October 18, 2002 | Center for Data Analysis Report on Regulation
Consideration of expensing stock options should take into account the following facts.
History of the Issue
The question of whether or not to expense employee stock options has been debated since the early 1970s. In 1972, the Accounting Principles Board (APB) adopted an accounting method - APB No. 25-that did not require option expensing and is still used today. The main reason that APB No. 25 did not require option expensing was that a reasonable method of valuing the options did not exist. In 1973, a model that still serves as one of the most widely used methods for valuing traded stock options-the Black-Scholes model- was published in the Journal of Political Economy. Through the years, the Black-Scholes model and a number of other models have attracted increased attention with a rise in the use of employee stock options.Companies began issuing employee stock options to their top executives more frequently in the mid- to late 1970s; by the early 1990s, many were actively issuing employee stock options to rank-and-file employees as well. By June 30, 1993, the use of employee stock options had become so widespread that the Financial Accounting Standards Board issued a proposed rule requiring that employee stock options be expensed.That proposal created a storm of debate and congressional lobbying that ultimately forced the FASB to rescind the proposed rule in late 1994. During this debate, as has been the case recently, the typical arguments in favor of expensing centered on the abuse of options and the exploitation of tax loopholes. In 1994, Senator Carl Levin introduced a bill that was designed to curb the use of employee stock options. Politicians are now using concerns regarding recent corporate scandals to renew the attack on employee stock options. In a recent speech on the Senate floor, Senator Levin, co-sponsor of S. 1940, claimed that options were "a driving force behind management decisions at Enron that focused on increasing Enron's stock price rather than the solid growth of the company." A spectrum of opinions have now been issued on this matter. Some have argued that options should be expensed because they would otherwise be a cost that companies could hide. Others have argued that they should not be expensed because doing so would lead to artificially depressed earnings. Still others have linked employee stock options directly to corporate scandals at Enron and other firms.
Amid this debate, the public can easily be confused about the value or dangers of employee stock options-especially given the technical nature of the accounting rules. In reality, employee stock options are simply a cost-effective way to compensate employees. Rather than formulating policies on the basis of largely untested notions, it would be better to conduct a careful examination of how employee stock options function.
How Employee Stock Options Work
The two key dates involved in analyzing the effect of an option on an employee's income are its grant date and its exercise date. The grant date is the date the option is awarded to the employee. The exercise date is the date the employee can exercise the right to "use" the option.Consider, for example, the following scenario. On January 1, company ABC grants an employee an option to buy one share of the company's stock. This option has a term of five years and an exercise price of $20. That is, on December 31, five years later, the employee may exercise the option to buy one share of company ABC's stock for $20. If, at the exercise date, the company's stock is selling on the market for less than $20, the employee can simply allow the option to expire. On the other hand, if the stock is selling for more than $20, the employee can make a "profit." For example, if the company's stock is selling for $30, the employee can buy a share for $20 and sell it for $30, thus gaining $10 in income. Even though nearly all employee stock option plans prohibit the employee from selling these shares for at least several years, the Internal Revenue Service requires the employee to pay taxes on the gain at the exercise date.
Since the firm has given the employee a form of compensation, it is allowed to take a tax deduction on the employee's gain-just as it would for normal salary expenses. However, the source of this compensation is very different from normal wages and consequently has contributed to the spread of misinformation in the media. The stock that is awarded to the employee can either be purchased on the open market or taken out of treasury stock. Treasury stock can be thought of as a "vault" where the firm holds shares of its own stock.
When the source of the option shares is the open market, the only cost to the firm is the cost of buying those shares-a cost that is accounted for in the body of the financial statements. In this case, since the number of shares on the market remains unchanged, there is no additional cost to the existing shareholders. Alternatively, when the source of the option shares is "the vault," there is an additional cost to the existing shareholders: The total number of shares on the market has been increased, thus diluting the value of each existing share.
This measure simply sums the abnormal returns for each stock for a given time period. This time period is called the "event window," referring to a period of time around each announcement date. For example, a -5/+5 event window examines abnormal returns for each stock in the sample from five days prior to the announcement date through five days after the announcement date, with "day zero" being the event/announcement date. Over any given event window, a positive CAAR for a sample of firms (at a significance level of at least 90 percent) is taken to indicate a favorable response to an event for those firms, while a negative CARR (at a 90 percent significance level) would be viewed as a negative response to an event.
We used 1992 year-end data for the 1993 announcement dates, 1993 year-end data for the 1994 announcement dates, and 1994 year-end data for the 1995 announcement date. The hypotheses tested are explained in the next section, and the subsequent section discusses the results from measuring the CAAR for five days prior to and for five days after each announcement date in Table 1.
Using the event-study methodology discussed above, there are several hypotheses that can be tested. These hypotheses and the responses providing support for or evidence against each one are as follows.
HYPOTHESIS 1: The market wants employee stock options to be expensed. Support for this hypothesis would be in the form of a positive response on Date 1 and, generally speaking, negative responses on Dates 2 through 6. Announcement Dates 3 and 4, both of which preceded the official FASB announcement on Date 1, introduce a source of ambiguity.
For instance, the information contained in Date 3 and Date 4 clearly shows that there was some public knowledge of the FASB's intention to propose a rule requiring options to be expensed. One view is that the signs of any responses on Date 3 and Date 4 should match the signs of a response on Date 1. This view holds that all three dates "announce" that FASB is considering the expensing proposal.
Alternatively, the information in Date 3 and Date 4 could be viewed as "announcing" that the expensing proposal is going to meet stiff resistance and, in all likelihood, will not go into effect. In this case, the signs on any responses of Dates 3 and 4 would be opposite of the sign of any response on Date 1. Since resistance to the proposed rule was clearly evident before the official announcement of the rule, and since the rule did not go into effect, we have taken the latter view.
Date 6 entails elements of vagueness. Most important, there is additional information in the formal announcement issued on this date, including the details of requiring a new footnote disclosure. Since it lowered information costs surrounding employee stock options, this new disclosure rule would have been likely to elicit a positive response around Date 6. Other issues surrounding Date 6 will be discussed below in the results section.
HYPOTHESIS 2: The market does not want employee stock options to be expensed. Support for this hypothesis would take the form of a negative response on Date 1 and positive responses on Date 2 through Date 6.
HYPOTHESIS 3: The market is indifferent to this accounting rule. Support for this hypothesis would be indicated if responses to most of the dates in Table 1 are found to be statistically insignificant.
On the full sample of 2,666 firms for June 30, 1993 (Date 1), when the Financial Accounting Standards Board formally announced the proposed rule requiring employee stock options to be expensed, both models show no statistically significant response. The results for the quartiles around Date 1 are similar.
In both Model 1 and Model 2, there was no significant response to the announcement in the highest, upper-middle, and lower-middle quartiles. For the lowest quartile, Model 1 found a positive 1.76 percent response at the 90 percent level, and Model 2 showed no significant response. These results provide some evidence for Hypothesis 3: that the market is indifferent toward the accounting rule. However, since a WSJ article on February 5, 1993, indicated that the Business Roundtable was trying to derail the FASB's expensing proposal, it is possible that the information contained in the Date 1 announcement was already valued in the market (as discussed above).
The results regarding February 5, 1993 (Date 4), are similar to those for Date 1. For the full sample of 2,551 firms for Date 4, as well as for all of the quartiles, both models show no significant response. These results appear to favor Hypothesis 3: that the market is indifferent to the accounting rule.Since the WSJ article on Date 4 indicated that the Business Roundtable had sent a private letter on January 18, 1993 (Date 5), to the FASB, we examined Date 5 as well. For the full sample of 2,776 firms on Date 5, Model 1 revealed a positive 1.73 percent abnormal return at the 90 percent level, and Model 2 revealed no significant response. Both models showed that there were no significant responses in the highest and lowest quartiles. However, a positive response was found for the upper-middle and lower-middle quartiles using both models (both at the 90 percent level). Given that the highest quartiles should show a more pronounced response than the lower quartiles, these results are somewhat peculiar. One possible explanation for these results is that this date, which is the date of a private letter (announced publicly almost one month later), did not contain any significant public information. Even if the results around Date 5 are taken to offer some evidence for either Hypothesis 1 or Hypothesis 2, most of the evidence thus far appears to support Hypothesis 3: that the market is indifferent to the accounting rule. The next date examined is December 14, 1994 (Date 2), when the Financial Accounting Standards Board officially rescinded the proposal that would have required option expensing. For the full sample of 2,749 firms, there was a negative 1.33 percent response at the 90 percent level using Model 1 and a negative 2.03 percent response at the 95 percent level using Model 2. The results for the quartiles around Date 2, however, are mixed. Model 1 shows that there was no significant response for the highest quartile, while Model 2 reveals a negative 3.67 percent response (at the 99 percent level) for the highest quartile. While the upper-middle and lower-middle quartiles all show a significant negative response under both models, the lowest quartile shows an insignificant response under either model. (See Table 3 in the Appendix.) When considered alone, this evidence appears to support Hypothesis 1: that the market wants options expensed. To be thorough, we also examined returns in the period around April 22, 1994 (Date 3), when a WSJ article announced that the FASB was likely to delay, for at least one year, any rule requiring that options be expensed. For the full sample of 2,540 firms on Date 3, and for the first three quartiles, both models reveal no significant responses. For the lowest quartile, Model 1 revealed no significant response and Model 2 revealed a positive 0.39% response at the 95 percent level. Even though more weight would be given to an official announcement than to news from a secondary source, it seems unlikely that investors concerned about this issue would not have responded at all when the WSJ announced there would be a likely delay. Indeed, when returns for the announcement day (Day 0) are examined, out of 10 possible responses (two models, each examining responses from the full sample and the four quartiles), there were five insignificant responses and five positive responses with no discernible pattern. (See Table 3.) Thus, the results for the period around Date 5 appear to contradict the support for Hypothesis 1 found with regard to Date 2. Since the results for these two dates seem to be conflicting, it is better to view all of the results together rather than separately.
The last date to be examined is July 6, 1995 (Date 6), when a WSJ article announced that it was likely that, within one week, the Financial Accounting Standards Board would announce the footnote disclosure that is still in use in 2002 (in FAS No. 123). Under both models, the full sample of 2,701 firms, the highest quartile, and the upper-middle quartile show significantly positive responses. (See Table 3.)
Norbert J. Michel is a Policy Analyst in the Center for Data Analysis at The Heritage Foundation. Paul Garwood is a Ph.D. candidate at the University of New Orleans.
Christopher Byron, MSNBC.com, July 17, 2002, at http://www.msnbc.com/modules/exports/ct_email.asp?/news/ 781188.asp. If this link is no longer active, the article can be obtained from the authors.
In fact, a widely accepted model to evaluate any type of options did not exist. In the early 1970s, standard put-and-call options, which are different from employee stock options, were not heavily traded.
Fischer Black and Myron Scholes, "The Pricing of Options and Corporate Liabilities," The Journal of Political Economy, Vol. 81, Issue 3 (1973), pp. 637-654.
The upward trend in issuing these options seems to have continued. According to Bear Sterns, the number of options granted by the firms in the S&P 500 in 2001 was nearly 7.5 billion, an increase of nearly 50 percent from the level granted in 1998. See Bear Sterns, "Accounting Issues" report, Employee Stock Option Expense, Is the Time Right For Change? July 2002.
This bill was defeated in the Senate by a vote of 88-9.
T. J. Rodgers, "Options Aren't Optional in Silicon Valley," The Wall Street Journal, March 4, 2002, p. A14.
While several accounting rules have been shown to be irrelevant to investors, very little work has been done on the rules for employee stock options. For information on other accounting rules that have been deemed irrelevant to the market, provided there is full disclosure, see R. S. Kaplan and R. Roll, "Investor Evaluation of Accounting Information: Some Empirical Evidence," Journal of Business, Vol. 45, April 1972.
According to Bear Sterns, the firms in the S&P 500 reported just under $80 billion in pre-tax option expenses in 2001, with the technology sector accounting for nearly half of the total. See Bear Sterns, Employee Stock Option Expense.
The shares could also come from "authorized but un-issued shares." When a company issues new equity, it frequently holds some of the new shares in reserve rather than placing all the shares on the market. Granting these shares for the options has the same effect as granting shares from treasury stock-additional shares are put on the market.
See Zvi Bodie, Robert Kaplan, and Robert Merton, "Options Should Be Reflected in the Bottom Line," The Wall Street Journal, August 1, 2002, p. A12.
The cost of equity capital is the return that investors require on their equity investment. Unlike the cost of debt capital (i.e., the interest paid on debt), there is no explicit cost for equity capital.
We can assume that there are no non-cash expenses, so that the net income is equal to the net cash flow.
Incidentally, the next best alternative forgone (i.e., the opportunity cost) for employee stock options would not be selling the options on the market; it would be the cost of paying the worker with cash. Absent perfect knowledge and risk neutrality, this amount is sure to vary from the estimated value of the options-yet another reason accounting statements should not include these costs.
Bear Sterns, Employee Stock Option Expense.
According to FAS No. 123, any method can be used to value the options as long as it "takes into account…the exercise price and expected life of the option, the current price of the underlying stock and its expected volatility, expected dividends on the stock and the risk-free interest rate…." See FAS No. 123, paragraph 19.
There are varying assumptions about time to expiration. In practice, nearly all employee stock options are exercisable over a range of years.
Statistical significance refers to the probability that a hypothesis is rejected when it is actually true (this is referred to as a Type I error). Typically, the significance level is set at 0.05 or 0.01, which means that the probability of a Type I error occurring is 5 percent or 1 percent, respectively. It is common, as in the results discussed below, to use the complement of the significance level. For example, reporting that a hypothesis is rejected, at the 95 percent level of significance, means that there is a 95 percent probability that a Type I error was not made.
Eventus® uses a t-statistic to test for significant abnormal returns. Basically, this sort of test checks to see whether the difference between the mean returns for the sample and the market index (during the event window) is statistically significant. Using the terminology in note 18, a statistically significant difference between the mean return of the sample and of the market, at the 90 percent level, is synonymous with rejecting the hypothesis that the returns are the same. The 90 percent significance level indicates there is a 90 percent probability that a Type I error has not occurred (that the mean returns are the same). For more information on the t-statistic, see Edwin Mansfield, Statistics for Business and Economics, 5th Edition (New York: W. W. Norton, 1994), Chapter 9.
Wall Street Journal articles were used because the Journal is one of the most widely read financial newspapers in the United States. While it is possible that an important announcement related to these accounting rules could have been omitted from the Journal, it is reasonable to assume that all major announcements were in this publication.
The citations for the articles from which the event dates were taken are as follows: Lee Berton, "FASB Is Likely to Postpone Requiring Stock Option Deduction From Earnings," The Wall Street Journal, April 22, 1994, p. A2; Lee Berton, "Business Chiefs Try to Derail Proposal on Stock Options," The Wall Street Journal, February 5, 1993, p. A2; and Roger Lowenstein, "The Cost of Employee Stock Options, Now Hidden, Might Earn a Footnote," The Wall Street Journal, July 6, 1995, p. C1. It should be noted that all four of the tables included herein are based on data developed by the tests explained in this paper.
The listing for item #215 is as follows: "This item represents shares reserved for stock options outstanding as of year-end plus options that are available for future grants. Prior to August 22, 1996, this item included: (1) Shares subject to shareholder approval, and (2) Stock appreciation rights attached to or associated with stock options. This item is not available for banks, utilities or property and casualty companies." Because item #215 is not collected for these sectors, they are omitted from our samples.
Each quartile consists of 25 percent of the sample, based on the ratio measure, less any firms for which CRSP could not find usable returns.
The quartile sample sizes vary because of unavailable stock returns in CRSP.
While Table 4 includes the results from 30-day and 60-day windows, these results are not discussed in the paper. According to standard methodology, using the wider event windows increases the probability of measuring a response to another event. For the sake of completeness, however, the tests were run using these windows as well. Even when the larger event windows are used, over 60 percent of the measured responses, for both the CARR and the announcement dates, are statistically insignificant.
At the very least, these results suggest that the market can tell how many options firms issue. Our tests also indicate that there was a statistically significant difference between the responses of the highest and lowest quartile on Date 2. Similar differences were found between the highest and lowest quartiles on all other announcement dates where statistically significant responses were found. These tests are available upon request.
To check for an endogeneity problem, the tests were re-run for Date 5 using an equally weighted portfolio, and all responses, for the full sample and all of the quartiles, were insignificant. The endogeneity problem, which does not appear to exist here, occurs when large, well-known firms in the sample realize a drop in their stock price. This drop, because the large firms represent a large portion of the value-weighted portfolio, causes a false positive response in the lowest quartile. This alternate test was performed when the lowest quartile was the only sub-sample to show a positive response. Our results were similar using both the equally weighted and value-weighted index.
The test with the equally weighted portfolio revealed similar results, with both models showing insignificant responses for the lowest quartiles.
Given that Compustat reported the number of options issued over the time period studied, it seems particularly difficult to argue that the market could not tell how many options firms issued.