The `repricing’ of executive stock options

Journal of Financial Economics 57 (2000) 129}154 The &repricing' of executive stock optionsq Don M. Chance!, Raman Kumar!,*, Rebecca B. Todd" !Department of Finance, Pamplin College of Business, 1016 Pamplin Hall, Virginia Tech, Blacksburg, VA 24061, USA "School of Management, Room 518-D, 595 Commonwealth Avenue, Boston University, Boston, MA 02215, USA Received 1 March 1997; received in revised form 17 March 2000 Abstract We examine a sample of "rms that reset the exercise prices on their executive options. These repricings follow a period of about one year of poor "rm-speci"c performance in which the average "rm loses one-fourth of its value. No other o!setting changes to option terms or compensation are made, and many "rms reprice more than once. Without repricing, a majority of the options would have been at-the-money within two years. We "nd that when faced with circumstances in which repricing might be chosen, "rms with greater agency problems, smaller size, and insider- dominated boards are more likely to reprice. ( 2000 Elsevier Science S.A. All rights reserved. JEL classixcation: G30; G32; J33 Keywords: Executive; Option; Incentives; Repricing q The authors would like to thank an anonymous referee, Bo Hiler, Kathy Ruxton, David Yermack, N. Prabhala, Dave Denis, Diane Denis, Jennifer Carpenter, and participants in the seminars at Vienna, Utah, Baruch, Virginia Tech, New Mexico, Alabama, and Boston University for valuable comments, as well as Senay Agca, Honghui Chen, Calin Valsan, and Wanying Lin for research assistance. This work was partially completed while Kumar was visiting the Yale School of Management, Yale University. Kumar acknowledges support from a Pamplin College of Business summer research grant. * Corresponding author. Tel.: #1-540-231-5700; fax: #1-540-231-3155. E-mail address: [email protected] (R. Kumar). 0304-405X/00/$ - see front matter ( 2000 Elsevier Science S.A. All rights reserved. PII: S 0 3 0 4 - 4 0 5 X ( 0 0 ) 0 0 0 5 3 - 2 130 D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 . . . the Board of Directors has determined from time to time that it is desirable to reprice certain outstanding options to bring their exercise prices into line with the then-current market price of the Company1s Common Stock. ¹ypically, this has occurred when market conditions have, in the view of the Board of Directors, arti,cially depressed the market price of the Common Stock for a protracted period, so that outstanding options are signi,cantly out-of-themoney for reasons not related to the Company1s performance. HealthSouth Corporation proxy October 28, 1994 1. Introduction Stock options are used widely in compensation and incentive plans of publicly traded corporations. A typical option grant gives the executive the right to buy a speci"ed number of shares at a "xed price, which is usually the stock price at the time of the grant, up to an expiration day. Ten years is a common time to expiration on the grant date. Many executive stock options are a combination of American- and European-style, permitting exercise at any time before expiration, with a waiting or vesting period of several years at the start. Although executive stock options normally have terms that are speci"ed explicitly in proxy statements, some corporations reserve the right to alter the terms of the option contract. One such feature is the right to change the exercise price.1 Firms can change the exercise prices of old options and/or cancel old options and reissue new options. The decision to change the exercise price normally is made by the compensation committee of the board of directors, though we shall refer to this as simply a board decision. The process of resetting the exercise price is commonly referred to as &repricing', and for consistency, we shall adopt that terminology.2 Repricing is a somewhat infrequent event. The Wall Street Journal (June 11, 1997, p. C11) reports on a survey of 250 high-tech "rms in which 21 reset exercise prices for employees and executives and an additional 13 reset exercise prices for some non-executive employees. Brenner, Sundaram, and Yermack (2000) "nd that 1.3% of executives they examine had options repriced between 1992 and 1995. Despite its infrequency, repricing is unquestionably receiving more attention. Many articles in the business, professional, and popular press have attacked the 1 Exercise prices of executive stock options are automatically adjusted in the event of a merger, stock split, or stock dividend. 2 Technically, &pricing' would refer to the process of determining a market value so the term &repricing' would not strictly be an accurate description of the process of resetting the exercise price of an option, but we use it here for consistency with the trade vernacular. D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 131 practice.3 In some cases, an entire news item is devoted to a repricing by a single company.4 In addition there is evidence that securities analysts are beginning to pay attention to repricing issues (Credit Suisse First Boston, 1998). In this paper, we examine a sample of "rms that reset the exercise prices on their executive stock options. We determine how the stock performs prior to repricing and whether that performance is "rm-speci"c or driven by market or industry factors. We compare our sample "rms with a carefully selected matched sample of "rms in the same industry that experience a similar price decline but choose not to reprice. We "nd that the poor performance prior to repricing is not driven by market or industry factors, that repricings are not accompanied by o!setting factors either in option terms or other cash compensation, and that many "rms reprice more than once. Investors do not react to the repricing, at least around the proxy "ling date. There is no abnormal performance subsequent to repricing. We also "nd that the majority of repriced options have substantial values prior to repricing. Using actual post-repricing performance, over half of the options would have been at-the-money without repricing within 19 months. The direct cost to shareholders of repricing is small, though the gain to an individual executive can be substantial. Finally, using a matched sample, we "nd that repricing is more likely for smaller "rms with insider-dominated boards and greater agency problems. Our paper proceeds as follows. Section 2 provides background information on the practice of repricing. Section 3 describes the data set and the tests. Section 4 reports the results of our tests, which examine the performance of the stock around the repricing event. Section 5 presents an analysis of why "rms reprice. Section 6 provides our conclusions. 2. Background information and previous research on option repricing A "rm might reprice its executive stock options for a number of reasons. One is to remove the loss in option value that could have resulted simply from poor market or industry performance. The argument, however, goes two ways. During periods of favorable market and industry performance, the stock price can rise even when "rm-speci"c performance is poor. An executive's options could, therefore, have considerable value not warranted by the executive's 3 See for example, USA Today (April 29, 1997, p. 12A), The Wall Street Journal (October 29, 1997, p. B1; April 9, 1998, p. R12; April 9, 1998, p. R4; October 30, 1998, p. B2; November 3, 1998, p. B20; April 8, 1999, p. R5; June 2, 1999, p. B1), The New York Times (July 15, 1998, p. D1), and Risk (October, 1998, p. 13). 4 See, for example, The Wall Street Journal (February 17, 1998, p. B6; November 23, 1998, p. B7; March 31, 1999, p. B2). 132 D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 performance. It follows that "rms might consider repricing their options upward after strong positive market- or industry-driven performance. As we show later, there is plenty of evidence that exercise prices are lowered but rarely, if ever, raised. When considering this point, however, one must ask whether a manager is responsible for "rm-speci"c performance or whether there are elements of randomness or, as some people would characterize it, luck. Experts are likely to disagree on this point, and we do not believe it can be resolved in this study. Fortunately, our "ndings are not dependent on a resolution of this issue, though the interpretations of certain results are. Some "rms argue that repricing is necessary to retain managerial talent. Another reason for repricing could be that management is simply too entrenched. Management might be able to convince the board of directors that it either deserves another chance to straighten out the problems or that the "rm's problems are market- or industry-driven. Another reason is given by Gilson and Vetsuypens (1993), who suggest that "rms in "nancial distress could be pressured by creditors to reprice to reduce the incentive to take on high-risk projects. Tax laws can partially explain why "rms reprice. To qualify for favorable tax treatment, executive stock options must be exercised sequentially. Thus, if options issued earlier are out-of-the-money while options issued later are inthe-money, the former might have to be repriced to qualify for favorable tax treatment on exercise.5 Alternatively, "rms might consider canceling old options and issuing new options. Firms might prefer repricing over cancellation and reissuance, however, because repricing would be modifying an existing contract and would not necessarily require shareholder approval, while cancellation and reissuance could require shareholder approval and, therefore, bring more scrutiny to the matter. 2.1. Academic research on repricing Gilson and Vetsuypens (1993) study a sample of "rms that "le for bankruptcy or privately restructure their debt during the years 1981}1987. Twenty-"ve of the 77 sample "rms reprice. The median stock price at the time of the repricing is about half the old exercise price, implying that the typical repricing is a 50% reduction in the exercise price. Firms that reprice signi"cantly underperform the 5 Consider a "rm that grants options with an exercise price of $50. The stock then falls to $40 and the "rm grants new options at $40, leaving the old options intact. If the stock rises to $48, the options issued at $50 have to be exercised "rst to qualify for favorable tax treatment. (The favorable tax treatment applies only to tax-quali"ed executive stock option plans; the plans covered in this study are tax-quali"ed.) D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 133 market for six years prior to the repricing date. Only 11 of the 25 "rms explain the repricing to their shareholders and only one "rm mentions the performance of the market as the reason. Saly (1994) develops a theoretical model and conducts some empirical tests of repricing after the October 1987 crash. She argues that repricing is an appropriate response after market-induced declines in the stock price. She examines repricing indirectly by comparing the number of options outstanding before and after the crash. She "nds that option grants increase signi"cantly following the crash and more so for "rms whose stock falls by the largest percentage. Although these results do not examine the act of repricing directly, they are consistent with the notion that many "rms reprice after market-induced declines in the stock. The question of whether "rms reprice after signi"cant marketinduced runups in the stock price is not addressed. Acharya et al. (2000) develop a theoretical model that argues that under some circumstances repricing can be optimal. The key determinant is the set of managerial compensation contracts that the "rm can o!er. With a full range of possibilities, Acharya et al. (2000) show that repricing is never in the shareholders' interests. Under typical compensation contracts, however, they show that repricing can be valuable. In one such scenario, managers are sensitive to economy-wide factors and managerial talent is relatively expensive to replace. Even without the Acharya}John}Sundaram model, it is not hard to create circumstances in which repricing has positive e!ects. Any of the previously cited justi"cations for repricing could conceivably bring positive bene"ts to shareholders. Whether they do or not is an empirical question. Brenner et al. (2000) build a model for valuing repriceable options and generate numerical estimates using hypothetical inputs. They conclude that repricing has a small e!ect on the ex ante value of the option, but the potential ex post value can be large. Using a sample of actual repriced options, they "nd that the occurrence of repricing is more likely for smaller "rms and for "rms with poor performance. Corrado et al. (1998) provide an alternative to the Brenner}Yermack} Sundaram model by using a utility-maximizing approach that re#ects the illiquidity of these options and by incorporating non-option wealth and vesting requirements. Their paper, however, does not provide any empirical results of speci"c repricings. In terms of its focus, our paper is closest to that of Brenner, Sundaram, and Yermack, but there are important di!erences. In their logit analysis of the incidence of repricing, their observations are at an executive-year level. In addition to including all executive-year observations for which there is a repricing, they use all executive-year observations in which the options are not repriced, without making an attempt to determine whether the options are likely to be out-of-the-money. Moreover, executive-year observations are unlikely to be independent. In our logit analysis, the observations are at the "rm level, and 134 D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 for our matched sample of non-repricing "rms, we use only those that perform as poorly as the sample "rms and are, therefore, likely to have options that are out-of-the-money. Later we suggest that our "rm-level approach is more appropriate. We also suggest that including all non-repricings without respect to performance in the logit analysis might not be appropriate to address the question of why some "rms reprice whereas others facing a similar situation do not. We also provide a number of additional results, including an event study around the proxy "ling date, an analysis on how the options might have performed had they not been repriced, a more precise quanti"cation of the loss in shareholder wealth that precedes the repricing and over which period this loss is incurred, and an examination of the incidence of multiple repricings. 3. The data and methods In 1993 the SEC began requiring "rms to provide information in proxy statements about any instances in which they reprice executive stock options. If a "rm reprices in 1992 or later, it must provide a ten-year history detailing any previous repricings for at least the CEO and the four highest-paid executives. The rule does not apply to employee stock options. We conduct a keyword search using the online National Automated Accounting Research System (NAARS) database created by Mead Data Central of Dayton, Ohio, and available on Lexis/Nexis. It comprises about 4000 of the largest publicly traded companies and is maintained by the American Institute of Certi"ed Public Accountants. The NAARS database contains ten years of annual reports, proxies, and 10Ks. After examining more than 300 "rms that mention certain key words, we obtain 40 companies and 74 repricing events with a ten-year repricing history. We require that the "rm have return data available starting at least 300 trading days prior to the event. This restriction provides su$cient data for estimation of parameters. After eliminating multiple repricings of any one "rm that are clustered so closely as to interfere with estimation of the market model parameters, we end up with a sample of 37 "rms and 53 events. Twenty-six "rms have one repricing, seven have two repricings, three have three repricings, and one has four repricings. For a given "rm-event, there are options held by di!erent o$cers and di!erent exercise prices due to their having been granted at di!erent times in the past. The repricings occur during the period of 1985}1994. 3.1. Descriptive statistics Table 1 contains descriptive statistics. One-hundred thirty distinct option issues are associated with the 53 repricings. Panels A and B provide information D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 135 Table 1 Descriptive statistics from a sample of "rms that reset the exercise prices of (&reprice') their executive stock options. A given event includes all issues repriced on a given date. The number of o$cers is the total number of top executives who have options repriced. The number of issues repriced is the total number of distinct options, as measured by the original exercise price and maturity, in which exercise prices are reset. The number of new options issued can be negative, because in some cases options are canceled and new options are issued with the former exceeding the latter. The sample is obtained by identifying "rms that provide ten-year repricing tables in proxy statements available on the NAARS (National Automated Accounting Research System) database of Lexis/Nexis. The sample consists of 37 "rms, 53 repricing "rm-events, and 130 distinct option issues. Variable Mean A. Repricing events [N"53] Number of o$cers 3.15 Number of issues repriced 2.21 Number of old options repriced 210,232 Number of new options issued !1,185 B. Issues repriced [N"130] Percentage change in exercise !41.33 price Change in months to maturity 9.21 Remaining months to maturity 66.62 (Number of options)](change 331,308 in exercise price) (New exercise price)/(closing 1.0002 price on day of repricing) Standard deviation 2.11 1.89 358,248 25,340 19.40 90th percentile 10th percentile 6 4 487,632 0 1 1 13,600 !2,320 !18.29 !66.65 27.02 32.91 536,644 45 113 903,500 0 25 9,759 0.20 1.09 0.875 Median 3 2 77,200 0 !38.61 0 60 100,098 1 on the events and the issues, respectively. On average, slightly more than three o$cers have exercise prices reset at each event. An average of slightly more than two distinct option issues are reset at each event. The average number of options repriced is over 200,000. The average reduction in the exercise price is about 41% with a standard deviation of almost 20%. The reductions range from 7% to 89%. No "rms in the "nal sample increase their exercise prices, although one "rm in the original sample increases its exercise price.6 The average change in the option maturity is just nine months, although there are actually only 12 issues by "ve "rms in which the maturities are extended; their average change in maturity is 76.5 months. As an initial estimate of the upper bound on the value of management's gain, we multiply the change in the exercise price by the number of options involved 6 This company raised its exercise price by only a small amount, from $5.20 to $6.00 on one issue and $5.50 to $6.00 on another. 136 D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 and obtain an average of $331,308 with a median of $100,098.7 The last row is the ratio of the new exercise price to the closing stock price on the repricing day. In most cases the new exercise price is extremely close to the closing stock price on the repricing day as reported by the Center for Research in Security Prices (CRSP), with 80% being within one-eighth on either side of the closing stock price. These sample statistics are very similar to the statistics from the full sample of 40 "rms and 74 events. The Wall Street Journal (November 3, 1998, p. B20) reports that high-tech "rms are frequent repricers. Therefore, we might expect our sample to be dominated by technology-related "rms. An examination of the SIC two-digit industry code reveals, however, that no single industry dominates, and "rms with nontechnical products or services make up almost one-third of the sample. An examination of the reasons given for the repricings reveals some interesting justi"cations. Only about one-third of the "rms even mention the repricing, though they provide the necessary tabular information. Evidently an explicit discussion about repricing is a rare event. As noted earlier, only 44% of the "rms in the Gilson and Vetsuypens (1993) study of companies in bankruptcy or reorganization give a reason for the repricing. The primary reason, cited in some form by 11 of 12 "rms, is that the existing exercise prices do not provide a su$cient incentive. Four "rms note that market and/or industry factors, which are outside of management's control, had driven the stock price down. Three "rms simply state that the company's recent performance had &adversely a!ected the stock'. One "rm justi"es its actions by noting that a competitor had repriced, and another indicates that repricing would have only a small cost to the "rm. The aforementioned Gilson-Vetsuypens study notes that 25 of their 77 "rms that had "led for bankruptcy or reorganization repriced. Based on their evidence, one might conclude that repricing is often associated with bankruptcy. Our sample, however, is quite di!erent. As we show in the next table, the average ratio of the book value of debt to total assets in the year of the event is only 25%, and almost one-"fth of the sample has 0}5% debt. The book value of leverage for these "rms does not change materially in the "ve years preceding the 7 It is easy to show that this estimate is an upper bound. From the Black-Scholes model, we have Lc/LX"!e~rTN(d ), and since we are referring to the exercise price change as a reduction, we 2 focus on the absolute value of the change. Prior to expiration, with a positive risk-free rate, and an in"nitesimal reduction in the exercise price, e~rTN(d )(1, so the call price will change by less than 2 the reduction in the exercise price. The actual reduction in the exercise price will be non-in"nitesimal but can be viewed as the sum of an in"nite number of in"nitesimal changes in X. The change in the call price can never catch up with the change in the exercise price unless at least one change in the call price relative to the in"nitesimal change in the exercise price exceeds unity. A typical but erroneous belief in practice is that the exercise price change is the gain in value. Clearly this is but an estimate and potentially a very bad one. We provide a more precise estimate of the value gain later in the paper. D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 137 repricing. As a check on whether our sample "rms are (or eventually go) into bankruptcy or reorganization, we conduct a keyword search on the Dow Jones News Retrieval over the period one year prior to and two years after each event. Even though there are several stories about poor performance and one story about the possibility of a future bankruptcy "ling, there are no stories of an actual bankruptcy "ling. As a further check, we examine the 1998 CRSP "les for up to four years after the last repricing date for each "rm and "nd that 29 out of the 37 "rms are still trading, seven were involved in a merger, and one was involved in an exchange o!er. We can therefore conclude that bankruptcy or reorganization is not a concurrent or imminent threat for the vast majority of our "rms and that all of our "rms survive at least 4 years after the last repricing. 3.2. Preliminary observations from accounting data We "rst examine the performance of these "rms in terms of standard accounting ratios. We collect the accounting data from COMPUSTAT for the event year and the 1}5 years prior to the event. Table 2 contains annual averages and medians (in parentheses) of various accounting measures of performance, leverage, and risk. The average measures of performance indicate a general decline in pro"tability over the "ve-year period prior to the event. Note that the average return on equity is negative for each of the six years, culminating in a value of !170% in the event year. Only the pre-tax return on assets shows some improvement over the "ve-year period, but it is only marginally positive in the event year. The median values, however, show much more stability, which suggests that only a small number of our sample "rms have deteriorating performance prior to the event and that the majority of the "rms do not experience problems until the event year. As we mention in the previous section, the debt to total assets ratio is relatively stable over the six-year period. The accounting descriptive statistics thus validate the pro"le of our sample as "rms with poor performance concentrated in the year of repricing but not in immediate danger of bankruptcy. In the next section we examine how the stocks of these "rms perform prior to and after the repricing. 4. Performance of the stock around the repricing event 4.1. Announcement ewects In recent years, repricing events are reported occasionally in The Wall Street Journal, but such stories primarily re#ect an increased journalistic interest in the practice and not an o$cial release of information. Indeed, we have reason to believe that most "rms would rather not make such an announcement. Repricing is, however, a partially observable event. Sometime after repricing, the "rm 138 D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 Table 2 Selected measures of performance, leverage, and risk from annual accounting data of the 53 "rm-events in the sample of "rms who reset the exercise prices of (&reprice') their executive stock options. Year zero is the "scal year in which the "rm reprices. Years !1 to !5 are the preceding "ve years. The number in each cell is the average with the median in parentheses. Data are obtained from the COMPUSTAT "les. Year !5 Year !4 Year !3 Gross pro"t margin (%) 34.71 (34.06) 20.58 (34.47) 32.95 (33.91) Operating margin before depreciation (%) 10.96 !15.51 (11.94) (12.11) Operating margin after depreciation (%) Pre-tax pro"t margin (%) Net pro"t margin (%) Return on sales (%) Year !2 Year !1 6.23 (34.96) Year 0 5.50 (34.06) 7.81 (31.62) 1.07 (11.28) !18.62 !17.88 (12.44) (10.25) !15.80 (8.00) 3.77 !23.63 (7.51) (7.59) !5.48 (6.69) !25.73 !24.17 (8.52) (6.20) !22.19 (3.43) 0.96 !28.90 (5.18) (5.23) !7.50 (3.02) !26.49 !43.08 (6.51) (5.26) !24.02 (0.05) !1.97 !31.27 (3.08) (3.81) !9.64 (1.98) !29.00 !45.54 (4.16) (3.23) !25.42 (0.03) 5.54 !24.99 (7.96) (9.18) !4.50 (5.62) !24.27 !42.52 (8.57) (7.19) !21.44 (1.62) 3.82 (9.88) Pre-tax return on assets (%) !30.44 (10.67) 4.42 (8.78) 5.51 (10.01) Return on equity (%) !18.46 !11.13 !113.34 (9.11) (10.81) (6.19) !33.54 (7.80) 5.20 (9.44) 0.82 (3.69) !4.65 !170.84 (9.97) (0.27) Book to market (%) 53.70 (42.62) 63.03 (48.22) 62.50 (50.06) 56.16 (44.02) 56.49 (45.31) 74.67 (61.49) Debt to total assets (%) 31.84 (29.81) 26.64 (25.11) 26.63 (18.30) 23.92 (18.16) 24.16 (19.70) 25.79 (25.18) "les a proxy with the SEC that contains the aforementioned repricing table. Typically this information is simply included with the "rm's next annual proxy. This raises the question of whether the market reacts in any way to the "ling and, therefore, to the public announcement of a repricing. We are able to identify the proxy "ling dates for 36 of the 53 "rm-events and conduct a standard event-study analysis, using both value- and equal-weighted CRSP NYSEAMEX-NASDAQ indexes.8 The results are shown in Table 3. For a window of 8 We are unable to obtain data for every event, because repricings prior to 1992 were not required to be reported in proxies. D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 139 Table 3 Percentage abnormal returns and corresponding t-statistics for a subsample of 36 repricing events for which SEC proxy "ling dates are available. Day 0 is the "ling date. The market model parameters are estimated over a 250-day period from day !500 to day !251. The CRSP NYSE/AMEX/ NASDAQ index is used for the market factor. Relative day !5 !4 !3 !2 !1 0 1 2 3 4 5 Value-weighted index Equal-weighted index Percentage abnormal return t-statistic Percentage abnormal return t-statistic 0.45 0.21 !0.18 0.61 0.98 !0.88 !0.33 !0.75 !0.06 0.97 !0.11 0.62 0.29 !0.24 0.83 1.35 !1.21 !0.45 !1.03 !0.08 1.33 !0.15 0.44 0.18 !0.12 0.69 1.07 !0.79 !0.33 !0.72 !0.02 1.03 !0.00 0.59 0.25 !0.16 0.94 1.45 !1.07 !0.45 !0.97 !0.03 1.39 !0.01 $5 days around the proxy "ling date, we "nd no signi"cant reaction. This "nding is not surprising. Proxy "lings, while technically public information, are not monitored carefully by the majority of investors, and the shareholders might not receive the proxy until a signi"cant time period has elapsed. Alternatively, the market does not perceive these events as providing signi"cant new information about the future performance of the "rm. The results are not consistent with a signi"cant wealth transfer from shareholders to management. 4.2. Market performance prior to and after the repricing Now we wish to determine how the repricing "rms perform prior to and after the actual repricing. Standard event-study methods are used to remove the e!ect of the market so that the remaining variation is attributed to "rm-speci"c performance. Again, we use both value- and equal-weighted CRSP NYSEAMEX-NASDAQ indexes as the market benchmark. The results are a!ected only slightly by the index portfolio-weighting scheme. Since some "rms mention the possibility of poor industry performance as the source of poor company performance, we also incorporate an industry factor into the event-study procedure. Beginning with the original work of King (1966), industry e!ects have been known to be a source of variation in stock returns. For a given "rm-event, we construct an index of all "rms with data available in the same two-digit SIC code as the sample "rm. These industry indexes are constructed as both value-weighted and equal-weighted combinations 140 D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 Table 4 Cumulative average residuals (CARs) starting on day !250 of stocks that reset the exercise prices of (&reprice') their executive stock options on day zero and average compound returns for selected days for the stocks as well as market and industry indexes. The sample consists of 37 "rms and 53 "rm-events. Regressions are estimated over days !500 to !251. The CRSP index is used for the market factor. The industry factor is an index constructed using all "rms having the same two-digit SIC code and available return data. When the CRSP value-weighted (equal-weighted) index is used, the industry index is also value-weighted (equal-weighted). The variables R4 , R. , and ~250,0 ~250,0 Ri are the average compound returns starting on day !250 on the stock, the market index, ~250,0 and the industry index, respectively, ending on day zero, and R4 , R. , and ~250,250 ~250,250 R* are the average compound returns starting on day !250 on the stock, the market index, ~250,250 and the industry index, respectively, ending on day 250. Student's t-statistic is provided in parentheses for the cumulative average residuals. R4 ~250,0 R4 ~250,250 CAR ~250,0 CAR ~250,250 R. ~250,0 R. ~250,250 R* ~250,0 R* ~250,250 Value-weighted indexes Equal-weighted indexes Market factor only Market and industry factors Market factor only Market and industry factors !24.49 !24.49 !24.49 !24.49 !10.02 !10.02 !10.02 !10.02 !48.49 (!4.47) !48.39 (!4.47) !43.61 (!4.00) !44.98 (!4.17) !46.81 (!3.06) !48.42 (!3.17) !40.80 (!2.65) !51.10 (!3.35) 8.49 8.49 10.72 10.72 23.43 23.43 26.00 26.00 9.06 8.51 22.93 21.71 corresponding to whether we use the value-weighted or equal-weighted market index. We conduct separate tests using the market factor only and then both the market and industry factors. The regression coe$cients are estimated over day !500 to day !251 relative to the event. The resulting estimates are used to remove the market and industry factors for days !250 to #250 relative to the event. Cumulative average residuals and average compounded returns are examined. Table 4 presents summary results for tests using both the value-weighted and equalweighted indexes. Fig. 1 illustrates the results for the value-weighted index when both the market and industry factors are removed. Similar "gures are obtained when removing only the market factor and when using equal-weighted indexes. D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 141 Fig. 1. Cumulative average residuals and average compounded returns on stocks on which the "rms reprice their executive stock options on day zero, average compounded returns on the CRSP value-weighted stock index, and average compounded returns on a value-weighted industry index consisting of "rms with the same two-digit SIC industry code. The risk adjustment is made by estimating a market model regression over days !500 to !251. Coe$cients from that regression are then applied to the returns over days !250 to #250. The sample is selected from among a larger sample of "rms identi"ed through a key work search on the NAARS (National Automated Accounting Research System) on Lexis Nexis. The sample consists of 37 "rms and 53 events. The results show that "rm-speci"c performance is signi"cantly negative during the period preceding the repricing. The cumulative average residual using the value-weighted market and industry indexes for the "rms is !48.39% on the repricing date, which is highly signi"cant with a t-statistic of !4.47. As Fig. 1 illustrates, "rm-speci"c performance falls steadily prior to the repricing. After the repricing, the CARs stabilize and 250 trading days later have a value of !48.42%. The average compounded stock return, starting from day !250, is !24.49% by the repricing date. The corresponding return is 8.49% for the market and 9.06% for the industry. As Table 4 shows, these results are robust to whether the value-weighted or equal-weighted indexes are used. Comparison of the compounded returns for our sample of events with those of the market reveals that our sample "rms on average start to underperform about 175 trading days before the repricing. An examination of the mean compounded returns four years prior to the repricing "nds no evidence that on average this underperformance starts any earlier. To verify the robustness of the results, we conduct some additional tests with slightly modi"ed samples. Elimination of multiple repricings of a given "rm or removal of "rms whose stock price is less than $5 on the repricing day results in essentially the same "ndings. We also examine the sensitivity of our results to the construction of the industry indexes by using a three-digit SIC code match instead of a two-digit match. Again, the results are essentially the same. 142 D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 4.3. Nonparametric analysis Given the possibility that these results could be biased by implicit assumptions about the distributions of the test statistics, we conduct the nonparametric sign-rank test. This procedure requires an explanation, however, because the process of accumulating residuals for the parametric t-tests is not appropriate for the nonparametric tests. Recall that the average residual for day t is created by "rst cross-sectionally averaging the residuals for each of the 53 events. Starting at day !250 the average residual is then accumulated by adding each day's average residual to the sum of the average residuals for the previous days, thus averaging crosssectionally "rst and then accumulating over time. For the nonparametric tests we obtain the time series from day !250 to #250 of cumulative residuals for each of the 53 "rms. Since none of our previous tests appear to be sensitive to whether the value-weighted or equal-weighted indexes are used or whether the market factor only or both the market and industry factors are included, we use only the value-weighted version of both indexes. The median cumulative residual on day zero is !41.69% and is signi"cant using the sign-rank test at better than the 1% level. 4.4. Cross-sectional analysis We also undertake a test of the relation between the magnitude of the reduction in the exercise price and the performance of the "rm, market, and industry. Speci"cally, we estimate the following regression: #e , (1) #b R4 #b R* %*X "b #b R. j 3 ~250,0,j 2 ~250,0,j j 0 1 ~250,0,j where %*X is the weighted-average percentage change in the exercise price of j the options repriced at event j, R. is the compounded return on the ~250,0,j is the compounded return market index over day !250 to day zero, R* ~250,0,j is the compounded on the industry index over the same period, and R4 ~250,0,j return on the stock for event j over the same period. Because for some events there are multiple options with di!erent exercise prices that are reset, the weighted percentage change in the exercise price is used. The weighted percentage change in the exercise price is calculated as a weighted average of the percentage change in the exercise price of each distinct repriced option. The weights are based on the number of options repriced. The results support our hypotheses. The percentage change in the exercise price is positively related to the compound return on the stock, a result consistent with the fact that the greater is the decline in stock price, the greater is the reduction in exercise price. The percentage change in the exercise price is not related to the compound industry return. The weighted percentage change in the exercise price is also not related to the equal-weighted market return but is D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 143 negatively related to the value-weighted market return.9 Note that if repricing were being driven by market or industry factors, the weighted percentage change in the exercise price would have been positively related to the market or industry factors. 4.5. Recidivism We examine the incidence of repeat repricings by using our full data set of 40 "rms and 74 events. We "nd that 22 "rms reprice once, ten "rms reprice twice, four "rms reprice three times, one "rm reprices four times, two "rms reprice "ve times, and one reprices six times. Thus, 18 "rms or 45% of our sample reprice more than once. Of the "rms that reprice more than once, the average time between repricings is 705 calendar days, or just barely under two years. The maximum is 2382 days, and the minimum is only 44 days. Three "rms reprice within 90 days, three reprice from 91}180 days later, six reprice from 181}270 days later, three reprice from 271}360 days later, and 19 reprice more than 360 days later.10 This evidence appears to be more consistent with entrenchment than with restoration of managerial incentives. 4.6. Valuation ewects on the repricing day In this section we provide option pricing model estimates of the valuation impact on the repricing day of the options in our sample. We use the BlackScholes model to compute the value of the options before and after the change in exercise price and calculate the di!erence. We are aware of the limitations and biases inherent in the Black-Scholes model, especially given the illiquidity of executive stock options. We believe that these biases, however, would likely cancel in computing the change in the value of the options. 9 Diagnostic tests reveal that the variance in#ation factors are small (maximum of 2.63) relative to the customary critical level (10), which suggests that the e!ect of multicollinearity is trivial. We also estimate the regression using an orthogonalizing procedure that removes the market factor from the industry and stock return. We "nd similar results, though the industry factor is positive and signi"cant. We "nd, however, that three outliers greatly a!ect the results, and when those are removed, the results are consistent with the original regressions. (Full details of the test results are available from the authors.) 10 The propensity of repeat repricings in our sample could be biased upwards because of the SEC requirement imposed in 1993 of providing a 10-year repricing history. As a result, all "rms that reprice before 1992 and do not repeat in the post-1992 period are not in our sample. In an attempt to adjust for this bias, we eliminate all repricings prior to 1992. For any "rm that reprices in 1992 or later, we require two repricings before counting it as a repeat repricing, even though the "rm could have repriced prior to 1992. This restriction leaves a sample of 40 "rms, of which ten reprice more than once. Though this is considerably less than in the full sample, it is still one-fourth of our sample and many of those "rms could yet reprice again. 144 D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 The exercise price and time to expiration are obtained from the proxy statement. An estimate of the risk-free rate is obtained by determining the constant-maturity Treasury rate for bonds with a maturity as close as possible to the option expiration. This yield is then converted to a continuously compounded rate. We use the average of the annualized standard deviation for each of the two years prior to the event as the volatility estimate. Only nine of the 37 "rms pay dividends, with many of those being extremely small and a!ecting the valuations of only 13 of the 53 events. Rather than make potentially erroneous assumptions about future dividends, we elect to drop those events from this exercise. We determine the dollar gain for each of the events by summing the changes in the values of each option issue that a "rm reprices. We also determine the overall percentage gain for all of the options of a "rm-event. The mean dollar value increase per "rm-event is about $141,000, with a median of about $65,000. The mean percentage gain is extremely skewed so the median of 16.5% is more useful than the mean. The largest dollar gain is slightly more than $800,000. These results seem to suggest that the direct economic impact to the shareholders is extremely small. This result is consistent with the daily event study results reported in Section 4.1 in which we "nd no signi"cant e!ect on the proxy "ling date. To determine whether these amounts are signi"cant for the executives, it is necessary to know how many executives share a gain and how large the gain is relative to the executive's wealth. Dividing the total gain per "rm-event by the number of executives sharing the gain gives an average gain per executive of about $46,000 and a median of about $25,000. In six of the "rm-events the average gain exceeds $100,000 with two exceeding $200,000; however, 11 of the "rm-events are a gain of less than $10,000 per executive. We obtain cash compensation information for a subsample of approximately half of the "rms. The average total cash compensation for the highest-paid executive is about $480,000 with a median of a little over $300,000. The estimated gain in value due to the repricing is at most around 10% of the annual cash compensation. 4.7. How high is the hurdle without repricing? Some "rms argue that if the options are so deep out-of-the-money, they e!ectively provide no incentive for managers. Therefore, an interesting question is what kind of performance is necessary for the options to become at-the-money under their original exercise prices. We examine the 130 unique option issues in the sample to determine the compound annual rate of return necessary for those options to become at-the-money. The median rate is only 11.3%. Fifty-four out of 130 require a return of less than 10%. Seventy-seven out of 130 require a return of less than 15%. These "gures suggest that the options are not so deep out-of-the-money as to make the hurdle insurmountable, at least for a large percentage of the "rms. It is particularly interesting to note that the required D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 145 annual rate of return corresponds roughly to the compound annual return on the industry index in the year following the repricing. The impression one receives from reading the popular press and hearing the comments of board members and management is that before repricing these options are practically worthless, but this is far from the truth. Eighty-three percent of the options are worth more than $1 and almost 40% are worth more than $4. The average value is $3.98. By comparison, in 1996 the average value of an equity option traded on the CBOE, though of much shorter maturity, was $3.76. To determine how well the options would have done had they not been repriced, we track the "rm's return performance following the repricing date through the end of 1996. This analysis assumes that the "rm's performance is not a!ected by the repricing. For 82 issues out of 130 options that are at-the-money by the end of 1996, the average number of trading days required to reach the exercise price is 232, but the median is only 138. Going back to the full sample of 130 options, had they not been repriced, one-third of them would have been atthe-money within eight months, and half of them would have been at-the-money within 19 months. Over half of the repriced options would have been at-themoney with the old exercise price in an average time span of two years just by earning the industry return. Note that in Table 1 the average repriced option has "ve and one-half years to go before expiration, suggesting that there would have been plenty of time for the options to become in-the-money without repricing. 4.8. Alternatives to repricing Even though we show that the incentives are still present for most of these options, a "rm concerned about restoring incentives by bringing the options back to at-the-money has other alternatives available to it. It can reprice and shorten the maturity, but we "nd no "rms that do this. It can also cancel old options and issue new at-the-money options such that the value of the old options equals or exceeds the value of the new options, leaving the repricing value-neutral.11 Our sample includes six events in which a "rm cancels old options. In one of those, it extends the maturity of the remaining old options. We estimate the Black-Scholes value under the old exercise price of the options canceled and those not canceled, de"ning this total to be the value of the options before the repricing. We then estimate the Black-Scholes value after the repricing, properly accounting for any extension of the maturity. The di!erence is the net gain or loss in option value from repricing. We are unable to do this calculation for one of the six "rms due to dividend payments that make estimation of long-term 11 The Wall Street Journal (April 8, 1999, p. R5) refers to this process as value-for-value repricing and suggests that more "rms are looking to do so, but its "ndings are essentially based on observations on only two "rms, Sunbeam and Cendant. 146 D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 option values problematic. For the remaining "ve "rms, there are a few options of one "rm in which an individual option issue ends up with less value after the repricing than before. After adding up the values for all options involved for a "rm, we "nd only one "rm that has an overall net reduction in option value due to the repricing. Thus, it is apparent that in the overwhelming majority of the "rm-events, the option holders receive a net gain from the repricing.12 It is possible that when a "rm reprices, thereby granting management something of value, it simultaneously takes something away in the form of a reduction in other compensation. It is impossible to determine what management's compensation would have been in the absence of the repricing. To address the question of whether a board makes o!setting reductions in other compensation, we examine the proxies of approximately half of the "rm-events in our sample, collecting information on salaries and other compensation for the event year and years surrounding it as well as scanning the text for obvious references to reductions in salary or other compensation. We "nd no references to a reduction in compensation to o!set the repricing.13 In a few cases, compensation in the event year is smaller than in the previous year, but there are about as many cases in which event-year compensation is larger than compensation in the previous year. Overall, we "nd no evidence in half of the sample that there are any obvious attempts to o!set management's gain with any reduction in value from some other source. Admittedly we cannot determine for certain that in exchange for repricing the options there are no penalties exacted by the board, such as granting a smaller salary increase or the awarding of fewer shares of stock or a reduction in future option grants. We "nd no evidence of it, however, and are doubtful that such penalties are imposed. 5. Why 5rms reprice An important question that remains to be answered is why some "rms choose to reprice while others facing a similar price decline do not. One way to address this question would be to collect a sample of all "rms that do not reprice, irrespective of their stock price performance, and compare them with "rms that do reprice, using various quantitative measures of characteristics and performance. This is the approach taken in Brenner et al. (2000), who use the "rm's stock performance over the last three years as an independent variable in their 12 Some "rms are even quite clever in creating value for management. One "rm reduced the exercise price on various options from an average of around 11 to 8 1/4 even though the stock was about 2 3/4, apparently leaving the options still signi"cantly out-of-the-money. It simultaneously announced a 3-for-1 reverse split, moving the options back to precisely at-the-money. 13 Interestingly, in one case we "nd reference to the repricing as compensation for a reduction in salary. D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 147 logit regression but do not use past performance as a matching criterion for selecting the sample of non-repricing "rms. Such an approach will include a large number of "rms that have improvements in shareholder wealth and would not even consider repricing. A logit analysis that includes repricing "rms, which we know have poor performance, with all non-repricing "rms irrespective of their performance will tend to "nd those factors signi"cant that a!ect the likelihood of experiencing poor performance, such as the volatility of stock returns. To determine which factors a!ect the likelihood of repricing, we should include only those non-repricing "rms that experience poor performance and, therefore, are likely to have options that are out-of-the-money. Such "rms could lower the exercise prices of their executive stock options but choose not to. It is the di!erence between only such non-repricing "rms and the repricing "rms that will tell us why some "rms reprice and others, facing a similar repricing decision, choose not to. In other words, for each sample repricing "rm, we need a matched "rm that experiences a similar decline in its stock price. Another important question is whether such an analysis should be done at a "rm level or at an executive-year level as in Brenner, Sundaram, and Yermack. Using the executive-year approach, a "rm that reprices its options for three of its top "ve executives in one year over the "ve-year period will be included three times as a repricing observation and for the remaining 22 times for the "ve executives over the "ve-year period as a non-repricing observation. The repricing decision is likely to depend on "rm characteristics such as size and agency problems, so including the same "rm as both repricing and non-repricing observations (in some cases for the same year) will bias the results away from "nding those characteristics signi"cant. Moreover, given that these characteristics are not likely to change signi"cantly from year to year, we believe that any "rm that reprices only once over a period of several years should not be included as a non-repricing observation for the surrounding years. We construct our matched sample following a carefully designed procedure. First, we match each sample repricing "rm-event with a set of candidate "rms selected from the same four-digit SIC industry code. To meet the requirement that the matched "rm has a similar incentive to reprice, it must experience a stock price decline similar to the percentage reduction in exercise price for the sample "rm.14 For each candidate "rm, we calculate each possible 250-day return over the 1992}1997 period.15 For each sample "rm and event, we then select a candidate 14 We also explore the possibility of matching using the repricing "rm's percentage price decline over the one-year period prior to repricing. This creates a problem, however, in that the declines for di!erent "rms occur over periods of di!erent length. Given that our objective is to "nd matched "rms that have a similar incentive to reprice, we believe that a more logical and simpler way is to match using the percentage decline in exercise price. 15 Since "rms have been required to report repricing only since 1992, we cannot choose matched "rms from the period prior to 1992, because we cannot determine if they repriced. 148 D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 "rm and its associated 250-day period in which the candidate "rm return is within $500 basis points of the sample "rm. In other words, if the sample "rm has a reduction in exercise price of 40%, then we search for matching candidate "rms with losses between 35% and 45% over a 250-day period. The "rm with the closest match on its return is selected and then examined for two pieces of information: use of executive stock options and no repricing by the time of the release of the next proxy. In addition, we require several other pieces of information, including proxies, certain accounting items, and market price data. If the information is available, the candidate "rm uses executive stock options and does not reprice, then that "rm is chosen as a match. If not, we move to the next candidate "rm, whose return is next closest. If no candidate "rms can be identi"ed as a match, we expand the de"nition of a potential match to a threedigit SIC code. If we are unable to match using a three-digit industry code, we go to a two-digit industry code. We obtained satisfactory matches for all but one repricing "rm-event, which we then discard from this stage of the analysis, leaving us with 52 "rm-events and the corresponding matched "rms. Of the 52 control "rms, 21 are matched on the four-digit SIC code, 17 on three digits, and the remaining 14 are matched on two digits. To identify the characteristics of "rms that reprice and to distinguish them from similarly performing "rms that do not reprice, we use variables that proxy for size, growth opportunities, dividend policy, and volatility. We hypothesize that "rms that reprice have a greater degree of agency problems than "rms that do not reprice. Thus, our logit analysis includes several measures of agency problems, such as inside ownership, insider domination of the board, and free cash #ow. The variables are shown and described below. Day 0 is the repricing date for sample "rms and the end of the 250-day period of similar performance for the matched "rms. SIZE "market value of equity, measured as of day !250. PAYOUT"dividend payout ratio, measured as the annual dividend divided by the price at the end of the last "scal year before day 0. FCF/TA "free cash #ow divided by total assets, measured as operating income before depreciation net of total income taxes, gross interest expense, preferred dividends, and common dividends, divided by the book value of total assets; all items are for the last "scal year before day 0. MKT/BK "the sum of the market value of equity and the book value of debt divided by the book value of total assets; all items are for the last "scal year before day 0. VOL "volatility of the stock, measured as the standard deviation of the daily return from day !500 to day !251. INSOWN"percentage inside ownership, measured as the number of shares held by insiders as of the last January 1 or July 1 before day 0 divided by the number of shares outstanding on the same D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 149 day; insider shares are obtained from CDA/Investnet Insider Holdings (formerly known as Spectrum 6) and shares outstanding are obtained from CRSP. INSDIR "percentage inside directors, measured as the number of directors who are o$cers, close family members, or retired o$cers divided by the total number of directors.16 5.1. Univariate tests We "rst conduct a univariate comparison of our sample and the matched "rms. The results are reported in Table 5. The most obvious "nding is that the repricing "rms are signi"cantly smaller than the matched "rms. Size could be a proxy for the extent to which information about a "rm is known, suggesting that "rms that are less known "nd it easier to reprice and have fewer negative repercussions. Size, however, could well re#ect data availability, making it more likely that large "rms would be selected for the matched sample. Thus, on a univariate basis size may not be very informative. In a multivariate test, however, we must include size, since it has the potential to be a relevant covariant. The percentage of inside directors is signi"cantly higher for the repricing "rms. This result suggests that the boards of repricing "rms are more insider dominated. Note, however, that the percentage of inside ownership is not signi"cantly di!erent. Repricing "rms have signi"cantly higher volatility. Using the nonparametric test, MKT/BK is signi"cantly lower for repricing "rms. These measures, however, could also re#ect size; thus, we must be careful in our interpretation of these univariate results. 5.2. Logit analysis We next conduct a multivariate test, a logit regression in which the dependent variable is one if the "rm reprices and zero if not. Because size, volatility, and payout rate tend to be highly skewed, we use a log transformation on these variables. The results for various combinations of the input variables are presented in Table 6. For all combinations, size is highly signi"cant, with smaller "rms more likely to reprice. As in the univariate tests, the percentage of inside ownership 16 We categorize directors as insiders, outsiders, or grey, using the criteria of Brickley et al. (1994). Grey directors include bankers, lawyers, investment bankers, and consultants, all of whom are considered to have potential business ties to the "rm. Insiders are o$cers, founders, and close family members of o$cers and founders. We run separate tests with grey directors included as insiders and with grey directors included as outsiders. The results of those tests lead to the same conclusions, so we report only the tests with grey directors included as outsiders. 150 D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 Table 5 Comparison of descriptive statistics for the sample of 52 "rms that reprice their executive stock options and a control sample of 52 "rms that do not reprice and are matched on SIC code and a percentage decline similar to the percentage reduction in the exercise price of the sample "rms' executive stock options. All measures are taken relative to an event day, which is the repricing day for the sample "rms and the end of the matched-return 250-day holding period for the matched "rms. SIZE is the market value of equity as of day !250. PAYOUT is the annual dividend divided by the price at the end of the last "scal year before day 0. FCF/TA is operating income before depreciation net of total income taxes, gross interest expense, preferred dividends, and common dividends divided by the book value of total assets. All items are for the last "scal year before day 0. MKT/BK is estimated as the sum of the market value of equity and book value of debt divided by the book value of total assets for the last "scal year before day 0. VOL is the standard deviation of return based on daily returns over the period !500 to !251. INSOWN is number of shares held by insiders as of the last January 1 or July 1 before day 0 divided by the number of shares outstanding on the same day with insider shares obtained from CDA/Investnet Insider Holdings, formerly known as Spectrum 6, and shares outstanding obtained from CRSP. INSDIR is the number of inside directors divided by total number of directors, where an inside director is an o$cer, a very closely related family member, or a retired o$cer. The di!erence between means is examined using the parametric matched pairs t-test. The di!erence between medians is examined using the nonparametric Wilcoxon sign-rank test (p-values are in parentheses). Variable SIZE Sample "rms Matched "rms Di!erence (sample}matched) Mean Median Mean Median 185.2 106.3 2847.7 1579.6 PAYOUT 8.22% 0.00% 12.63% 0.00% FCF/TA 0.0663 0.0933 0.0833 0.1089 MKT/BK 2.55 1.57 2.94 2.28 VOL 3.93% 3.45% 2.74% 2.63% INSOWN 22.9% 19.7% 20.7% 8.0% INSDIR 52.4% 54.5% 33.2% 28.6% Mean !2662.5 (0.001) !4.41% (0.254) !0.0170 (0.546) !0.39 (0.424) 1.19% (0.001) 2.2% (0.614) 19.2% (0.001) Median !1487.0 (0.001) 0.00% (0.077) !0.0139 (0.229) !0.55 (0.033) 1.05% (0.001) 7.1% (0.541) 17.5% (0.001) has no signi"cant e!ect. A possible reason that inside ownership has no signi"cant e!ect is that there are two opposing e!ects. As insider ownership increases, the greater is the cost of repricing borne by the insiders, making repricing less desirable. As inside ownership increases, however, management becomes more entrenched and a repricing proposal is more likely to win approval. The percentage of inside directors has a signi"cant e!ect on repricing in the "rst three models, suggesting that the greater the extent of insider control, D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 151 Table 6 Parameter estimates and p-values in parentheses of logit regressions of the occurrence of repricing for the sample of 52 "rms that reprice their executive stock options and a control sample of 52 "rms that do not reprice and are matched on SIC code and a percentage decline similar to the percentage reduction in the exercise price of the sample "rms' executive stock options. All measures are taken relative to an event day, which is the repricing day for the sample "rms and the end of the matched-return 250-day holding period for the matched "rms. SIZE is the market value of equity as of day !250. PAYOUT is the annual dividend divided by the price at the end of the last "scal year before day 0. FCF/TA is operating income before depreciation net of total income taxes, gross interest expense, preferred dividends, and common dividends divided by the book value of total assets. All items are for the last "scal year before day 0. MKT/BK is estimated as the sum of the market value of equity and book value of debt divided by the book value of total assets for the last "scal year before day 0. VOL is the standard deviation of return based on daily returns over the period !500 to !251. INSOWN is number of shares held by insiders as of the last January 1 or July 1 before day 0 divided by the number of shares outstanding on the same day with insider shares obtained from CDA/Investnet Insider Holdings, formerly known as Spectrum 6, and shares outstanding obtained from CRSP. INSDIR is the number of inside directors divided by total number of directors, where an inside director is an o$cer, a very closely related family member, or a retired o$cer. Intercept LN(SIZE) Model 1 Model 2 Model 3 Model 4 27.0753 (0.001) 25.0923 (0.001) 26.2231 (0.001) 35.7618 (0.001) !2.1694 (0.001) !2.3645 (0.001) !2.2558 (0.001) !2.9648 (0.001) 1.9141 (0.529) 0.9654 (0.812) LN(1#PAYOUT) FCF/TA 7.9231 (0.016) MKT/BK !0.2051 (0.186) !0.0255 (0.883) !1.2143 (0.387) !0.4658 (0.763) !0.0732 (0.969) !0.0324 (0.119) !0.0359 (0.102) !0.0385 (0.127) !0.0066 (0.792) 0.0397 (0.068) 0.0472 (0.049) 0.0625 (0.035) 0.0447 (0.137) LN(VOL) INSOWN INSDIR the more likely is the "rm to reprice. This variable becomes insigni"cant in the fourth model, which includes another signi"cant agency variable, free cash #ow divided by total assets. Given the relative stability of the coe$cient on percentage of inside directors, we suspect that the reduction in signi"cance of the 152 D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 percentage of inside directors is due to the higher standard errors resulting from multicollinearity between these two variables. Though not shown in the table, when grey directors are included as insiders for both sample and matched "rms, the percentage of inside directors is signi"cant in all four models, and the free cash #ow variable remains signi"cant. Interestingly, volatility is not signi"cant, which is consistent with our earlier observation that the higher volatility of the sample "rms may simply be a result of their smaller size. The payout ratio and MKT/BK are not signi"cant, again con"rming that the signi"cance of MKT/BK in the univariate test may be just picking up the size e!ect. Our results are di!erent from those of Brenner et al. (2000) in two important ways. First, we "nd stronger evidence of agency problems for "rms that reprice executive stock options, based on the signi"cance of free cash #ow and insider domination of the board. Brenner et al. (2000) do not use these variables, possibly because their analysis is at the executive-year level and not at the "rm-event level. Second, with size and past performance as independent variables in their logit regression, they "nd that "rms in more volatile industries are more likely to reprice, while we "nd that volatility has no e!ect on the incidence of repricing. A possible explanation is that we match on industry and past performance in selecting our control "rms while they do not, and that for "rms within the same industry with similarly bad performance, volatility has no e!ect on the probability of repricing. Their "nding could imply that "rms that are more volatile have a greater probability of experiencing poor performance and consequently a greater probability of repricing, while our "nding implies that conditional on industry and poor performance, volatility has no e!ect on the probability of repricing. Our evidence on the free cash #ow variable and the percentage of inside directors is consistent with the notion that agency problems arising out of the shareholder}manager con#ict are a prime factor in motivating a decision to reprice. 6. Conclusions In this paper we observe that repricing occurs after a period of overwhelmingly poor "rm-speci"c performance. Over the one-year period before repricing, the average "rm loses 25% of its value. The average reduction in the exercise price is about 40%. Investors do not react to the repricing, at least around the proxy "ling date. Firm performance following the repricing is normal. We have no direct evidence, however, that the poor "rm-speci"c performance prior to repricing is attributable to management or to factors outside of managerial control. We estimate that the average gain to the executives and loss to the shareholders is less than $150,000 but the gain is about 10% of total compensation to D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 153 the average executive. At least 25% of the repricings are repeat events. If "rm performance is not a!ected by the repricing, the typical "rm would need to earn a compound return of only 11% over the remaining life of the option to become at-the-money without repricing or extending the maturity. Using actual postrepricing performance, over half of the options would have been at-the-money without repricing within 19 months. There is no evidence that "rms retire options in such a manner as to leave the repricing value-neutral. A logit analysis of repricing "rms matched with "rms that undergo similar market performance but choose not to reprice reveals that repricing "rms are signi"cantly smaller, have higher free cash #ow, and have a greater percentage of directors that are insiders. Firm volatility, dividend yield, growth opportunities, and insider ownership have no e!ect on the incidence of repricing. Repricing has created numerous controversies in the "nancial press in recent years. No fewer than 11 articles on repricing appeared in The Wall Street Journal from 1997 through 1999. Yet repricing is a relatively infrequent event, and the wealth transfer from shareholders to management is very small. This media frenzy over a relatively infrequent event that results in little cost to the shareholders is consistent with Jensen's (1991) &politics of "nance' view, wherein he cites the negative media attention to the LBO/takeover wave of the 1980s. While it is certainly true that the period preceding repricing is almost always characterized by a substantial loss in shareholder wealth, the repricing itself could be nothing more than a symptom of poor performance and agency problems associated with small "rms. Given the preponderance of news stories, however, some "rms apparently are becoming more sensitive to the negative perceptions of repricing.17 References Acharya V.V., John K., Sundaram R.K., 2000. On the optimality of resetting executive stock options, Journal of Financial Economics, forthcoming. Brenner M., Sundaram R.K., Yermack, D., 2000. Altering the terms of executive stock options. Journal of Financial Economics, forthcoming. 17 Interestingly, HealthSouth Corporation, from whose proxy we quote at the beginning, issued the following statement in its proxy about one year later: The 1995 Plan prohibits any reduction of the exercise price of outstanding options granted under the plan except by reason of merger, business combination, recapitalization or similar change in the capitalization of the Company. The 1995 Plan likewise prohibits the cancellation of outstanding options accompanied by the reissuance of substitute options at a lower exercise price. HealthSouth Corporation proxy May 13, 1995 154 D.M. Chance et al. / Journal of Financial Economics 57 (2000) 129}154 Brickley, J.A., Coles, J.L., Terry, R.L., 1994. Outside directors and the adoption of poison pills. Journal of Financial Economics 35, 371}390. Corrado, C.J., Jordan, B.D., Miller Jr., T.W., Stansfeld, J.J., 1998. Repricing and employee stock option valuation. Unpublished working paper. University of Missouri, Columbia. Credit Suisse First Boston, 1998. Employee stock options in the new Americas. Equity Research, November 2, (1998) 21}23. Gilson, S.C., Vetsuypens, M.R., 1993. CEO compensation in "nancially distressed "rms: an empirical analysis. The Journal of Finance 48, 425}458. Jensen, M.C., 1991. Corporate control and the politics of "nance. Journal of Applied Corporate Finance 4, 13}22. King, B.F., 1966. Market and industry factors in stock price behavior. The Journal of Business 34 (Part 1), 139}190. Saly, P.J., 1994. Repricing executive stock options in a down market. Journal of Accounting and Economics 18, 325}356.