The cigarette market has long been controversial in this country and elsewhere. Proposals to ban (or at least severely curtail) the advertising and promotion of cigarettes have long been supported by major health organizations, the U.S. Surgeon General, and key members of the U.S. Congress (Warner et al. 1986). In August, 1995, the Food and Drug Administration (FDA) proposed strong restrictions on tobacco advertising, such as limiting most print advertising to “tombstone” advertisements containing no images and little information other than warnings. Sweeping restrictions on tobacco advertising have already been adopted in many other nations. These events have produced lively discussion of the workings of advertising and other aspects of the cigarette market. This debate has brought forth at least two major issues. The first is the probable effect of current and further restrictions on advertising. Some observers have predicted that a complete ban would bring a reduction in smoking, whereas others have argued that similar measures in this and other countries have had little or no effect on smoking while possibly interfering with beneficial market changes (e.g., Laugesen and Meads 1991; Stewart 1992). A second point of contention focuses on unexpected results. Many observers have argued that previous interventions in the cigarette market suffered from the “law of unintended consequences,” in the sense of having done more to protect the industry than promote the welfare of consumers. As is out-lined subsequently in a review of the literature, studies have often concluded that major interventions did not bring the market changes presumably sought by regulators and legislators. If previous hopes were frustrated by an inadequate understanding of the dynamics of the cigarette market, it would be wise to investigate these dynamics further before launching a new and more vigorous round of interventions.

The effects of information and regulation on the cigarette industry’s welfare form the focus of this article. Our research takes an approach that differs from most previous studies by directly examining how regulation and other events have affected the welfare of the cigarette industry itself. We do this partly because of the obvious desire to address the issue of unintended regulatory effects, but there are additional reasons for taking a nontraditional approach. These are rooted in the specific type of regulation that has prevailed in this particular market. 

Regulation of Cigarettes

Cigarettes occupy a peculiar regulatory status. Tobacco products have been treated as neither food nor drug, with the effect that political intervention has mainly taken the form of informational regulation.[1] Essentially, regulation has sought to modify the information environment by providing information directly, requiring sellers to provide information, or restricting the competitive use of information in advertising. Within these constraints, buyers and sellers have been left free to adjust to changing consumer preferences and competitive forces.

Market dynamics have been dominated by informational shocks and regulatory responses since the “cancer scare” of the early 1950s. These shocks (some created by government itself) include: (1) the first widely accepted medical reports linking smoking and cancer, which appeared in the early 1950s; (2) two episodes (in 1953-54 and 1957-60) in which health information induced rapid changes in advertising content, only to have these changes reversed by Federal Trade Commission (FTC) intervention; (3) two major government reports on smoking and health, first from the United Kingdom in 1962 and then from the U.S. Surgeon General in 1964; (4) major antismoking campaigns led by the Surgeon General and other federal officials, beginning in the late 1970s; and (5) a steady succession of updates from the Surgeon General and others on new developments in the analysis of health effects from smoking. Even today, as local regulation has sought directly to restrict smoking in public places (a noninformational form of regulation), the centerpiece of regulatory initiatives in the United States and elsewhere is the proposal to ban or severely restrict cigarette advertising and promotion (FDA 1995; Warner et al. 1986).

Regulation through information, rather than through direct control over production and allocation of resources, presents policy analysts with an interesting set of problems. The informational approach offers several theoretical advantages. These include preservation of individual freedom of choice; accommodation to individual differences in preferences and circumstances; the ability of market participants, including sellers, to change markets without the inefficiencies inherent in command-and-control regulation; and (within limits) the opportunity for market outcomes (rather than post hoc studies of what “would have been”) to reveal whether the assumptions and predictions that underlie intervention are valid (cf. Beales, Craswell, and Salop 1981; FTC 1979).

On the other hand, the effects of informational regulation are sometimes difficult to specify, predict, or assess. The freedom of buyers and sellers to adjust to changed conditions makes for ambiguous theoretical predictions about observable market changes. For example, behavioral changes depend strongly on how new information compares with old, but policymakers often have a poor understanding of consumer information before the intervention and may fail to anticipate that information mandated by regulation is not actually “new” to consumers. Also, sellers could adjust their products and advertising in ways that leave equilibrium market sales relatively unchanged despite major changes in, say, the tar and nicotine yield of cigarettes or consumer attitudes toward smoking.[2]

Assessing the Effects of Regulation

These properties of regulation through information complicate the task of assessing regulatory effects. The usual approach has been to analyze actual market changes in advertising, information, and sales. Several statistical studies have measured the effects on cigarette consumption of the cancer scare of the early 1950s and the series of government actions that began with 1964 Surgeon General’s report and continued through the label warnings required in 1966 and the 1971 ban on broadcast advertising. The results are roughly as follows: The cancer scare apparently depressed sales for at least the short term and perhaps for several years (Ippolito, Murphy, and Sant 1979; Schneider, Klein, and Murphy 1981; Porter 1986; Warner 1979). Health information after 1964 apparently had an impact in the sense of reducing smoking to much less than would reasonably have been predicted for the years after 1964, with some of this probably due to the 1964 Surgeon General’s report (see Bishop and Yoo 1985; Ippolito, Murphy, and Sant 1979; Porter 1986; Schneider, Klein, and Murphy 1981; Warner 1979; Warner and Murt 1982; for a review of previous studies, see Franke 1994; Mc Auliffe 1987). However, the 1966 warnings apparently have had little independent effect (McAuliffe 1988). In addition, the 1971 broadcast advertising ban, which also ended free antismoking messages broadcast under the Federal Communication Commission’s (FCC) fairness doctrine, was probably counterproductive because its net effect was to reduce competition in the market but not to reduce smoking (Eckard 1991; Franke 1994; Simonich 1991; Warner 1979).

Additional insights have come from nonstatistical analyses of market outcomes. The 1965 legislation, which required label warnings but prohibited the FTC from requiring warnings in advertisements, was expected by some observers to benefit the industry (Drew 1965). This and subsequent legislation that mandated warnings were found by federal courts to have preempted liability suits for failure to warn (Cipollone v. Liggett Group, Philip Morris, and Lorillard 1992). A few analysts, looking at the neglected era before the 1964 Surgeon General’s report, have suggested that FTC advertising regulation in the 1950s and early 1960s may have protected cigarette sellers from their own competitive, self-destructive tendency to seek market share by exploiting fear of smoking (Calfee 1985, 1986; McAuliffe 1988; Ringold and Calfee 1989, 1990).

These results suggest taking a different approach to analyzing the effects of regulation on the cigarette industry. A researcher can employ methods that address directly the question of how information and regulation affected the financial state of the industry itself. We here look at stock returns of the major cigarette sellers, using what has become known as event analysis to isolate unexpected or unusual changes in the returns of cigarette sellers.[3] This method has two advantages: The first is that stock prices are the surest guide to the extent to which events were expected to confer benefits and costs on individual firms or industries. This is important in view of the suggestion from previous research that regulation may have done little to promote the putative goals of improving information and reducing smoking, while protecting the interests of the cigarette industry itself (Calfee 1987). A second advantage is that market valuations may anticipate later events that are important to consumers or policymakers. In some circumstances, for example, a decline in stock returns could presage a decline in sales that will occur after information known mainly to industry analysts later becomes known to smokers; but we must be careful to avoid ascribing more knowledge to market participants than was available at the time. For example, it is far from clear that anyone could forecast the reaction of consumers to the new health information emerging in the 1950s, nor could anyone easily forecast how individual firms would respond to the complex mix of new health information, new advertising themes, and changed consumer perceptions.

The only other work we are aware of in this genre is the study by Mitchell and Mulherin (1988). Their event analysis found that actions leading to the 1971 broadcast ban--which the industry had supported--were associated with positive returns to the cigarette industry.[4] This finding, combined with the results of other statistical studies indicating that the ban was associated with an increase rather than a decrease in smoking, strongly suggests that the net effect of the 1971 broadcast advertising ban was to benefit the cigarette industry.

Background and Hypotheses

The cigarette market in the years from 1950 to 1965 was dominated by several remarkable events: publicity about the first widely credited academic reports linking smoking with cancer, a brief outburst of controversial advertising that sought to exploit smokers’ fears, federal intervention in 1954-55 to halt this advertising, a vigorous “tar derby” in 1957-59 during which some sellers advertised tar and nicotine content, federal intervention in 1960 to halt the tar and nicotine advertising, the appearance in 1962 of a major British government report finding that smoking causes cancer, the arrival in January 1964 of the first major U.S. government report on smoking and health, and finally, in 1965, legislation requiring warning labels. Previous analysis has looked mainly at cigarette sales in assessing the effects of these and later events. It is clear from a cursory examination that though the cancer reports buffeted the industry during 1950 through 1952, sales at first continued to rise and then abruptly fell during 1953 and 1954. Per capita sales then resumed a strong upward course that was not disturbed until the 1964 Surgeon General’s Report, after which sales quickly rebounded in 1965 and then fluctuated around the 1964 level until commencing a steady decline in 1975 (Maxwell 1986).

The Cancer Scare: 1950-54

In 1950, two medical journals published studies demonstrating a strong statistical link between smoking and lung cancer.[5] A follow-up study appeared in 1952 (Doll and Hill). In 1953 came the results of a laboratory experiment in which cigarette tar applied to the skin of a mouse caused cancer, followed in 1954 with yet more epidemiological results (i.e., Doll and Hill 1952, 1954; Wynder, Graham, and Croninger 1953). All these reports received extensive publicity and aroused discussion about the possibility that the cancer scare would adversely affect cigarette sales. The magazines, Consumer Reports and especially Readers Digest, with its vast readership, also played an important role. Consumer Reports published articles on smoking and health in June 1952, February 1953, and December 1953, and Readers Digest, which had long campaigned against smoking, published articles in December 1952 (Norr 1952), December 1953 (Lieb 1953), and July 1954 (Miller and Monahan 1954).[6] In particular, in the July 1954 Readers Digest article, Miller and Monahan summarize the new epidemiological and laboratory evidence in a manner that attracted widespread attention. yet the evidence on smoking and cancer was far from overwhelming in the eyes of the public health community; the American Cancer Society, for example, did not take a position on this issue until 1957, and even then it advised smokers only that there was a “suspicion” that smoking “[increases] the likelihood of developing lung cancer” (Consumer Reports 1957; Miller and Monahan 1954). The U.S. government took no position at all until a brief statement by the Surgeon General in 1957 (Burney 1958).

Unfortunately, it is difficult to specify precise “events” that might have a distinct effect on the market. What was emerging was not merely evidence on smoking and health, but also evidence on how the press treats medical results and on smokers’ reactions to such information. Nonetheless, we begin our analysis with the most obvious hypothesis, namely, that new adverse health information reduced the market valuation of cigarette firms during the period of the first early and significant reports in 1950. 

Fear Advertising: 1953-54

Another, perhaps unexpected development during these same years was a new departure in product development and cigarette advertising (for a more general treatment of the content of cigarette advertising over time, see Ringold and Calfee 1989, 1990). Before 1953, Brown & Williamson’s Viceroy brand was the only filtered brand from a major manufacturer, and its market share was negligible. Late in 1952, Lorillard, one of the smallest of the six major cigarettefirms, introduced a new brand called Kent--with a filter that was remarkably effective in removing nicotine and tar from cigarette smoke. Liggett & Myers, another of the smaller firms, introduced its filtered L & M brand in 1953, and filter brands from the other four major firms followed in 1954 (Miles 1982). Especially notable were the new advertising campaigns devised by the smaller four firms (the two largest firms, American Tobacco and R. J. Reynolds, controlled approximately 60% of the market.) These firms advertised their filter brands (and some nonfilter brands) in ways that blatantly appealed to health fears: “The cigarette [Philip Morris] that takes the FEAR out of smoking: (Business Week 1953a, p. 54); “Here’s how you can get the protection you need against nicotine and tars” (Consumer Reports 1953a, p.66); “[Kent] takes out more nicotine and tars than any other leading cigarette--the difference in protection is priceless” (Wagner 1971, p. 82); Viceroy’s “double-barrelled protection” (Business Week 1954, p. 83); and the new L & M filter was “just what the doctor ordered” (Wagner 1971, p. 83).[7]

This kind of advertising became known as fear advertising. It was controversial, not least of all because many industry insiders and observers were certain the advertisements increased rather than allayed fears, and helped only the advertised brands (building from small market shares) and the small firms selling them, while hurting larger sellers and, therefore, overall cigarette sales. Business Week, Consumer Reports, and other magazines attributed part of the unprecedented sales decline of 1953 and 1954 to this outbreak of fear-based advertising.[8]

The advent of fear advertising is another important event in the cigarette market. Fear advertising was widely noted in the popular and trade presses during 1953 and 1954. It came to an end as a result of a set of cigarette advertising guides from the FTC. First circulated in late 1954 and formally issued in 1955, the guides made clear that the Commission would prosecute any seller whose advertising contained references “to either the presence or absence of any physical effect of smoking” (FTC 1955).[9] The trade press celebrated the cessation of health-related advertising as a factor in the rapid upturn in sales in 1955 and through the rest of the decade (Fortune 1953, 1963; Wootten 1956, 1957, 1958).

The 1960 FTC Ban on Tar and Nicotine Advertising

In 1957 came additional health shocks in the form of more rigorous epidemiological studies relating smoking and lung cancer, a public recognition (without a formal report) by the Surgeon General that the link was probably causal in nature, and recommendations by nongovernment cancer experts that smokers use cigarettes with substantially reduced tar and nicotine yield (Burney 1958; Wynder 1988). This expert advice, amplified by two striking stories in Readers Digest (Miller and Monahan 1957a, b), promptly led to an outburst of tar and nicotine advertising, which similar to the fear advertising five years earlier, was confined to the smaller four firms (Calfee 1985). A tar derby raged for two and a half years, during which time tar and nicotine yield declined substantially in virtually all brands, filtered and nonfiltered, and the market share of filtered brands increased dramatically, In 1959, five firms (all but R. J. Reynolds) introduced new filtered brands, most of which had less measured yield than existing brands. As of December 1959, most of these new brands were still in an introductory phase of a brand life cycle and were heavily dependent on tar and nicotine advertising.

During December 1959 and January 1960, the FTC secretly negotiated an agreement among all six cigarette firms to institute a “voluntary” ban on all tar and nicotine claims in advertising, justifying this action by the fact that no studies had yet shown health benefits from reduced tar and nicotine (FTC 1960; Kinter 1960). The ban was announced on February 5, 1960. Press reports made clear that the smaller firms strongly objected to the ban because it would impede acceptance of the newly introduced generation of low-tar brands. On the other hand, some industry analysts believed that the end of explicit tar and nicotine claims would help the market overall.[10]

The February 1960 tar and nicotine advertising ban forms a fairly precise event. Our hypothesis is that the smaller firms were hurt relative to larger ones. This is apparently consistent with industry expectations at the time the ban was instituted.[11]

The 1962 Report of the Royal College of Physicians

Early in March 1962, the Royal College of Physicians (1962, p. 57), a British organization, released Smoking and Health, a report that concluded that smoking is a cause of lung cancer and bronchitis and “probably contributes to the development of coronary heart disease and various other less common diseases.” This report was well-publicized in the United States and was a factor in President Kennedy’s decision in June 1962 to ask the Surgeon General to convene a special committee to produce a report on smoking and health (Wagner 1971, Chapter 8). The Royal College report was not the only significant development in smoking and health during 1962 and 1963. Readers Digest published articles in June (1962) and August (1963a, b) describing not only the British report but also antismoking activity of various kinds in the United States and abroad. The U.S. Air Force ended its practice of providing free cigarettes at its hospitals, the American Heart Association began its campaign against smoking, and most cigarette advertisements were pulled from U.S. college newspapers. We hypothesize that these events, particularly the introduction of the Royal College report, had an adverse effect on cigarette stock returns. The Royal College report has hardly been noticed in subsequent studies of the cigarette market, perhaps because there was no immediate change in sales trends and perhaps alsobecause much thinking about the cigarette market in this country has been dominated by the successive Surgeon General’s reports, each of which looks back fondly on the 1964 report. As with the original cancer studies, it is difficult to specify a period over which the report would have an effect, because no one knew how consumers and the industry would react to this document, but we expect the impact to appear mainly in the short run (i.e., in 1962). 

The 1964 Surgeon General’s Report and the FTC’s Proposed Trade Rule

In January 1964, the Surgeon General released a report by its select committee on smoking and health. The report, whose conclusions mainly mirrored those of the 1962 Royal College of Physicians report, was carefully guarded from the public before its tumultuous release. The FTC immediately announced its intention to construct a trade rule that would, among other things, require health warnings in advertisements (FTC 1964; Miles 1982). The FTC action was eventually halted by Congress in the 1965 law that brought the Surgeon General’s warning to package labels (Wagner 1971). 

Methods

The seminal paper by Fama and colleagues (1969; hereinafter termed FFJR) defined the basic event study methodology used in numerous studies to examine the effect of new information on asset prices. This approach estimates a market model.

(1) R[it] = alpha[i] + beta[i]R[mt] + epsilon[it],

where

R[it] = return of company i’s stock during time period t and R[mt] = return on the market index.

The model is estimated with data that includes the period of the event of interest. The residuals from Equation 1 are used to measure abnormal returns for the firms in the sample during the event period. These abnormal returns are cross-sectionally averaged across firms during the event period. The null hypothesis that the average abnormal return equals zero can then be tested.

This approach, however, is not suitable for the problem at hand. The FFJR method assumes that the residuals are independent and identically distributed. Binder (1985) describes three problems with this assumption:

1. Abnormal returns are likely to differ across firms.
2. Previous studies suggest that the variance of the residuals will differ across firms.
3. The residuals will not be independent if the event occurs during the same calendar time period for some firms and these firms are in the same or related industries. The dependence is especially severe when these two conditions hold for all the firms in the sample.

In addition, many of the events of interest in this study represent periods of distinct regulatory regimes rather than single points in time. Information is continually assessed and reassessed throughout such time windows. A method is required that addresses Binder’s concerns while also measuring the effects of regulatory regimes over a complete event period. Such an approach is described by Binder (1985) and Rose (1985), where

(2) R[it] = alpha[i] + beta[i]R[mt] + A / SIGMA / a = 1 delta[ia]D[at] + u[it]. The variable delta[ia] measures the average excess return over the particular regulatory regime period a, and the dummy variables D[at] equal one during the a