Rubin_00_08_paper.pdf

The Effects of Product Liability Litigation on the Value of Firms1
Correspondence:Paul H. RubinDepartment of EconomicsEmory UniversityAtlanta, GA 30322-2240Voice: 404-727-6365Fax: 630-604-9609Email: prubin@Emory.eduhttp://www.Emory.edu/COLLEGE/ECON/Rubi.htmApril 15, 2000 1 We would like to thank Owen Beelders, John Calfee, James Cooper, Chris Curran, Sherry L. Jarrell, and JohnYun, for helpful comments, Barbara Maaskant for granting access to the CRSP data, and Ron Harris for technicalsupport in obtaining the CRSP data. The usual disclaimer applies.
The Effects of Product Liability Litigation on the Value of Firms: Abstract We use event study methodology to examine the effects of product liability litigation on firms inthe automobile and pharmaceutical industries. Others have examined verdicts and found noeffect. We find that the filing of lawsuits leads to significant losses in firm value. These lossesare approximately equal to the upper bound of the direct loss in value of the firms involved; thereappears to be relatively little loss in reputation from product liability events. We find that in theautomobile industry, competitors lose when one firm is sued, but in the pharmaceutical industry,a lawsuit against one firm leads to an increase in value of other firms.
The Effects of Product Liability Litigation on the Value of Firms 1. Introduction
The law and economics literature on product liability is enormous. Research has examined standards for liability and also optimum forms of damage payments. However, there is nowhere in the literature a comprehensive event study of product liability litigation. This paper uses event study methodology to examine the effects of this form of litigation on the value of firms and also of competitors. Studies using stock returns data are superior to tests using accounting data because they reflect the social costs of a lawsuit that are not recognized in accounting data. Explicit measurements of damage payments and legal costs are an inadequate estimate of the costs of product liability litigation because they do not take into account other potentially important costs such as reputation costs and future damage payments. While there are event studies of interfirm litigation (e.g., Hertzel and Smith, 1993), and studies of aspects of the product litigation process (e.g., Viscusi and Hersch, 1990; Gerber and Adams, 1998) there is no comprehensive study of product liability. In this paper, we measure the costs to firms in two industries of product liability litigation. It is particularly useful to perform this measurement since the most comprehensive study in the literature (Gerber and Adams, 1998) looks only at verdicts and finds no significant effects. Our study looks at all litigation related events and we do find significant effects. Since such studies are relevant for policy (as discussed by Gerber and Adams and also by their commentators, Peltzman, 1998, and Rubinfeld, 1988) it is important to A second goal of this research is to examine the impact of product liability lawsuits on the value of competitors. The literature currently fails to consider the potential externalities that lawsuits have on the value of competitors. Thus, an examination of the effects on competitors will yield a more accurate estimate of the true social costs of the lawsuits. A third goal is to determine whether firms suffer reputation costs as a result of lawsuits. A finding of excessive capital market losses (i.e. negative returns greater than expected damage payments) surrounding the initial public announcement of a potential liability problem would indicate that firms suffer reputation costs. It is well known that firms suffer losses in reputation from government sanctions (e.g., Peltzman, 1981; Jarrell and Peltzman, 1985; Rubin et al., 1988) but the effect of private litigation is not fully understood. If firms do not suffer the same reputation losses from private as from government actions, (which is what we find) this might mean that consumers believe that private product liability litigation provides less information about product quality than does government action. This might be interpreted as implying that consumers and investors view such litigation as rent seeking, rather than as quality improving. This would be consistent with the views of many commentators, who view the current product liability system as being inefficient; see, for examples, Calfee and Rubin (1992); Huber (1988); and Priest The main findings are that firms do suffer significant negative returns surrounding the announcement of lawsuit event dates. Capital market losses very nearly approximate a worst case scenario for out of pocket costs in both the automobile and pharmaceutical industries. Thus, the results indicate that investors may assume a worst case outcome upon learning of a potential product problem. The negative returns are approximately equal to the expected losses associated with the problem, but do not seem to include additional reputation losses. An interesting corollary to this last result is that the capital markets seem to respond differently to private market incentives than to regulatory incentives (e.g., recalls). There seems to be a much greater reputation loss from government ordered recalls than from private lawsuits. Finally, competitors also exhibit significant effects surrounding event dates. However, the returns are negative for competitors in the automobile industry and positive for competitors in the pharmaceutical industry. Possible explanations for the difference in the industries are discussed Section two contains a review of the literature of important event studies in law and economics that are relevant to this research. Section three discusses the research agenda in detail. Section four discusses the methods used for performing event studies used to estimate the results. Section five discusses the data. Sections six and seven provide a discussion of the results for the automobile and drug industries respectively. Section eight summarizes the research and discusses its significance, and provides some suggestions for future research.
2. Relevant Literature
Two papers use event study methodology to examine the effects of product liability related events on the value of corporations, but neither paper attempts to systematically estimate the costs of product liability lawsuits. Viscusi and Hersch (1990) recognize that important events in product liability litigation may affect the value of firms, but they do not reach any general conclusions about the costs of product liability lawsuits. Garber and Adams (1998) examine only verdict dates and make no effort to focus in on the date of the initial news release Garber and Adams provide the first comprehensive event study of lawsuit verdicts involving automobiles. They examine verdicts because they believe that verdicts are “prominent” and “contain elements of surprise” (i.e., new information).2 They found 116 verdicts in favor of a domestic automobile manufacturer between January 1985 and July 1996 in the Automotive Litigation Reporter. They extended their search to find plaintiff verdicts. Some plaintiff verdicts were found in the Automotive Litigation Reporter. Others were found by calling attorneys listed in the Automotive Litigation Reporter to ask for information on unpublished verdicts. The authors further expanded their plaintiff verdict data set by searching Jury Verdicts Weekly and newspaper databases. The search yielded 64 plaintiff’s verdicts against domestic automobile manufacturers.
Although the data set is fairly large, they did not follow a systematic approach to finding data points, particularly for the plaintiff’s verdicts. Some of the data points were found by calling plaintiff’s lawyers to discover unreported verdicts. Furthermore, the bulk of the data points come from the Automotive Litigation Reporter, which had a circulation of only 150 in 1994. Garber and Adams admit that the cases published there are an “unsystematic sample of unknown completeness” and that “almost all of the articles are based on unsolicited reports from attorneys who send information to the publisher.”3 The results for both the plaintiff and defendant verdicts are insignificant. In fact, all the results are signed oppositely from the expected sign (positive for defendant verdicts and negative for plaintiff verdicts). Garber and Adams state that the insignificant results indicate that lawsuits have little or no effect on automobile companies as compared to other important news events affecting the companies. In his discussion, Peltzman agrees with this position.4 An alternative interpretation is that investors have already incorporated the expected damage costs into the stock price. If this is so, then trial verdicts do not contain the “element of surprise” claimed by Garber and Adams and needed for an event study to be meaningful.5 Theoretically, the second explanation seems more plausible. The authors focus their research on events that come at the 2 Garber and Adams (1998), p. 2.
3 Garber and Adams (1998), p. 9, footnote 15.
4 “I think that this conclusion would stand up even if court cases were analyzed from their beginning instead of, asGarber and Adams have done, from their conclusion.” Peltzman, 1998, p. 45.
end of the litigation process. According to the efficient markets hypothesis, the market has already incorporated investor expectations concerning the lawsuit into the security price at this time. Thus, verdicts are news only to the extent that they differ from what investors expected.
Consequently, positive abnormal returns for plaintiff’s verdicts may simply indicate that the damage award was not as high as expected, rather than that the lawsuits have no effect on Viscusi and Hersch (1990) examine the stock market reaction to the announcement in The Wall Street Journal of 21 events related to defective products filed between 1970 and 1985.
They also provide a brief examination of two large class action lawsuits, those involving agent orange and DES. They find that the announcement of events related to product liability lawsuits can have a significant impact on the value of firms. However, they do not attempt to make any generalizations about their results, nor do they draw any meaningful conclusions from their findings. Their neglect to provide an in-depth analysis of the results probably arises from the fact that their sample is very small and the events that they examine are very diverse.6 Another group of event studies examines the effect of product recalls on the value of firms. The first and most prominent recall event study is Jarrell and Peltzman (1985) on the effect of product recalls on the value of producers. They examine the stock market effects of product recalls on the value of firms, and on competitors, in the automobile and pharmaceutical industries.
Their sample consists of 26 drug recalls between 1974 and 1982 and 116 “major” automobile 5 In his comments, Rubinfeld (1998) makes this point as well.
6 The types of events that they examine differ considerably (p. 226-227). Examples of the events that they chooseinclude the following: 1) “Nader group and others file Federal suit seeking ban of Oraflex as an imminent hazard;”2) “Ohio Attorney General sues Penn Central for $14.1 million (in) damages and correction of ‘unsafe conditions;’”3) “Rockwell International involved in Pullman suit since it designed and mounted undercarriage of subway cars;”and 4) “Shareholder sues company and executives alleging awareness of DC-10 defects leading to a crash.” Thesample also includes five different events dealing with the Dalkon Shield litigation. Less than half the sampleconsists of events that announce the filing of a claim. None of the events deal with a verdict or settlement.
recalls between 1967 and 1981.7 Jarrell and Peltzman find that the loss in the value of a firm surrounding a product recall is substantially larger than the direct costs of the recall. (Direct costs are primarily the cost of repairing or replacing defective products.) Based on this difference, they conclude that the cost of a recall includes substantial losses of goodwill. They also find that the announcement of a product recall exerts a negative externality on the value of competitors in that competitors experience negative abnormal returns during the interval surrounding the recall.
The Jarrell and Peltzman study spurred two groups of researchers to reexamine their work. Hoffer, Pruitt, and Reilly (1988) revised Jarrell and Peltzman’s automobile data set by eliminating events with overlapping windows and other events that did not meet the criteria of a NHTSA recall as discussed by Jarrell and Peltzman. Under the revised data set, Hoffer et al. find that recall announcements do not significantly affect the value of automobile firms. They also fail to find significant negative returns for the competitors of firms announcing recalls. Barber and Darrough (1996) also revise Jarrell and Peltzman’s data selection method. Barber and Darrough include all recalls in their sample, while Jarrell and Peltzman and Hoffer et al. examine only large recalls. Furthermore, they also examine the effects of recalls on three Japanese manufacturers. Barber and Darrough find that firms do suffer statistically significant losses surrounding recalls. However, they do not find significant negative effects on competitors.
Rubin, Murphy, and Jarrell (1988) examine recalls of a broad array of consumer products. The study covers 48 recalls by the Consumer Product Safety Commission involving 31 firms over the period 1977 to 1981. The average loss in equity value from a recall was 6.9 percent and the abnormal returns were negative in 73 percent of the cases. Although the authors 7 The events for the drug recalls are found in either The Wall Street Journal or the weekly reports of FDA Recallsand Court Action in the Food, Drug, and Cosmetic Reporter, an industry newsletter. Five of the twenty-six recallsinvolve multiple event dates. As a result, the sample consists of thirty-two event dates, nineteen of which werereported in The Wall Street Journal. All 116 events dealing with automobile recalls were reported in The Wall did not examine in detail the direct costs of the recalls, they believed that the effects were much greater than any plausible value for these effects.
3. Examination of Product Liability Using Event Study Methodology
There are several issues regarding product liability that may be examined using event study methodology. First, no one has provided a comprehensive analysis of the effects that the lawsuits have on the value of firms. How do firms react to different types of events? Do firms experience negative returns surrounding the filing of a claim or do the losses occur when firms lose lawsuits? Garber and Adams (1998) fail to adequately answer these questions because they do not examine filing events. This is key to event studies because the market immediately forms expectations about the effect of a verdict at the time of the initial case filing or the initial verdict.
Another question deals with the effects of product liability lawsuits on the value of competing firms. Rubin and Bailey (1994) state that “other manufacturers may be ambivalent about the outcome of a suit.” They note that “while a loss will change the legal rule in an unfavorable direction, it will also harm the reputation of the defendant firm, and therefore may generate business for other firms in the industry.”8 Thus, competing firms may experience a shift in their demand curves to the right because they produce a substitute for a product that is now perceived as more costly or less desirable to consumers. An alternative hypothesis is that competitors suffer negative returns surrounding the date of a lawsuit filed against a competitor.
This would occur if investors view lawsuits against any firm in an industry as likely to increase the likelihood that all firms in the industry will be implicated in future lawsuits, thereby raising costs to all firms in the industry. The lawsuit may also cause the demand curves of competitors Street Journal. A recall is included if it exceeded 50,000 autos for GM, 20,000 for Ford, and 10,000 for Chrysler.
The numbers were chosen to roughly approximate the market shares of the firms.
8 Rubin and Bailey (1994), p. 811.
to shift to the left as consumers change their preferences for a product that is now perceived as dangerous or unsafe regardless of who actually manufactured the defective good.
Neither Viscusi and Hersch (1990) nor Garber and Adams (1998) compare damage payments to the estimated loss in the value of firms implicated in lawsuits. If decreases in firm value surrounding the filing of a lawsuit are substantially greater than expected damage payments, then one may infer that firms involved in product liability lawsuits suffer reputation costs. The primary result of the Klein and Leffler (1981) analysis of corporate incentives to sell quality goods is that the existence of above market prices along with the presence of nosalvageable capital (e.g., expenditures on brand name recognition) are sufficient to provide firms with an incentive to supply quality goods. If a firm fails to honor its commitment to supply quality goods, then consumers will refuse to pay a quality-assuring price premium and the firm will lose quasi-rents from future sales. Klein and Leffler find that the lost quasi-rents are approximately equivalent to the value of the firm’s brand name. Thus, lawsuits that are damaging to a firm’s brand name are likely to be associated with reputation costs due to lower quasi-rents from future sales. Losses in firm value in excess of damage payments may indicate that estimates of the direct cost of product liability litigation do not reflect the true cost. On the other hand, if losses are no larger than expected damage payments, this may indicate that the market does not view the litigation as providing any useful information about firm reputation.
4. Event Study Methodology
Event studies dealing with stock market data all rely on the validity of the efficient markets hypothesis. The hypothesis states that the stock market is informationally efficient. This means that security prices reflect all available information.9 Thus, any new information that affects investors’ valuations of a security is captured immediately in the price of the security. In the rest of this section we discuss the event study methodologies used to generate the results. We begin with a description of a methodology that uses dummy variables to capture abnormal returns.
Next we discuss a methodology known as the market adjusted return. We also describe a non- 4A. The Dummy Variable Method
The use of dummy variables to calculate abnormal returns is common in the law and economics literature. Both Mathios and Plummer (1989) and Hertzel and Smith (1993) use the dummy variable method rather than the market model10 to calculate abnormal returns. Tests using the dummy variable method yield results identical to tests using the traditional market model.11 The advantage of the dummy variable technique is that results are obtained in a single regression rather than the two steps required in the market model. To calculate abnormal returns using the dummy variable approach, the following equation is estimated using ordinary least squares for 150 trading days before the event date through 50 days after the event date:12,13 where Rjt = the continuously compounded daily rate of return for firm j in period t; MKTt = the continuously compounded daily rate of return on the equallyweighted market portfolio; 9 See Fama (1991) for a detailed summary of the efficient markets hypothesis.
10 The traditional market method involves regressing a firm’s return on the market return over a pre-event period toobtain estimates of the regression coefficients. This is accomplished by performing the following regression: Rt =c1 + c2MKTt + et (where Rt is the firm’s return and MKTt is the market return). The regression coefficients are thenused to calculate abnormal returns (ARt = Rt – c1 – c2MKTt, where ARt = the firm’s abnormal return on day t).
11 Karafiath (1988), p. 353.
12 The estimation period is consistent with the literature. Mathios and Plummer (1989) use 200 days before theevent through 50 days after the event. Karpoff and Lott (1993) use 100 pre-event window days and 100 post-eventwindow days. Mitchell and Maloney (1989) estimate their equations over 50, 100, and 150 day intervals and obtainconsistent results for each period. Mitchell (1989) uses 253 days before the event. Hersch (1991) and Viscusi andHersch (1990) both use 70 days days before the event.
13 If fewer than 200 days separate event dates, then we use the same regression coefficients and we add additionaldummy variables. For example, suppose events one and two occur on days 300 and 325 respectively and that eventsthree and four occur on days 550 and 575 respectively. In this situation we would use two regression equations.
Thus, to obtain coefficients for the first and second events, the regression period would cover day 150 through day375 with separate dummy variables for the two event dates. Similarly, for the third and fourth events, the regressionperiod would cover days 400 to 625.
Dt = 1 for days inside the event window, 0 otherwise; andejt = the disturbance term for security j in period t.
The disturbance term is assumed to be N(0, σ2). The coefficient bj estimates the systematic risk of security j. The coefficient cj is the average abnormal return for the security on each day of the We examine two different event window lengths. One is three days long, beginning on day t = -1 and ending on day t = +1. We define the event date, t = 0, as the day preceding the appearance of the information in the news media.14 We chose one very short event window for two reasons. First, many of the event dates involve jury verdicts. Information leakage is unlikely for these types of events because investors do not have access to juries. Second, a short window minimizes the likelihood that other news contaminated the estimates of the abnormal returns. This is particularly important for automobile manufacturers because they are in the news very frequently. Other event studies that involve many events in a relatively short time We also provide results for a ten-day event window beginning on day t = -5 and ending on day t = +4.16 A benefit of using a wider event window is that it picks up some of the information “leakage” that may occur prior to the public announcement of the event. Including a few days after the announcement date in the event window may also be important because sometimes the full implications of an event are not immediately clear. Thus, the additional days 14 Barber and Darrough (1996) also define day t = 0 as the day before the appearance of the information in the newsmedia.
15 Barber and Darrough (1996) use a two day event window in their examination of auto recalls. Hertzel and Smith(1993) also use a two day event window in their examination of the Pennzoil-Texaco litigation.
16 Jarrell and Peltzman (1985) use a ten-day event window for both their auto results and their drug results. Jarrelland Peltzman do not use a larger window for the automobile firms because it would result in serious event-windowoverlap problems. They do examine larger windows for the drug firms, but find that ninety percent of firms’abnormal returns occur in the ten days surrounding the event date. Specifically, CAPE(-49,50) = -6.742 percentwhile CAPE(-4,5) = -6.132 percent. However, CAPE(-29,30) = -5.479 percent is less than CAPE(-4,5). Thus, itdoes not appear that there is anything to be gained by examining larger event windows.
give investors more time to accurately determine the effect of the event. As indicated above, the downside to a wide event window is that other non-product liability information is more likely to Since we use a multi-day event interval, results are provided in the form of cumulative prediction errors (CPE). CPEs are derived from the following equation: where L = the length of the event window (number of days); and cj = the estimated coefficient on the event period dummy in equation (1); The significance of the CPE for an individual security is tested using the t-statistic for the dummy coefficient (cj). A test of the reaction of several firms to product liability litigation events requires the determination of additional statistics. The cumulative average prediction error (CAPE) for a group of securities is the average of the individual CPEs. The significance of CAPEs is measured by summing the test statistics for all the individual CPEs and dividing the total by the square root of the number of securities included in the CAPE. We calculate the abnormal change in the value of the firm during the event interval, by multiplying the firm’s market value immediately preceding the event window by the abnormal return. Thus, the change in the value of the firm reflects only the information that is made available during the event 4B. Market Adjustment Test
The second method used to estimate abnormal returns is called the market adjustment method. It is the method used by Barber and Darrough (1996) in their reexamination of Jarrell and Peltzman (1985). The market adjustment test yields results that are well specified and just as powerful as the traditional market model.17 One of the primary differences between the market adjustment method and the dummy variable method is that the market adjustment method provides a unique abnormal return for every day in the event window while the dummy variable method only provides an average daily abnormal return for the event window. Thus, the market adjustment model permits the researcher to determine the period in which the bulk of a firm’s The abnormal return is calculated using the following equation: where PEjt = the abnormal return for firm j on date t; Rjt = continuously compounded rate of return for firm j on date t; andMKTt = continuously compounded rate of return on the market on date t.
The significance of the abnormal returns is calculated by dividing the event window abnormal return by the standard deviation of the abnormal returns in the estimation period.
CPEs are calculated by simply summing up the daily abnormal returns for a security and the test statistics for CPEs are calculated by summing the t-statistics for the individual days and dividing by the square root of the number of days included in the CPE. CAPEs and their t-statistics are calculated in the same way as under the dummy variable method 4C. Nonparametric Sign Test
In addition to calculating CPEs and their significance, we perform sign tests. The sign tests, which are based on the sign of the abnormal returns, require two assumptions. First, we assume that abnormal returns are independent over time and across securities. Second, we assume that there is an equal probability that an abnormal return has a positive or negative sign (i.e., the probability that the abnormal return is negative equals 0.5). The test determines whether the percentage of abnormal returns that are negative (or positive) is significantly different from 50 percent. The test statistic is standard normal with mean zero and variance one and is derived from the following formula: where N = the total number of abnormal returns; and N- = the number of abnormal returns that are negative.
5. Data Set Construction
We constructed the data set of events from Dow Jones News Service for the years 1985- 1995. The search included all sources listed under the heading: Major News Sources,18 and also The Chicago Tribune.19 The following search commands were used to discover product liability cases for the years 1985 to 1995: products liability or product liabilityverdict and (liability or litigation or lawsuit)ford and (liability or litigation or lawsuit)general motors and (liability or litigation or lawsuit)chrysler and (liability or litigation or lawsuit)drug and (liability or litigation or lawsuit) After finding a case using the search commands listed above, we performed an individual search for each case in order to discover the initial news article mentioning the alleged product problem.
For example, for a news article dealing with a legal problem involving Eli Lilly’s Prozac, we would perform a very broad search such as “Prozac and Eli Lilly.” Broad searches such as these often yielded over one hundred articles used to find the initial event date. Dow Jones News 18 The Major News Sources heading includes the following publications: The Wall Street Journal, The New YorkTimes, The Los Angeles Times, The Washington Post, DJNS, Financial Times, and Barron’s.
19 Although, it is common for researchers performing event studies to only use events reported in The Wall StreetJournal, many researchers extend their samples by using events announced in other sources as well. Jarrell andPeltzman (1985) and Viscusi and Hersch (1990) use events that are not published in The Wall Street Journal.
Karpoff and Lott (1993) rely solely on events announced in The Wall Street Journal. However, they also examine29 cases carried on Dow Jones News Service, but not publicized in The Wall Street Journal. The results for the 29cases reported only on Dow Jones News Service are consistent with the sample of cases taken from announcements Service only goes back to 1985 for most sources. As a result, we used Lexis-Nexis to discover lawsuits that occurred before 1985. We performed the same searches on Lexis-Nexis as on the Dow Jones News Service.20 6. Results for the Automobile Firms
The search for product liability related events yielded a total of 44 initial “bad” events for the automobile firms (Chrysler, Ford, and General Motors). We define initial “bad” events as those that involve the release of negative news about a product problem for the first time. We placed the events into one of the following event categories: filing, losing, and uphold. Filing and losing events are simply event dates for which the initial publication of news about the product problem involves, respectively, the filing of a lawsuit or the loss of a lawsuit. The data set contains 15 filing events and 25 losing events. The uphold events are cases for which we did not find an event date for the case filing or the verdict. All uphold events involve negative news for the firms (i.e., none of the uphold events deal with a verdict that was favorable to the firm being upheld). The data set contains four uphold events.
We assume that Chrysler, Ford, and General Motors are the only competitors in the industry. Alternatively, we could have selected competitors based on their two, three, or four digit Standard Industrial Classification number. However, this method includes many parts suppliers that are not in direct competition with automobile manufacturers. We follow the custom in the literature and examine only manufacturers.21 in The Wall Street Journal. Karpoff and Lott cite this as evidence that their results are not sensitive to their sampleselection method.
20 Not all of the sources that are included in the Dow Jones News Service search are available on Lexis-Nexis. TheChicago Tribune, The Los Angeles Times, DJNS, and Barron’s are not available on Lexis-Nexis prior to 1985. TheWashington Post is available back to 1977. The Wall Street Journal is available beginning in 1975 while The NewYork Times is available beginning in 1968. We are not sure how far back The Financial Times is available, but weonly found one event date (1984) in The Financial Times.
21 Jarrell and Peltzman (1985), Hoffer, Pruitt, and Reilly (1988), and Barber and Darrough (1996) all confine theirdata set for automobile competitors to firms engaged in automobile assembly.
The event dates of greatest importance are the filing dates because these dates are earlier in the litigation process than losing and uphold events. Thus, filing dates are the “cleanest” because they are the least likely to have been contaminated by prior information. If the filing event dates are clean, we can expect a negative impact on the value of a firm surrounding the filing of a lawsuit. This is in part because the filing of a lawsuit indicates that plaintiff attorneys believe that a favorable verdict is sufficiently probable to justify what is often a substantial investment. When interpreting the results for the losing and uphold events, one must keep in mind that there may be significant information leakage prior to the event date. Thus, while one would generally expect an initial losing event date to exhibit negative returns surrounding the announcement of the verdict, it is possible that the market has already incorporated expected damage payments into the firms’ valuation.
Results for all 44 individual initial bad events are provided in appendix A. The discussion of the results in the following tables focuses on the aggregate results.
Table 6.1 provides the results of sign tests performed on the CPEs listed in Appendix A.
Table 6.1: Sign Tests for Automobile CPEsCAPE Variable * = significant at 10%; ** = significant at 5%; *** = significant at 1% The sign tests for all first party22 initial events are correctly signed, but insignificant. The sign test for all competitor initial events are significant at the ten percent level under both estimation methods for the three-day event windows. For filing events, despite the small number, the sign test is significant under the three-day market adjustment test. Under every test for the first party filing events at least 60 percent of the observations are negative. Three of the competitor filing event tests are also significant. None of the tests for the losing events or uphold events are significant and several of the losing event tests are incorrectly signed.
Table 6.2 summarizes general results for initial bad events across all firms for the two methods of calculating abnormal returns.
22 The results for first party events include only CPEs for firms directly involved in a suit.
Table 6.2: CAPEs for Initial Bad Events23CAPE Variable * = significant at 10%; ** = significant at 5%; *** = significant at 1% The CAPEs for all initial bad events are correctly signed, but insignificant. The more interesting results are found when the CAPEs are broken down by event type. The CAPEs for first party filing events are significantly negative for both tests and both windows with CAPEs ranging from –1.77 percent to –2.38 percent and t-statistics ranging from –1.90 for the ten-day dummy test to –3.32 for the three-day market adjustment test.
The CAPEs for first party losing events are insignificant and incorrectly signed. The positive sign on many of the CAPEs could indicate that investors expected a larger damage award. Thus, the announcement of the verdicts could actually be good news from investors’ perspectives. In the alternative, the results could indicate that for losing events, investors have already discounted the expected damage awards. The CAPEs for the uphold events are also insignificant, though correctly signed. Again, this could indicate information leakage.24 The 23 Jarrell and Peltzman (1985), Karpoff and Lott (1993), Rubin, Murphy, and Jarrell (1988), and Mathios andPlummer (1989) all present their general results in CAPE form.
24 There are four instances where event windows overlap for the ten-day tests. In one of the cases the same firm isinvolved with the overlapping windows. Excluding the overlapping events actually improves the results for thefiling events; therefore, we do not exclude the events.
results for the losing events confirm the findings of Garber and Adams (1998). More importantly, however, the significant results for the filing events indicate that Garber and Adams examined the wrong events and their finding of no significant effect on firms as a result of product liability lawsuits was due to this error.
Competitors also exhibit significant negative abnormal returns surrounding filing announcements under both tests and for both windows with t-statistics ranging from –2.10 for the ten-day dummy test to –3.70 for the three-day market adjustment test. Thus, it appears that the filing of a lawsuit exerts a negative externality on competitors. This lends support to the notion that investors may be more worried about the possibility of the competitors being sued than they are encouraged by any potential competitive advantage the competitors may have acquired as a result of the sued firm’s misfortune.
To determine when the firms experience the bulk of their negative returns, we create plots of the daily CAPEs for days t = -5 to t = +4 under the market adjustment model.
All Initial Events
Filing Events
Losing Events
Uphold Events
The plot for the filing events illustrates a stark contrast from the plot for all initial bad events and the plot for losing events. The plot for the filing events shows a definite drop surrounding the event date while the plots for all initial events and losing events remain relatively stable. The plots clearly indicate that the market views initial filing events much differently than initial Table 6.3 provides the average change in the value of firms and competitors surrounding Table 6.3: Average Change in Value of Firms and their Competitors25 (in millions)CAPE Variable * = significant at 10%; ** = significant at 5%; *** = significant at 1%All figures are in 1995 dollars.
25 The asterisks indicate the significance of the average changes in firm value. They correspond to the significanceof the CAPEs listed in Table 6.2.
Table 6.3 indicates that the values of automobile firms and their competitors fall dramatically surrounding the announcements of lawsuits. The firms facing the lawsuits fall in value anywhere from 276.45 million dollars for the three-day dummy variable test to 499.22 million dollars for the ten-day market adjustment test. The value of each competitor falls almost as much as the value of the firm facing the lawsuit for both tests under the ten-day window.
Total industry losses as a result of a lawsuit filing range from 638.47 million dollars to 1.43 billion dollars. The average of the four estimates for industry losses exceeds one billion dollars.
The estimate for the ten-day market adjustment test is the best estimate because it is based on CAPEs for the firms and their competitors that are both significant at the one percent level.
Thus, the best estimate indicates that the automobile industry loses an average of almost 1.50 billion dollars whenever a firm faces a new liability problem.
Do capital market losses exceed out of pocket costs? This is a difficult question to answer because insufficient data exists to exactly determine the out of pocket costs for every filing event. However, reliable data does exist for estimating the out of pocket costs that Ford experienced as a result of the Pinto litigation. We chose the Pinto because this episode has been well studied in the literature and information about this automobile is widely available. Thus, a generous estimate of out of pocket costs associated with the Pinto should provide an adequate upper bound for out of pocket costs.
A total of 219 people died in Pintos in an accident where a fire was also present.26 We do not have an exact figure for the number of non-fatal burn victims in Pinto fires, but it is likely to be similar to the number of deaths. We base this statement on a report by Ford in 1978 that it was aware of 21 fatalities and 23 burn injuries, while at the same time, the National Highway 26 National Highway Traffic Safety Administration, Office of Defects Investigation. See Appendix D for the year byyear break down.
and Traffic Safety Administration knew of 27 fatalities and 24 burn injuries in Pinto fires.27 Thus, it seems reasonable to assume that approximately the same number of people suffered non- fatal burn injuries as died. To estimate Ford’s damage payments, we obtained annual average jury verdict awards for compensatory damages in wrongful death cases.28 We also obtained an estimate for average damages paid in lawsuits involving burn injuries.29 The damage award figures are provided in Appendix C. Based on the number of injuries and deaths and the award figures, we estimate that Ford paid out 310.6 million dollars for wrongful death suits and burn suits.30 One shortcoming of this estimate is that the award figures do not include punitive damages. However, punitive damages are rarely awarded. In one study of 23,129 civil verdicts from 1981 to 1985 covering 42 counties in ten states, punitive damages were awarded in only 1.5 percent of the cases.31 Furthermore, in those cases in which punitive damages are awarded they are very often reduced or eliminated on appeal. The 310.6 million dollar figure is likely too high for three reasons. First, the figure assumes that every person burned or killed in an accident involving a Pinto fire sues. Second, the estimate assumes that every plaintiff wins his suit or receives money in a settlement.32 Finally, the figure does not take into account the fact that damage payments are deductible business expenses. Thus, the overall damage figure should be reduced by the amount that is saved in taxes. Assuming a marginal corporate tax rate of 40 percent, the expected post-tax out of pocket costs amount to only 186.4 million dollars.33 An 27 Birsch and Fielder (1994), p. 928 Statistical Reference Index. Current Award Trends in Personal Injury, 1992, 1995, 1997 editions. Compiled bythe Congressional Information Service, Inc.
29 This estimate is derived from a table in Viscusi (1991), p. 103 (citing Insurance Services Office (1997)).
30 All figures dealing with out of pocket costs are stated in 1995 dollars.
31 Daniels and Martin (1987), p. 13.
32 In addition, those who settle usually receive less than the average damage payment because the main purpose of asettlement is hedge the risk of receiving nothing at trial.
33 The 40 percent corporate tax rate is a reasonable estimate for the time period in question. From 1987-1994, themarginal corporate tax rate was 34 percent. From 1980 – 1986, the corporate tax rate was 46 percent. U.S. MasterTax Guide, 1980-1994 editions.
upper estimate of the cost of recalling the cars is 93.5 million dollars.34 Including the post-tax costs of the recall, out of pocket costs rise to 242.5 million dollars.35 The upper bound estimate for damages associated with the Pinto litigation very nearly approximates the estimated loss in market value of the automobile firms surrounding filing events for the three-day event windows under both the dummy and market adjustment tests (276.5 and 310.6 million dollars respectively). However, the 242.5 million dollar figure falls short of the estimated change in value of the firms for the ten-day window (450.1 and 499.2 million dollars for the dummy and market adjustment tests respectively). Furthermore, the estimate for out of pocket costs falls significantly short of all estimates of changes in the value of industry as a whole (ranging from 638.5 to 1,431.4 million dollars).
One possible explanation for such an industry effect greatly in excess of the upper bound for damage payments is that investors worry that since automobile models have many design aspects in common, a suit over a problem with one model is easily translatable to another model.
The data set indicates that there is some support for the notion that once a firm is sued for a particular product flaw, similar lawsuits are filed involving other makes and models. For example, after Ford was sued for deaths occurring in Pinto fires, similar lawsuits involving vehicular fires were filed with respect to four other models in the data set of initial events.
Similarly, four of the initial events involve allegations of brake defects on various models. This could explain not only the significant losses suffered by competitors, but also may explain any 34 Birsch and Fielder (1994), p. 5. The figure is adjusted for inflation.
35 The figure ignores higher insurance premiums as a potential out of pocket cost. In the 1980s approximately 30percent of corporations self-insured. Viscusi (1991), p. 25 (citing Alternative Commercial Lines InsuranceMechanisms. 1987. Hartford: Conning and Co.; Alternative Markets Update. 1987. Hartford: Conning and Co.).
Ford is among those self-insuring. Viscusi (1991), p. 26 (citing The Wall Street Journal, May 12, 1986, p. 14).
Futhermore, inclusion of insurance coverage is problematic for two reasons. First, higher (lower) premiums couldsimply indicate that a firm is purchasing more (less) coverage. Viscusi (1991), p. 29. Second, insurance premiumsfluctuate with interest rates. When interest rates increase, insurance companies may lower premiums because thepremiums will earn more interest before claims must be paid. Id.
excess losses that firms suffer with respect to their own lawsuits, since other models of the target firm may also involve the same problem. Thus, the reason that the industry as a whole suffers losses in excess of out of pocket damages with respect to an individual model may be attributable to the fact that the all firms use a similar design in many models.
The finding that firms suffer capital market losses roughly equivalent to a worst case scenario stands in contrast to the literature on the effect of regulatory events. Jarrell and Peltzman (1985) find that automobile firms may suffer capital market losses surrounding the announcement of recalls as much as ten times as great as the out of pocket costs; Rubin et al.
(1988) also find losses much larger than expected costs of CPSC recalls. Thus, it appears that the capital market responds more accurately to private sector than to public sector initiatives.
Alternatively, it may be that capital markets anticipate that consumers will punish firms more for government mandated recalls than for private lawsuits. This would in turn imply that the market does not place much faith in the ability of private lawsuits to detect undesirable behavior of firms. This would be consistent with a view that product liability litigation is more a form of rent seeking than an efficient method of punishing firms that do not honor their implicit 7. Discussion of Drug Data, Methodology, and Results
The data set for pharmaceuticals includes 40 products and 35 firms. Six products were omitted because CRSP returns are not available for the firms involved. 36 Two other products were omitted because publicity surrounding their alleged defects was available prior to the data collection period.37 The data set covers a wide range of products including contraceptives, breast 36 The products are butazolidin (an arthritis drug), advil, sudafed, an anti-nausea patch, parlodel, accutane (an acnedrug). CRSP returns are not available because the manufacturer is either a foreign firm or is not traded publicly inthe United States.
37 The drugs excluded due to prior publicity are enovid and thalidomide.
implants, antibiotics, painkillers, anti-depressants, nutritional formulas, and pacemakers. No firm is involved in problems with more than six different products. For each product we collected the initial event related to a potential product problem. The data set includes observations for 62 initial bad events.38 In 38 cases, the initial bad event is a filing event. In nine cases, the initial bad event is a firm losing event. In one case, the initial bad event is an upholding of an adverse verdict. In the other 14 cases, the initial bad event consists of a pre-suit announcement. We include pre-suit events as a category because problems associated with drugs are often publicly announced well before any lawsuits are filed or lost. The abnormal returns surrounding a filing event are not likely to capture the true costs of a potential products liability problem if information about the problem was previously released. Investors realize that the probability that a firm will face liability for a product increases dramatically if the product is associated with medical problems prior to the filing of a lawsuit. Consequently, much of the true costs of product liability lawsuits are incurred surrounding the pre-suit announcement.
We use an industry portfolio to determine the effect of lawsuits on the value of competitors. The industry portfolio consist of all firms with Standard Industrial Classification number 2834 (Pharmaceutical Preparations). The number of firms in the industry portfolio ranges from 35 to 70 over the years 1968 to 1995.
The table in appendix B provides individual firms’ abnormal returns for all 62 initial bad events. The results presented in the following tables focus exclusively on aggregate results.
Table 7.1 provides sign tests for the data in Appendix B.
38 The data set does not overlap much with Jarrell and Peltzman’s data set. Our data set includes only five of their26 products. One reason is that twelve of their drug recalls were reported in the Food, Drug and Cosmetic Reporter,but not The Wall Street Journal. Another three products involved firms for which CRSP data were not available.
Finally, some of the recalled drugs in their sample were recalled before the batch of the product before it hit storeshelves.
Table 7.1: Sign Tests for Drug CPEsCPE Variable * = significant at 10%; ** = significant at 5%; and *** = significant at 1% At least 61 percent of all initial information events are negative under the four tests. In addition, all four figures for all initial information events are significant at the ten percent level or better.
All the tests for the filing and losing events are insignificant while all the tests for the pre-suit events are significant at the one percent level. In fact, at least 13 out of 14 pre-suit events are negative under all four tests. The lack of significance for the filing and losing events indicates that the market may have incorporated the information about these product problems before the Table 7.2 provides summary statistics for the results reported in Appendix B. The weighted figures in the table treat lawsuits involving more than one firm as only one observation.
For example, if one suit named three firms, we average the abnormal returns for the three firms and use the average as one observation to avoid giving excess weight to one lawsuit. In addition, the unweighted figures are heavily influenced by the DES lawsuits, which comprise approximately 20 percent of the unweighted figures. As the table illustrates, the use of weighted or unweighted figures generally does not affect significance levels.
Table 7.2: CAPEs for Initial Bad Events for all Drug FirmsCAPE Variable * = significant at 10%; ** = significant at 5%; *** = significant at 1% The weighted CAPE for all initial bad events is significant at the one percent level under both tests and both window lengths.40 The negative returns range from –3.569 percent for the ten-day dummy variable test to –4.407 percent for the three-day dummy variable test. The weighted CAPEs for pre-suit events are much higher ranging from –8.971 percent for the ten-day dummy variable test to –11.528 percent for the three-day dummy variable test. All four weighted CAPEs for pre-suit events are significant at the one percent level. The weighted CAPEs for filing events are much lower than the pre-suit CAPEs. CAPEs are significant only for the three-day dummy variable test (-1.050 percent) and three-day market adjustment test (-0.859 percent). The weighted CAPEs for the losing events fall even lower, ranging from +0.46 percent for the ten-day dummy variable test to –1.16 percent for the three-day market adjustment 39 There are fewer weighted all initial events than the sum of the weighted pre-suit, filing, and losing events becausesometimes different firms are subjected to different types of initial events for a particular problem. For example,Dow Corning lost a breast implant suit on 12-10-91. However, the first time that a news article linked Bristol-MyersSquibb to potential liability for breast implants was on 1-7-92. Thus, for breast implants, one firm faced an initiallosing event while another faced an initial pre-suit event. For the weighted statistic for all initial events, theindividual CPEs are averaged; but for the weighted pre-suit and losing events, each firm’s CPE is included in theproper category.
40 The discussion of the results focuses on the weighted statistics for the reasons stated immediately before the table.
test. None of the weighted test statistics for losing events are significant. Depending on the abnormal return estimate used, the abnormal returns surrounding weighted pre-suit events are 7 to 13 times greater than the abnormal returns surrounding weighted filing events and 10 to 25 times greater than the abnormal returns surrounding weighted losing events.41 These results indicate that information leakage is likely for both the filing and losing events even though the event dates represent the first time that information about the problem is published in the news media.42 In addition, the abnormal returns surrounding the filing events are larger on average than the abnormal returns surrounding losing events. Thus, there appears to be more information leakage for initial losing events than for initial filing events.
We screened the event windows for unrelated negative news stories. We found 11 instances where other unambiguously negative news was released during the event window.
Excluding these event dates does not affect the significance of any statistics and the exclusion of the contaminated events actually improved the results. Therefore, we keep all the contaminated The following plots illustrate the daily CAPEs for the market adjustment tests.
41 This statement ignores the positive CAPE for losing events under the ten-day dummy variable test.
42 The discussion of information leakage assumes that on average the expected liability damages are the same for thedifferent types of initial events.
All Initial Events - Weighted
Pre-Suit Events - Weighted
Filing Events - Weighted
Losing Events - Weighted
The plots are fairly flat in the days preceding the event date indicating the absence of information leakage in the days leading up to the event date.43 A comparison of the plots illustrates the dramatic difference between the impact of the pre-suit events and the filing and losing events.
Table 7.3 shows the reaction of an equally-weighted portfolio of drug firms to the initial bad news events of their competitors.
43 Jarrell and Peltzman (1985) base their analysis of the drug firms on a ten-day window. They examine largerwindows, but find that 90 percent of firms’ abnormal returns occur in the ten days surrounding the event date.
Table 7.3: Reaction of Competitor Portfolio to Initial EventsCAPE Variable * = significant at 10%; ** = significant at 5%; *** = significant at 1% The industry portfolio reacts positively to bad new events that affect competitors. For all initial events (weighted), the portfolio experiences a statistically significant CAPE for both the three-day (+0.274 percent) and ten-day (+0.646 percent) market adjustment tests. The portfolio CAPE is significantly positive (1.349 percent) for weighted pre-suit events and weighted losing events (1.168 percent) under the ten-day market adjustment test. The results clearly indicate that investors believe that drug firms benefit from the misfortunes of their competitors.
These results stand in stark contrast to the results for the automobile firms. The automobile firms experience significantly negative returns when a competitor is sued. One potential explanation is that automobile manufacturers make products that are often much more similar than the products of competing drug manufacturers. For example, if an automobile manufacturer is sued because of a design defect associated with brakes, investors may believe that the likelihood of competitors being sued over brakes increases because all automobiles are equipped with brakes. However, if a drug firm is sued over its contraceptive, investors may believe that consumers are likely to switch to the products of competitors that are completely different. Contraceptives and other pharmaceuticals come in many different forms; consequently, bad news for one design may be good news for other designs. Of course, lawsuits against a drug firm may also lead to a general fear of more lawsuits in the industry as a whole, but the results indicate that potential substitution effects may outweigh this effect.
Table 7.4 shows statistics for the average change in the value of firms surrounding initial events. One must keep in mind that the true social costs are lower because of the significantly positive abnormal returns exhibited by the industry portfolio surrounding the event dates.
Table 7.4: Average Real Change in the Value of Firms Surrounding Initial Events44CAPE Variable * = significant at 10%; ** = significant at 5%; *** = significant at 1%;All figures are in 1995 dollars.
The figures of importance are the results for the pre-suit events because only the pre-suit events are generally significant. The eight figures for the weighted and unweighted estimated change in the value of the firms surrounding pre-suit events fall within a fairly narrow range (-492.6 million dollars to –641.4 million dollars).
Do drug firms experience capital market losses in excess of estimated damage payments? Based on estimates of the upper bound for out of pocket damage payments found in Table 7.545, it appears that the capital market losses very nearly approximate the estimated upper bound for 44 The asterisks refer to the significance of the CAPEs provided in Table 7.2.
45 Table 7.5 is reproduced in Appendix D with footnotes describing the sources of all figures and the method ofcomputing the damage payments.
Table 7.5: Estimated Deaths, Non-Fatal Injuries, and Damage Payments for Pre-Suit EventsProduct The estimate for average out of pocket costs for all products for which data are available (Average-1) indicates that firms suffered losses ranging from 7.9 to 17.5 billion dollars (or 4.7 to 10.5 billion dollars post-tax). However, if the estimated damages for Carter Wallace’s Felbatol and Medtronic’s pacemaker are excluded, the estimate for out of pocket costs falls (Average-2) to 741.7 to 998.7 million dollars (or 445.0 to 599.2 million dollars post-tax). We think that Average-2 is a better estimate than Average-1 because it is almost impossible to believe that Medtronic paid 83 billion dollars in damages and Carter-Wallace paid upwards of 116.7 billion dollars. The basis for this statement is that Medtronic had a market capitalization of approximately 3.1 billion dollars immediately before the event. Similarly, Carter Wallace had a market capitalization of only 557.5 million dollars immediately before the event. The Carter- Wallace and Medtronic lawsuits were also brought in jurisdictions that permit users to sue even in the absence of present harm. Thus the lawsuits involving Carter-Wallace and Medtronic include everyone who has used the product even if the product has no negative effects on the consumer. These users are highly unlikely to collect significant damage payments. As stated earlier, estimates for capital market losses ranged from 492.6 to 641.4 million dollars and the estimate for average post-tax (Average-2) damage payments ranges from 445.0 to 599.2 million dollars. The figures are strikingly similar and indicate that investors may assume a worst case scenario upon learning of a potential problem with a drug product.
8. Conclusion and Implications for Further Research
Viscusi and Hersch (1990) and Garber and Adams (1998) are the only papers to consider the stock market effects of product liability litigation. Their treatment of the subject is incomplete and this paper fills many gaps in the literature. The primary contribution of the research is the estimation of the costs to manufacturers of product liability litigation. Contrary to the findings of Garber and Adams, the results indicate that lawsuits do have a statistically significant negative effect on the value of firms. Garber and Adams erred in examining verdicts; it is the announcement of a problem or the filing of a lawsuit that leads to losses. This result Another important result is the determination of the effects that product liability lawsuits have on the competitors of the firms facing the lawsuits. The question of whether the effects are positive or negative was previously unanswered. The results indicate that the effects on competitors can be either positive or negative depending on the industry. One possible explanation is that investors may view the products of automobile firms as much closer substitutes, indicating that similar lawsuits may more easily be brought against competitors, than A third contribution relates to the reputation costs of product liability lawsuits to the firms directly involved in the suit. Neither Viscusi and Hersch nor Garber and Adams adequately address this issue. The results indicate that firms facing lawsuits for their products suffer capital market losses approximately equal to a worst case scenario associated with the litigation. Thus, we cannot definitively conclude that individual firms suffer reputation costs as a consequence of product liability lawsuits. Since studies of government mandated recalls have uniformly found losses greater than the costs of the recalls, this evidence indicates that the market puts more credibility on the ability of government agencies to detect firm cheating than on the ability of private lawsuits to detect such cheating.
This research raises several questions. First is the explanation for the difference in the automobile and pharmaceutical industries of the effects on competitors of lawsuits. Further analysis of this issue would be appropriate. Second, we have in some cases attempted to measure the costs to the firm of the sequence of litigation that could be expected to follow the lead case. However, a more accurate measure of these costs would be useful. Third, while we have examined two industries where product liability is common, extensions to additional industries would also be useful. Since effects of litigation on the value of firms appear to be real and significant, further examination is clearly warranted.
References
Barber, Brad M., and Masako N. Darrough. 1996. “Product Reliability and Firm Value: The Experience of American and Japanese Automakers, 1973-1992.” Journal of PoliticalEconomy 104(5): 1084-1099.
Birsch, Douglas, and John H. Fielder. 1994 The Ford Pinto Case: A Study in Applied Ethics, Business, and Technology. Albany, New York: State University of New York Press.
Brown, Stephen J., and Jerold B. Warner. 1985. “Using Daily Stock Market Returns: The Case of Event Studies.” Journal of Financial Economics 14(March): 3-32.
Calfee, John E., and Paul H. Rubin. 1992. “Some Implications of Damage Payments for Nonpecuniary Losses.” Journal of Legal Studies 21(2): 371-441.
Daniels, Stephen, and Joanne Martin. 1987. Empirical Patterns in Punitive Damage Cases: A Description of Incidence Rates and Awards. American Bar Foundation Working Paper#8705.
Fama, E. F. 1991. “Efficient Capital Markets: II.” Journal of Finance 46(5): 1575-1617.
Garber, Steven, and John Adams. 1998. “Product and Stock Market Responses to Automotive Product Liability Verdicts.” Brookings Papers on Economic Activity: Microeconomics:1-44.
Hersch, Joni. 1991. “Equal Employment Opportunity Law and Firm Profitability.” The Journal of Human Resources 26(1): 139-153.
Hertzel, Michael G., and Janet Kiholm Smith. 1993. “Industry Effects of Interfirm Lawsuits: Evidence from Pennzoil v. Texaco.” The Journal of Law, Economics, and Organization9(2): 425-444.
Hoffer, George E., Stephen W. Pruitt, and Robert J. Reilly. 1988. “The Impact of Product Recalls on the Wealth of Sellers: A Reexamination.” Journal of Political Economy 96:663-770.
Huber, Peter W. 1988. Liability: the Legal Revolution and its Consequences. New York: Jarrell, Greg, and Sam Peltzman. 1985. “The Impact of Product Recalls on the Wealth of Sellers.” Journal of Political Economy 93(3): 512-36.
Karafiath, Imre. 1988. “Using Dummy Variables in the Event Methodology.” The Financial Karpoff, Jonathan M., and John R. Lott Jr. 1993. “The Reputational Penalty Firms Bear from Committing Criminal Fraud.” Journal of Law and Economics 36(October): 757-802.
Klein, Benjamin, and Keith B. Leffler. 1981. “The Role of Market Forces in Assuring Contractual Performance.” Journal of Political Economy 89(4): 615-641.
Mathios, Alan, and Mark Plummer. 1989. “The Regulation of Advertising by the Federal Trade Commission.” Research in Law and Economics 12: 77-93.
Mitchell, Mark L. 1989. “The Impact of External Parties on Brand-Name Capital: The 1982 Tylenol Poisonings and Subsequent Cases.” Economic Inquiry 27(4): 601-618.
Mitchell, Mark L., and Michael T. Maloney. 1989. “Crisis in the Cockpit? The Role of Market Forces in Promoting Air Travel Safety.” Journal of Law and Economics 32(2): 329-355.
National Highway Traffic Safety Administration, Office of Defects Investigation.
Peltzman, Sam. 1981. “The Effects of FTC Advertising Regulation.” Journal of Law and Economics 24 (December): 403-48.
Peltzman, Sam. 1998. “Comments on Garber and Adams.” Brookings Papers on Economic Activity: Microeconomics: 45-48.
Priest, George L. 1987. “The Current Insurance Crisis and Modern Tort Law.” The Yale Law Journal (96): 1521-1590.
Rubin, Paul H., and Martin J. Bailey. 1994. “The Role of Lawyers in Changing the Law.” Journal of Legal Studies 23 (June): 807-831.
Rubin, Paul H., R. Dennis Murphy, and Gregg Jarrell. 1988. “Risky Products, Risky Stocks.” Regulation (Number 1): 35-39.
Rubinfeld, Daniel L. 1998. “Comments on Garber and Adams.” Brookings Papers on Economic Activity: Microeconomics: 48-53.
Statistical Reference Index. Current Award Trends in Personal Injury, 1992, 1995, 1997 editions. Compiled by the Congressional Information Service, Inc. Call number C5180-1.
U.S. Master Tax Guide. 1980-1994 editions. Chicago, IL: Commerce Clearing House Inc.
Viscusi, Kip W. 1991. Reforming Products Liability. Cambridge: Harvard University Press.
Viscusi, Kip W., and Joni Hersch. 1990. “The Market Response to Product Safety Regulation.” Journal of Regulatory Economics 2:215-230.
Appendix A
Table A: Abnormal Returns for Individual EventsProblem General Motors
Corvette construction
Ford
Semi-truck clutch assembly
Chrysler
Stalling Valiants, Darts,
* = significant at 10%; ** = significant at 5%; *** = significant at 1% Appendix B
Table B: Abnormal Returns for Individual Firms46Product ***(-4.06177) ***(-3.26574) ***(-3.82950) ***(-2.95477) 46 Some of the firms involved in lawsuits are units of another firm. In these cases we also provide the name of theparent corporation in parentheses. Since the subsidiaries are not traded independently, the results are derived fromthe returns of the parent companies.
***(-4.49990) ***(-3.25357) ***(-4.33525) ***(-3.18341) ***(-4.56238) ***(-2.50441) ***(-4.56014) ***(-2.64589) * = significant at 10%; ** = significant at 5%; and *** = significant at 1% Appendix C
a The figures for 1975 to 1980 are estimates based on the data from 1981 to 1995. Awards increased an average of9.5 percent over the fifteen years 1981 to 1995. Thus, I assumed the same trend for the years in which data weremissing (1975 to 1980).
b These figures are all based on the figure from 1977. Over the period 1971 to 1988 product liability awards grew anaverage of 12.9 percent a year. (See Viscusi (1991), p. 96) I assume that awards continue to increase at the samerate from 1989 to 1995.
Appendix D
Estimated Deaths, Non-Fatal Injuries, and Damage Payments for Pre-Suit EventsProduct (post-tax)a If available, the figures for both deaths and adverse reactions are based on the highest available estimate from anunbiased government source (for example, the FDA or NHTSA).
b We include all non-fatal complaints no matter how minor in the figure for adverse reactions.
c With two exceptions the figures in this column are based on the wrongful death average damage payments listed inAppendix D. We treat all adverse reactions the same as deaths for purposes of calculating damage payments. Wealso assume that all damage payments are made at the time of the initial news event. The figures for the DalkonShield and Shiley’s heart valve are based on reported damage payments.
d One article states that some children died as a result of the baby formula. “Metzenbaum seeks new Syntex probe;cites company memo as evidence for infant formula inquiry.” The Washington Post. 12-29-85, p. D1. However,the number of deaths is likely to be very low because another article states that there were 141 reported cases ofadverse reactions, but no deaths. “The doctor’s world; recovery gauged in baby formula deficiency.” The New YorkTimes. 7-9-91, p. C3.
e See The New York Times article in note d supra. The figure is based on reports received by the Center for DiseaseControl. The reports concern hypochloremic metabolic alkalosis, a disorder in which insufficient intake of chloridemay cause lethargy, vomiting, and a failure to grow.
f See The Washington Post article in note d supra. The figure is based on a 1983 Justice Department memo allegingproblems ranging from death to minor ailments.
g “Criminals by any other name; corporate executives.” Washington Monthly. January, 1986. The product waslinked to kidney and liver failure.
h Diller, Wendy. “Strong medicine: drug firm’s hard sell; it’s an RX for profit, but is the public served?” TheRecord. 7-16-89, p B1.
i “Hired guns’ aim to keep veil of secrecy on product damages.” San Diego Union-Tribune. 5-4-91, p. A3. Oraflexallegedly caused liver damage in users.
j Kurtzman, Joel. “Business diary / August 16-21” The New York Times. 8-23-92, sect. 3, p. 2.
k “Plaintiffs’ motion seeks California class for pacemakers recipients.” Mealey’s Litigation Reports: Drugs andMedical Devices. 5-16-97, vol.2, no. 2. The figure is based on the number of members in class action suit.
l “Optical Radiation may face criminal probe.” UPI. 3-24-92. Article reports that 33 users required eye surgery and116 individuals reported elevated eye pressure.
m “First Michigan case filed against Abbott Laboratories.” PR Newswire. 2-8-94. The figure is based on FDArecords.
n “Drug’s toxic side effects call U.S. testing into question.” The Sun-Sentinel. 1-9-94, p. 3F. The figure includes 54cases of acute kidney failure and 113 cases of hemolytic anemia (a blood disease). The figure is based on adversereactions reported to the FDA.
o Palosky, Craig S. “Generic asthma drug blamed for illnesses; hundreds sue pharmaceutical company.” The TampaTribune. 3-13-95. The figure is based on adverse reactions reported to the FDA. The complaints allege that bacteriacause pneumonia, other respiratory problems, and some deaths.
p “Outbreaks: (HCV) Gammagard responsible for IGIV-associated infections.” Blood Weekly. 1-27-97. The figureis based on reports to the Center for Disease Control. The product is alleged to cause Hepatitis C.
q “FDA urged to continue use of new epilepsy drug.” Chicago Tribune. 9-28-94, p. 7. The drug is alleged to causeaplastic anemia.
r The class action suit was allowed to include all 100,000 current and former users. “Class action lawsuit involvinganti-epilepsy drug Felbatol allowed to proceed.” PR Newswire. 12-18-95. However, only 3,000 adverse reactionswere reported to the FDA. “Plaintiffs contend Felbatol substantially injured thousands.” Pharmaceutical LitigationReporter. March, 1996, p. 11135.
s Schmeck, Harold M. “FDA head says 2 drugs linked to 32 deaths.” The New York Times. 1-30-75, p. 20. Figureswere reported by the FDA.
t “Antibiotics under fire.” Chemical Week. 1-29-75, p. 19. Figures were reported by the FDA.
u Bacigal, Ronald J. The Limits of Litigation: The Dalkon Shield Controversy. Durham, North Carolina: CarolinaAcademic Press, 1990, p. 3 (citing Lord, “The Dalkon Shield Litigation: Revised Annotated Reprimand by ChiefJudge Miles W. Lord,” 9 Hamline Law Review 7, 26 (1986)).
v Mintz, Morton. At Any Cost: Corporate Greed, Women, and the Dalkon Shield. New York: Pantheon Books,1985, p. 3-4. This figure includes women who became pregnant while using the Dalkon Shield. An estimated66,000 of the women miscarried.
w When Robins filed for bankruptcy on August 21, 1985, Robins and its insurer Aetna Casualty & Surety Companyhad already paid out 530 million dollars for claims. Couric, Emily. “The A. H. Robins Saga,” 72 A.B.A. Journal 56,56 (July 1, 1986). When American Home Products acquired Robins, it agreed to contribute 2.4 billion dollars to atrust to cover all remaining claims. The Executive Letter, “A.H. Robins’ Reorganization Plan Moves Forward,” July25, 1998. At the time Robins filed for bankruptcy, the company’s equity value was only 249.7 million dollars.
Bacigal, Ronald J. The Limits of Litigation: The Dalkon Shield Controversy. Durham, North Carolina: CarolinaAcademic Press, 1990, p.47. Thus, one could argue that Robins’ true out of pocket costs could not surpass theremaining stockholder’s equity value. Therefore, We list two estimates for Robins out of pocket costs.
x We was unable to find reliable estimates for the out of pocket costs for minor players in the breast implantlitigation.
y The average-1 figure includes the estimates for all of the drugs.
z The average-2 figure excludes the estimates for Medtronic’s pacemaker and Carter Wallace’s Felbatol.

Source: http://economics.emory.edu/home/assets/workingpapers/rubin_00_08_paper.pdf