Journal of Financial Economics 00 (2002) 000-000 When a buyback isn't…

Journal of Financial Economics 00 (2002) 000-000

When a buyback isn't a buyback: Open market repurchases and
                     employee options

                                       Kathleen M. Kahle*

    Katz Graduate School of Business, University of Pittsburgh, Pittsburgh, PA 15260, USA

                       (Received 20 September 2000; accepted 6 June 2001)

Abstract

This paper examines how stock options affect the decision to repurchase shares. Firms announce
repurchases when executives have large numbers of options outstanding and when employees have
large numbers of options currently exercisable. Once the decision to repurchase is made, the amount
repurchased is positively related to total options exercisable by all employees but independent of
managerial options. These results are consistent with managers repurchasing both to maximize their own
wealth and to fund employee stock option exercises. The market appears to recognize this motive,
however, and reacts less positively to repurchases announced by firms with high levels of nonmanagerial
options.

JEL Classification:G30, G32

Key Words: share repurchase, executive stock options, employee stock options

 I thank Ken Lehn, Frederik Schlingemann, Kuldeep Shastri, René Stulz, Shawn Thomas, Cynthia von
Skansen, Ralph Walkling, and an anonymous referee for helpful comments and suggestions. Tomas
Jandik and Gang Hu provided excellent research assistance.


* Tel.: 412 648 1519, Fax: 412 648 1693
E-mail address: kkahle@katz.pitt.edu

0304-405X/00$-see front matter © 2002 Elsevier Science S.A. All rights reserved
1. Introduction




        Early studies of open market stock repurchases document positive abnormal returns of 3-4% at

the announcement (Vermaelen, 1981; Dann, 1981). The two most commonly accepted interpretations

of this reaction are the signaling theory and the free cash flow theory. The signaling theory posits that

the repurchase constitutes a revelation by management of favorable new information about the value of

the firm's future prospects. Several empirical studies find evidence consistent with this theory.

Comment and Jarrell (1991) find that the announcement- day return is positively associated with the

percent of outstanding shares repurchased and negatively associated with the firm's recent stock

returns. Ikenberry, Lakonishok, and Vermaelen (1995) examine the long-run performance of

companies following open market repurchases, and find that firms that are more likely to be

repurchasing shares because of undervaluation exhibit positive abnormal returns of 45.3% in the four

years after the announcement. Further support for the signaling theory comes from the companies

themselves, who often cite `undervaluation' as the motive for their open market repurchases (The Wall

Street Journal, April 1, 1998, p. T2, and Sept. 9, 1998, p. NE2).

        A second explanation for the positive market reaction to repurchase announcements is the free

cash flow hypothesis. According to this theory, open market repurchases mitigate agency conflicts by

returning free cash flow to shareholders. Other methods of distributing cash, such as debt-for-equity

swaps, leveraged recapitalizations, and dividends also alleviate agency problems. Repurchases,

however, are more flexible and efficient than major leverage-increasing transactions such as debt-for-

equity swaps and leveraged recapitalizations. Compared to dividends, repurchases are tax



                                                                                                            2
advantageous to shareholders and do not imply the future commitment to returning cash to shareholders

that is commonly associated with dividend increases. Special dividends would also not imply a future

commitment, but DeAngelo, DeAngelo, and Skinner (2000) document that special dividends have been

rare in recent years; in 1995, only 1.4% of NYSE firms paid special dividends. However, there is no

evidence that specials have been displaced by common stock repurchases.

        While these traditional motives for repurchases still exist, neither the signaling nor the free cash

flow hypotheses explains the surge in buybacks during the 1990s. During this period, the number of

firms repurchasing stock, as well as the dollars spent on repurchases, increased drastically. Fig. 1

shows the total dollar value and number of open market repurchases reported by Securities Data

Corporation (SDC) from 1980 through 1997. In 1996, a record 1,475 companies announced plans to

buy back $177 billion in stock. In contrast, 600 companies announced repurchases totaling less than

$40 billion in 1992 and 1993 combined.

        One explanation for the increasing popularity of buybacks in the 1990s is that recent innovations

in compensation policy, in particular the growing use of stock options by companies, have caused

changes in payout policy. ShareData, a California-based research firm, reports that the number of

companies granting stock options to all employees has increased substantially in recent years. They find

that 45% of companies with 5,000 or more employees have option plans, while in smaller companies,

74% offer option plans to all employees. The value of stock options and grants grew from $8.9 billion

to $45.6 billion between 1992 and 1997 (Strege, 1999). Further, according to Executive

Compensation Reports, "mega" option grants of 250,000 shares or more are now doled out by one of

four companies (The Wall Street Journal, Mar. 27, 1997, p. C1). Fig. 2 graphs the average number

of executive options outstanding and exercisable from 1992 through 1997 by firms on Standard and


                                                                                                               3
Poor's Execucomp database. Consistent with the increase in repurchases, the average number of

options outstanding and exercisable by top executives has tripled during this time period.

        I examine open market share repurchases from 1993-1996 to determine the effect of options

on the firm's decision to repurchase stock, the actual repurchases made, and the market reaction to the

announcement of a repurchase. I begin by examining firms' motives for repurchasing. Consistent with

the signaling and free cash flow hypotheses, I find that firms are more likely to repurchase shares than to

increase dividends when recent stock performance has been poor or when free cash flow is high. I also

examine two hypotheses relating growth in stock options to the increasing popularity of buybacks: the

option-funding hypothesis and the substitution hypothesis. The option-funding hypothesis predicts that

repurchases are intended to fund the exercise of employee stock options. Thus, the decision to

repurchase should be related to options recently exercised and to options expected to be exercised in

the near future. Since the latter variable is not available, I use outstanding and exercisable options as

proxies.

        The substitution hypothesis predicts that executive options create different incentives than do

employee options. While employee options provide firms with incentives to repurchase shares to avoid

the earnings dilution that could be caused by option exercise, executive options create an incentive not

to pay dividends, since payment of dividends reduces the value of both exercisable and unexercisable

options held by managers.

        My results support both the option-funding and the substitution hypothesis. Consistent with the

hypothesis that in the 1990s firms repurchase shares to fund the exercise of employee stock options, I

find that firms are more likely to announce a repurchase when total options exercisable, as a percentage

of shares outstanding, are high and when many options have recently been exercised. However, the


                                                                                                            4
decision to repurchase is positively related to the number of executive options outstanding, even after

controlling for total options outstanding. When options outstanding are divided into exercisable and

unexercisable options, the repurchase decision is positively related to total options exercisable, but

unrelated to total unexercisable options. Unexercisable executive options, however, have a positive

effect on the repurchase decision. Overall, my results provide evidence that firms repurchase shares to

fund employee option exercises, but beyond that, firms are more likely to repurchase if managerial

wealth would be negatively impacted by the payment of dividends.

        For firms that announce a repurchase, I also examine the determinants of the number of shares

actually repurchased. I find that repurchases are more likely in large firms with low market-to-book

ratios, high free cash flow, and low capital expenditures. Executive options increase the likelihood of

repurchasing, but once that decision is made, the fraction of the market value of equity repurchased

depends only on total options exercisable. Executive options and unexercisable options have no

additional explanatory power. Likewise, the decrease in shares outstanding after the repurchase

announcement is positively related to total exercisable options. These results are consistent with

managers substituting repurchases for dividends. However, once the decision to repurchase is made,

managerial options provide no additional incentive beyond that of employee options in determining the

number of shares repurchased.

        The option-funding hypothesis also provides an explanation as to why many companies have

more shares outstanding months after announcing buybacks than they had before the announcement

(The Wall Street Journal, Feb. 10, 1997, p. C1). Repurchases are widely publicized by companies,

whereas offsetting dilutive actions, such as option exercises and halts in buyback programs, are not.

Stephens and Weisbach (1998) show that managers take advantage of the flexibility inherent in


                                                                                                          5
repurchases, and are more likely to follow through with a repurchase under two situations: (1) after poor

stock performance, which is consistent with the undervaluation motive, and (2) after positive cash flows,

which is consistent with the free cash flow argument. They propose several alternative measures of

actual shares repurchased, and show that only a fraction of announced open-market repurchases ever

take place. My results suggest that the method used to estimate actual share repurchases should be

considered in light of whether the firm has large numbers of stock options. In addition, the results

indicate that the notion that stock buybacks are a signal of undervaluation could be overstated in recent

years, so it is important for investors to determine the reasons behind individual buybacks before

deciding whether a repurchase is a positive sign.

        Finally, the option-funding hypothesis predicts a different market reaction to the announcement

of a repurchase than does the signaling hypothesis. If companies are repurchasing shares to fund

employee stock option exercises, then in an efficient market, the announcement- period return should

not be as positive as if the repurchase were due to undervaluation or free cash flow considerations. For

my sample of firms, I find an abnormal announcement-period return of 1.6%, which is lower than the 3-

4% abnormal return found in previous studies. I also find that the announcement return is negatively

related to nonexecutive options outstanding. This finding holds even after controlling for other factors

that affect returns, such as firm size, the percent of shares sought in the repurchase, prior stock returns,

payment of dividends, and measures of free cash flow. Further, although both repurchases and stock

options are endogenous variables, my results are robust to several methods to control for endogeneity.

Overall, my results indicate that although the motives for repurchases have changed in the 1990s, the

market recognizes this change and adjusts its reaction accordingly.




                                                                                                               6
        The rest of the paper proceeds as follows. Section 2 provides further detail on the substitution

and option-funding hypotheses, and discusses previous literature on repurchases and stock options.

Section 3 discusses the sample collection procedure and methodology. Results on the determinants of

repurchases, the market reaction to their announcement, and endogeneity issues are examined in Section

4. Section 5 concludes.




2. Hypotheses and literature review




2.1. The substitution and option-funding hypotheses

        During the 1990s, the popularity of repurchases drastically increased, but shares outstanding in

many of these companies did not correspondingly decrease. One potential explanation for this

surprising fact is the growing use of stock options as a method of compensation. Growth in stock

options could increase the popularity of buybacks for two non-mutually exclusive reasons: the option-

funding hypothesis and the substitution hypothesis.

        The first way in which an increase in stock options can affect repurchases is through the need

for firms to have shares available to fund employee option exercises, which I refer to as the option-

funding hypothesis. This hypothesis has several implications. First, firms are more likely to repurchase

shares than to pay dividends if there are many employee options outstanding that could be exercised in

the near future. Previous studies show that open market repurchases give management the option to

repurchase if the stock is undervalued (Ikenberry and Vermaelen, 1996). They also give management

the flexibility to adjust the number of shares repurchased, depending on the number of options exercised

by employees. Second, once the decision to repurchase is made, the amount repurchased should be


                                                                                                           7
positively related to options recently exercised and options currently exercisable. Finally, a greater

number of exercisable options makes it more likely that firms are repurchasing to fund these options and

less likely that a repurchase is a signal of undervaluation. In an efficient market, repurchases to fund

employee options do not provide the positive signal that repurchases for undervaluation do. Thus, the

announcement-period return should be less positive for firms that are more likely to be repurchasing to

fund options, i.e., for firms with more exercisable options.

         The economic rationale for the option-funding hypothesis is unclear. One potential explanation

is dilution. A repurchase reduces shares outstanding, while the cash used to repurchase the shares

reduces paid-in capital, but is not deducted from earnings. Thus, by repurchasing shares in conjunction

with option exercises, firms are able to avoid the dilution of basic EPS that would occur if shares

outstanding increased.1 In a traditional corporate finance framework, of course, earnings dilution

should not be important. Only cash flow should matter, and "cosmetic" changes in reported earnings

should be irrelevant. However, the common view among practitioners is that reported earnings do

matter, even above and beyond cash flow. Anecdotal and empirical evidence tends to support this

view:


             a. One commonly cited reason for the concern about EPS is that managers believe that
                analysts and investors focus on EPS and EPS growth. If these individuals blindly apply
                multiples, such as P/E ratios, to the reported earnings of growth stocks, EPS dilution is
                important. In fact, Microsoft recently announced that, at the urging of analysts, it would
                resume buying back its stock to provide for the company's huge pool of employee
                stock options and to counteract potential dilution of its shares (The Wall Street
                Journal, Aug. 8, 2000, p. B6).

1
 Firms also report fully diluted EPS, which adds interest payments that would not have been made upon conversion
back to earnings, and increases the denominator to reflect the effect of dilutive securities such as stock options and
convertible securities. Options are considered dilutive if their exercise price is below the market price. If firms are
worried about diluted EPS, then they would likely buy back shares as options come in-the-money rather than as they
are exercised.


                                                                                                                      8
              b. In addressing merger accounting methods, Bruce Wasserstein states that, "with many
                 investors focused on earnings, companies often hesitate to take on dilutive transactions."
                 (Wasserstein, 1998)
              c. After the SEC issued a guideline stating that companies involved in stock repurchase
                 programs would not qualify for pooling treatment if they acquired another company,
                 several large corporations, including Cisco Systems, 3Com Systems, and Gillette,
                 canceled their repurchase programs.
              d. Andrade (1999) finds that EPS accretion has a positive and significant effect on
                 acquirer abnormal performance, both at the announcement and for 18 months following
                 completion of the deal. The magnitude of this effect is higher for firms with a larger
                 percentage of unsophisticated investors.
              e. Burgstahler and Dichev (1999) and Kang (1999) show that firms manipulate both cash
                 flows from operations and changes in working capital to avoid earnings decreases and
                 losses.


        It is true that by repurchasing shares, managers give up financial assets that could be put to

work elsewhere in the firm, so repurchasing can affect the firm's ability to generate future earnings.

However, if the funds used to repurchase stock would have been distributed to shareholders in another

form, such as dividends, then the level of earnings will not be changed. Alternatively, if the share

repurchase is financed by a reduction in cash holdings or other liquid assets, the interest forgone should

be minimal.

        There are other possible explanations as to why firms fund employee stock option exercises by

repurchasing rather than by issuing new shares. For example, management might not want to dilute its

ownership stake. The firm could have a target capital structure, and issuing new shares would move

them away from this target. Issuing new shares might require an increase in authorized shares, which (a)

requires a shareholders' vote and (b) has tax implications in some states.

        Regardless of the motive, I show that a strong correlation exists between stock options and

both the decision to repurchase and the amount repurchased. Moreover, my research indicates that the

market realizes that shares repurchased as a result of stock options do not have the signaling impact of


                                                                                                             9
other repurchases: the announcement-period return is lower for firms with high levels of nonmanagerial

stock options.

        The second hypothesis relating growth in stock options to an increase in buybacks is the

substitution hypothesis. The substitution hypothesis predicts that managers have excess cash flow that

they want to return to shareholders. They can accomplish this goal through either a stock repurchase or

a dividend increase. If the excess cash is transitory, a repurchase is preferable since there is no future

commitment involved. Guay and Harford (2000) demonstrate empirically that the choice between a

dividend increase and a repurchase depends on the permanence of the cash flow shock. The post-

distribution cash flows of dividend-increasing firms do not revert back to pre-distribution levels,

whereas those of repurchasing firms tend to settle below the pre-distribution levels. However,

managers who hold options on their company's stock have further reasons for preferring repurchases.

Unlike dividends, repurchases do not dilute the per-share value of the firm since the cash outflow is

matched by a proportionate reduction in shares outstanding. Since the value of stock options is

negatively related to future dividend payments (unless the options are dividend protected), managers

maximize the value of their options by substituting repurchases for dividend growth.2 Managers could

also choose to retain earnings, which would not affect the value of their options as long as agency

problems did not increase as a result. However, if managers plan to exercise options in the near future,

the positive announcement return associated with repurchases could further increase managerial wealth,


2
 Murphy (1998) reports that only 1% of CEOs with options have dividend protection. Firms have incentives not to
offer dividend-protected options due to their accounting treatment. Under current accounting standards, dividend-
protected options are considered variable-plan options because the strike price is contingent upon future events.
According to FASB rules, the cost of these options must be recognized on the income statement as compensation
expense. Fixed-plan options do not result in a compensation expense as long as the option is granted with an
exercise price greater than or equal to the current market price.



                                                                                                                10
resulting in a preference for repurchases rather than retention. Since the main focus of this paper is how

options affect the decision to repurchase, I am interested in why a firm chooses to increase its dividend

rather than repurchase, not why a firm chooses to pay out rather than retain earnings.

        The substitution hypothesis has several testable implications. First, if the substitution hypothesis

is true, then managerial options provide additional incentives beyond employee options for firms to

repurchase. Managers will be more likely to choose a repurchase over a dividend increase when the

decision has a greater effect on their wealth, i.e., when there are more managerial options outstanding.

The more options managers hold, the more likely they will be to choose a repurchase over a dividend

increase. This relation should hold for both exercisable and unexercisable options held by management,

since dividends decrease the value of both. Second, once the decision to repurchase is made,

managerial options provide no additional incentive beyond that of employee options in determining the

number of shares actually repurchased. Finally, the announcement return to the repurchase could be

positively related to managerial options if options align managers' incentives with those of shareholders

in such a way that the repurchase creates value for shareholders.




2.2. Literature review

        Several other studies examine stock options and the payout policies of firms. Their conclusions

as to the effect of options on payout are mixed. Bartov, Krinsky, and Lee (1998) examine

management's decision to repurchase stock versus increase dividends. Using Compustat's common

shares reserved for conversion of stock options (data item #215) as a proxy for management options,

they find that managers choose a repurchase over a dividend increase when (a) management views




                                                                                                           11
shares as undervalued, (b) management has options that are not dividend protected, and (c) institutions

own a large fraction of equity.

        Jolls (1998) also examines the dividend increase versus repurchase decision. Using option

grants in the preceding year as a measure of employee options outstanding, and collecting executive

options outstanding from proxy statements, she finds that executive options, rather than employee

options, play a role in repurchase behavior. Her results support the substitution hypothesis, but not the

option-funding hypothesis.

        Fenn and Liang (2000) examine the payout policy of firms in the 1990s. Using Execucomp

data on executive stock and options, they find that firms with the greatest potential agency problems (as

proxied by high free cash flow and a low market-to-book ratio or low managerial ownership) have the

highest payouts. Further, managerial stock options are associated with a lower probability of using

dividends to return cash to shareholders and a higher probability of using repurchases. Consistent with

the substitution hypothesis, they conclude that growth in managerial options could explain the shift

towards repurchases in recent years.

        Weisbenner (1999) finds that the size of a firm's option program and the number of option

exercises during the year are strong predictors of actual share repurchases. Unlike previous papers,

Weisbenner collects data on both total and managerial options outstanding. However, after controlling

for the overall size of a firm's option program, he finds no correlation between the option holdings of

executives and stock buybacks. Weisbenner does find that executive options increase the probability of

a firm retaining earnings versus paying them out.

        My paper differs from previous research in several respects. First, rather than using proxies, I

collect data on both total and executive options, which provides more accurate measures. Bartov,


                                                                                                           12
Krinsky, and Lee (1998) use Compustat data to proxy for managerial options. Using the same variable

they do, I find that the correlation between this variable and executive options outstanding or exercisable

is less than 0.30 for my sample of firms. Similarly, Jolls (1998) uses option grants to proxy for

employee options outstanding. However, options granted in the previous year might not be a good

measure of exercisable options since there is often a multiyear vesting period associated with grants.

        Second, I collect data not only on options exercised in the year of the repurchase but also on

the total number of options outstanding and exercisable in the year before, the year of, and the year after

the repurchase for all employees of the firm. I collect the same data for the top management of each

firm, so that I know the fraction of total options held by the firm's decision makers. Previous work,

such as Jolls (1998) and Fenn and Liang (2000), has trouble separating the option-funding hypothesis

from the substitution hypothesis. By collecting data on total options and options held by executives, and

by further separating each into exercisable and unexercisable options, I am better able to disentangle the

substitution effect from the option-funding effect.

        Finally, unlike Fenn and Liang (2000) and Weisbenner (1999), I am interested in how stock

options affect the initial repurchase decision and the market reaction to the announcement of a

repurchase. Consequently, I examine repurchase announcements reported in SDC, as well as actual

repurchases that take place during the year. Using this information, I study not only the effect of total

and executive options on a firm's decision to repurchase versus increase dividends, but also the effect of

stock options on the market's reaction to the repurchase announcement.




3. Data and methodology




                                                                                                            13
        The sample of repurchases is from Securities Data Corporation's Mergers and Acquisitions

database. I begin by collecting all open market repurchases with original announcement dates between

January 1, 1991 and December 31, 1996, resulting in a sample of 5,147 repurchases. Restricting the

sample to firms on the Center for Research in Security Prices (CRSP) NYSE, Amex, or Nasdaq tapes

with returns available over the period of interest reduces the sample size to 4,661. Standard and Poor's

Execucomp database contains 1,792 of the repurchasing sample. However, data on managerial options

outstanding and exercisable for the year before, the year of, and the year after the repurchase

announcement are only available on the Execucomp database for 1,063 observations. Compustat data

availability on book equity, free cash flow, and total assets further reduces the sample to 755

repurchases. For these 755 observations, I collect data from the firm's annual report on options

exercised, total options outstanding, and total options exercisable held by all employees in the year

before, the year of, and the year after the repurchase announcement. For 43 of the observations, I am

unable to obtain 10-Ks. Thus, the final sample consists of 712 repurchases.

        To examine the characteristics that lead a firm to choose one payout method over the other, I

also collect a sample of firms that increase their dividends during this time period. I begin with 1,104

firms listed in the Execucomp database that are not part of my repurchase sample. For these firms, I

collect data from CRSP on regular quarterly dividends paid during fiscal years 1992-1996. If the firm

increases its dividend during the year, I retain that observation. This results in a total of 1,038 firm-

years with dividend increases. Since data collection is time-intensive, I select a random sample of one-

fifth of these observations, for a final dividend-increasing sample of 205. For this sample, the

announcement day is considered to be the date on which the board of directors declares the increased

dividend.


                                                                                                            14
        Executive stock option data are from Standard and Poor's Execucomp database. This

database contains information on executive compensation and ownership for the S&P 1500 companies,

beginning in 1992. Using these data, I calculate the total number of options held, the number of

exercisable and unexercisable options held, and the shares owned by top executives. By combining the

Execucomp data with the data on options held by all employees collected from the annual reports, I

separate options outstanding (exercisable and unexercisable) into executive versus nonexecutive options

by subtracting options held by management from total options held by all employees. Due to different

report dates (the Execucomp data are from proxy statements rather than annual reports), a handful of

companies have more executive options outstanding than total options outstanding. For these

companies, I assume that executive options must account for almost all of total options, and set

executive options equal to total options. The results are not affected if I assume that nonexecutive

options are missing, however.

        The number of repurchases increases steadily during the sample period, with 72 occurring in

1993, 177 in 1994, 208 in 1995, and 255 in 1996. However, the sample size in 1993 is potentially

reduced due to my requirement that Execucomp data be available in the year prior to the repurchase

announcement. The dividend-increasing sample is relatively evenly divided across the sample period,

with 60 occurring in 1993, 50 in 1994, 43 in 1995, and 52 in 1996.




4. Results




4.1. Characteristics of repurchasing versus dividend-increasing firms




                                                                                                       15
         Panel A of Table 1 provides summary statistics for the dividend-increasing and repurchasing

firms. The focus is on variables intended to control for other motivations for repurchases, such as to

signal undervaluation or to alleviate free cash flow problems. The average dividend-increasing firm

increases its dividend by 9.8% (measured by the dividend increase as a percent of the last dividend

paid), and earns an abnormal return of 0.5% over the three-day announcement period. Repurchasing

firms announce that they will buy back an average of 6.4% of shares outstanding, with an abnormal

announcement-period return of 1.6%.3 This return is lower than that found in earlier studies, which is

consistent with the market recognizing that repurchases in the 1990s are less likely to signal

undervaluation.

         If undervaluation is a motive for repurchases, then repurchase announcements should follow

poor stock returns. Management is less likely to consider firms with large recent runups to be

undervalued. On the other hand, positive stock returns increase the number of in-the-money options,

which increases the likelihood of option exercise and the dilution of EPS. In the 40 days prior to the

announcement, dividend-increasing firms experience no abnormal price movements. Repurchasers

exhibit a ­3.6% abnormal return, consistent with undervaluation. Note, however, that although the

average performance of repurchasing firms is poor immediately prior to the repurchase announcement,

long-run performance must be such that the stock price is above the exercise price of the firms' options,

since most of these firms exercise options in the year they in which announce the repurchase.

         Size has been used in previous studies as a proxy for financing costs and information

asymmetry. Larger firms should have lower financing costs, more stable cash flows, and less

3
 SDC contains information on the total dollar value of the transaction and the percent of shares outstanding sought
by the firm. In some instances, the dollar value of the repurchase is reported, but the percent sought is missing. In



                                                                                                                   16
information asymmetry. Lower financing costs enable a firm to distribute more cash to shareholders,

since if they need to raise money in the future, the funds will be relatively inexpensive. The repurchasing

sample is larger than the dividend-increasing sample, with an average market value of equity of $5,847

million, compared to $2,632 million. The medians are smaller, at $1,391 and $1,114, respectively.

These numbers are larger than reported by Guay and Harford (2000). However, my sample comes

from a more recent time period, and restricting my sample to firms on Execucomp results in larger firms.

         Market-to-book asset ratios proxy for investment opportunities. Agency costs of payouts

predict a negative relation between payouts and market-to-book, since firms with good investment

opportunities maximize shareholder value by using cash flow to finance investment, rather than

distributing cash flow to shareholders. The repurchasing sample has market-to-book asset ratios similar

to those reported by Guay and Harford (2000), and higher than those of the dividend increasers.

         Debt can also influence a firm's payout decision. High debt levels can proxy for financial

distress. Debt is also a substitute for payouts to shareholders, since it alleviates free cash flow

problems. Firms with high debt are thus less likely to distribute cash. Bagwell and Shoven (1988)

show that highly leveraged firms are less likely to repurchase. I find that dividend-increasing firms have

more debt than repurchasing firms, when measured as total debt to assets or long-term debt to assets.

These numbers are consistent with leverage proxying for financing costs. Firms with high leverage have

high financing costs, and would be expected to repurchase fewer shares. It is also consistent with

dividend-paying firms having more stable cash flows, and thus being able to take on more debt.




these cases, I define the percent sought as the total dollar value divided by closing stock price four days prior to the
announcement, divided by shares outstanding (Comment and Jarrell, 1991).


                                                                                                                      17
         Firms with high capital expenditures should have both better investment opportunities and less

free cash flow, and thus should pay out less. Capital expenditures do not differ significantly between

repurchasing and dividend-increasing firms. Finally, firms with high levels of free cash flow derive

greater benefits from distributing cash to shareholders, since they are at greater risk of overinvesting.

The repurchasing firms have higher ratios of free cash flow to assets than do the dividend-increasing

firms.

         Since I am interested in the effects of stock options on repurchases, Panel B of Table 1

provides statistics on total and executive options outstanding and exercisable in the year after the

announcement, as well as total options exercised in the year of the announcement. Both the substitution

hypothesis and the option-funding hypothesis predict that repurchasers should have more options

outstanding and exercisable than dividend-increasing firms. Panel B confirms this hypothesis.

Compared to dividend increasers, repurchasing firms have almost twice the number of executive and

employee options outstanding and exercisable in the year of and the year after the event. Similarly,

repurchasers have significantly more options exercised in the year of the announcement.

         Table 2 contains the industry distribution of the dividend-increasing and repurchasing firms,

relative to the universe of all firms on the Compustat tapes during 1992-1996. A variety of industries

are represented. Repurchases occur most often in high-tech industries such as chemicals (81),

machinery and computer equipment (59), finance & insurance (58), and electronics (47), and in service

and retail-related industries. As a percentage of the total number of firms in each industry available on

Compustat, however, repurchases occur predominantly in tobacco products (31%), printing and

publishing (21%), food and kindred products (16%), and wood and paper products (16%). In the




                                                                                                            18
dividend-increasing sample, the utilities industry accounts for the greatest number of observations and

for the greatest percentage of observations relative to the size of the industry (8%).




4.2. The decision to repurchase versus increase dividends

         I first examine motives for repurchases. In particular, do stock options appear to influence a

firm's decision to repurchase stock rather than increase dividends? Table 3 provides a logit model of

the decision to repurchase shares versus increase dividends. Consistent with the univariate results,

larger firms with higher ratios of free cash flow to assets are more likely to repurchase than to increase

dividends. Likewise, the probability of a repurchase is negatively related to a firm's stock price

appreciation prior to the announcement. The first regression also indicates that repurchases are more

likely both when more options have been exercised in the year of the repurchase and when the firm has

more total options outstanding in the year after the repurchase.4 These results are consistent with the

option-funding hypothesis.

         The second regression examines whether executive options have explanatory power beyond

total options. If firms are buying back stock to fund employee options, repurchases should be positively

related to the number of options outstanding. However, managerial options provide an additional

incentive, since paying dividends will decrease the value of unprotected options. Consequently, if

managers are acting in their own best interests, executive options could have additional explanatory

value beyond total options. Consistent with this hypothesis, in the second regression, the coefficients on

both total options outstanding and executive options outstanding are positive and significant. Similar

4
 The correlation between options exercised in year 0 and options exercisable (outstanding) in year +1 is 0.366 (0.417).
The mean (median) ratio of options exercised during the year to options exercisable at the beginning of the year is



                                                                                                                     19
results are obtained if I decompose total options outstanding into executive and nonexecutive options.

Here, the positive coefficient on executive options is consistent with both the substitution hypothesis and

the option-funding hypothesis. The positive coefficient on employee options, however, is only consistent

with firms buying back stock to fund employee option exercise.

        The third regression decomposes options outstanding into exercisable and unexercisable

options. In this specification, repurchases are significantly positively related to total exercisable options,

but not to total unexercisable options. This result is consistent with firms buying back stock to fund

options that they expect to be exercised in the near future. Exercisable executive options have no

explanatory power beyond that of total exercisable options. Unexercisable executive options, however,

are significantly positively related to the decision to repurchase. This result is consistent with the

managerial wealth hypothesis. Total exercisable options determine whether a firm needs to repurchase

stock to fund upcoming option exercises. Paying out dividends, however, decreases the value of all

managerial options, whether they are currently exercisable or not. Consequently, if managers own large

numbers of unexercisable options, they will be more likely to repurchase shares than increase dividends,

even if they do not need to fund employee option exercises. Overall, my results support both the

option-funding and the substitution hypotheses. Firms announce repurchase programs when they need

shares to fund option exercises among employees and when managerial wealth would be negatively

impacted by a dividend increase.

        The parameters of the logistic models developed in Table 3 reflect only a subsample (205) of

the population of 1,038 dividend increasers that could have been included in the study. As a result, the



50% (30%). However, excluding options exercised during the year from this and later regressions does not affect the
sign or significance of other independent variables.


                                                                                                                 20
population probability of a repurchase (p) cannot be used to compute the likelihood function associated

with the sample. Instead, as shown in Palepu (1986), the logistic regression should be estimated using

the conditional probability of observing a repurchase given that a dividend increase is included in the

sample (p'). Because the sample includes 19.7% of the dividend increasers, the conditional probability

(p') of observing a repurchase is

                                                        (1)( p)
                                        p'=                              .                                    (1)
                                              (1)( p) + ( 0.197)(1 - p )

For a logistic specification,

                                                      1
                                           p=                 .
                                                 (1 + e - x )

                                                                                                              (2)

Therefore,

                                                         1
                                        p'=            ln(0 .197 )- x
                                                                            .                                 (3)
                                              (1 + e                    )

Because the likelihood function maximized in the regressions in Table 3 uses the above expression p', all

of the parameters associated with p are unaffected other than the constant term. The constant term

differs by ln (0.197) or ­1.625. Thus, if the purpose of the model is only to determine whether the set

of variables has a significant statistical relation to the repurchase probability, the bias is unimportant.

However, to classify firms that announce a repurchase into those predicted to repurchase and those

predicted to increase their dividend, I make the above correction to the intercept to account for the

subsample of dividend increasers that were excluded from the logistic regression. Using regression 3 of

Table 3 with this correction, the threshold value of p that minimizes the sum of the probabilities of type I

and type II errors is 0.42. At this threshold, 35 firms repurchase when expected to increase their


                                                                                                              21
dividend. These firms have a mean (median) announcement return of 2.58% (1.92%). Firms that

regression 3 predicts to repurchase, and that do repurchase, have a mean (median) announcement

return of 1.56% (1.25%). The medians are significantly different from each other at the 10% level. The

results are consistent with the market already partially impounding the good news of a repurchase into

the price of firms it expects to repurchase. Firms that are not predicted to repurchase, based on the

model, have an additional surprise component. Alternatively, news of a repurchase for firms predicted

to increase their dividend could be positive due to tax differences between repurchases and dividends.




4.3. Determinants of the amount repurchased

        The results in Table 3 indicate that options influence a firm's decision to repurchase its stock.

As documented in Stephens and Weisbach (1998), however, a firm that announces a repurchase does

not always follow through with it. Table 4 examines the determinants of the dollar amount actually

repurchased by firms announcing a repurchase program. The dependent variable is equal to the dollars

spent on repurchases in the year of and the year after the announcement, defined as Compustat's dollars

spent on repurchases (annual data item #115) minus any decrease in the par value of preferred stock

(annual data item #130), divided by the market value of equity.5 Jagannathan, Stephens, and Weisbach

(2000) state that the Compustat measure overstates open market repurchases because (a) it includes

Dutch auctions, privately negotiated deals, and self-tender offers and (b) it includes repurchases of other

securities and conversions of other securities into stock. Since I focus only on firms that have

announced repurchase programs, the first problem is minimized, and Jagannathan et al. show that this

5
 I measure actual repurchases over a two-year window since Stephens and Weisbach (1998) show that firms acquire
67-79% of shares announced within a two-year period. In unreported results, I also examine the determinants of the
percent sought in the repurchase announcement. The results of those regressions are similar.


                                                                                                                22
measure is relatively accurate. To minimize overstatements of repurchases caused by the second

problem, I subtract decreases in the par value of preferred stock. This correction should make the

repurchase measure more reflective of repurchases of common stock only. The control variables used

in this table are the same as those used in Table 3, and the interpretations are similar. The exception is

the measure of stock returns. In Table 3, I use the excess returns in the 40 days prior to the

announcement. In Table 4, I measure the buy-and-hold return in the one year after the announcement in

order to determine if the firms' actual stock returns after the announcement determine the amount

actually repurchased.

        Regression 1 of Table 4 indicates that the amount actually repurchased is positively correlated

with firm size and with the percentage of shares outstanding that firms initially announce that they are

buying back. As discussed earlier, if larger firms have lower financing costs or information asymmetry,

and more stable cash flows, these firms can afford to pay more out to shareholders. The coefficient on

free cash flow, standardized by book assets, is positive and significant, consistent with firms with high

free cash flow having more money available for repurchases. Firms with high market-to-book asset

ratios or high capital expenditures repurchase a lower percentage of shares. If these variables proxy for

investment opportunities, then this result is consistent with agency considerations. Large levels of debt

result in lower repurchases, consistent with free cash flow arguments or financial distress costs. Finally,

high post-announcement returns lead to lower levels of repurchasing, consistent with the firm no longer

being undervalued. However, when I interact the post-announcement return with a dummy equal to one

if the firm's exercisable options are above the sample mean, the coefficient is positive and significant.

Stock price increases generally reduce the amount repurchased, consistent with the signaling theory.




                                                                                                            23
However, if the price increases result in more exercisable options, the firm repurchases more, which is

consistent with the option-funding hypothesis. After controlling for these factors, the coefficient on total

options outstanding is positive and significant. This indicates that firms with more options outstanding

repurchase more shares, which further supports the option-funding hypothesis.

        Regression 2 of Table 4 examines whether executive options have explanatory power beyond

total options outstanding. The coefficients on the control variables are similar to those in regression 1.

The coefficient on total options outstanding is positive and highly significant in this regression, but

executive options have no additional explanatory power. Combined with Table 3, this regression

provides evidence that while executive options influence the decision to repurchase, actual repurchases

following the decision depend only on total options outstanding. Regression 3 replicates regression 2,

but includes industry dummies, with the result that the coefficient on debt levels loses its significance.

The coefficient on total options outstanding remains positive and significant.

        Regression 4 is similar to regression 2, but divides total options into exercisable and

unexercisable options. The coefficients on the control variables are similar to the previous regressions.

The coefficient on total options exercisable is positive and significant, while unexercisable options are

positive but insignificant. Executive options, whether exercisable or unexercisable, are not significant.

        In summary, firms repurchase more shares when they are large, have low growth opportunities

(proxied by the market-to-book ratio), and are undervalued (proxied by one-year post-announcement

returns). As free cash flow increases, the amount repurchased also increases. The key results in Tables

3 and 4, however, indicate that executive options influence the decision to repurchase rather than pay

dividends, but once the decision to repurchase is made, the amount actually repurchased depends only

on the number of exercisable options held by all employees.


                                                                                                             24
        While Table 4 examines the actual repurchases of firms, it does not show the net effect of option

exercises versus repurchases. To measure this effect, I examine the change in shares outstanding in the

year after the repurchase announcement in Table 5. Shares outstanding are obtained from CRSP, and

adjusted for any stock splits or dividends that occur during this period. The dependent variable is the

change in shares outstanding in the year after the repurchase announcement, divided by the number of

shares outstanding at the time of the announcement. A positive (negative) number indicates that shares

increased (decreased). The independent variables are the same ones used in Table 4. Consistent with

size proxying for financing costs or information asymmetry, larger firms repurchase more and thus have a

greater decrease in shares outstanding. Higher debt levels are associated with greater share dilution

(i.e., a greater increase in shares). This relation is consistent with firms with a higher chance of financial

distress not wanting to distribute cash through a repurchase. It is also consistent with debt substituting

for payouts, since firms with more debt having lower agency costs. Higher capital expenditures and high

post-announcement stock returns are also associated with greater share dilution. However, the

interaction of the post-announcement return with a dummy equal to one if exercisable options are above

the sample mean has a negative and significant coefficient, indicating that when positive returns result in

more exercisable options, the firm repurchases more stock.

        After controlling for the above factors, the coefficient on total options exercised during the year

is close to one, indicating that for every option exercised, shares outstanding increase by almost one. If

firms were repurchasing shares at the same time option exercises were occurring, this coefficient should

be close to zero; however, if firms repurchased last year in anticipation of this year's exercise, shares

outstanding would have decreased last year and the coefficient on exercises this year would be one.

Thus, my results provide evidence that firms are repurchasing stock in anticipation of future exercises.


                                                                                                              25
The results also support the idea that firms are repurchasing to avoid earnings dilution. Since options

are considered dilutive if their exercise price is below the market price, firms worried about dilution

should buy back shares as options come in the money rather than waiting until they are exercised. In

contrast to the coefficient on options exercised, total options outstanding and exercisable in the year

after the repurchase announcement are significantly negatively related to the change in shares

outstanding. Thus, while there is a one-to-one correspondence between options exercised and the

increase in shares outstanding in the year of the repurchase, firms that have more options outstanding,

and in particular more exercisable options, are more likely to repurchase to fund upcoming exercises

and prevent dilution. As in Table 4, executive options do not provide additional explanatory power

beyond total options.

        These results have implications for the best method to use to calculate actual shares

repurchased. Jagannathan, Stephens, and Weisbach (2000) propose that examining monthly decreases

in CRSP shares outstanding, adjusted for new stock issues, provides a lower bound on shares

repurchased since this measure does not adjust for shares reissued to employee benefit plans or through

the exercise of stock options. Alternatively, using purchases of common and preferred stock from

Compustat overstates common stock repurchases since it includes conversions of other classes of stock

into common stock, purchases of Treasury stock, and redemptions of preferred stock. The problems

with the Compustat data can be minimized by focusing on firms with announced repurchase programs

and subtracting decreases in the par value of preferred stock, and Jagannathan et al. suggest that this

measure is likely to be the most accurate. My results support this conclusion, providing evidence that

option exercises have a significant effect on the CRSP measure of the change in shares outstanding,

which could seriously bias this estimate of actual share repurchases.


                                                                                                          26
4.4. Announcement returns and employee options

        The results so far indicate that options affect both the decision to repurchase and the amount

repurchased. Further, the univariate results in Section 4.1 show that the announcement-period return to

repurchases is lower in the 1990s than it is in earlier periods. If some firms in the 1990s are

repurchasing to fund employee option exercises rather than to signal undervaluation or to return free

cash flow to shareholders, then the announcement return should be less positive for firms for which this

motive is more likely. Table 6 examines this hypothesis by regressing announcement-period returns on

firm and repurchase characteristics.

        Consistent with the signaling hypothesis and with Comment and Jarrell (1991), regression 1

shows that the announcement return is significantly negatively related to stock price increases in the 40

days prior to the event and positively related to the percent of shares the firm seeks to buy back. The

return is also negatively related to a dummy equal to one if the firm pays a dividend, which is consistent

with a repurchase announcement being more of a surprise or sending a stronger signal for firms that

historically have not returned cash to shareholders through dividends. It is also positively related to the

ratio of free cash flow to assets, again indicating that the repurchase is better news if the firm is also

performing well. After controlling for these factors, the announcement return is significantly negatively

related to total options outstanding. The market appears to recognize that the firm is buying back shares

to fund employee stock option programs. However, the coefficient on executive options outstanding is

positive and significant. Since the result that announcement returns are lower for firms with many

exercisable options could also be driven by the market anticipating and pricing the repurchase




                                                                                                             27
announcement, I include a dummy for firms expected to repurchase (based on the logit model in Table

3) and interact this dummy with the options variable. The variable is not significant.

         Since large numbers of executive options could align managers' incentives with those of other

shareholders, and thus have a positive effect on the announcement return, regressions 2 and 3 divide

options into executive and nonexecutive options outstanding and exercisable to more clearly separate

the two effects. In these regressions, the coefficients on nonexecutive options are still negative and

significant, but the coefficients on executive options, while positive, are insignificant.

         The final regression shows that the announcement-day return is significantly negatively related to

the increase in total options exercisable from year ­1 to year +1.6 The change in executive options has

no additional explanatory component. In unreported results, the regressions are repeated with industry

dummies included. The results are similar.

         If the option-funding hypothesis is true, then firms with many options outstanding will repurchase

more frequently to avoid dilution. The end result could be that repurchases by these firms are more

anticipated and thus create less of a price reaction. To address this issue, I reestimate the above

regressions (a) for only those repurchases that are the first repurchase the company makes in the 1990s,

and (b) including a dummy variable equal to one if it is the first repurchase. The sign and significance of

the coefficients on the option variables do not change.

         If firms are repurchasing shares in the 1990s to fund employee stock option exercises, then

firms should begin repurchasing at the same time employees begin exercising options. In my sample,


6
  In collecting the data on options, I find no cases in which firms grant options that are immediately exercisable. The
typical option grant is made with an exercise price equal to the fair market value on the date of the grant. The option
becomes exercisable in regular installments, commencing one to two years after the date of the grant, and unless
exercised, the options terminate ten years after the date of the grant. Consequently, this change in options
exercisable is due to new vesting over time and increases in stock prices.


                                                                                                                    28
there are nine firms that had no options exercisable in the year prior to the repurchase, but for which

options became exercisable in the year of or the year after the repurchase announcement. For seven of

the nine firms, the repurchase was the first undertaken in the 1990s. An additional 17 firms had options

exercisable in the year prior to the repurchase, but the first exercise took place in the year of or the year

after the repurchase. For 13 of these 17 firms, this was the first repurchase of the 1990s. These results

provide additional evidence that firms time repurchases to fund employee option exercises.




4.5. Endogeneity/causality

        It should be noted that both repurchases and stock options are endogenous variables.

Consequently, caution must be used when interpreting any relation between the two. Both options and

repurchases can be affected by external factors such as increased cash flow, positive stock returns, etc.,

and I attempt to control for these factors in the regression analysis.

        In an attempt to control for possible endogeneity issues in the regressions in Tables 3, 4, and 5,

I examine the subset of firms in which endogeneity should be less of an issue. Certain types of firms,

such as young, high-tech firms that do not pay dividends, are more suited to using stock options as part

of their compensation scheme. If these firms have investment opportunity sets and cash flow

characteristics that make repurchasing the most valuable or efficient way to distribute cash, there will be

a natural relation between exercisable options and repurchase activity for these firms. For example, a

period of good performance will result in stock price increases, in which case more of a firm's options

are likely to be in-the-money and exercisable. This same period of good performance is also when the

firm is generating sufficient cash flow to warrant a repurchase. Consequently, I rerun the regressions in

Tables 3, 4, and 5 for only those firms that do pay dividends. The results are virtually identical.


                                                                                                           29
        Endogeneity can affect the announcement-period return regressions of Table 6 in two ways.

First, firms with more stock options tend to be young, high-tech firms, and information asymmetry is

higher for these firms than for firms with low levels of stock options. Repurchases should alleviate the

information asymmetry, and thus firms with more stock options should have higher announcement-

period returns. This relation would bias against finding a negative relation between stock options and

announcement returns, as predicted by the option-funding hypothesis. Second, firms with more stock

options should tend to have high growth opportunities. In this case, a repurchase could signal reduced

growth opportunities, which would lead to a negative relation between stock options and announcement

returns. If this relation holds, there could be a negative relation between stock options and

announcement returns, regardless of whether the option-funding hypothesis is true. Consequently, I

examine capital expenditures and research and development expenses for the repurchase sample. For

the entire sample, capital expenditures and R&D increase from the year before to the year after the

repurchase. The increase is insignificant, however. I also examine the subset of repurchasers whose

options outstanding or options exercisable are greater than the sample mean. For this subsample, the

increase in capital expenditures and R&D is positive and significant. The results indicate that the

repurchase is not a signal of decreased growth opportunities, and