficients on COMPX * are less negative than before. They are positive, but insignificantly different from zero in all three years. This shows that the expected incentive effects of stock-based compensation and price offset the cost of dilution. There is little effect on BV and NI. Finally, ABK also prove that an increase in stock-based compensation expense is associated with a negative effect on share price, or return RET it = ã 0 + ã 1 NI it + ã 2 ÄNI it + ã 3 ÄCOMPX * it + ã 4 ÄLTG it + å it where RET is annual share return. The variables are deflated by share price at the beginning of each year t. The coefficient on the change in stock-based compensation, ÄCOMPX * is significantly negative in both 1996 and The coefficients on net income, NI, change in net income, ÄNI, and change in expected long-term earnings growth, ÄLTG are significantly positive in both 1996 and 1997 (expect on ÄLTG in 1997). After controlling for changes in 14

15 expected long-term earnings growth, net income, and changes in net income, the findings show that increases in stock-based compensation expense are related to decreases in share prices. In the 2003 version of their articles, Daniel, Kale, Naveen hypothesize that stock-based compensation expense serves as an informational signal. They find that option expense recognition engenders a significantly positive price effect, [and] the greater the decrease in the dispersion of analysts forecasts after the announcement, the more positive is the announcement return (Daniel, Kale, and Naveen 2). DKN assume that stock-based compensation expense (known as option expense recognition in DKN) has no effect on firm s cash flows. In other words, they assume that option values reported in footnotes are reflected in the firm s share price. Hence keeping them out of the income statement does not prevent a firm s cash flows from decreasing. DKN s sample consisted of 129 firms that announced in 2002 that they would be expensing employee stock options. They went on to examine the share price reaction to option expense recognition announcement by presenting abnormal returns for various event windows (Daniel, Kale, and Naveen 14). Using EVENTUS software to estimate abnormal returns, DKN computed market-adjusted and market model returns where the parameters of the market model are estimated over a 150-day window [-210, -61] (Daniel, Kale, Naveen 14-15). The CRSP equally-weighted index returns and the CRSP value-weighted index returns were used as benchmarks. The daily returns for the sample firms are the average of the abnormal returns, and the abnormal returns over different event windows are the sum of the average abnormal returns over the period. The standard error is computed using the time-series of portfolio mean abnormal return during the event window. 15

16 DKN focus their interpretation of their results on the five-day window (-3, 1). The unadjusted returns in the (-3, 1) window are significantly positive. In addition, the number of positive return firms is significantly higher than the number of negative return firms in this window. Rees and Stott (1998) find a significant association between the disclosed compensation expense using the fair value method and firm value that is in the opposite direction from other income statement expenses (Rees and Stott 1) The result implies that disclosed stock option expense is a value-relevant measure, and that employee stock options provide benefits to the granting firms through their incentive inducing character. RS also find the positive association between the employee stock option expense and firm value is greater for firms with more growth opportunities because high-growth companies have a high demand for cash. ESOs provide talented employees a compensation package while concurrently preventing employers from using up large cash outlays in the short-term (Rees and Stott 1). In addition, this sends a positive signal to the market about the growth potential of the firm. RS argue that two factors associated with ESOs show that their impact on firm values differs from other earnings components. ESOs motivate employees to work harder for the betterment of their respective firms; hence ESOs send a positive signal to the market. Also, findings show that the Black-Sholes pricing model overstates the cost of ESOs to employers when exercised early. Because this compensation expense is overstated, weight placed on this earnings component will be relatively lower than on other earnings components. Hence compensation expense contains only noise, [and] its association with firm value will not be significant (Rees and Stott 10). To test the value-relevance of the ESO expense, RS use the following model 16

17 R i = ã 0 + ã 1 ESP i + ã 2 ESO i + å i where R is the twelve-month return for company i (ending three months after the fiscal year end); EPS is the primary earnings-per-share before non-operating items for company i deflated by price at the beginning of the return cumulation period; ESO is the after-tax effect of the fair value expense on earnings-per-share as reported in the companies footnotes deflated by price at the beginning of the return to cumulation period. All the observations are from the 1996 fiscal year (the first year of the ESO expense disclosure). The data was gathered from the footnotes of all companies that filed their 10-K with the SEC using EDGAR (Electronic Data Gathering, Analysis, and Retrieval) system. The companies that were chosen for the sample satisfied three criteria: a December 31 fiscal year; disclosed compensation expense related to ESOs in 1996; and a cusip available on the Disclosure/SEC database. The coefficient on ESO presents the information about the value-relevance of the disclosed ESO expense. A positive value will indicate that the incentive effects of employee stock options dominate; a negative value will indicate that the dilutive effects of ESOs dominate. The mean ESO for the firms in the sample is relatively small. However it represents about 65% of the mean EPS, earnings-per-share, and the ESO values around the 90 th percentile are large and can have a very significant impact on reported EPS. The coefficient on EPS is significantly positive as is the coefficient on ESO. This implies that the incentive benefits ESOs provide outweigh the dilutive effects. RS hence find that employee stock option expense is significantly associated with firm value. However unlike other typical operating expenses, the association between the ESO expense and annual returns is positive (Rees and Stott 18). RS conclude that the benefits derived from employee stock options plans outweigh the costs. 17

18 Bell, Landsman, Miller, and Yeh (2001) concentrate their study on profitable computer software firms; however their results are applicable to this paper. Using the Ohlson and Feltham- Ohlson valuation models and concentrating on 85 profitable computer software companies, BLMY study the effect of employee stock options on firm value. BLMY find that the market does not treat employee stock options as an expense, but as an intangible asset. In addition, BLMY find a conflict between the positive manner in which investors appear to value ESO expense, and the negative relation between current ESO expense and future abnormal earnings (Bell, Landsman, Miller, and Yeh 7). BLMY use the Ohlsonand Feltham-Ohlson valuation models to compare the APB 25, Exposure Draft, and SFAS123 methods of expensing employee stock options in regards to their effect on the market s firm valuation. None of the samples used by BLMY recognize ESO expense. The benchmark APB 25 abnormal earnings and valuation equations are A AEARN it = ù 0 + ù 1 AEARN i, t-1 + ù 3 BVE i, t-1 + å it MVE it = á 0 + á 1 AEARN it + á 3 BVE it + å it M where AEARN t is abnormal earnings, and equals NI t (net income before extraordinary items and discontinued operations per fiscal year t ) minus rbve t-1 (book value of common equity for the same period). MVE t is the market value of the common shares outstanding at end of year t (Bell, Landsman, Miller, and Yeh 11). The application of SFAS 123 abnormal earnings and valuation equations are A AEARNESO it = ù 0 + ù 1 AEARNESO i, t-1 + ù 3 BVE_ADJ i, t-1 + å it MVE it = á 0 + á 1 AEARNESO it + á 3 BVE_ADJ it + å it M 18

19 where AEARNESO t is abnormal earnings, and equals NI t - rbve t-1 - ESOEXP t (ESO expense equals reported net income less SFAS 123 defined pro forma net income). BVE_ADJ t is adjusted book value; it equals BVE t less ESOEXP t. The abnormal earnings and valuation equations due to Exposure Draft rules are AEARNESO i = ù 0 + ù 1 AEARNESO i, t-1 + ù 2 ESOEXP i, t-1 + ù 3 BVE_ADJ i, t-1 + ù 4 ESOASSET i, A t-1 + å it MVE it = á 0 + á 1 AEARNESO it + á 2 ESOEXP it + á 3 BVE_ADJ it + á 4 ESOASSET it + å it M where ESOASSET is the prepaid compensation component of BVE_ADJ. Looking at the results for the benchmark (APB 25) models, BLMY show that profitable software firms exhibit persistence in abnormal returns similar to firms in other industries. In addition, BLMY find that the coefficients on abnormal earnings and equity book value are significantly positive. For the SFAS 123 models, the coefficient on the lagged abnormal returns adjusted for ESO expense (ù 1 ) is significantly positive; the coefficient estimates for adjusted equity book value (ù 3 ) are insignificant. The coefficients on adjusted abnormal earnings are all positive, but insignificant except for the 1997 pooled sample. While the coefficients on adjusted equity book value are all significantly positive. The coefficients on ESOEXP are significant for all three years. This shows that investors value ESOEXP as an asset instead of an expense. For the Exposure Draft models, ESOEXP is significantly negatively associated with next year s abnormal earnings for the pooled sample. ESOASSET does not help predict abnormal earnings, and its inclusion does not have an effect on the ESOEXP coefficient in the abnormal earnings regression. For the valuation equation, ESOEXP is significantly positive, and drops by about one-third when ESOASSET is included into the equation. ESOEXP and ESOASSET are 19

20 both value relevant. These findings show that investors not only value ESOASSET as a firm asset, by value it more highly than other net assets. Contrary to theory (the authors provide explanation for why this might have occurred), ESOEXP is negatively related to future abnormal earnings, although investors positively value ESOEXP. IV. RESEACH DESIGN Although many firms are opposed to expensing employee stock options according to SFAS 123, many other firms have chosen to do so. It can be concluded that firms that recognize SFAS 123 expense find it beneficial to do so: expensing employee stock options signals positive information to capital markets. This paper therefore tests to see whether firms announcing their intention to recognize SFAS 123 expense observe a positive abnormal stock return. In addition, the results will also show if there is a negative abnormal stock return, as companies claim will happen if they have to expense. Over all, expensing effects will be observed. For each firm, the model is estimated using RETURN ft = â 0f + â 1f MARKET_RETURN t + â 2f ANNOUNCEMENT ft + å ft where RETURN ft is firm f s daily stock return; MARKET_RETURN is the daily market return (value-weighted); ANNOUNCEMENT f is a dummy variable equaling one the day firm f announces it will recognize SFAS 123 expense and equaling zero on all other days; å ft is of course the error term. Because a large number of firms announced their decision to recognize SFAS 123 expense in 2002, the data for each firm is collected for the 252 trading days in that calendar year. The data for firms that made their announcement in early 2003 was also included and collected for 252 trading days ending on March 31, Hence t represents each of the trading days. The main coefficient that is observed in this study is on ANNOUNCEMENT, â 2f, because it estimates the impact on firm f s return after announcement. 20

21 After running the regression 156 times, for each of the firms in the data sample, the coefficients on ANNOUNCEMENT f are averaged. In other words, the announcement effect for each of the firms is averaged to create an announcement effect for all the expensing firms. V. DATA The starting sample consists of 156 firms that announced their intention to recognize SFAS 123 expense starting in the summer of 2002 and through early The year 2002 was chosen because before then few firms had been expensing stock options, while 2002 experienced a surge of willing announcers who made their decision public. The list of firms was obtained from a variety of sources, but specifically from the Standard & Poor s website. Eventually all the press releases were located for each of the firms, as verification, either through on-line search, or by directly contacting the company. Daily stock returns for each of the firms and daily market returns for the 252 trading days in 2002 were obtained from a Bloomberg machine (dividends were taken into account). A variety of companies decided to expense during the summer of Although the companies that made an announcement to expense self-selected themselves, a large number of companies representing varies industries is included in the list. Below are the daily returns for a month before and a month after the announcement date of two random firms that decided to announce their intentions to expense during the observed time period. Sovereign Bancorp announced its intention to expense employee stock options on July 17, 2002, being one of the first firms to do so. The month prior to the announcement, its daily stock returns were decreasing. However after the announcement, daily stock returns began to increase. 21

22 Sovereign Bancorp Inc: Daily Stock Returns a Month Before and After Announcement Date Daily Stock Return Date Pulte Homes Inc announced its intention to expense employee stock options on March 17, 2003, being one of the last firms to do so in the observed time period for this paper. The month prior to the announcement, its daily stock returns were decreasing. However after the announcement daily stock returns began to increase even though it was one of the last firms to expense and hence the market should have been anticipating its announcement. 22

23 Pulte Homes Inc: Daily Stock Returns a Month Before and After Announcement Date Daily Stock Return Date VI. FINDINGS Table 1 presents the mean announcement return for all the recognizing firms, 0.12%. To measure whether the estimate differed from zero, two Z-statistics were used 1 : Z1 = (1/ sqrt(f))ó F f = 1 (t f / sqrt(k f /(k f -2)) Z2 = mean(t)/stdev(t/sqrt(f-1)) Where F is the total number of firms, t f is the t-statistic on the estimated coefficient of ANNOUNCEMENT for firm f, and k f is the degrees of freedom for firm f. Z1 equals 1.30 and Z2 equals 0.66, so the mean announcement return is not significantly different from zero. However this could be attributed to the fact that as more firms announce that they will recognize employee stock option expense, the market begins to anticipate firms announcements. Earlier 1 At the time of its conception in August 2003 the model developed in this paper was original. However in October 2003 Aboody, Barth, and Kasznik released a paper with a similar model that was brought to my attention in February 2003 by Joanne Young, a PhD candidate at the Stanford Economics Department. Although no other aspect of that paper was used for this one, at Young s suggestion, the same significance test, to see whether a mean differs from zero, was used. The test is in White (1984) and Bernard (1987), but I first learned of it from the ABK article. 23

24 announcements would not have been fully anticipated by the markets because they were first to announce their intention to expense. To test this hypothesis, the firms were divided in two groups, early and late announcers. Firms that announced their intention to expense in July were considered early announcers, and all other firms in the sample were considered late announcers. Of the 156 firms, 32 made their announcements in July, while 124 firms made their announcements subsequently. Table 1 presents the mean announcement return for the early announcers, 1.24% and late announcers, -0.25%. Early announcers have significantly positive announcement returns (Z1 = 4.27 and Z2 = 2.02), while late announcers have announcement returns that are not significantly different from zero (Z1 = and Z2 = -0.44). These results are consistent with the hypothesis: the market did not anticipate announcement by the early announcers, and reacted positively to the announcements. However, the late announcers experienced neither positive nor negative announcement effects; investors were anticipating their news. It may be assumed that the results were achieved due to self-selection, after all the announcing firms voluntarily decided to recognize ESO expense using the method outline in SFAS 123. In addition, many financial institutions announced simultaneously their decision to recognize. Therefore the firms were once again divided into groups by industry. Reviewing the list of announcing firms, many came from a variety of industries so that no one industry dominated. A division was established between financial institutions and non-financial institutions. Table 2 and table 3 present the findings. Once again, when the late and early announcers are combined, the mean announcement return is not significantly different from zero for both the financial and non-financial institutions (Z1 = and Z2 = for financial institutions, and Z1 = 1.62 and Z2 = 0.83 for non-financial institutions). The mean 24

25 announcement return is also not significantly different from zero for both the late financial and late non-financial announcing firms (Z1 = and Z2 = for financial institutions, and Z1 = and Z2 = for non-financial institutions). However, the mean announcement return is significantly positive for both financial and non-financial firms that are early announcers, except for financial firms due to Z2 (Z1 = 2.40 and Z2 = 1.06 for financial institutions, and Z1 = 3.55 and Z2 = 1.78 for non-financial institutions). My findings show that firms deciding to expense their employee stock options send a positive signal to investors. Firms that announced their intention to expense employee stock options in their coming income statements had a mean announcement return that was significantly positive. However, the findings only show this for early announcers because the market anticipates the announcements from late announcers. The importance of these results is that investors rewarded companies that were first to announce their decision to expense. Clearly investors do pay attention to which companies are expensing, and unlike the claims some non-expensing companies make, the investors do not react negatively to this announcement even though expensing will mean that earnings will appear smaller in future financial statements. VII. THE REASON FOR EXPENSING As the findings show, when late announcers reveal their intent to expense employee stock options, the market does not react in either direction to this news. The market only reacted favorably when the early announcers made their announcements. Although late announcers do not show a positive announcement return (because investors anticipate their announcements), perhaps an increase in the company s return is reflected prior to the announcement and hence unobservable. Non-expensing companies however would argue that this is not the case, claiming 25

26 they have no incentive to expense. Ultimately however, if there are no effects to expensing employee stock options (now that no one who announces to do so will be an early announcer), then why expense? The answer revolves around the concept of employee stock options as incentive instruments. Importantly, the findings show that companies do not experience negative effects from expensing, and therefore companies cannot use that as an argument against doing so. Actually, the reason most companies do not expense employee stock options is because they want to inflate their bottom line, even if the final earnings figure is a bit contrived. From the findings of the previous section we can deduct that investors, or at least those who advise them, read footnotes. They are aware that stock options have value and are a cost to the company that issues them. If this were not the case then investors would have reacted negatively to the companies that announced their decision to expense because of their potentially lower earnings (relative to past quarters) due to expensing. Companies that rely heavily on ESOs, which include many technology firms, continue to argue that employee stock options attract and retain talented employees. By eliminating the accounting benefits of stock options, these companies claim they will lose top employees because they will no longer be able to provide them with incentives to stay with the company and perform well. I would have to disagree with this argument because there are other incentive instruments that companies can use to attract talented employees. I will discuss these alternative incentive instruments in a following section. Companies can use incentive instruments other than employee stock options; however they do not do so because other incentive instruments would have to be expensed. Patricia Dechow, Amy Hutton, and Richard Sloan found that top executives of firms submitting comment letters opposing mandatory expensing receive a greater 26

27 proportion of their compensation from options, receive higher levels of total compensation, and their firms use options relatively more intensively for top-executive compensation than for other employees (Dechow, Hutton, and Sloan 2). Companies have found a loophole by avoiding expensing employee stock options, and today because a model does not exist by which ESOs can be properly valued, companies continue to argue in favor of the status quo. If in fact employee stock options provide for the best incentive compensation programs, they will continue to be used irregardless of whether companies have to expense them. However, when all incentive instruments have to be expensed, then the best one will clearly prevail. It cannot be judged if ESOs have an advantage over other instruments in regards to efficiency or effectiveness, because their accounting benefit seems to be the main reason companies use them. Otherwise, employee stock options do not seem to be ideal. Employee stock options have a large effect on the number of shares a company has in circulation. In a 2002 study, Core, Guay, and Kothari concluded that the existing FASB treasury-stock method of accounting for the dilutive effects of outstanding options systematically understates the options dilutive effects, and thus overstates earnings per share (Core, Guay, Kothari 2). They find that dilution due to employee stock options is twice what companies claim it is. In addition, whenever options are exercised, not only is the value of the shares held by shareholders diluted because more shares are in circulation (knowledge.wharton.upenn.edu 1), but shareholders rights are also weakened. Although a diluted earnings per share calculation exists, to calculate it, the number of common shares outstanding is increased by figuring out how many shares can be purchased with the in-the-money options at the current market price (the current stock price divided by the difference between the exercise price and the stock price). However since options holders tend to postpone exercising until share prices rise further, the 27