Research Support School of Business Administration EMPLOYEE STOCK OPTIONS EXERCISES: AN EMPIRICAL ANALYSIS Working Paper #9401-42 Steven Huddart & Mark Lang Dec. 1994

I~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~ ~ ~~ ~ I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ I- - - - - - -- - - - ---- -~ ----~ --- — ~

Employee Stock Options Exercises: An Empirical Analysis* Steven Huddart Mark Lang Graduate School of Business Stanford University August 1994 WORK IN PROGRESS Please do not cite or quote without permission. This project is at an early stage. Comments are appreciated. We thank Nicholas Reitter and Rick Lambert for several helpful discussions. Participants at workshops at Wisconsin, Stanford, and the Financial Accounting Standards Board provided many valuable comments. We are grateful to ShareData, Inc. for furnishing us with some of the data used in this study and to Debashis Bhattacharya for able research assistance. Send correspondence to: STEVEN J. HUDDART VISITING ASSISTANT PROFESSOR ACCOUNTING DEPARTMENT THE UNIVERSITY OF MICHIGAN SCHOOL OF BUSINESS ADMINISTRATION 701 TAPPAN STREET, RM. 3260 ANN ARBOR, MICHIGAN 48109-1234 313 936-2771 FAX 313 763-5688 Steven-Huddart@ccmail.bus.umich.edu

I I i I I i I i I i 1 1 1 1 j i I i i I I i I i I i i I i i t i I i i I i i I

Introduction This paper documents the exercise behavior of employees who hold non-transferable long-term options on the stock of their employer. We report patterns in exercise behavior found in thedetailed grant and exercise records of seven corporations. Our data span the last decade and cover more than 40,000 employees. Understanding employee exercise behavior is essential to resolving the debate over the valuation of employee stock options in corporate financial statements. Over the last decade, the importance of stock options in employee compensation packages has increased markedly.1 Reasons for the popularity of options include the absence of a charge against accounting income for most option compensation, favorable tax treatment and the positive incentive effects of linking employee compensation to share price. In response to perceived inconsistencies between accounting for employee stock options and other forms of compensation and the resulting problems in comparing earnings across firms with different compensation strategies, the FASB proposed a change in accounting for stock options. The change would require the recognition of an expense in the income statement based on an estimate of the options' cost to the employer corporation. Reaction was dramatic and often negative. Beyond decrying the lack of relevance and reliability in the resulting financial statements, opponents cite the proposed standard's potential effect on firms' willingness to issue employee options, the viability of emerging companies, the welfare of employees and, in the extreme, the competitiveness of US firms in world markets. Many of the accounting issues raised in response to the exposure draft hinge on the actual exercise practices of employees who own options. Despite the potential importance of employee stock options to earnings under the proposed rules, little is known about option programs or employee exercise decisions. Principally, the lack of research follows from the lack of publicly available data on options for employees other than top executives. In this study, we provide descriptive evidence on option programs and exercise history for six companies, one large industrial company, one large high1 See, for example, Coopers and Lybrand (1993). 1

technology company, one medium financial services company and three small high-technology corporations which recently began trading publicly. Our study is based on detailed data on employees, option grants, and exercise. Our goal in considering a small sample of companies is to provide descriptive detail which could be lost in aggregation. The choice of very dissimilar companies permits comparison across a range of environments. The evidence in this study should be of interest to the FASB in its deliberations because it bears directly on the economic cost to the corporation of issuing options. As discussed in more detail in the next section, the current FASB proposal is based on option pricing models that were developed to value publicly traded options. As many commentators have noted, employee stock options (ESOs) differ from publicly traded options (TSOs) because ESOs cannot be sold by the employees who hold them. From a valuation perspective, the effect of this difference between TSOs and ESOs hinges on the effect of non-transferability on exercise decisions. Empirical evidence from option exercise is central to issues like whether valuation can be estimated with sufficient reliability to be recognized on the income statement, what valuation approach should be used and what underlying assumptions should be made in valuing options. Further, once requirements have been established, evidence on exercise practice will be important to corporations in deciding what value to assign to options for financial reporting purposes. Evidence on the implications of exercise practice for ESO valuation is also important in contexts other than the FASB's activities. The SEC requires that proxies issued since 1993 include the value of options granted to the five most highly compensated employees. To date much of the valuation has been based on the Black-Scholes formula. However, the same issues that face corporations in complying with the FASB proposal are also present with the SEC requirements.2 Further, corporations and compensation consulting firms also estimate stock option value for internal decision-making purposes. Again, an understanding of employee exercise patterns is necessary for accurate estimation. 2 The SEC disclosure requirements include descriptive data on options outstanding in addition to estimates of option values so that investors can assess the value of options outstanding for themselves. However, knowledge of option exercise patterns is necessary for such estimation. 2

Beyond the implications for valuation, the research has potential implications for our understanding of compensation more generally. One often-cited motivation for increased use of stock options is their effect on employee incentives. However, the incentive effects depend on when the options are exercised and whether the employee retains share ownership. If employees generally exercise options for cash immediately on the vest date, the incentive effects are clearly different than if they hold them until expiration, especially since most options have long lives (e.g., ten years) but vest relatively early. An analysis of actual exercise policy provides information useful in anticipating the effects of a given option grant, in designing optimal future option grants and in anticipating the likely effect were option use to decrease as a result of the FASB (or other) action. Finally, the data provide the opportunity to assess the effects of taxation on exercise decisions. Previous research provides evidence on the importance of taxation on corporate income shifting over time. Similar incentives exist for shifting by individuals around changes in tax regulations through option exercise decisions. Anticipated changes in tax regulation during 1986 and 1992 provided incentives to shift option exercise between periods. In on-going research, we are investigating the importance of taxbased incentives to exercise decisions. In the next section, we discuss stock options, the FASB's proposal, and reaction to it in more detail. Next, we present the issues addressed in this paper. While the paper is primarily descriptive, existing theoretical and empirical research provides guidance as to the likely effects to be observed in the data. In the third section, we discuss the data used in the study and, in the fourth section, present the results. Finally, we present conclusions and extensions. Background On June 30, 1993, the FASB issued an Exposure Draft of a Proposed Statement of Financial Accounting Standards, "Accounting for Stock-based Compensation." In understanding current accounting under Accounting Principles Board Opinion No. 25 (APB 25), Accounting for Stock Issued to Employees, and the changes proposed in the Exposure Draft, it is useful to differentiate between fixed stock options and variable stock options. With 3

fixed stock option plans terms are known at the grant date and a strike price, expiration date and vesting dates are generally specified at the grant. With variable stock option plans, terms like the number of shares to be issued and the strike price are not known at the grant date and are determined by events which occur after the grant date (like performance). Under APB 25, compensation expense associated with issuing options is determined on the measurement date, the date at which the terms of the option (number of shares to be issued and strike price) are known. Compensation expense is the difference between the market price of the stock and the strike price on the measurement date. For fixed options, therefore, the measurement date is generally the grant date and, as long as the strike price is no less than the market price (i.e., the option is at or out of the money), no compensation expense is recognized. For variable options, on the other hand, the measurement date is the date on which the terms become known (generally the vest or expiration date) and, because the strike price may not be greater than or equal to the market price at the time, compensation expense must often be recognized. In practice, variable option plans are relatively rare, a fact which many have attributed to their accounting treatment.3 The most controversial aspect of the current proposal is a requirement to recognize as compensation expense the value of fixed stock options issued to employees based on an option pricing formula. The new requirements are based on the argument that: 1. Employee stock options have value, 2. The value of employee stock options is reasonably estimable, 3. The value of stock options represents compensation, and 4. Compensation expense should be recognized in the income statement. The method used to estimate the value of stock options would be a modified version of the Black-Scholes formula. Because the FASB recognized that 3 Consistent with that notion, compensation consultants have begun designing variable ESOs in anticipation of the FASB's proposed changes in accounting for ESOs. Variable ESOs will be relatively more attractive under the proposed rules because the compensation cost of fixed ESOs will increase. 4 -- -I"'-~~I"-~ --- —--— plllCI ---

ESOs are often exercised before the expiration date, the term of the option would be the expected exercise date rather than the option term. To the extent that the actual exercise pattern differed from expectations, compensation expense would later be adjusted to the actual life of the option. Opposition to the standard has been pronounced. Major issues raised by critics include: 1. The potential effects of the changes on companies willingness to issue options and, ultimately, on the competitiveness of US companies, 2. The difficulty in accurately measuring the cost of the options to the company at the grant date, 3.- Conceptual issues as to whether an option grant represents and expense or a capital transaction, and 4. The lack of clearly identified demand for including option value on the income statement (as opposed to disclosure in the notes). To date relatively little direct evidence has been brought to bear on these issues. Most evidence to date has been on the magnitude of the adjustment to compensation expense as a percentage of net income. One recent study is Coopers and Lybrand (1993) which quantifies the likely effect of the pronouncements (and the sensitivity to changes in assumptions) on 27 emerging and mature companies to provide evidence on the likely magnitude of the effect of the proposal on companies' net income and shareholders' equity.4 Their results suggest that the proposal would have had substantial effect on reported profitability in the period studied. The estimated average reduction in net income after the phase-in period would have been 3.4% for mature companies and 26.5% for emerging companies. Further, their evidence suggests that the valuation of options is sensitive to the underlying assumptions, particularly the term of the option and the expected stock price volatility, both of which may be difficult to estimate. 4 Related research includes Foster, Koogler and Vickrey (1991) who estimate executive option value as a percentage of 1986 operating income for a sample of companies based on data from proxy statements. We focus on the Coopers and Lybrand (1993) results because they include all employee stock options issued by the sample companies and are based on more recent data. 5

Our paper takes a different approach, focusing not on the significance of the resulting effects relative to net income, but on describing the actual exercise patterns. Our attempt to characterize exercise patterns reflects concern expressed about the difficulty in valuing stock options. There are at least two issues. First, there is no public market for employee stock options to use in assessing the accuracy of option costs recorded on the income statement. ESO values cannot be directly inferred by observing prices. Moreover, the accuracy of potential option pricing formulas cannot be assessed by comparison to observed prices. The FASB's proposal hinges in part on the assertion that "during the last 20 years mathematical models to estimate the fair value of options have been developed to meet the needs of investors" and that "software available for personal computers reduces the application of those models to a fill-inthe-blank exercise." However, the models cited were developed for publicly traded options and presume transferability of the option and a particular exercise schedule (generally exercise at expiration). Anecdotal evidence of actual exercise decisions suggests that the exercise policy for publicly traded options differs from the actual exercise policy for employee stock options. Without empirical evidence as to actual exercise practice, it is not clear that the assumptions underlying the valuation of traded stock options apply to employee stock options. While other models exist which take into account actual exercise practice (e.g., exercise as a function of time to expiration, market-to-strike price ratios and employee wealth), evidence on actual exercise patterns is necessary to evaluate the need for and best approach to alternate valuation approaches. There has been some recent effort to value employee stock options. Huddart (1994) proposes an alternate approach to valuing employee stock options which explicitly incorporates the effects of inalienability and risk aversion on employee exercise decisions. In that model, as in the BlackScholes formulation, risk neutral employees exercise options on nondividend-paying stocks only at expiration. However, risk averse employees generally exercise options prior to expiration, even for stocks which pay no dividends. The intuition is that in certain price ranges (particularly when the market-to-strike ratio is large) employees prefer to exercise options in order to reinvest the proceeds in other assets. Employees would prefer not to have too large a portion of their wealth concentrated in options on their employer's 6

stock. The model is consistent with the received wisdom that employees generally exercise options well before expiration. Therefore, an option valuation formula (like the Modified Black-Scholes approach described by the FASB) based on exercise as a function of time alone is not descriptive. Incorporating the expected time to exercise into the analysis in the manner suggested in the exposure draft would result in a misvaluation of the underlying options by a potentially substantial amount because it does not reflect the decision rules used by employees. Other factors are likely effect valuation as well. For example, Cuny (1993) argues that employees are more likely to remain with their current employer as the stock price rises. Since employee turnover is correlated with stock price movements, adjustments to valuation for turnover should capture these effects. Further, tax-based incentives, liquidity constraints and beliefs of future stock price movements held by individual employees are all likely to affect employee exercise decisions in ways which are not captured by simply adjusting the Black-Scholes model for expected exercise based either on past history or easily estimated anticipation of the future. In this paper, our approach is primarily to describe the actual exercise activity that occurs and to attempt to relate it to potentially important factors suggested by theoretical research and intuition. While our results neither answer the question of whether option value should be expensed on the income statement (or disclosed in the notes) nor provide the most appropriate option valuation approach, they do provide indirect evidence on those issues by indicating the extent to which the assumptions underlying the proposed approach are consistent with observed behavior. Data Our analysis is based on option data for a sample of six companies which agreed to participate under conditions of confidentiality. All six companies issue stock options to a large number of employees. Table 1 provides descriptive data on the sample companies. Companies A-C are NYSE listed companies with 1992 market capitalization in excess of $1 billion, net income in excess of $50 million and more than 10,000 employees. Company A is diversified, Company B is an electronics company and Company C is in the financial services industry. All have been trading 7

publicly for at least a decade. Companies D-F are smaller companies in the computer industry which began trading publicly in the last decade and are included on the NASDAQ. All three had 1992 market capitalization of less than $1 billion, net income of less than $50 million and fewer than 10, 000 employees. Company G is an employee owned corporation. The stock of company G is traded only (i) among employees and (ii) between the coporate treasury and employees at a price established by formula. The market capitalization indicated by the formula price in 1992 is between $1 billion and $500 million. In 1992 net income was less than $50 million. Company G has more than 10,000 employees. Our sample includes a range of companies, which allows us to investigate the robustness of the results across a variety of settings. Further, the presence of two firms that recently went public allows us to investigate the issues for the subset of firms which has caused the greatest concern on the part of the many commentators-emerging high technology firms for whom a change in accounting requirements for options could have a significant effect on reported results and, allegedly, the continued use of option based compensation and capital markets access. Coopers and Lybrand (1993), for example, find in their sample that options as a percent of compensation and, hence, the implications of the proposed standard are much more substantial for emerging than for mature firms. Because we believe that important differences may exist across companies, we present our results on an individual company basis and then present conclusions more generally. Empirical Design Our empirical analysis is designed to provide evidence on four primary questions: 1. To what extent are options exercised prior to maturity? 2. Assuming that significant exercise does occur prior to maturity, is it clustered around specific, predictable points during the option's life? 3. To what extent is exercise predictable based on past history or comparison across companies? 8

4. What factors other than time to exercise appear to be important determinants of exercise? Our first set of analyses is a description of exercise patterns as a function of time to expiration. To the extent that exercise seems to cluster around certain dates and seems consistent across grants, time since grant may be an important determinant of exercise policy. Next we consider the relation between exercise and calendar time. To the extent that events occurring at the company level change incentives to exercise options, patterns should be evident in exercise activity as a function of calendar time. Finally, we examine exercise as a function of other factors including the market to strike ratio. If crossing a hurdle stock price level triggers exercise, that pattern should be evident in the relation between recent stock price movements and exercise. Description of Options The options for all six sample companies are fixed stock options, generally with strike price equal to the market price on the date of the grant. Therefore, under current accounting, no expense would have been recognized for the options issues. Company A issued non-qualified ten-year options that vest annually in increments of 25% over 3-4 years. Company B issued incentive stock options prior to 1986, but for most of the period issued only non-qualified options.5 For Companies B and C, options included in the analysis are all ten-year options vesting ratably at 25% per year.6 Companies D and E issued non-qualified stock options with somewhat more complex terms and vesting provisions. Company D issued ten-year options that vest monthly over four to five years. Company E issued ten-, six- and 5 Incentive stock options differ from non-qualified options in their tax treatment. In general, incentive stock options provide no deduction for the employer but allow employees to defer gain recognition until the underlying stock is sold, at which time profits are taxed at capital gains rates. Non-qualified options provide a deduction for the employer at the exercise date for the difference between exercise and strike price and are recognized at ordinary income rates when exercised. While the tax differences may affect incentives to exercise options, the direction of the effect is not clear. In our empirical analysis, conclusions for the incentive stock options are consistent with those for the non-qualified stock options. 6 For comparability, a small number of options with other terms, issued primarily to nonUS employees of Company B were excluded from the analysis. Also excluded for Company B are a small number of options with cliff vesting. Conclusions are not sensitive to their exclusion. 9

five-year options that vest over various schedules. Company F issued five year options which vest ratably over three to four years. Company G issued only five year options that vest annually in increments of 10% to 40% over four years. Data on the employees granted options indicate option grants extended deep into the organization. For Companies A, B, C, D, E, F, and G, 14,832, 24,126, 608, 575, 729, 3,232, and employees, respectively, were granted options at some time during the period. Table 2, Panel A presents descriptive data on the grants for the sample companies. We do not have data on number of employees in the company on a continuous basis and therefore base analysis on number of employees at year-end. Taking the mean across years in the sample, the percentage of employees receiving options ranges from 0.8% for Company C to 70.2% for Company F. The other computer companies, D and E both also issue shares to a substantial percentage of employees, with medians of 32.4% and 55.2%, respectively. While it is difficult to disentangle the effect of industry from size, the relative high percentage for Company B, 9.2%, is consistent with a tendency for firms in high-technology industries to issue options to more employees. Table 2, Panel A also presents data on the range of percentages of employees receiving options across years. There is a substantial differences on a year-by-year basis for most sample companies. Further, for the large companies A and C there is evidence of increasing use of options over time, with the percentage of employees receiving options increasing monotonically over the sample period. Table 2, Panel B presents data on options granted in a given year as a percentage of options outstanding at year-end. The mean percentage ranges from 0.5% for Companies A and C to 8.6% for Company F. In general, the pattern across companies and over time is similar to that for percentage of employees receiving options. Since options are typically outstanding for several years, the data suggest that options outstanding are a large fraction of shares outstanding, particularly for Companies D-F. Table 2, Panel C presents descriptive evidence on the number of grants during a year by company, data which are important for structuring the empirical tests. The number of grants per year illustrate the companies policy of grant distribution. At the extremes, Company B has a general option grant once per year, with additional grants in some years based on merit, while 10

Company F issues options almost daily in some years based on employee anniversary and promotion dates. Because strike price is generally equal to market price at issuance and vesting and expiration are based on grant date, companies which issue options more frequently have a wider variety of options outstanding at any point in time. Table 2 also presents data on the number of options granted on a grant date, the number of employees receiving options and the number of options per employee. The data suggest that most grants represent a substantial number of shares granted to a large number of employees, with each employee receiving options on a relatively large number of shares. However, there is also a substantial range across companies. In particular, Companies A, D and F often grant options to only a few employees on a given grant date. The number of options per employee across all companies is quite tightly clustered for all companies based on the median and tenth and ninetieth percentiles, but each company's distribution has a substantial right tail: the maximum value is at least 80 times as large as the median value. Univariate Analysis of Exercise Activity Table 3 presents data on exercise activity by employee. In general, employees exercise a significant percentage of their shares owned at a time. The median percentage exercised ranges from 13% for Company E to 100% for Company G. Further, for all companies but Company E, the ninetieth percentile of exercise is for 100% of shares granted. It should be noted, however, the data are potentially biased downward because the sample includes a significant number of options which had not expired as of the end of the sample period. Table 3 also presents data on the percentage of life elapsed at exercise. The median ranges from 21% for Company F to 92% for Company D, suggesting that an assumption that employees exercise primarily at expiration is not supported by the data. Data for the tenth percentile range from 5% to 39% suggesting that a substantial proportion of employees exercise within the first year or two. Further, the ninetieth percentile ranges from 39% of option life for Company E to 99% for Company G. Additional evidence on exercise patterns is available in Figure 1 which plots frequency of exercise as a function of exercise percentage and life of the 11

option on an employee/grant basis. The distribution across exercise percentages suggests that most exercise takes place at 25%, 50%, 75% or 100% of shares granted. Comparing across the life of the option, the maximum points for 25% exercise occurs in the first six months of the second year of option life, for 50% exercise occurs in the first six months of the third year of option life, for 75% exercise occurs in the first six months of the fourth year of option life, and for 100% exercise occurs in the first half of the fifth and sixth years of option life. The highest peak occurs at the beginning of the second year of option life and 25% of shares granted. This may reflect a tendency for a significant subset of employees to exercise immediately on the first vesting anniversary. The other peaks suggest that some employees wait past this anniversary, but then exercise all available options at a later vesting date. Further, the graph confirms that much of exercise takes place well before expiration. The left-hand panel of figures 2a-g present cumulative monthly exercise as a percentage of shares issued, averaged over grants made by Companies A-G in a particular year as a function of time since grant. In general, the conclusions for average grant exercise policy are similar to those from Figure 1. For Company A, approximately fifty percent of options granted were exercised in the first half of the option life. That fact is striking given that options vest annually over the first three to four years of the option life. The fact that option exercise is generally spread over time also has implications for the valuation for valuation. As discussed in Hemmer, Matsunaga and Shevlin (1994), the option pricing model suggested in the FASB Exposure Draft assumes that all exercise occurs at the expected exercise date. To the extent that exercise is spread over time, the proposal ignores the concavity of the valuation function with respect to time to expiration and, hence, assigns too high a value to options. For Company B and C, exercise occurs somewhat later but still generally takes place well before expiration. Analysis of Company D and E is limited by the facts that the company began issuing options late in the sample period and Company D issues options with a variety of lives. Nevertheless, the evidence for the first three or four years suggests as much of a tendency toward early exercise as for the more established companies. Company F issues five year options and, for many grants, exercise appears to occur relatively early in the options life. 12

In addition to providing evidence on the average time to exercise, the figures also provide insight into the estimability of option exercise based on past experience. The Exposure Draft suggests options be valued based on their expected time to exercise, with a catch-up adjustment later in the options life to adjust option life based on actual exercise experience. To the extent that exercise is difficult to predict ex ante, the resulting valuation estimates (and the charge against net income) will be unreliable at the time the options are granted and the later catch-up adjustment will tend to be large and increase the variability of reported net income. The dispersion across grant/years in Figure 2 suggests that exercise patterns can vary substantially. Even for the largest company, examining the time at which the median option from a grant is exercised (i.e., cumulative option exercise reaches fifty percent), average option life varies from approximately four years to almost seven years. The variability for the other large companies is comparable and, for the smaller companies, even greater. In addition, the differences across companies suggest that using one company as a basis for another company's expected exercise experience (as has been suggested, for example, with newly public companies) is not likely to improve estimates substantially. Finally, the dispersion across grants suggests that, absent more clear guidance as to estimating expected lives, a potentially wide range of times to maturity could be justified. Estimates that later proved high of low would be corrected by the catch-up adjustment under the current proposal, reducing incentives to underestimate the life of the option (and, hence, compensation expense at the grant date). However, were the catch-up adjustment to be dropped (as many commentators have suggested) incentives to choose a low expected option life to maximize net income would exist. The fact that exercise patterns appear to differ substantially across companies and grants within a company begs the question: What factors appear to determine exercise activity? One obvious possibility is that factors that are company-wide (e.g., stock price history) or economy-wide (e.g., changes in tax regulations or macro-economic conditions) might influence exercise policy. An approach to isolating those types of effects is to examine the data in calendar time (as opposed to time elapsed since the grant date). The right-hand panels of figures 2a-g present cumulative plots of shares exercised as a percentage of shares granted as a function of calendar time, along with a plot of the stock price path over the options' lives. Starting with 13

figure 2a, there is evidence of clustering in exercise across grants at detectable points in time. In particular, there are months in which increased exercise activity appears to occur for most grant/years simultaneously. Further, those months appear to coincide with periods of substantial stock returns. This result is not sensitive to exclusion of options with market to strike ratios of less than one, suggesting that even for options in the money, stock price movements are an important determinant of exercise. This observation stands in sharp contrast to what we would expect to see for TSOs. For TSOs, exercise before expiration is sub-optimal and is not generally a function of the market-to-strike ratio or recent share price movements. That result obtains because a publicly traded option's market price is greater than the difference between the market value of the stock and the strike value of the option. In the case of employee stock options, risk aversion limits an employee's willingness to hold an option that is deep in the money. In these cases, non-transferability precipitates exercise. Therefore, exercise is likely to be a function of the market-to-strike ratio. Further, option exercise is likely to be a function of recent stock price movement because exercise is likely to take place when stock price first crosses a threshold market to exercise ratio. The patterns for the other companies in figures 2b-f lead to similar conclusions. Movements in stock price seem clearly important in determining exercise. For example, figure 2b suggests substantial clustering of option grants in the last two years of the sample period, a period of substantial stock price appreciation. Similarly, clustering in exercise activity is observed for the other sample companies. Figure 2g stands in sharp contrast to figures 2a to 2f. Most employees at Company G wait until just before expiration to exercise the options they hold. The exercise pattern for options granted on different dates at Company G are highly correlated. Also, exercise at Company G on average occurs much closer to the expiration date of the stock option than at any other company. Huddart and Lang (1994) and Kulatilaka and Marcus (1994) argue that increases in stock price volatility lead to earlier exercise. Since Company G's stock price is substantially less volatile than the stock price of any other company, the finding that exercise occurs much later at company G is consistent with the predictions of the analytical models. 14

Regression Analysis The preceding analysis suggests that both time to expiration and stock price movements may be important determinants of employee exercise activity. To address those issues formally, we conduct regressions of the percent of options exercised as a function of various explanatory variables including stock price measures and time to expiration. The unit of observation is a grant/month and the dependent variable is percentage of shares exercised for a specific grant in a given month divided by the number of shares originally issued in that grant. Because grants vary substantially, both within and across companies, in the number of employees who received them and the total number of shares issued, small grants will tend to have a higher variance in percent exercised in a given month than large grants. Therefore, we estimate weighted least squares regressions with weights based on the number of employees receiving shares. By-company results weighting based on shares issued are very similar, but pooling across companies is hampered by differences in shares. We also estimated ordinary least squares regressions excluding the smallest grants with similar results. 15

In terms of explanatory variables, we consider four returns variables and two variables based on the market to strike ratio: 1. Monthly return ending fifteen days prior to the month 2. Return over the fifteen days prior to the month 3. Return over the month 4. Return for the following month 5. Market to strike ratio on the first day of the month 6. Market to strike ratio squared Use of these stock price measures is motivated by studies like Huddart (1994) who suggests that, given risk aversion, exercise will be a function of market to strike ratios, with risk averse employees exercising when some threshold market to strike ratio is reached. Assuming that the sample of employees have a distribution of risk tolerances, one would expect exercise to generally increase in market to strike ratio. Further, the relation would probably be nonlinear with little exercise at market to strike ratio slightly above one as the most risk averse employees exercise, increasing more rapidly as one approaches the market to strike ratio for the average employee and then leveling off for high market to strike ratios. In addition, recent return is lagged returns are likely to be associated with exercise since, loosely speaking, exercise will tend to occur at the time a market to strike ratio threshold is first crossed. Inclusion of future returns allows us to examine whether exercise is primarily a function of knowledge about the future rather than a reaction to past performance. We also include three other variables: 7. Percentage of grant available 8. Years of life remaining 9. Standard deviation of return. Percentage of grant available captures the extent to which the grant was not fully vested and the cumulative effect of past exercise. Percentage of life remaining captures the fact that as expiration approaches, the value of the option component decreases, increasing the tendency to exercise for a given market to strike ratio. Finally, the standard deviation of returns captures the 16 ------- ~ ~A. _. B -~I~ --- —-— r-'xI,_I'.... s.

notion that an increase in price variability may cause increased exercise by risk averse employees. Regression results are included in table 4, beginning with pooled analysis and then company by company. Observations with market to strike ratios less than one and cases in which grants have been fully exercised are excluded. Rank regression results are included in addition to parametric results as some of the relations may be nonlinear and distributional assumptions underlying ordinary least squares may be violated. Conclusions are generally consistent across companies and between ranked and ordinary least squares regressions. In terms of the returns variables, recent returns are strongly positively associated with exercise, particularly contemporaneous returns. The relation is negative for the longer lagged returns which, together with the positive association for contemporaneous returns, suggests that exercise activity is greatest in periods in which the stock price is rebounding from a previous fall. There is little evidence of a negative relation between exercise and future returns, suggesting that exercise does not occur primarily in anticipation of future returns. The relation with market to strike ratio is positive, suggesting more exercise when market to strike ratios are high even after controlling for recent returns.7 However, the coefficient on the market to strike ratio squared is negative, suggesting that the relation is increasing, but at a decreasing rate.8 The coefficient on percentage of grant available is generally positive, suggesting that more exercise occurs when more options are available. The coefficient on percentage of life remaining varies across companies, suggesting that the relation between option life and exercise may be more complex than we have captured. Finally, the standard deviation of returns is generally negative, providing little evidence that increased share price variability increases exercise activity. The regression R2's range from 0.05 to 0.46 for the parametric weighted least squares specification (0.08 for the pooled regression) and from 0.17 to 0.67 7 Significance levels of coefficient estimates should be viewed with caution because of potential violations of independence across observations. We are currently investigating the extent to which this is an issue. 8 A problem with the quadratic formulation for market to strike is that the relation can become negative for high market to strike ratios. Based on the coefficient estimates, that would occur for market to strike ratios greater than seven. We are currently considering alternate specifications. 17

for the rank regression (0.28 for the pooled regression) suggesting that, while a significant proportion of the variation is explained, much of it remains. Extensions Our analysis to date is clearly limited in the analysis and variables we consider and the conclusions that we draw and is at the work in process stage. We are currently expanding the analysis in several ways. First, we are considering alternate empirical specifications. The regressions are likely to suffer from a variety of econometric issues, several of which were discussed earlier. In addition, the unit of observation used above, the grant month, is only one possible measure. Perhaps a measure computed at the employee level is more appropriate. That would allow us to explicitly incorporate knowledge of employee characteristics (e.g., salary and rank data) into the analysis. Second, we are considering other variables to include in the analysis including some that capture tax considerations. For example, substantial tax law changes occurred during the sample period which may have changed incentives to exercise. In addition, we have not incorporated variables like vesting month which appeared to be important in figure 1. Finally, there may be (and a cursory examination suggests that there are) seasonalities in the data related to factors like liquidity needs which have not been incorporated. Third, we are expanding the sample to include additional companies as more data becomes available. While we will probably never have access to a large sample of companies, additional data will enhance generalizability of the results. 18

References Coopers and Lybrand, 1993, Stock Options: Accounting. Valuation and Management Issues, New York. Cuny, C. and P. Jorion, 1993, "Valuing Executive Stock Options with a Departure Decision." Working Paper, University of California —Irvine. Foster, T. III, P. Koogler and D. Vickery, 1991, "Valuation of Executive Stock Options and the FASB Proposal." The Accounting Review 66 July): 595-610. Hemmer, T., Matsunaga, S. and T. Shevlin, 1994, Correspondence with the Financial Accounting Standards Board, March 11. Huddart, Steven, 1994, "Employee Stock Options," Journal of Accounting and Economics, (forthcoming). Huddart, Steven and Mark Lang "Roundtable on Stock Option Valuation: Presentation to the Financial Accounting Standards Board, April 18, 1994," mimeo Kulatilaka, Nalin and Alan J. Marcus, 1994, "Early Exercise and the Valuation of Employee Stock Options" Working Paper, Boston College. 19

I I I II Table 1: Descriptive Statistics on Sample Companies and Option Terms Company A B C D E F G Panel A: Company description Industry Market Capitalized Employees Net Income Exchange Diversified Electronics > $50B $10 - 50B > 100K 50- 100K >$1B $.5 - 1B NYSE NYSE Financial $1 - 10B 10- 50K $.05 -.5B NYSE Computer <$1B <1K <$.05B NASDAQ Computer <$1B <1K < $.05B NASDAQ Computer <$1B 1 - 10K < $.05B NASDAQ Service <$1B 10 - 50K < $.05B unlisted Panel B: Option terms Strike Term (Years) Vesting Schedule = Market 10 straight line = Market 10 straight line = Market 10 straight line = Market 10, 6, & 5 straight line Monthly 4 to 5 = Market = Market = Market 10 5 5 straight straight straight line line line or sum of years' digits Various Annually Annually 4 to 5 3 to 4 4 Vesting AnniversariesAnnuallyAnnually Annually Years to Full Vest 3 to 4 4 4 Notes: Data in Panel A are from Compustat for fiscal year 1992.

Table 2: Descriptive Statistics on Option Grants Company A B C D E F G Panel A: Percentage of employees (sample year in parentheses) who receive options by year Minimum Year Median Year Maximum Year 1992 Mean 0.1 (1) 0.4 (5) 3.3 (9) 3.3 (9) 1.1 4.5 9.1 15.1 8.5 9.2 (1) (5) (6) (8) 0.0 (1) 0.6 (6) 1.9 (11) 1.9 (11) 0.8 3.4 (6) 48.5 21.5 (5) 59.4 100.0 (4) 62.5 4.3 (9) 62.5 32.4 55.2 (2) (2.5) (4) (4) 31.1 82.0 100.0 31.1 70.2 (11) (1) (3) (11) 15.7(1) 23.9(8) 30.5(10) 29.5(12) 22.8 Panel B: Mean number of options (sample year in parentheses) granted per year as a percent of shares outstanding Minimum Year Median Year Maximum Year 1992 Mean 0.2 (1) 0.5 (6) 0.8 (9) 0.8 (9) 0.5 0.4 0.9 2.3 0.9 1.1 (1) (6) (6) (8) o.o (1) 0.6 (6.5) 1.1 (11) 0.9 (12) 0.5 1.0 4.5 21.3 1.4 7.3 (7) (3) (1) (10) 2.7 (2) 5.2 (3.5) 14.8 (1) 6.5 (4) 7.0 3.1 6.0 32.4 4.5 8.6 (4) (10) (9) (11) 3.5 5.1 7.8 6.6 5.1 (6) (4) (12) (13) Panel C: Number of grants per year Minimum Year Median Year Maximum Year Mean 4.0 9.0 9.0 8.0 1.0 1.0 3.0 1.5 1.0 1.0 4.0 1.6 5.0 10.0 15.0 9.9 6 7 7 6.7 1 110 224 103.6 3 7 12 6.3

Table 2: Descriptive Statistics on Option Grants (continued) Company A B C D E F G Panel D: Number of options granted on a grant date (thousands) Minimum 2 11 1 0 1 0 1 10% 3 313 4 2 19 0 13 Median 11 1,944 33 12 59 2 275 90% 2,109 2,552 783 179 205 17 678 Maximum 4,408 3,471 840 1,898 1,068 23,987 2,052 Mean 507 1,787 231 70 110 17 325 Observations 72 12 23 119 40 1,450 152 Panel E: Number of employees receiving options on a grant date Minimum 1 17 1 1 1 1 1 10% 1 100 1 1 5 1 1 Median 3 6,958 7 3 17 2 67 90% 518 7,941 272 44 52 13 872 Maximum 7,005 11,824 387 364 287 3,225 1,743 Mean 390 6,020 88 16 30 13 152 Panel F: Number of options granted per employee Minimum 100 25 200 8 100 1 15 10% 500 100 1,000 375 100 50 50 Median 500 175 1,696 1,500 1,800 379 500 90% 3,000 500 4,848 9,000 8,500 2,676 2,000 Maximum 65,000 65,000 62,200 120,000 150,000 300,000 400,000 Mean 1,301 297 2,626 4,348 2,626 1,300 1,204

Notes: Data in Panel A are computed as the number of employees receiving options during a sample year divided by total employees as of the end of the year as reported in Compustat. Numbers in parentheses represent the year in the sample for that company for which the value obtained (e.g., for a company in the sample from 1984 -1992, 1984 is year 1 and 1992 is year 9). Data in Panel B are computed as the number of option shares granted during the sample year divided by total shares outstanding as of year end as reported in Compustat. Data in Panel C are based on the number of option grants per year. Data in Panels D-F are based on the empirical distribution of all option grants in the sample.

I II I II Table 3: Exercise Activity by Employee Company A B C D E F G Panel A: Fraction of options granted that are exercised by an employee at one time 10% Median 90% Mean 0.25 0.30 1.00 0.45 0.25 0.63 1.00 0.64 0.20 0.40 1.00 0.48 0.05 0.31 1.00 0.41 0.02 0.13 0.40 0.18 0.24 0.50 1.00 0.55 0.14 1.00 1.00 0.72 Panel B: Fraction of life elapsed (all observations) at the time of exercise 10% Median 90% Mean 0.11 0.27 0.51 0.30 0.13 0.35 0.71 0.40 0.11 0.31 0.64 0.35 0.17 0.38 0.67 0.40 0.09 0.22 0.39 0.23 0.05 0.21 0.60 0.26 0.39 0.92 0.99 0.79 Panel C: Fraction of life elapsed (expired options only) at the time of exercise 10% Median 90% Mean 0.21 0.68 0.99 0.64 0.24 0.46 0.83 0.49 0.20 0.48 0.83 0.51 0.49 0.96 1.00 0.84 Notes: Data are based on employee-by-employee exercise activity. Panel A presents options granted that an employee exercises at one time, Panel B presents the fraction of the exercise for all options, and Panel C presents the fraction of the life elapsed at exercise for whose expiration date had passed by the end of the sample period. the percentage of life elapsed at only those options

Table 4: Regression of Percentage of Options Exercised on Returns, Market-to-strike, Options Available, Time, and Volatility (coefficients and standard deviations x 1,000) Intcp R-3 R-2 R-1 RO R1 R2 MTS MTS2 Vest PGA LL V-1 V-2 R2 n Public Tobit Std. Err. -15.078 -10.792 -16.079 27.974 45.082 13.610 3.082 14.784 -2.159 0.026 0.026 0.038 0.041 0.038 0.041 0.028 0.015 0.003 2.437 17.371 -0.149 -2.287 -7.139 0.054 0.014 0.001 0.019 0.019 22072 OLS Std. Err. -10.358 -9.966 -17.435 26.845 41.665 11.931 2.915 11.969 -1.712 0.716 0.727 1.039 1.136 1.048 1.119 0.761 0.420 0.070 3.062 13.392 -0.382 1.529 0.398 0.036 0.910 -2.762 0.2084 22072 0.509 0.496 Rank Std. Err. 17148000.000 -22.986 -11.964 162858.675 3.631 3.742 8.738 48.542 23.281 3.470 3.673 3.587 4.349 89.989 3.843 3.886 -197.847 141.463 126.406 -73.178 -96.180 0.3005 22072 5.439 3.198 4.332 4.336 4.083 Company A Tobit Std. Err. -29.533 -12.931 -18.033 87.857 86.694 52.331 -10.352 39.852 -9.942 24.910 13.915 -0.580 -6.563 -0.596 0.163 0.113 0.164 0.147 0.153 0.154 0.122 0.105 0.026 0.103 0.048 0.010 0.107 0.095 3450 OLS Std. Err. Rank Std. Err. -23.549 -10.698 -15.013 83.772 75.686 50.419 -7.351 32.655 -8.123 20.673 14.036 -0.693 1.450 4.997 3.513 5.124 4.599 4.721 4.844 3.791 3.170 0.784 3.217 1.487 0.299 2.896 0.978 0.3144 2.795 3450 3450 1844081.848 -9.858 -52.268 37.770 109.123 43.834 -8.619 192.468 62.959 82133.472 6.673 6.719 5.998 6.318 6.492 7.182 17.096 15.785 -24.596 -10.182 -28.162 38.210 59.221 15.332 1.935 28.358 -6.297 -1.432 0.034 0.021 0.033 0.033 0.032 0.033 0.023 0.031 0.008 0.040 161.684 169.662 -23.590 -65.181 0.2681 8.938 17.893 7.814 8.006 Company B Tobit Std. Err. 9.224 -0.899 10.285 0.012 0.001 0.019 4.980 0.017 637 OLS Std. Err. -24.394 -10.204 -28.187 38.279 59.023 15.150 1.917 28.182 -6.258 -1.425 4.962 3.042 4.930 4.818 4.789 4.920 3.364 4.506 1.197 5.971 9.222 -0.899 10.239 1.800 0.177 2.808 4.981 0.4572 2.569 637 637 Rank Std. Err. 84545.892 -128.748 -157.155 174.851 351.777 95.184 -51.192 365.878 31884.063 19.430 20.202 19.689 19.781 19.793 20.412 21.412 57.239 337.561 -147.624 40.599 -6.340 0.6732 54.377 26.004 25.976 21.405 19.486

i i i I I i I I I I I I I I I i Intcp R-3 R-2 R-l RO R1 R2 MTS MTS2 Vest PGA LL V-1 V-2 R2 n Company C Tobit Std. Err. OLS Std. Err. 48.616 -37.942 -188.008 86.298 89.436 28.064 -30.474 46.679 -11.200 35.608 0.204 1.369 1.333 2.169 1.880 2.051 2.038 1.283 1.109 0.267 1.690 0.686 0.891 -1.803 -36.906 0.059 0.903 1.166 1196 1196 -11.785 -10.014 -73.829 27.701 25.780 12.625 -13.354 19.838 -3.981 35.520 7.270 -0.487 -3.199 1.569 0.0546 8.586 8.892 14.093 12.875 13.667 14.003 8.528 6.572 1.483 12.208 4.583 3.997 6.204 6.208 Rank Std. Err. 846507.838 -124.325 -216.376 168.412 80.352 19.675 -54.558 -70.983 123.081 54596.842 22.744 23.394 22.166 24.337 22.850 23.586 35.436 36.863 -45.972 -6.475 17.008 -22.926 14.988 0.362 -2.412 29.141 -3.666 -84.202 0.180 0.175 0.213 0.270 0.232 0.254 0.159 0.114 0.018 0.919 -82.453 85.893 -54.059 -81.229 0.2158 29.907 28.659 24.627 24.488 1196 Company D Tobit Std. Err. 8.898 2.400 -10.772 -15.491 0.157 0.012 0.112 0.117 4088 OLS Std. Err. -9.808 -5.007 5.047 -2.067 2.035 2.038 2.415 3.037 9.145 6.388 -2.067 2.719 2.851 1.856 8.993 -1.003 -13.825 10.536 0.659 -1.516 -3.267 0.1056 1.306 0.206 8.750 1.813 0.141 1.234 1.299 4088 Rank Std. Err. 2284728.382 -4.104 23.381 -49.866 27.752 0.962 -10.963 238.161 96883.821 12.670 12.507 13.241 12.702 12.506 12.627 14.386 1.715 15.256 214.715 -47.180 -71.974 0.1738 20.513 15.240 15.462 13.178 12.761 4088 Company E Tobit Std. Err. OLS Std. Err. -111.653 -31.024 -25.042 8.866 84.922 29.198 14.548 45.043 -5.714 74.202 -16.703 0.440 0.224 0.292 0.340 0.289 0.309 0.239 0.167 0.025 1.319 0.294 2.080 15.477 15.969 0.035 0.171 0.175 946 -18.082 -5.728 -9.966 11.920 41.103 17.654 3.615 12.244 -1.516 65.633 -13.466 -0.331 12.055 8.562 4.575 5.965 7.109 5.921 6.436 4.888 3.066 0.467 28.384 5.837 0.742 3.668 3.932 3.643 0.111 946 Rank Std. Err. 193724.784 -127.301 -112.742 14.762 195.832 10.272 101.025 475.889 58301.970 26.671 26.315 25.841 26.288 29.442 28.992 38.499 7.329 -50.099 176.770 2.597 61.611 0.2043 23.431 41.463 43.012 28.316 27.179 946

Intcp R-3 R-2 R-1 RO R1 R2 MTS MTS2 Vest PGA LL V-1 V-2 R2 n Company F Tobit Std. Err. -83.987 5.738 0.443 0.306 7.244 -9.868 36.485 -20.226 -2.664 38.787 -6.191 70.527 28.598 0.416 0.480 0.400 0.483 0.313 0.268 0.045 1.430 0.295 5.102 -15.293 -18.648 0.026 0.290 0.293 11755 OLS Std. Err. -9.839 -0.904 1.368 1.071 1.202 -1.737 13.696 -5.557 1.474 1.689 1.404 1.703 1.035 13.241 -1.938 23.937 11.290 1.102 0.918 0.152 4.747 1.073 0.105 -4.965 -5.203 0.0505 11755 0.072 0.953 0.946 Rank 3127515.915 67.505 -30.311 -42.945 99.620 -80.955 14.023 55.814 Std. Err. 136980.760 6.419 6.257 6.042 6.037 6.082 6.497 7.293 485.426 31.803 213.290 -128.541 -137.536 0.3374 11755 11.897 6.292 10.272 7.943 8.107 Company G Tobit Std. Err. 192.846 -16.072 136.520 58.224 -34.718 127.233 25.563 3.662 -3.018 65.686 -200.066 -50.085 -17.455 -31.746 0.310 1.020 2.731 2.394 2.782 2.475 1.028 0.274 0.059 0.264 0.180 0.037 0.709 0.684 3315 OLS Std. Err. 186.312 -13.130 28.397 -24.209 83.370 27.056 20.829 4.140 -2.757 78.468 -195.317 -47.163 3.196 -11.596 0.3429 10.900 35.520 92.045 80.330 93.543 80.730 35.419 9.592 2.072 9.197 6.365 1.308 23.166 22.917 3315 Rank Std. Err. 2923864.596 8.044 112.268 94.761 -142.936 138.335 64.918 -55.030 120815.875 15.094 29.789 25.498 30.406 25.289 14.315 18.800 -19.599 -9.841 -12.041 22.775 47.261 16.671 5.572 22.054 0.035 0.051 0.075 0.081 0.075 0.079 0.054 0.027 -48.948 -25.077 -533.447 -41.601 -60.960 0.4589 20.140 18.102 24.050 26.737 26.617 3315 All Combined Tobit Std. Err. -3.315 -11.398 18.774 -0.415 -6.094 -13.342 0.005 0.084 0.023 0.002 0.034 0.033 25387 OLS Std. Err. -9.320 -9.117 -16.646 22.613 44.004 13.172 1.015 1.472 2.106 2.290 2.126 2.240 5.064 16.930, -2.458 -5.526 11.050 -1.002 -0.549 -6.085 0.0875 25387 1.535 0.775 0.132 2.474 0.665 0.055 0.927 0.917 Rank Std. Err. 18836000.000 -26.278 -21.207 21.148 57.980 26.284 -1.662 107.325 158022.776 3.527 3.618 3.389 3.538 3.462 3.654 3.757 -197.156 166.976 126.773 -76.012 -105.145 0.2752 25387 5.084 2.940 3.429 4.800 4.632

Intcp R-3 R-2 R- R R R2 MTS MTS2 Vest PGA LL V-1 V-2 R2 Notes: The unit of observation is a grant-month. Grant-months where the market-to-strike ratio is less than one are omitted. Regressions are estimated for all observations in the combined regression and then on a company-by-company basis. The first set of results for the combined and by-company regressions are for tobit regressions. The second set are parametric regressions. The third set are for rank regressions. Regressions are estimated applying weighted least squares with weights proportional to the square root of the inverse of the number of employees issued options. The dependent variable is percent of a grant exercised in a given month. Intcp is the intercept from the regression. R-3 is the return over days -60 to -31, R-2 is the return over days -30 to -16, and R-1 is the return over days -15 to -1, all relative to the beginning of the month in question. RO and RI are the returns over the first and last half of the month in question, respectively, and R2 is the return over days +1 to +15 relative to the end of the month. MTS and MTS2 are the market-to-strike ratio and market-to-strike ratio squared for the grant at the end of the month. Vest is the fraction of the total grant that vests in the month. PGA is the percentage of the grant available to be exercised as of the beginning of the month, adjusted for the percentage of options which are not yet vested, those which have already been exercised, and those that have been cancelled. LL is the remaining option life to expiration in years. V-1 and V-2 are the standard deviation of daily stock price changes over one and two months prior to the month in question. R2 is the adjusted R-squared from the regression and n is the number of observations over which the regression is estimated. n

Frequency 10126 -1 6751 3376 2 ' 10.00 27 ~4 ~C~9-~ Li fl jdiZ;-~ci~p~s~h~ If jJ I) IJ iJ-Y 7.63 EXPCT LIFE 102 0.50 Figure 1 This figure plots the frequency of employee exercise decisions over the life of the option and the fraction of the grant exercised by the employee at one time. EXPCT 0>9~ ~is the percentage of shares granted to the employee that are exercised together. LIFE is the elapsed life of the option in years. There are 83,615 events in which employees exercise options in our data set. In the majority of cases, employees exercise at one time exactly 25%, 50%, 75%, or 100% of the options granted to them. Exercise of 25% of the grant is clustered at the first anniversary of the grant date. Similarly, exercises of 50%, 75%, and 100% of the grant are clustered at the second, third, and fourth anniversaries of the grant. This pattern suggests many employees exercise options soon after they vest.

, 1: -::.......................................................................................... P~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 100t~'''''`''~'' '''''':''':''''~`''''' ''~'''' ~''......~'''` ~ 100.................. P................... e P r, e c r e 70..C n e t,o4: ~~~~~~~~......:......:................. —, —! —. —......i......i... n i ~~~~~~~~~~~~~~~ ~ ~ ~ ~~~~~~~~~~~......i......i..... I....!.......!......... t..............n (A) Eso FAE E o......:............ 7 x x.... e 8 r e, a -:...............................:.,,,,...........C...............................:: i.....'.,....:........,,:~.!,,~..,.....-...........:......:......i...... ~~~~~~~~e., 20......i......... d e '....................:..............:.......... i 10 T,........................d...... 0 12 24 36 41 6o 72 4 96 106 1i 2 144la Months since grant 0.12 Grut m 194 ------ 196 - - --- 106 12 24 36 48 s 72 84 W 106 120 132 144 1 197!19Ii! I 067 1 --- 00WM Months since January 1982 o~~~~~~s d 5 1 o 100 100 P ~ ~~~~onh ic grant........................................................................... P P 0 1410 Yw 1 ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~........1.......1. '!.....!........!..... ' '! ' '!....i'!....!......i...... 110:.....:.......... I....:........:......:.....:........:.... ei P': e. c r e 70 C n ~ ~~~ ~ ~~~~~~~~:::;:: e 0: n e t so~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~~~~~............-...........'.....'.....!......'......-...........:.....i....! t 60 '~i ~ il~-~i~~~ —~i~~- ~~ ~i n ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~,i......!......i......i......i......!-.....!......i......!......!......i....i ~~~~(B) SQ~~~~...E.... 840 X — 'L~ ~: 40 ~~'*' ' ' ' ' '.I............................. C 7~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~''"......... i~? i...........3........ ~~~~~~~~~~~~~.......:.............:......:......:......:..:....:.......:...: J:::D: ~~-: ~~~:~-.jl:,::::: r::e: 04...............:............... d 10..................... i......... Months since grant 5 if3~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Gr*Yw16 ---- In ----- 960 12 24 26 46 s0 72 84 96 106 120 132 144 4 19 6 190 1896 1966 132Months since January 1982 P. ]:~ ~ ~~~~~~~~~~~~~~~~~~~~~P9......:.....:........................... eP r.....e......e c. r t i i i... i I n..: i t:~~~~~~~~~~~~~~~~~~~~~~~~~~ts.....................................r: i i i i1 1 11i i i,; 1 1i.....,,.......... E 0I.....I...........; 5.....I.....I,.'.'!,-,.'1.';;'''' ' ' ' ''' ' '.....i7 -..... c r 0 12 24 26 46 60 72 4 86 106 126 126 144 20 s~~~~~~~~~~~~~~~~~~~ Months since grant '" ]'...... - i.. 1..... 0.12 24.. 4 6..1 1 2 4:: r:C:.,,...:......:....... n e 70 'r'~~~~~Ai '; ~I~- ~~~~~-~~~~~c......:......:.........: '::-:,o i-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1.....i......i......'...........:............ —: t " ~r~ ~~~~~~~~:~~~~~~~-:-:....: Months since grant 0.......:......... ~~~~~~~~~~~~.:~:::::~,,.j. ni i:........... li 7:."',fi i, —...:;.-...... '........-; t.:..[.i:................ S~~~~~~~~~~~~~ e ~~~~~~~::::::,::I::!,oi...::r::1:.i:: e:2:4:6:,:o:_,.-:r:s:o la 3::: 'lionhsicgrn.l.....i:..............i...i......:..i... Grt Yr 193 ----- 19,4 — "- INS 0 12 24 M 4< 10 72 <4 go Jo 120 132 144 19 1987 - 196M 1, Months since January 1982 ---- - -- -- - i~~ --- ---------------------— ~~~I~''-""

4 P e r c e n t (D) E x e r c i s e d 90 -80" 70 -60 -40 40............~..............~.............,....... If '...................,........................... I I...............*....... r" ".. 'T'T". 0 12 24 36 48 0 72 64 96 106 120 132 144 Months since grant Gt Yewr --- ------- 16 1... 1 --- --- 195 ----- 198 --- 198 -- 1 ---- i --- 197 -- 192 too 0~ (E) P e r c e n t E x e r c s e d 90 -so' 80 -70 -80 -40 -30 -20 -10-.......................................;!............................................. q. ~..~.................~........ 106 12 132 1,6.4 P e r c e n t E x e r c i s e d P c e E x e r c i s e d 1C p e r c e n t E s x e r c i e d. 70". w - 500 40"30 - 20" 20;10 o................................... OD......................:.........:...... so ~-.....~-.....'......:-.....'.....-.....~-.....'.....'.....:......:......:.. 7; i......!......!............!............i......!.............!......!......!..... so.....'.....'.".....-..... i......-..... ~'.....:-.....:......:......!......i.. ~! ~~~~~ ~~......!......!.................!...!..!...!...!... so ~ ~~~~~~~~~~~~~~~~....... 40 ~.....:-.....?......:-.....:-.....'.....'.....'.....:.-............... 3 i -......!......!.............!......!......!......!...... i-..... - i... 2 iD......:......:.......:......:'- --.....:'.....:'.....:".....i....i... m -.........i...!...!..!................ '...,o:..:.......:..:.-......:......:......:..:...:'...... 30 L 0 12 24 36 41 60 72 6 91 106 t20 132 144 Months since January 1982 r"...........................................~..... L. n-i w -1 -- ------- -------------------- I.. ..... I. -... II...i 0 12 24 38 48 s 72.....! r Months since grant GrtYm r - - - ~ --—. IM ---—. -i -—. 1 t1 ----- 1 t~ P e r c e n t (F) E x e r c s e d Sk Ik 7 a 41 SI 3 2 1I I I,...............i- C --.............................. so-.................. — -- --..........:.....:......:......:......:...... 0,....:. a 12 24 x 4I& so? 2 4 go 10 12 0 132 00 30 10 06 06 O6 I6 —.................................. 0 12 24 36 M4 60 72 14 96 101 120 132 144 Months since January 1982 I Gr Months since grant -- IYew19 ----- 19 ----- 19 - --- i0 - -- IN2 ----, 0 12 24 36 41 0o 72 &4 96 16 132 144 Months since January 1982

in. --- —:............:..............................`....... n:1 P................................................... e p r e.j.| r t i20....J. n t ' 1 - s J 18 e.....i'.....::......[.....['! J'i......i.............................. f........: (G) x E.o e x - - ii i i i i i ~_~~i~ _~ i_ ____ix i:: M nhi nce8 i: year and dividing by the total options granted. Right-hand Panels: Thin lines plot C r S d e 0 12 24 6 4 80 72 6 96 106 12D 132 144 Months since grant o Gran Yawr ------- 1...1963......-6 1964 0 12 24 36 so 72 4 96 106 1906 IM 1997 19 - - - 960, 1967 -- 1961 - wMonths since January 1982 Figure 2 Companies A to G Left-Hand Panels: Plot of cumulative options exercised since the grant date as a percentage of options granted. All options granted in one year are treated as a single grant. Data were computed by cumulating the options exercised each month by grant year and dividing by the total options granted. Right-hand Panels: Thin lines plot the cumulative options exercised as a percentage of options granted over calendar time. Gray circles overlaid on these lines indicate periods when the options were out of the money. All options granted in one year are treated as a single grant. Data were computed by cumulating the options exercised each month by grant year and dividing by the total options granted. The thick black line plots mid-month stock price. I . --------------------— "~-"~I —~-`-~I'-