Tobacco Documents Online

KING PA. ER THE POLITICAL ELEMENT IN SCIENCE AND TECHNOLOGY: SAMMEC II AND THE ANTI-SMOKING LOBBY CENTER FOR STUDY OF PUBLIC CHOICE DEPARTMENT OF ECONOMICS GEORGE MASON UNIVERSITY TIMN 445029

THE POLITICAL ELEMENT IN SCIENCE AND TECHNOLOGY: SAMMEC II AND THE ANTI-SMOKING LOBBY by Richard W. Ault and Robert B. Ekelund, Jr. Auburn University March 1993 TIMN 445030

TABLE OF CONTENTS Executive Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I. Introduction ........................................ 1 II. SAMMEC and Economic Cost Computations ................... 3 What Is SAMMEC II? ................................. 3 How SAMMEC Has Been Used ........................... 4 Economic Project Evaluation and the Concept of Cost .............. 5 III. The Methodology of SAMMEC II .......................... 10 Indirect Mortality Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . • • • • • 12 Direct Morbidity Costs ................................. 16 Indirect Morbidity Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . • • • • • . . 24 IV. Technical Flaws in SAMMEC II Methodology. .................. 27 V. Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . 34 Footnotes .............................................. 37 References ............................................. 39 Attachment A ........................................... 42 Section I .......................................... 42 Section II ......................................... 43 Section III ......................................... 44 Section IV ......................................... 45 Section V ......................................... 46 i TIMN 445031

LIST. OF TABLES Table 1 Diagnosis Set .................................. 11 Table 2 Relative Risk Elements ............................ 14 Table 3 Smoking Prevalence Rates .......................... 15 Table 4 Years of Potential Life Remaining Relative to Life Expectancy: Males .............................. 17 Table 5 Years of Potential Life Remaining Relative to Life Expectancy: Females ............................. 18 Table 6 Present Value of Future Earnings (4%) .................. 19 Table 7 Present Value of Future Earnings (6%) .................. 20 Table 8 Present Value of Future Earnings (10%) .................. 21 Table 9 Relative Rates for Hospital Days Per 100 Persons and Physician Visits Per 100 Persons by Smoking Status ................. 23 Table 10 Relative Rates for Work-Loss Days Per 100 Employed Persons Per Bed-Disability Days Per 100 Females Keeping House by Smoking Status ................................. 25 11 TIMN 445032

EXECUTIVE SUMMARY Software with the acronym SAMMEC II (Smoking-Attributable Mortality, Morbidity, and Economic Costs) was developed in 1990 for purposes of evaluating populations (state, city, and national) in order to determine the so-called "social costs" of smoking. The software SAMMEC II, developed initially by James M. Schultz as SAMMEC in 1985 and 1986 in conjunction with the Minnesota Department of Health, has been used primarily in one well publicized nationwide study -- the National Status Report on Smoking and Health, 1990 (the "Sullivan Report") -- and in various states in the United States. The purpose of this research paper is to evaluate the second generation software (based on the initial SAMMEC) from the related perspectives of logic and statistical adequacy. We note, from the outset, that any "costs" of smoking (to the extent that such costs may exist beyond the price of the product) are all rzp 'vate in nature -- that is, they are borne and internalized by individuals in private actions. We also note that, in any generally accepted methodology, both "benefits and costs" must be evaluated. All rational individuals, in making choices, evaluate both the benefits of consumption and the putative costs. Property rights and property dispositions belong to the individual alone in voluntary exchange. As Ezra Mishan, an authority on economic cost-benefit analysis, who is misinterpreted by SAMMEC II developers, argues, "Insofar ... as additional risks associated with the service or facility [or product] are all voluntarily assumed, there is no call for ui TIMN 445033

intervention in the allocative solution to which the market tends" (see pp. 8-10 of our report). The fact that individuals actually purchase tobacco products means that perceived, calculable benefits flow to individuals as they do from tobacco production and sale. Any such benefits are totally ignored in SAMMEC II's calculations, however, further vitiating their implied warranty as to having produced a scientific impact study. While the exclusion of benefits is a critical omission in SAMMEC II, our critique evaluates SAMMEC II chiefly on its own grounds -- that is, by its assumption that the "social costs" identified do actually exist. Within the framework of the SAMMEC II estimates, three broad categories of "social costs" for smoking are said to exist: (1) indirect mortality costs consisting of lost income resulting from "premature" deaths; (2) direct morbidity costs, including all costs of hospitalization and health care due to illnesses which are claimed to be associated with smoking; and (3) indirect morbidity costs which encompass productivity losses due to worker absenteeism. At the heart of all of these tripartite cost estimates is what SAMMEC II developers call the Smoking Attributable Fraction (SAF) which is a function of: (a) smoking prevalence and (b) "risk ratios." Even assuming that smoking prevalence data are accurate, the SAMMEC calculation of SAF is rendered unreliable by its calculation of risk obtained from a 1989 Surgeon General's Report (Reducing the Health Consequences of Smoking: 25 Years of Proaress. A Report of the Surgeon General, 1989) for the following two reasons: (1) Reported disease incidence among smokers and nonsmokers is based upon no other factors besides sex, age, and smoking history. These data cannot explain any systematic differences in disease incidence since it does not control for factors such as health history, prior medical care, exercise and food intake, or genetics that are all iv TIMN 445034

identified as significantly related to disease (see Attachment A, Section I); and (2) Reported differences in disease incidence between smokers and nonsmokers are biased since data come from death certificates on which physicians systematically bias reportage (see text, p. 22). This means that the smoking-attributable impact estimates are calculated by mathematical formulae and are not based on actual disease incidence or actual costs. While these problems riddle the entire SAMMEC II study, other key difficulties pertain to specific calculations. Three of the most significant include: (1) The use of mean income to estimate claimed future losses in calculating indirect mortality and morbidity costs. Assuming any losses to be social (which they are not), a mean income calculation must bias the result since a higher percentage of smokers have been widely reported to be found in the lower-income groups (see Attachment A, Section II); (2) The lack of data on health care expenditures which forces SAMMEC II to treat all forms of health care as perfect complements. Where statistics do not exist -- as in nursing home care -- SAMMEC II makes apparently unsupported assumptions and uses a proxy. The application of meaningless disease incidence data (see 1 above) to several biased health cost data creates multiplicative errors in SAMMEC II estimates; and (3) The calculation of indirect morbidity costs also is unreliable and useless since it is based on relative differences in absenteeism rates between smokers and nonsmokers that do not account for age, health, worker status, or income (see Attachment A, Section V). These problems mean that, evaluated on its own, cost evaluations based on SAMMEC II software are unreliable. This conclusion applies to the Sullivan Report and to state (and city) studies of the costs of smoking and to any other investigation that uses the SAMMEC II program as its foundation. When the data show what the SAMMEC II authors expect, they accept it; v TIMN 445035

when the data fail to show what they expect, they make ad hoc adjustments to the data. This makes a mockery of scientific inquiry. Finally, we note difficulties in SAMMEC II's cost calculations which are central to our evaluation and critique and which, incredibly, are recognized by SAMMEC II developers as well. Apparently, recognizing the scientifically inconclusive nature of their own methodology, SAMMEC II authors note that the technology implied by SAMMEC II software "enhances the credibility and perceived authority of disease impact estimates" while on the same page of the studv (Schultz, Novotny and Rice, p. 9) they admit that ". . . the issue of whether smokers cost nonsmokers money, or vice versa, is unresolved." Earlier, on page 8 of their study, the authors argue that the net economic effect of cigarette smoking is "speculative." The dangers inherent in such "politically correct science" should be obvious to all. This manuscript was produced under a grant from the Tobacco Institute. The views expressed are those of the authors and not necessarily those of the Institute or its member companies. vi TIMN 445036

I. INTRODUCTION Politics and science are more interrelated now than at any time in human history. A vast arsenal of modem science is directed and totally or partially funded by local, state, and federal s governments and millions of small donations. Much of this so-called scientific activity comes from not-for-profit entities such as the American Lung Association, the American Heart Association, the National Institutes of Health (NIH), the Centers for Disease Control, 50 state departments of health and public health services, and myriad local health organizations. Obviously, the profit motive is the driving force for profit-maximizing entities while non-pecuniary (for example, "religious") motives may explain organizational characteristics of purely voluntary activity. The behavior of not-for-profit entities -- like those enumerated above which pay salaries and benefits, and are directly financed by government or rely on government subsidies and donations -- are the most difficult and interesting to analyze. These entities are widely perceived to be in the public interest. However, they may actually lack clearly identified ownership and control, providing management and employees with less incentives to use resources in an economical, socially desirable manner.' One of the most important aspects of such ill-defined accountability assignments in "goodwill entities" is the encouragement of such entities and organizations to use or mold science into politically popular (and politically "sellable") projects and activities. Science is then used to expand "expense preference behavior" -- expanded bureaucracies, higher managerial and 1 TIMN 445037

employee salaries, and so on -- and to form the basis for direct political activities by the goodwill entities themselves or by the politicians that support and nurture them. These activities may be at the expense of science and at the cost of the public interest. Indeed, the aims of science and economic calculation may be twisted to political ends. One example of such questionable activity is the development, dissemination and spreading use of a spreadsheet software that purportedly estimates the so-called "social cost" of smoking in a population. The software, which operates within Lotus 1-2-3, is called the Smoking-Attributable Mortality, Morbidity, and Economic Costs (SAMMEC II) Computer Software and Documentation, authored by project staff James M. Schultz, Thomas E. Novotny, and Dorothy P. Rice (henceforth SNR).2 SAMMEC II and periodic updates of the program are maintained by the Office on Smoking and Health of the U. S. Department of Health and Human Services. The central purpose of this research is to critically evaluate the scientific and economic foundations of SAMMEC II methodology. We explore both conceptual issues and methods of calculation which show that the developers and users of the software appear to be virtually inventing an economic "cost" of smoking -- and are developing exaggerated estimates of the so-called cost. After a more detailed discussion of SAMMEC II and what it is used for in Section II, we will evaluate the methodology of SAMMEC II in Section III, present conceptual and calculation biases in Section IV, and estimate the size of these biases and draw conclusions concerning the role of science in politics and the role of politics in science in a concluding section (Section V). 2 TIMN 445038

II. SAMMEC AND ECONOMIC COST COMPUTATIONS Since SAMMEC II has been recently developed, it is important to understand exactly what the software contains and how it has been used. More importantly, perhaps, some fundamental assumptions about the calculation of economic costs attend SAMMEC II and virtually all health cost assessment programs. While these assumptions -- relating to the economic concept of costs -- are fairly well known, an examination of them provides a foundation for analyzing the SAMMEC II methodology. What is SAMMEC II? SAMMEC II employs mortality statistics, economic "cost" data, and smoking prevalence estimates to calculate the alleged disease impact of smoking in "large" populations to make calculations specific to these populations. While large municipalities might be able to use the software, SAMMEC developers stress that the technology is primarily intended for use in states since "small communities and counties will produce unreliable estimates of disease impact if their populations are less than a few hundred thousand persons" (SNR, SAMMEC, p. 1). Similar caveats are issued by the project staff for developing nations or countries with non-representative smoking prevalence data, poor quality mortality data, and economic data "not similar to the United States." 3 TIMN 445039

The authors also issue warnings about data intervals in the program. Since SAMMEC II's smoking-attributable disease impact estimates are not a "surveillance system," they warn that data intervals should be large (3 to 5 years) because changes in disease incidence and in spending for diseases which are claimed to be associated with smoking are calculated (that is, by a mathematical formula) and are not based on any actual disease incidence or actu costs. Thus, users of the software (states and their public health departments, primarily) must supply raw mortality data for the most recent year available by disease category, five-year age group, and sex. Users must also supply current and former smoking prevalence data, either developed in their own state health departments or from national surveys that include state-specific information. SAMMEC II supplies (see section IV of the SAMMEC manual: Schultz et al. 1990) data by appropriate age group, smoking status, and sex from the 1985 Current Population Survey and from the 1988 Behavioral Risk Factor Surveillance System (BRFSS). SAMMEC II also provides users an option on cost data: provide personal health-care cost data (from states or local sources) or use SAMMEC's raw economic cost estimates for the states. (Cost data is provided for 1987, but users are invited to adjust it by multiplying the state per-capita cost in 1987 by population in later years.) How SAMMEC Has Been Used The most publicized use of the SAMMEC methodology has been the so-called Sullivan Report which was splashed across the pages of virtually all U. S. newspapers in 1990. According to Secretary of Health and Human Services Louis W. Sullivan in his second National Status Report on Smoking and Health to Congress (1990), "Each and every American, including 4 TIMN 445040

those who don't even smoke, is paying a hidden tax of approximately $221 per person per year for the consequences of smoking -- a tax that adds up to $52 billion annuallx. "3 Use of the SAMMEC II program led to the following breakdown of cost estimates in the Sullivan Report: DIRECT MEDICAL COSTS (hospitalization, physician's services, etc.) $23.7 billion INDIRECT MORBIDITY COSTS (essentially an economic "cost" of reduced labor productivity and increased absence) $10.2 billion INDIRECT MORTALITY COSTS (forfeited future earnings assuming "premature" deaths of smokers) $18.5 billion TOTAL COSTS $52.4 billion These numbers were developed as a direct application of the SAMMEC methodology. This means that the government relied on historical records on illness and mortality developed for SAMMEC and computed gross differentials between smokers and nonsmokers. These differences were then used as the basis for computations of costs. Computational adequacy of any and all estimates depends critically upon the methods by which estimates are developed and on the quality of the data employed. Of even more significance is the question of the conceptual logic that undergirds the entire SAMMEC-software-Sullivan Report output. Economic Project Evaluation and the Concept of Cost An immediate and critical issue is raised by the Sullivan Report estimates (based on SAMMEC II): correctly interpreted, does the assumption that P--y "social costs" exist with respect to smoking have actual or logical merit? A related question is whether SAMMEC should 5 TIMN 445041

be examined and exposed for the scientific leaps and lapses it contains given that the answer to the first question may be "no"? First, consider the issue of social and private costs, and the related issue of cost-benefit calculations. A private cost is one that is borne directly by the choosing individual. Benefits and costs are attributable to these rational choices. "Social costs" are those that are imposed on other members of society by the actions of individuals. In accepted economic analysis, all of the "costs" contemplated by SAMMEC II and those appearing in the Sullivan Report (discussed above) are private in nature. Costs and benefits, if any, of smoking behavior are all borne by the individual. Any reduced income from medical expenses or reduced labor input or productivity would all be experienced by the individual who smokes. Direct mortality costs -- those due to forfeited future earnings due to alleged "premature" death -- are all paid by the individual who dies. In spite of Sullivan Report claims to the contrary, all insurance costs and risk costs the Report associates with smoking are individual costs and not "social costs." The SAMMEC/Sullivan anti-smoking activists admit that the issue of whether [in the aggregate] smoking creates net costs (presumably, after benefits are taken into account) is unresolved (SNR, SAMMEC, p. 9), but they persist in discussing "social costs" from smoking behavior where no "social costs" appear to exist.4 For example, consider the logic used in calculating full foregone income from "premature" death which SAMMEC developers treat as a cost. They argue that "the individual, not just the output he or she contributes in excess of consumption, is valued by society. Economists today generally agree that consumption should not be deducted" (SNR p. 13) and they cite Mishan (1971). But rational individuals, in calculating risks, assess not only their own consumption foregone, but that of the foregone 6 TIMN 445042

savings. This does not mean that benefits would not have flowed to "heirs and assigns" but that all property rights and property dispositions belong to the individual alone in voluntary exchange. Here Mishan's actual argument (not one attributed to him by SAMMEC II developers) is valid: "Insofar ... as additional risks associated with the service or facility are all voluntarily assumed, there is no call for intervention in the allocative solution to which the market tends. As for project evaluations, insofar as benefits are calculated by reference to estimates of consumers' surplus, no allowance need be made for additional risk of loss of life. For the sum each person is willing to pay for the services provided by the project is net of all the risks associated with them." (1971, p. 698) The quotation from Mishan raises yet another critical and related issue for the fundamental adequacy of the SAMMEC/Sullivan study -- the fact that no benefits are assumed to flow to the individual or societ,y from smoking.5 While we assess the SAMMEC study on its own grounds (no benefits), it is worthwhile considering the conceptual inadequacy of such a project. With respect to (perceived) "social costs" only, SAMMEC inventors admit that "the net economic effect of cigarette smoking in future scenarios is speculative" (SNR, p. 8) but, simultaneously, they refuse to consider any possible benefits from smoking. From an economic perspective, there are, of course, enormous benefits from tobacco production and sale. All manner of production and marketing inputs are employed as billions of dollars in income, jobs, and taxes are generated. But benefits from tobacco (excluding taxes) even considering~its, association with certain health conditions flow to individuals. Again, more careful attention to Mishan (1971), who is completely misinterpreted by SAMMEC developers, would have given some scientific validity to the software. Mishan, in a provocative discussion, emphasizes a net benefit to individuals from smoking. Within a framework of Pareto estimation in the presence 7 TIMN 445043

of any claimed risk, Mishan emphasizes that only involuntarv risks need be counted and he uses tobacco as an example: "If [hypothetically] smoking tobacco causes 20,000 deaths a year, no subtracting from the benefits, on account of this risk, need be entered in a cost-benefit analysis of the tobacco industry inasmuch as smokers are already aware that the tobacco habit is [assertedly] unhealthy. And if, notwithstanding their awareness, they continue to smoke, the economist has no choice but to assume that they consider themselves better off despite the risks" (p. 996). Mishan goes on to say that the benefits to smokers, net of any risk are reflected in the demand schedule for tobacco. He notes that once the area under the demand curve has been estimated and used as an approximation of the benefit smokers derive from the use of tobacco, any further subtraction for such claimed risks would entail "double counting" (p. 996). There are, in short, two good reasons why one might consider the SAMMEC software to be of questionable scientific value and unreliable at the outset: (1) SAMMEC makes no distinction between private and "social" costs and, if it did, none of the costs supposedly it attributes to smoking could be found to be "social" in nature; and (2) Any social and individual benefits from smoking and tobacco production and sale are assumed to be non-existent. This omission means that no cost/benefit study of the commonly-accepted type -- where benefits to individuals and society are included as in product or project evaluations -- is being conducted. These considerations raise an important question: if, logically and in generally accepted practice and analysis, there are no "social costs" due to smoking, should SAMMEC methods and software be analyzed for the conceptual errors it contains, based upon SAMMEC II's assumptions that "social costs" exist? We believe that the answer is yes. Even if policymakers at state and other levels believe that smoking creates "social costs," they and the community at 8 TIMN 445044

large should be aware that SAMMEC-based computations are unreliable. Such an exposure shows just how far organizations will go in order to further policy conclusions. Further, it is noteworthy that SAMMEC developers openly invite policymakers or state or not-for-profit ("goodwill") organizations to employ the software results "packaged in understandable and concise terms" to influence politics in a community or state. As noted above, SAMMEC II users are exhorted to explicitly eschew any possible benefits from smoking or from tobacco production and sale. Specifically, SAMMEC II states that "the smoking-attributable economic cost data produced by SAMMEC II should not be used in cost-benefit analysis of the value of tobacco versus illness in society." Presumably, this is because the health community (federal or state agencies and goodwill, not-for-profit entities) might lose some support, financial and otherwise, if benefits and actual costs were fairly considered. Scientific objectivity appears to be a secondary concern to SAMMEC developers. As they readily admit, "disease impact estimates are a form of argument and have a political as well as a scientific reality." The technology implied by SAMMEC II software "enhances the credibility and perceived authority of disease impact estimates" (SNR, p. 9). Incredibly, on the same page (SNR, p. 9), the authors admit that ". .. the issue of whether smokers cost nonsmokers money, or vice versa, is unresolved." It is our purpose, in what follows, to investigate the basis for SAMMEC II's estimate of "scientific reality." 9 TIMN 445045

III. THE METHODOLOGY OF SAMMEC II In attempting to quantify the "costs" of smoking, SAMMEC II (Smoking-Attributable Mortality, Morbidity, and Economic Costs) identifies three categories of costs: (1) indirect mortality costs, (2) direct morbidity costs, and (3) indirect morbidity costs. In the view of SAMMEC II, the indirect mortality costs of smoking are the lost earnings which result from "premature" death which SAMMEC II attributes to smoking. The second category of costs, direct morbidity costs, are the costs of prevention, detection, treatment, and rehabilitation of diseases claimed to be linked to smoking. Finally, the indirect morbidity costs are the productivity losses due to illness and disability which they attribute to smoking. SAMMEC II treats the sum of these three components as the economic "costs" of smoking. To generate estimates of these estimates, SAMMEC II identifies certain illnesses which are claimed to be smoking-related. This list is summarized in Table 1 (SNR Table 6A, p. 24) and is compiled from illnesses identified by the American Cancer Society plus certain perinatal conditions identified by McIntosh (McIntosh, 1984, pp. 141-148). Users of SAMMEC II can base their cost estimates on this list of illnesses, or they can add additional illnesses at their discretion. 10 TIMN 445046

Table 1 Diagnosis Set SANIlVIEC II Software ICD-9-CM Code I. Adults: Male and Female; Ages 35-85 + (5-year age groups) Infectious Diseases 010-012 Neo lasms --~ - - 140-149 150 157 161 162 180 188 189 Cardiovascular diseases 410-414 380-398, 401-404, 415-417, 420-429 430-438 440-448 Respiratory Diseases 490-492,496 Respiratory tuberculosis Diagnoses Lip, oral cavity, pharynx Esophagus Pancreas Larynx Trachea. lung, bronchus Cervix uteri Urinary bladder Kidney, other urinary Coronary heart disease Other heart disease Cerebrovascular disease Other arterial disease Chronic obstructive pulmonary disease Other respiratory disease 890-899 Burn deaths 11. Children: Male and Female; Age: < 1 Perinatal Conditions 765 Short estation-low birth wei ht 769 Respiratory distress syndrome 770 Res iratorv conditions of the newborn Si ns and Svm toms 798.0 Sudden Infant Death Syndrome 11 480-487, 493 Injuries TIMN 445047

SAMMEC II is designed to facilitate attempts to quantify the costs of smoking for a particular city, county or state. Therefore, it enables users to supply data which are particular to the area under study. In particular, the user must supply: (1) mortality statistics -- deaths from the purportedly smoking-related illnesses identified in Table 1 by gender and five-year age category for ages 35-85+, (2) smoking prevalence rates -- the percentage of the population who are current and former smokers for ages 35-64 and 65+, (3) population data -- the number of persons in the group under study by gender and five-year category for ages 35-85+, (4) health care costs -- total personal health care expenditures for hospitalization, physician fees, medications, nursing home costs, and other professional services, and (5) earnings data. In addition, the user has the option of providing several other types of data relating to relative risks, life expectancy, earnings, and labor force participation. In cases where this optional data are not provided by the user, SAMMEC II provides the relevant data on a national and state-by-state basis and allows the user to select the data set on which cost estimates are to be based. Indirect Mortalitv Costs In the view of SAMMEC II, the indirect mortality cost of smoking consists of the lost income due to "premature" death. In order to generate such estimates, SAMMEC II first attempts to determine the number of deaths assertedly due to smoking. The critical element, upon which SAMMEC II's estimates rest, is the so-called smoking-attributable fraction (SAF). The SAF is defined as the "maximal proportion" of deaths (or disease cases) that purportedly SAMMEC II regards as linked to cigarette smoking. SAF is defined as: 12 TIMN 445048

SAF =(Po + Pi(RRi) + Pz(RR2) - 1)/(Po + Pt(RRt) + pARR,)) where: Po = percent of never smokers in the group under study p, = percent of current smokers in the group under study P2 RRl percent of former smokers in the group under study relative risk of death for current smokers compared with never smokers RR2 = relative risk of death for former smokers compared with never smokers The relative risks (RR, and RR,) are defined as follows: RRt = mortality rate in current smokers/mortality rate in never smokers RR2 = mortality rate in former smokers/mortality rate in never smokers The relative risk estimates which are used by SAMMEC II are reported in Table 2 (SNR Table 2 is also Table 7A, p. 26). They are obtained from the 1989 Surgeon General's Report. SAMMEC II also provides smoking prevalence rates which are reported in Table 3 (SNR, Table 7B, p. 27). As an alternative, the user can supply smoking prevalence rates which are specific to the group under study. SAF is therefore to be interpreted as the proportion of all deaths (or illnesses) that SAMMEC II attribute to smoking. It is fairly obvious that the SAF is a positive function of both relative risks and of smoking prevalence rates. Therefore, for males under 65, the SAMMEC II estimate of the SAF for neoplasms of the trachea, lung and bronchus is .91, while the corresponding SAF for respiratory tuberculosis is .35. According to SAMMEC II, 91 percent of all deaths due to cancer of the lung, trachea, or bronchus may be 13 TMN 445049

Table 2 Relative Risk Estimates SAMMEC II Software ICD-9-CM NEOPLASMS Diagnoses Current Males Former Females Current Former 140-149 Lip, Oral Cavity, Pharvnx 27.48 8.80 5.59 2.88 150 Esophagus 7.60 5.83 10.25 3.16 157 Pancreas 2.14 1.12 2.33 1.78 161 Larynx 10.48 5.24 17.78 11.88 162 Trachea. Lung, Bronchus 22.36 9.36 11.94 4.69 180 Cervix Uteri NA NA 2.14 1.94 188 Urinarv Bladder 2.86 1.90 2.58 1.85 139 Kidnev, Other Urinarv 2.95 1.95 1.41 1.16 CARDIOVASCULAR DISEASES 390-398 Rheumatic Heart Disease 1.85 1.32 1.69 1.16 401-404 Hypertension 1.85 1.32 1.69 1.16 410-414 [schemic Heart Disease Ages 35-64 2.81 1.75 3.00 1.43 Ages 65+ 1.62 1.29 1.60 1.29 415-417 Pulmonary Heart Disease 1.85 1.32 1.69 1.16 420-429 Cardiac Arrest/Other Heart 1.85 1.32 1.69 1.16 430-438 Cerebrovascular Disease Aees 35-64 3.97 1.38 4.80 1.41 I Ages_65 + 1.94 1.27 1.47 1.01 440 Atherosclerosis 4.06 2.33 3.00 1.34 441 Aortic Aneurvsm 4.06 2.33 3.00 1.34 442-448 Other Arterial Disease 4.06 2.33 3.00 1.34 RESPIRATORY DISEASES 010-012 Respiratory Tuberculosis 1.99 1.56 2.18 1.38 480-487 Pneumonia. Influenza 1.99 1.56 2.18 1.38 490-492 Bronchitis. Emphysema 9.65 8.75 10.47 7.04 493 Asthma 1.99 1.56 2.18 1.38 496 Chronic Airwavs Obstruction 9.65 8.75 10.47 7.04 PERINATAL CONDITIONS' 765 Short Gestation/Low Birth Weight 1.76 1.76 769 R irato Distress Syndrome 1.76 1.76 770 Re iracorv Conditions-Newborn 1.76 1.76 798.0 Sudden Infant Death S ndrome 1.50 1.50 OTHER CONDIT IONS 890-899 Burn Deaths * Perinatal conditions: deaths among infants < 1 year. ** Burn deaths determined from injury surveillance studies. 14

Table 3 Smoking Prevalence Rates United States, All Races, 1987 Males Females Age Group Current Former Current Former 35-64 34.8 34.9 29.0 20.2 65 and over 17.2 53.4 13.7 19.8 Source: 1987 National Health Interview Survey (OSH. unpublished data). attributable to smoking, while 35 percent of tuberculosis deaths may be linked to smoking. This difference is due to the much smaller relative risks that are assigned to tuberculosis. Correspondingly, the SAMMEC II SAF for females under 65 for hypertension (0.17) is lower than the corresponding male SAF (0.29) primarily due to the lower smoking prevalence rates for females. (For this illness, the estimated relative risks provided by SAMMEC II are nearly equal.) The SAF's then become an important ingredient in calculating the number of deaths that SAMMEC II attributes to smoking. For each such illness (say, hypertension), deaths (by gender) occurring within a five-year category are multiplied by the SAF for that diagnosis, age category, and gender: SMOKING-ATTRIBUTABLE MORTALITY = DEATHS x SAF. For example, if the population in question experiences 1,000 male deaths in a year due to hypertension, 290 of them (1,000 x 0.29) would be attributed to smoking by SAMMEC II. If 100 deaths occurred in males age 45-49, then 29 of such deaths would be attributed to smoking. 15 TIMN 445051

SAMMEC II also purports to calculate the number of years of potential life assertedly due to smoking. For each five-year age category, SAMMEC provides the number of years to age 65, age 75, or to full life expectancy. The separate tables of life expectancy used for males and females are reported in Table 4 (SNR Table 9B) and Table 5(SNR Table 9C). SAMMEC II also translates the death statistics into claimed future earnings losses. For each person whose "premature" death SAMMEC II attributes to smoking, a future stream of earnings is calculated assuming a one percent annual increase in productivity and an average annual inflation rate. The present value of future earnings (PVFE) is calculated the same way, but with future earnings discounted to present value. SAMMEC II allows the user to select discount rates of 4 percent, 6 percent, or 10 percent. Therefore, SMOKING-ATTRIBUTABLE INDIRECT MORTALITY COSTS = DEATHS X SAF X PVFE. Tables 6, 7, and 8 report present values of future incomes for workers in different age groups for each of the three discount rates respectively. Direct Morbiditv Costs In the SAMMEC II methodology, the direct morbidity costs assigned to smoking consist of the expenditures for the prevention, diagnosis, and treatment of smoking-related diseases and medical conditions claimed to be associated with smoking. To calculate these supposed costs, SAMMEC II attempts to determine the portion of those costs which it attributes to smoking. To do so, it assigns all such costs to one of five distinct cost centers. Specifically, all the costs are classified as: (1) hospitalization and outpatient clinical care 16 TIMN 445052

Table 4 Years of Potential Life Remaining Relative to Life Expectancy: Males United States, 1985 Age of Death All Races Whites Total Non-white Blacks 0-4 70.1 70.7 60.5 64.7 5-9 65.3 65.3 61.7 59.9 10-14 60.3 60.9 56.8 55.8 15-19 55.5 56.1 51.9 50.2 20-24 50.9 51.4 47.3 45.6 25-29 46.3 46.3 42.8 41.2 30-34 41.7 47.2 38.4 36.3 35-39 37.1 37.5 34.1 32.6 40-44 32.5 32.9 30.0 28.6 45-49 28.1 28.4 26.0 24.7 50-54 23.9 24.2 22.2 21.1 55-59 20.1 20.2 18.8 17.8 60-64 16.5 16.6 15.7 14.9 65-69 13.3 13.3 13.0 12.3 70-74 10.5 10.5 10.5 9.9 75-79 8.1 8.0 8.3 7.9 80-84 6.1 6.0 6.5 6.2 85 + 5.1 5.1 5.9 5.7 Source: National Center for Health Sciences, 1988. 17 TIMN 445053

Table 5 Years of Potential Life Remaining Relative to Life Expectancy: Females United States, 1985 Age of Death All Races Whites Total Non-White Blacks 0-4 77.0 77.4 74.1 72.3 5-9 72.1 72.5 69.3 68.0 10-14 67.2 67.6 64.4 63.1 15-19 62.3 62.7 59.5 58.2 20-24 57.4 57.8 54.6 53.4 25-29 52.6 53.0 49.8 48.6 30-34 47.7 48.1 45.1 43.9 35-39 42.9 43.4 40.4 39.3 40-44 38.2 38.5 35.9 34.8 45-49 33.6 33.9 31.5 30.4 50-54 29.1 29.4 27.3 26.3 55-59 24.9 25.0 23.3 22.5 60-64 20.8 21.0 19.7 19.0 65-69 17.1 17.2 16.3 15.7 70-74 13.6 13.6 13.2 12.7 75-79 10.5 10.5 10.4 10.8 80-84 7.7 7.7 8.0 7.8 85 + 6.4 6.4 7.0 6.9 Source: National Center for Health Sciences, 1988. 18 TIMN 445054

Table 6 Present Value of Future Earnings (1985 -- Discount Rate: 4%) Age Males Females < 1 $421,225 $341,574 1-4 454,561 368,388 5-9 519,459 420,790 10-14 602,092 487,557 15-19 689,576 552,141 20-24 745,680 578,481 25-29 749,695 558,019 30-34 717,630 513,796 35-39 653,498 454,897 40-44 561,016 388,555 45-49 450,452 319,279 50-54 331,478 249,422 55-59 223,719 181,151 60-64 108,880 117,333 65-69 42,879 67,346 70-74 19,176 36,593 75-79 9,383 18,847 80-84 4,698 9,164 85+ 1,442 2,311 Source: R...., K..... and D....., 1988. 19 TIMN 445055

Table 7 Present Value of Future Earnings (1985 -- Discount Rate: 6%) Age Males Females < 1 $208,631 $173,738 1-4 236,117 196,515 5-9 293,977 244,559 10-14 374,790 311,678 15-19 468,782 384,026 20-24 541,021 425,804 25-29 568,546 424,982 30-34 565,043 402,176 35-39 532,289 364,873 40-44 471,190 319,090 45-49 389,462 268,529 50-54 294,646 214,826 55-59 194,878 159,614 60-64 101,085 105,272 65-69 39,713 61,103 70-74 17,802 33,574 75-79 8,789 17,531 80-84 4,457 8,655 85 + 1,406 2,257 Source: R... , K... and D..... 1988. 20 TIMN 445056

Table 8 Present Value of Future Earnings (1985 -- Discount Rate: 10%) Age Males Females < 1 $ 60,306 $ 53,331 1-4 75,494 66,797 5-9 111,042 98,205 10-14 170,371 150,622 15-19 251,439 217,161 20-24 324,215 264,945 25-29 363,750 277,578 30-34 381,607 272,561 35-39 377,822 255,422 40-44 350,066 230,617 45-49 302,591 200,822 50-54 239,410 166,546 55-59 165,416 128,340 60-64 38,714 87,226 65-69 34,679 51,510 70-74 15,581 28,792 75-79 7,802 15,372 80-84 4,837 7,773 85 + 1,344 2,155 Source: R..., K... and D..., 1988. 21 ,rIMN 445057

costs, (2) nursing home care and home health care costs, (3) the costs of the services of primary physicians and specialists, (4) the costs of services of non-physician health practitioners, and (5) medication costs. However, there are no direci data on the relative rates of utilization for each of these categories of costs by smokers and former smokers compared to those who have never smoked. To solve this problem, SAMMEC II uses survey data from the National Health Interview Survey. Only persons 35 years of age and older who had conditions which are considered by SAMMEC II to be smoking-related were included. From this sample, the relative rates of utilization for smokers and for former smokers relative to nonsmokers were computed. These calculations were made both for usage of hospitalization and for usage of physician services. The relative usage rates are reported in Table 9 (SNR, Table 11A, p. 41). Since no reliable data on relative usage rates for nursing home services are presented -- for care by non-physician health professionals, or for medication -- the assumption was made by SAMMEC II authors that the relative rates of usage that pertain to physician services hold as well for medication. Similarly, the usage rates of nursing home care and of care by non-physician professionals is assumed to be identical to that for hospitalization. From these relative rates of utilization, the program again calculates smoking-attributable fractions (SAF's) which are similar to those discussed above. Namely: SAF =(Po + Pi(RRi) + Pz(RRa) - 1)/(Po + Pi(RRt) + P2(RRz)) where po, pt, and P2 are defined as before but where RR, and RR2 differ slightly. Here, RR, is the relative rate of utilization of health care for current smokers compared with never smokers, and RR, is similarly defined for former smokers. The SAF's used to estimate the 22 TIMN 445058

Table 9 Relative Rates for Hospital Days Per 100 Persons and Physician Visits Per 100 Persons By Smoking Status Males Females Both Sexes 35-64 65- 35-64 T 65- 35-64 65- Hospital Days Per 100 Persons Never Smokers 42.6 155.9 64.1 125.0 56.5 133.5 Former Smokers 82.0 191.5 73.6 168.3 78.8 186.6 Current Smokers 84.3 183.9 69.7 76.2 77.3 127.7 RR (current) 1.98 1.18 1.09 0.61 1.37 0.96 RR (former) 1.93 1.23 1.15 1.35 1.37 1.40 Physician Visits Per 100 Persons Never Smokers 323.7 454.2 421.2 583.4 386.8 547.8 Former Smokers 370.5 501.4 529.3 671.1 432.0 559.1 Current 377.2 456.0 463.8 428.0 418.4 441.4 RR (current) 1.17 1.00 1.10 0.74 1.08 0.81 RR (former) 1.14 1.10 1.26 1.15 1.12 1.02 direct morbidity costs which SAMMEC II attributes to smoking differ across cost centers, age groups, and sexes. As is evident in Table 9, SAMMEC II reports that certain groups of smokers use less hospitalization and physician care than do corresponding groups of nonsmokers. However, in calculating the SAFs, SAMMEC II assigns a relative risk of 1.0 to these cases on the 23 TIMN 445059

grounds that ". . . cigarette smoking would never be protective for health care utilization." (SNR, SAMMEC II, Module 2, p. 40.) Indirect Morbidity Costs The final step in the calculation is to determine the costs of treating the population in question for illnesses which are claimed to be associated with smoking and then multiply by the appropriate SAF. The final category of costs which SAMMEC II estimates is the productivity loss due to disability and illness assertedly caused by smoking-related diseases. The methodology used is essentially the same as that used to compute indirect morbidity costs and direct mortality costs. Specifically, SMOKING-ATTRIBUTABLE INDIRECT MORBIDITY COSTS = DISABILITY DAYS x SAF x MEAN PER CAPITA INCOME This calculation is gender-specific. The number of disability days is the total work-loss days for employed persons and bed-disability days for non-employed homeworkers which SAMMEC II attributes to smoking. The SAFs for work-loss days were calculated by: (1) The work-loss days per 100 currently employed persons and bed-disability days per 100 females keeping house (combined) were calculated for current smokers, former smokers, and never smokers who responded to the 1987 NHIS. Only persons 35 and older, who reported illness due to the diseases SAMMEC II treats as smoking-related, were included in the calculation. 24 TIMN 445060

(2) The number of persons unable to work or keep house per 1,000 persons were calculated for current smokers, former smokers, and never smokers who were respondents to the 1987 NHIS. (3) Relative rates of disability for current smokers and former smokers compared with never smokers were calculated. The calculations are reported in Table 10 (SNR, Table 12A, p. 45). Table 10 Relative Rates for Work-Loss Days Per 100 Employed Persons Per Bed-Disability Days Per 100 Females Keeping House By Smoking Status Males Females Both Sexes 35-64 65- 35-64 65- 35-64 65- Never Smokers 237.8 239.9 495.9 573.1 407.1 632.4 Former Smokers 287.1 517.5 542.2 785.5 392.2 691.0 Current Smokers 346.7 749.0 631.0 730.0 489.0 734.0 RR (current) 1.46 3.12 1.27 1.08 1.20 1.16 RR (former) 1.21 2.16 1.09 1.17 0.96 1.09 (4) The relative rates of disability are used with smoking prevalence rates to calculate the SAF of total disability days reported by persons with diseases SAMMEC II attributes to smoking. SAF was calculated by the following formula: P2 SAF =(Po + Pi(RRI) + Pa(RR,) - 1)/(Po + Pt(RRi) + Pa(RRz)) = percent of never smokers in the group under study = percent of current smokers in the group under study = percent of former smokers in the group under study 25 TIMN 445061

RR, = relative risk of disability for current smokers compared with never smokers RR, = relative risk of disability for former smokers compared with never smokers SAMMEC II computes per capita income by the following formula: Mean per capita income = (((wages + supplements) x (full-time employed persons)) + (.5 x (wages + supplements) x (part-time persons)))/(number of persons in the population) This approach treats each work-loss (or bed-disability) day as if it impacted a person with average income. SAMMEC II identifies three specific categories of costs which it attributes to smoking -- lost income due to "premature" death assertedly caused by smoking, lost income due to illness or disability, assertedly due to smoking, and the health care costs of disease and illness assertedly attributable to smoking. For each of these three categories of costs, SAMMEC II estimates a smoking attributable fraction (SAF) to measure the proportion of the costs which it attributes to smoking. By multiplying the SAFs by the total costs of illnesses which SAMMEC II treats as smoking-related, the program then computes the total so-called "costs" of smoking. 26 TIMN 445062

IV. TECHNICAL FLAWS IN SAMMEC II METHODOLOGY In Section II we examined many of the conceptual difficulties inherent in the traditional attempts to estimate the so-called costs of smoking. However, in addition to conceptual problems, SAMMEC II is seriously flawed by several important technical errors. As a consequence, even if one ignores the conceptual problems involved in SAMMEC II cost estimates, there are other good reasons.to challenge the conclusions of SAMMEC II. In the following three sections we will describe several of those flaws as they relate to each of the three main categories of costs considered in SAMMEC II. A major ingredient used in SAMMEC II to calculate the "costs" of smoking is the smoking attributable fraction which is a function of smoking prevalence and the risk ratio. In estimating the risk ratios, SAMMEC II simply computes the ratio of the incidence of "premature" death, disability, or illness for smokers and former smokers, comparing it to those who have never smoked. Separate relative risks are used for men and for women and for those over and under 65. However, what this means is that the confounding effects of some important factors other than sex, age, and smoking history on deaths and illnesses due to certain illnesses are totally ignored. For this methodology to be correct, one of two things must be true. Either it must be the case that these other factors (diet, exercise, heredity, environment, occupation, level 27 TIMN 445063

of health care) have no effect on the incidence of these diseases or that there are no systematic differences between smokers and former smokers compared to those who have never smoked with regard to important health factors. Obviously, neither of the two cases is credible. There appears to be overwhelming evidence that diet, exercise, and other factors are significantly related to certain of the diseases which SAMMEC II treats as smoking related (see Attachment A, Section 1). In addition, since smokers and nonsmokers differ markedly with regard to occupation, race, income, education level and so on, it is almost certain to be the case that they differ markedly with regard to many important health-related factors. Had these other factors been taken into account, it is quite likely different relative risks estimates would have emerged. It is quite possible, for example, that inclusion of these factors would have produced no significant differences between smokers, former smokers, and nonsmokers. Undoubtedly, the decision to ignore the effects of these important factors made it easier to produce estimates of the "costs" of smoking; this may, in fact, explain this omission. However, it is clear that this omission led to the capacity to make biased calculations. There is at least one additional reason that the relative risks reported in Table 3 lack credibility. Records pertaining to the cause of death are based on death certificates, where the cause of death is determined by a physician. However, when an autopsy is ordered, there reportedly are numerous discrepancies between the cause of death initially indicated and the actual cause determined by the autopsy.b C. Yoe has lots of support for the proposition that death certificates are quite unreliable, including smoking-specific entries. 28 TIMN 445064

Further, the discrepancies appear to be highly correlated with the patient's smoking history. Eysenck (1986) reports data which indicates that physicians systematically err by determining that a " smoking -related disease" is the cause of death in an excessive number of cases involving smokers. For nonsmokers, physicians are inclined to ignore diseases which they perceive as being "smoking-related" which leads them to under-assign such diseases as the cause of death in the case of nonsmokers. This bias could result in relative risk estimates in excess of 1.0 even if the true incidence of "smoking-related disease" is identical for smokers and nonsmokers (see Attachment A, Sections III and IV). In addition to the biases in estimating relative risks, SAMMEC II errs in its method of placing a dollar value on the "premature" deaths which it attributes to smoking. In calculating the present value of future earnings which are lost due to "premature" death, SAMMEC II assigns a mean future income to each person who dies "prematurely". The authors of SAMMEC II argue that: "The indirect costs for diseases that affect a population with above average productivity will be underestimated if mean earnings are used to measure foregone output" (SNR, 1990, p. 12). However, this also implies that the methodology will overestimate the costs of diseases which affect populations with below average incomes, -- smokers reportedly being one such group (see Attachment A, Section II). Since smoking-attributable fractions are an important element in SAMMEC II's method of estimating the excess health care costs which it attributes to smoking, the criticism described above pertains as well here. In computing relative rates of usage of hospital rooms and physician services, factors other than age, sex, and smoking history are ignored. Therefore, the relative risks (and therefore the smoking-attributable fraction) estimated here 29 TIMN 445065

are almost certain to be biased upward. In addition, there are two other serious problems with the relative risk estimates used to calculate the supposed "costs" of smoking. First, to calculate smoking attributable fractions to apply to each of the five categories of medical costs (hospitalization, nursing homes, physicians, non-physician health professionals, and medication), SAMMEC II requires estimates of the relative rates of usage (between smokers and nonsmokers) for each category of costs. However, no usage-specific data are presented to enable relative usage rates to be calculated for three of the five categories -- nursing home care, care by non-physician health professionals, and the costs of medication. SAMMEC II solves this problem by using estimates of the relative rates of usage for the other two cost categories and applying those relative rates to the three categories in question. Specifically, medication is assumed to be used at the same relative rates by smokers and nonsmokers as are the services of physicians; nursing home care and care by non-physician professionals are assumed to be used at the same relative rates as the costs of hospitalization. This methodology treats various types of health care as perfect complements. Specifically, since no specific nonduplicative data sets are available on nursing home vs. hospital care, nursing home care is treated as a perfect complement to hospitalization; treatment by non-physicians is treated as a perfect complement to treatment by physicians. However, it would seem that for many users the two types of care are substitutes -- not complements. What this implies is that in groups for whom use of hospitals is relatively high, the use of nursing homes would be relatively low. However, SAMMEC II methodology arbitrarily assigns the same relative rate to both nursing care use and hospital 30 ,rIMN 445066

use. For this reason, the direct morbidity costs estimated by SAMMEC II methodology are not credible. The relative risks used by SAMMEC II are flawed in a second serious way. As Table 9 shows, SAMMEC II reports that certain groups of smokers use hospitalization and physician services at a lower rate than corresponding groups of nonsmokers. For example, according to SAMMEC II for women over 65 who currently smoke, the rate of hospital use is only 61 percent as high as the rate of use by women over 65 who have never smoked. However, SAMMEC II arbitrarily assigns a relative rate of 1.0 to these cases. In other words, when the data show what the SAMMEC II authors expect, they accept it; when it doesn't, they reject the data out of hand. This procedure is clearly not a part of an objective scientific inquiry. Any unbiased estimating procedure will underestimate in some cases and overestimate in other cases. However, when attempts are made to adjust some estimates upward (on the grounds that SAMMEC II authors perceive them to be too low) without adjusting other estimates downward, a bias is introduced into the results. In this particular case, let's suppose the following -- that smokers and nonsmokers use hospital care at exactly the same rate. If this were the case, a large, random sample of smokers and nonsmokers will produce an estimate of relative rates of usage which is very close to 1.0. However, if the large sample is broken down into subcategories (men over 65, women over 65, etc.) the estimates of the relative rates for some sub-groups will turn out greater than 1.0 and others will be less than 1.0. The decision to adjust upward relative rates for sub-groups which are less than 1.0 only introduces a bias into the results. 31 TIMN 445067

This decision reveals SAMMEC II methodology for what it is -- politics based on questionable science. The authors apparently have decided what results they want, and they are willing to adjust calculations to ensure that they find what they are looking for. The most serious flaw in estimating indirect morbidity costs (lost income due to illness and disability) again pertains to the use of biased relative rates. As Table 10 shows, SAMMEC II uses relative rates which report about 30 percent more work-loss days for smokers than for nonsmokers. These rates are developed from productivity loss estimates by Rice et al. (1986), which were similar to those reported in Rice and Hodgson (1985). The latter study claims that smoking causes smokers to experience 32 percent more work loss than nonsmokers. SAMMEC II, in line with these studies and, indeed, employing their results, attributes the entire difference in absenteeism rates between smokers and nonsmokers to smoking. Research studies such as those by Leigh (1986) and Allen (1981a, 1981b), which do consider the effect of significant factors, fail to account for potential interdependencies between smoking characteristics and other behavior. Even the popular press links smoking with other lifestyle choices such as drinking, breakfast consumption, hours of sleep and exercise (see, for example, Otten 1988). (See also Attachment A, Section I.) The results obtained by SAMMEC II based on productivity studies of Rice et al., as a recent study shows, may be inaccurate and unscientific. These work-loss studies employ only a simple means difference test as a statistical basis for their deductions. Thus, they are incapable of determining whether absence rates are influenced by smoking since they cannot account for factors common to smokers as a group. However, a paper directly relevant to 32 TIMN 445068

V. SUMMARY AND CONCLUSIONS SAMMEC II software and the logic underpinning it are not simply faulty. As noted earlier in this paper, SAMMEC's political aims were to enlist government and "goodwill" not-for-profit entities to develop exaggerated and unreliable estimates of the "social costs" of smoking. SAMMEC II developers violated fundamental and long-recognized social and economic principles of cost-benefit analysis by loosely interpreting costs and then by complete elimination of benefits from the calculation. First, it is highly questionable to argue that any "social costs" could apply to smoking. If any such "costs" do exist, all consequences of individual behavior are p •nvate costs in any reasoned calculation of "costs." This means that if there are any lost earnings due to illness or "premature" death plus all medical expenses, they would be borne by and must be attributed to individuals. This principle has been given scientific status in all competent studies of costs and benefits. But even SAMMEC's authors readily admit that -- on SAMMEC's own grounds, i.e., on calculating costs as "social costs" -- any "social costs" of smoking may be non-existent or negligible when all costs of the health care system are taken into account. As they argue, the "net economic effect of cigarette smoking in future scenarios is speculative" and "movement toward a nonsmoking society may be recessionarv in the long term" (SNR, p. 8, [emphasis ours]). 34 TIMN 445069

In addition to the questionable attribution of costs, a second important flaw in the SAMMEC II study is that no benefits from tobacco are calculated either for individuals or in terms of the generation of economic growth. Since individuals exhibit a positive demand for cigarettes despite the claimed risks, there obviously are perceived net benefits attached to the purchase and use of cigarettes. In a broader context of costs and benefits, society benefits from a tobacco industry through direct and indirect employment of labor and other inputs and through the multiplier effects on income from tobacco production and sale. Any individual and social benefits are entirely absent from the SAMMEC II study. Beyond these serious conceptual problems and omissions, the quality and reliability of the SAMMEC II study may be analyzed on its own technical merit. Here, we argue in the present study, the problems inherent to the study should make any user or potential user totally skeptical of its value. As we have shown in the analysis of the SAMMEC II software, "seat-of-the-pants" result-driven estimates of risk -- the fundamental basis of all SAMMEC II calculations -- riddle the entire study of indirect mortality, direct morbidity and indirect morbidity "cost" estimates. We find that other problems are endemic to SAMMEC II taken on its own grounds as well. Specifically, there are five major problems in the SAMMEC II calculation: (1) Systematic differences in disease incidence between current and former smokers and nonsmokers cannot be reliably determined based upon the data used in the study. Since no factors other than sex, age and smoking history are used, the so-called disease differential provides insufficient and incomplete information, let alone sound estimates of differential disease incidence. Many factors (and most medical researchers would admit the fact) such as prior health care, exercise, and diet have been reported to be related to the incidence of disease (see Attachment A, Section I). It is elementary that failure to account for these factors must at best provide severely biased calculations of risk ratios in SAMMEC; 35 ,rgMN 445070

(2) The risk ratio may also be biased due to the inaccuracies of death certificates; (3) Economic calculations of future income losses =- another basis for estimating indirect mortality costs -- are biased on the upside, since a large percentage of past and present smokers reportedly are in lower-income groups (see Attachment A, Section II); (4) SAMMEC II, in the calculation of direct morbidity costs, treats all forms of health care as perfect complements. In fact, lack of data in this area seemingly was not a hindrance to utilization of questionable 'statistics by SAMMEC II inventors; and (5) Indirect morbidity costs also are calculated with clearly questionable estimates of worker absenteeism. The best data in this area show that there are no systemic differentials of worker "productivity" between smokers and nonsmokers (see Attachment A, Section V). A careful study of SAMMEC II produces only one conclusion: the study is so fundamentally flawed as to be unreliable for assessing the alleged "social costs" of smoking. 36 TIMN 445071

FOOTNOTES 'The standard property rights view of the not-for-profit firm predicts inefficient activity due to an inappropriate assignment of property rights (Alchian and Kessel, 1962). Examples of this kind of firm are the administratively regulated utilities. One might expect the og_odwill not-for-profit firm -- government supported or financed and/or supported by large numbers of small donations -- to engage in an even more attenuated manner. That is, one would expect to find: (a) unrestrained research director-managers to be distributing expenditures; (b) higher labor/capital ratios and lower output in such firms; and (c) inefficient use of huge quantities of largely donated resources. ZSchultz is a member of the department of epidemiology and public health at the University of Miami School of Medicine; Novotny, who holds an M.D., is employed in program services activity for the Office on Smoking and Health in the Public Health Service of the U. S. Department of Health and Human Services; Rice is with the School of Nursing at the University of California, San Francisco. SAMMEC, as a microcomputer software, was developed for the Minnesota Department of Health to calculate so-called smoking-attributable disease impacts for "local" (i.e:, non-national) populations. The software, initially created by Shultz in 1985 and 1986 (see Shultz, 1985, 1986a), was applied by him to New York City, (1986b), formed the 37 TIMN 445072

basis of a doctoral dissertation (Shultz, 1988a), and was refined into SAMMEC II at the Minnesota Department of Health in 1988 (Shultz, 1988b). 3The authors, perhaps in awareness of the political nature of the research presented in the Sullivan Report, issue a disclaimer buried deeply in the text: "They (the data) do not describe a net cost effect nor do they indicate the potential savings if tobacco use were eliminated in the United States" (National Status Report, 1990, p. 40). ' SAMMEC II, it should be noted, provides citations to literature that reports lower lifetime (aggregate) medical costs for smokers rather than nonsmokers, but subsequently ignores this literature. See, in particular Leu and Schaub, (1983); Schelling, (1987); and Warner (1987). SMishan's actual objection to using a net cost method of calculating the reduced income effects of loss of life is that a simple calculation of income minu consumption "has no regard for the feelings of the potential decedents. It restricts itself to the interest only of the surviving members of society; it ignores society ex ante and concentrates wholly on society ex post" (1971, p. 690). 6This is based on material from Eysenck, Hans J., "Smoking and Health," in Smokina and Societv, Robert Tollison, ed., D. C. Heath, 1986. 38 ,ygMN 445073

REFERENCES Alchian, A. and R. Kessel, "Competition, Monopoly, and the Pursuit of Money," in Aspects of Labor Economics, (Princeton: National Bureau of Economic Research, 1962). Allen, S. G., "Compensation, Safety, and Absenteeism: Evidence from the Paper Industry," Industry and Labor Relations Review, Vol. 34 (1981a), pp. 207-218. Allen, S. G., "An Empirical Model of Work Attendance," Review of Economics and Statistics, Vol. 63 (1981b), pp. 77-78. Ault, Richard W., and Ekelund, R. B., Jr., "A Preliminary Critique of the Sullivan Report," Manuscript, (1990). Ault, Richard W., Ekelund, R. B., Jr., Jackson, J. D., Saba, Richard S., Saurman, David S., "Smoking and Absenteeism," Applied Economics, Vol. 23 (April 1991), pp. 743-754. Blinder, A. S., "Wage Discrimination: Reduced Form and Structural Estimates," Journa of Human Resources, Vol. 11 (1973), pp. 9-22. Eysench, Hans J., "Smoking and Health," in Robert Tollison (ed.) Smoking and Societv, (Boston: D. C. Heath, 1986). Leu, Robert B. and Schaub, Thomas, "Does Smoking Increase Medical Care Expenditure?" Social Science and Medicine, Vol. 17 (1983), pp. 1907-1914. McIntosh, I. D., "Smoking and Pregnancy: Attributable Risks and Public Health Implications," Canadian Journal of Public Health, Vol. 75 (1984), pp. 141-148. Mishan, E. J., "Evaluation of Life and Limb: A Theoretical Approach," Journal of Political Economy, Vol. 79 (July/August 1971), pp. 687-705. Otten, A. "Smoker's Problems go well Beyond Smoking," USA Today (August 23, 1988), p. 29. 39 TIMN 445074

Rice, D. P., and Hodgson, T. A., "Economic Costs of Smoking: An Analysis of Data for the United States," Unpublished Paper Presented at the Allied Social Science Association Annual Meetings, San Francisco (December 1985). Rice, D. P., Hodgson, T. A., Sinsheimer, P., Browner, W., Kopstein, A. N., "The Economic Costs of the Health Effects of Smoking, 1984," Milbank Ouarterlx, Vol. 64 (1986), pp. 489-547. Schelling, T. C., "Economics and Cigarettes," Preventive Medicine, Vol 15 (1987), pp. 549-560. Shultz, J. M., Moen, M. E., Pechacek, T. F., et. al., "The Minnesota Plan for Nonsmoking and Health: The Legislative Experience," Journal of Public Health Policy 7 (1986a), pp. 300-313. Shultz, J. M., Smokina-Attributable Mortality. Morbidity. and Economic Costs - Comouter Software and Documentation, Center for Nonsmoking and Health, Minnesota Department of Health (Minneapolis, Minnesota, August, 1986b). Shultz, J. M., "New York City: Smoking-Attributable Mortality, Morbidity, and Economic Costs," Renort of the Mayor's Committee on Smoking and Health, Mayor's Committee on Smoking and Health, City of New York, (July 1, 1986c). Shultz, J. M., SAMMEC: Smoking-Attributable Mortality. Morbiditv. and Economic Costs (Computer Software and Documentation) Minnesota Center for Nonsmoking and Health, Minnesota Department of Health (Minneapolis, Minnesota, 1988). Shultz, J. M., Quantifying the Disease Impact of Cigarette Smoking: The Development and Application of Computer Software for Estimating the Health and Economic Costs of Smoking (Doctoral dissertation, University of Minnesota, June, 1988). Shultz, J. M., Novotny, Thomas E., and Rice, Dorothy P., SAMMEC II: Smoking-Attributable Mortality Morbidity and Economic Costs (Computer Software and Documentation), Office on Smoking and Health, U. S. Department of Health and Human Services (April, 1990). U. S. Department of Health and Human Services, Reducing the Health Consequences of Smoking• 25 Years of Progress A Report of the Surgeon General, U. S. Department of Health and Human Services, Public Health Service, Centers for Disease Control, Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health, DHHS Publication No. (CDC) 89-8411, (1989). U. S. Department of Health and Human Services, Smoking and Health: A National Status Report, 2nd edition (1990). 40 TIMN 445075

Warner, K. E., "Health and Economic Implications of a Tobacco-Free Society," Journal of the American Medical Association, Vol. 258 (1987), pp. 2080-2086. 41 TIMN 445076

ATTACHMENT A ADDITIONAL REFERENCE MATERIALS Section I There are several studies which suggest that there are other factors, in addition to smoking, which have been statistically associated with, for example, lung cancer and heart disease. Such factors include occupational and environmental exposures, lack of exercise, diet, and/or genetics. For details on these studies,, please refer to the following articles. • Blot, William J. and Fraumeni, Joseph F. Jr., "Arsenical Air Pollution and Lung Cancer," Epidemiology Branch, National Cancer Institute, Bethesda, Maryland 20014, from The Lancet, pp. 142-144, July 26, 1975. • Blot, William J. and Fraumeni, Joseph F. Jr., "Geographic Patterns of Lung Cancer: Industrial Correlations," (National Cancer Institute), NIH, A521 Landow Building, Bethesda, Maryland 20014, from American Journal of Epidemiology, Vol. 103, No. 6, pp. 539-550, 1976. • Minowa, M., Shigematsu, I.; Nagai, M.; and Fukutomi, K.; "Geographical Distribution of Lung Cancer Mortality and Environmental Factors in Japan," Department of Epidemiology and Department of Public Health Statistics, Institute of Public Health, Tokyo, Japan, from Soc. Sci. Med., Vol. 15D, pp. 225-231, Pergamon Press Ltd., 1981, printed in Great Britain. • Fletcher, Gerald F. MD; Blair, Steven N. PED; Blumenthal James PhD; Caspersen, Carl PhD; Chaitman, Bernard MD; Epstein, Stephen MD; Falls, Harold PhD; Sivarajan Froelicher, Erika, S. PhD, MPH, RN; Froelicher, Victor F. MD; and Pina, Ileana L. MD; "Statement on Exercise" Benefits and Recommendations for Physical Activity Programs for All Americans; A Statement for Health Professionals by the Committee on Exercise and Cardiac Rehabilitation of the Council on Clinical Cardiology, American Heart 42 TIMN 445077

Association, AHA Medical/Scientific Statement, Position Statement, approved by the American Heart Steering Committee on February 19, 1992. Requests for reprints should be sent to the Office of Scientific Affairs, American Heart Association, 7272 Greenville Avenue, Dallas, Texas 75231. • Dargie, Henry J. and Grant, S., "Exercise," Department of Physical Education and Sports Science, Glasgow University, Glasgow, England, from British Medical Journal, Vol. 303, pp. 910-912, October 12, 1991. • Van Saase, Jan L.C.M.; Noteboom, Willy M.P.; and Vandenbroucke, Jan P.; "Longevity of men capable of prolonged vigorous physical exercise: a 32 year follow up of 2259 participants in the Dutch eleven cities ice skating tour," Department of Clinical Epidemiology, Leiden University Hospital, P.O. Box 9600, 2300 RC Leiden, The Netherlands, from British Medical Journal, Volume 301, pp. 1409-1411, December 22-29, 1990. • Koo, Linda C., "Dietary Habits and Lung Cancer Risk Among Chinese Females in Hong Kong Who Never Smoked," Department of Community Medicine, University of Hong Kong, Hong Kong, from Nutrition and Cancer, pp. 155-172, 1988. • Renaud, S. and DeLorgeril, M., "Wine, alcohol, platelets, and the French paradox for coronary heart disease," INSERM, Nutrition and Vascular Physiopathology Research Unit (Unit 63), 22 avenue du Doyen Lepine, CP 18, 69675 Bron Cedex, France, from The Lancet, Vol. 339, pp. 1523-1526, June 20, 1992. • Lynch, Henry T.; Fain, Pamela R; Albano, William A; Ruma, Thomas; . Black, Lynn; Lynch, Jane; and Shonka, Michael; "Genetic/Epidemiological Findings in a Study of Smoking-associated Tumors," Department of Preventive Medicine and Health, Creighton University School of Medicine, 2500 California Street, Omaha, Nebraska 68178, pp. 163-169. Section II For further information regarding the idea that smokers are more likely to be in blue collar occupations, one can review the articles listed here. • Sterling, T.D., Ph.D., and Weinkam, J.J., D.Sc., "Smoking Characteristics by Type of Employment," Faculty of Interdisciplinary Studies, Simon Fraser University, Burnaby, B.C., Canada, from Journal of Occupational Medicine, Vol. 18, No. 11, pp. 743-754, November 1976. 43 TIMN 445078

• Sterling, Theodor D., "Does Smoking Kill Workers or Working Kill Smokers? OR The Mutual Relationship Between Smoking, Occupation, and Respiratory Disease," prepared for the School of Workers, University of Wisconsin, short course entitled "Institute for Expertise in Handling Occupational Disease Claims," held from April 18-22, 1977, at Madison, Wisconsin, from International Journal of Health Services, Vol. 8, No. 3, pp. 437-452, 1978. • Sterling, T.D., "The Mutual Relationship Between Smoking, Occupation and Disease," Faculty of Interdisciplinary Studies and of Computing Science, Simon Fraser University, Burnaby, B.C., Canada, from Arh. hig. rada toksikol., Vol. 30, pp. 79-89, 1979. • Weinkam, James J, DSc, and Sterling, Theodor D., PhD, "Changes in Smoking Characteristics by type of Employment From 1970 to 1979/80," School of Computing Science, Faculty of Applied Sciences, Simon Fraser University, Burnaby, British Columbia, Canada V5A 1S6, from American Journal of Industrial Medicine, Vol. 11, pp. 539-561, 1987. • Miller, William J. Jr., MPA, and Cooper, Richard, MD, "Rising Lung Cancer Death Rates Among Black Men: The Importance of Occupation and Social Class, " Department of Community Health and Preventive Medicine, Northwestern University Medical School, 303 East Chicago Avenue, Chicago, Illinois, 60611, from Journal of the National Medical Association, Vol. 74, No. 3, pp. 253-258, 1982. Section III Differences in the reported disease patterns of smokers and nonsmokers may be biased due to the fact that epidemiologic data are derived from death certificates; this phenomenon is called "detection bias". Such bias occurs when a disease is diagnostically sought more vigorously in individuals who are exposed to the suspected cause than in individuals without such exposure. In other words, smokers are more likely to receive more screening and other diagnostic procedures for lung cancer, and therefore, to be diagnosed with the disease than nonsmokers. For further reading on detection bias, please refer to the articles below. • McFarlane, Michael, J., MD; Feinstein, Alvan R., MD; and Wells, Carolyn K., MPH; "Necropsy Evidence of Detection Bias in the Diagnosis of Lung 44 'rIMN 445079

Cancer," Department of Medicine and Epidemiology and the Robert Wood Johnson Clinical Scholars Program, Yale University School of Medicine, New Haven, Connecticut; and the Cooperative Studies Program Coordinating Center, Veterans Administration Medical Center, West Haven, Connecticut; Dr. McFarlane is now with the University of Kansas Medical Center; reprint requests to Department of Internal Medicine, University of Kansas Medical Center, 39th and Rainbow Boulevard, Kansas City, Kansas 66103, from Arch Intern Med, Vol. 146, pp. 1695-1698, September 1986. • Wells, Carolyn K. and Feinstein, Alvan R., "Detection Bias in the Diagnostic Pursuit of Lung Cancer," Department of Internal Medicine and Epidemiology, Yale University School of Medicine, 333 Cedar Street, 1-456 SHM, New Haven, Connecticut 06510, from American Journal of Epidemiology, Vol. 128, No. 5, pp. 1016-1026, 1988. • Feinstein, Alvan R. and Wells, Carolyn K., "Cigarette Smoking and Lung Cancer: The Problems of ' Detection Bias' in Epidemiologic Rates of Disease," from the Cooperative Studies Program Support Center and the Department of Medicine of the West Haven Veterans Administration Hospital, and the Departments of Medicine and Epidemiology of the Yale University School of Medicine, from Transactions of the Association of American Ph siy cians, Vol. 87, pp. 180-185, 1974. Section IV Death certificates may not accurately reflect the actual cause of death. The following articles report that scientists, who have examined autopsy populations and compared the results of the autopsies with the causes of death placed on the death certificates, have found serious error rates in death certificate information, especially for lung cancer. • Carter, John R., M.D., "The Problematic Death Certificate," University Hospitals of Cleveland, Cleveland, Ohio 44106, editorial from The New England Journal of Medicine, Vol. 313, No. 20, pp. 1285-1286, November 14, 1985. • Engel, Linda W.; Strauchen, James A.; Chiazze, Leonard Jr.; and Heid, Marian; "Accuracy of Death Certification in an Autopsied Population with Specific Attention to Malignant Neoplasms and Vascular Diseases," Laboratory of Pathology, Building 10, Room 1A30, National Cancer Institute, 45 TIMN 445080

NIH, Bethesda, Maryland, 20205, from American Journal of Epidemiology, Vol. 111, No. 1, pp. 99-112, 1980. • Rosenblatt, Milton B, M.D.; Teng, Peter K., M.D.; and Kerpe, Stase, M.D.; "Diagnostic Accuracy in Cancer as Determined by Post Mortem Examination," from Progress in Clinical Cancer, Vol. 5, pp. 71-80, 1973. • Rosenblatt, Milton B., M.D.; Teng, Peter K., M.D.; Kerpe, Stase, M.D.; and Beck, Irene; "Causes of Death in 1,000 Consecutive Autopsies," Medical and Pathology Departments, Doctors Hospital, New York, New York, from New York State Journal of Medicine, pp. 2189-2193, September 15, 1971. • Schottenfeld, David, M.D.; Eaton, Muzza, Ph.D.; Sommers, Sheldon C., M.D.; Alsonso, Daniel R., M.D.; and Wilkinson, Carol, M.D.; "The Autopsy as a Measure of Accuracy of the Death Certificate," Memorial Sloan-Kettering Cancer Center, 411 East 69th Street, Room 320, New York, New York 10021, from Bull. N.Y. Acad. Med., Vol. 58, No. 9, pp. 778-794, December 1982. Section V For further reading on the subject that there are factors other than smoking which are related to absenteeism, one can refer to these articles. • Ault, Richard W.; Ekelund, Robert B. Jr.; Jackson, John D.; Saba, Richard S.; and Saurman, David S.; "Smoking and Absenteeism," Department of Economics, Auburn University, Auburn, Alabama 36849 and Department of Economics, San Jose State University, San Jose, California, from A lied Economics, Vol. 23, pp. 743-754, 1991. • Bonilla, Carlos E., "Determinants of Employee Absenteeism," published by the National Chamber Foundation, 1615 H Street, N.W., Washington, D.C. 20062, 1989. 46 TIMN 445081