Tobacco Documents Online

Plusieurs auteurs soutiennentclue les fumeurs sont A ksourcedecoats eonsiderables pour la sociLt6: frais d'hospitalisatioii-et de services medicaux et production pe=due_par le d6c&s prematur6 des. fumeurs._Pour 1'ann6e 2986, au Canada, noas_estimons & s669 millions les coats supplementaires relids & 1'usage du tabac. Farr contre, cornme la rd4iuction des_touts futurs de sant6 s'eleve it $462 mdlions, Ies cofits externes. nets n'atteignent-que $207 millions. Ceaconts donnent lieu a un transfert, ntaia its sont surswmpensds par dautres #ransferts tels que lea taxes payees parr lea fitmeurs et Ies r6duetions des prestations des rdgimes de retraite:: En iAalit6,. il se produit ainsi un transfert net global des fumeurs en faveur des noa_fumeurs de 1'ordre do $4.3 milliards. Meme en considerant un eventad trPs large d'hypothsw m6dicalPs, le sens de cette conciusion-ne change pas. Several authors maintain that smokers impose a considerable burden on society through hospital- ization and medical costs and lost output due to premature death. In this paper, supplementary costs related to smoking are estimated at 669 million dollars for.the year 1986 in Canada. However, since future health cost reductions reach 462 million, the net external costs generated do not exceed 207 million. These costs giv'e_ rise to transfers, but these in turn are more than compensated,by other transfers such as taxes-paid by smokers and reduetions in pension benefits which lead to a net flow overall of 4.3 billion dollars- in favour of non-smo3cers. The direction of this conclusion remains unchanged even considering a wlde range of medical hypotheses. I Introduct"ion = It is the purpose of this article to deter- mine- if observations such as this one are n here is a strong current ofopinion to the_.' - valid or not. The analysis -is based on two L_effect that tobacco beingharmful ta the~ _ fundatriental. -eoncepts or Eriter,ia: exter- heaiih,-emok~rsunpose`eiitraexpenditures= :_ nalities and_ inter-group transfera.. A oa society and becor»e a-_ f`i_naiicial burden = smoker may bee responsible for hospitatixa fio~r noxt-smokers. To _tluo~e only one repre- -_ tion costs viwhic.b become an external cost in sentative esamp -1._- $aitlett- E.19$91,. Canada given - the public nature of the ~ svfites: health_ care system. In. this` instance, the smoker imposesacost on others. The ques- The net result iofsmokirig in Canadal is that it _ tion then arises: who pays for what and who coats eaeh Canadian _$18* annually in heslth.-benefits between smokers and- non- - care related eustss so that 35% aP their iountry= smokers. This calls for an examination of men can smoke. transfers between the two groups. In the previous example, non-smokers finance a Ca+sadio4Public policy=dnatyae de Politiques. lCVTII:3:300-,317 1992 Printed inCana+ielimprimtkau Canads ~ ~~att!570s6

Table 1 Relative risks coefficients (44 Disease Lowest value Highest value• Men Women Men Women Malignant neoplasm of oral cavity 2.76 1.22 esophagus 1.82 4.89 stomach 1.39 L31 pancreas L50 L42 larynx 6.52 3.25 trachea, bronchus and lung 3.76 2.03 cervix uteri L72 bladder 1.40 L66 kidney 1.20 n/a Ischaemic heart disesses 1.58 1.30 Bronchitis and emphysema 12.07 n/a 13.00 3.25 - 6.43 4.89 1.80 2.30 3.10 2.50 13.59 6.52 14.20 5.00 3.00 2.89 2.80 2.66 2.66' 1.71 L78 24.70 24.70t 'zis table reads as follows: taking for example the highest value for men, the risk of death from oral cavity can- cer is 13 times higher for smokers than for non-smokers. souHCS: US Department of Health and Human Services (USDHHS, 1982,1983,1984). ' The highest values are used as our basic assumptions, as stated above. t A relative risk coefficient was not established for kidney cancer and pulmonary diseases in women; therefore, we adopted the coefficients given for men. good proportion of supplementary health costs, so that transfers go from non- smokers to smokers. On the other hand, smokers pay taxes on tobacco which bene- fit non-smokers to a large extent. In this case, the transfers go in the opposite direc- tion. A complete balance sheet of revenues and expenditures accruing to smokers and non-smokers is necessary before one can conclude whether or not smokers are a bur- den on non-smokers. -? Basic Assumptions This study is based on three major assump- tions:l/ smoking is harmful to the health of smokers; 2/ smoking is not harmful to the health of non-smokers; and, 3/ smokers know that smoking is detrimental to their health. Smoking is Harmful to Smokers Numerous medical studies show that con- sumption of tobacco is harmful to the health of smokers. Since our expertise does not lie in the medical field, we simply use the findings published by the Surgeon General of the United States2 as working hypotheses without endorsement 3 They take the form of relative risk coefficients (4.) associated with a broad range of dis- eases.4 These coefficients give the relative risk of dying if one is a smoker as opposed to a non-smoker. Since several estimates are often given for the same illnesses, we have always chosen the highest values for our calculations so as to avoid any sugges- tion that we have selected hypotheses favourable to our conclusions. It will be ob- vious that with higher risk coefficients, re- lated health cost estimates will also be higher than otherwise. Table 1 gives the highest and lowest estimates of risk coeffi- cients as has been found in the Reports of the Surgeon General. The risk coefficients will be used to esti- mate the proportion of costs imputable to tobacco; those costs are broken down ac- cording to the International Classification of Diseases (ICD-9). As a result, a cor- respondance had to be established between the diseases given in Table 1 and those of ICD-9. For cancer of the oral cavity, we used the Smokers' Burden on Society 301

broader category Malignant neoplasms of the lip, oral cavity and pharynx. This is con- sistent with our practice of always opting for the least favourable choice. For cancers of the oesophagus, stomach, pancreas, larynx, trachea, bronchus and lung, cervix uteri, and bladder, ICD-9 provided a corre- sponding listing of diseases under the same names. For bladder cancer, we chose Malig- nant Genito-urinary Organ Tumors, Kid- ney and Other Urinary Organs n.o.s., a sub- category that includes kidney cancer but extends beyond it. For ischaemic heart dis- ease, a corresponding category of the same name was chosen from ICD-9. Finally, for bronchitis and emphysema, we selected the very broad category Respiratory System Diseases, Chronic Obstructive Pulmonary Disease and Allied Conditions; as usual, we made the choice to bear most unfavourably upon our findings. To the 11 diseases of Table 1 we as- sociated 30 diseases, when considering the three digit disaggregation of the ICD-9 classification. The Health of Non-smokers There is a widespread opinion to the effect that smoking is harmful to the health of non-smokers. It is based mainly on strong statements made by the Surgeon General of the United States, such as the following one taken from his 1986 report (p.13): Involuntary smoking is a cause of disease, in- cluding lung cancer, in healthy non-smokers. We found however that the evidence in sup- port of this opinion is far from conclusive. Concerning lung cancer among spouses, the 1986 review of the US Surgeon General is largely an examination of three prospec- tive and ten retrospective studies.5 High- lights of these findings, provided in his ta- bles 7, 8 and 9, can be found in our Table 2. In our judgment, Table 2 does not sup- port the Surgeoii General's thesis, since two thirds of the studies tabled - nine out of 13 - do not provide relative risk coeffi- cients statistically different from one. Of 302 Andre Raynauld and Jean-Pierre Vidal the four studies which would support the Surgeon General's thesis that smoking is dangerous for non-smokers, the findings of two are highly tenuous. With reference to the study by Trichopoulos, the Surgeon General (USDHHS, 1986) observes that some biases may have arisen in the selec- tion and interview process. As well 'the di- agnosis of cancer was not confirmed for 35 per cent of the cases'. In the study by Cor- rea et al. the sample was limited to 30 cases overall. Finally, this comment found in the Surgeon General's Report (USDHHS, 1986) and concerning the study of Gar- finkel would make anyone suspicious of the results obtained in this area of research: Among the published studies on involuntary smoking, this is the only one involving inde- pendent verification of the diagnoses of all cases. This verification showed that 13 percent of the cases classified as lung cancer were not primary cancers of the lung. This study showed that 40 percent of the women with lung cancer who had been classified as non-smokers (or smoking not stated) on hospital records had actually smoked, compared with 9 percent of the controls. The in- clusion of lung cancer patients who had actually smoked, would have substantially increased the odds ratio with inuoluntary smoking, because 81 percent of the potentially mis-classified cases had husbands who smoked, compared with 68 percent of the 'true' non-smoking patients with lung cancer.s Other remarks by the Surgeon General deserve attention. For example, on page 91 of the same report, we find this observation concerning tobacco use by parents: None of the studies with data on parental smok- ing had sufficient numbers to examine the ef- fects of parental smoking on non-smokers. Similarly, we find this observation about the use of tobacco in the workplace: The workplace, an important source of tobacco smoke exposure, was not considered in the early studies on involuntary smoking. Later case-con- 2501357038

Table 2 Summary results on passive smoking relative risk coefficients for lung cancer Prospective studies Study Sample Cancers Significant results at 5% Hirayama 1981, 1983, 1984 91,450 Garfinkel 1981 176,739 Gillis et al. 1984 2,744 200 153 14 ik = 1.9 for women whose husbands smoked more than 20 cigarettes daily None None Retrospective studies Study Cancers Significant results at 5% Trichopoulos 1981, 1983, 1984 77 41' = 2.5 for women whose Correa et al. 1983 30 husbands smoked more than 20 cigarettes daily V = 3.1 for thoee whose Chan and Fung 1982 84 partners smoked more than 40 packs yearly None Koo et a1.1983, 1984 88 None Iiabat and Wynder 1984 78 None Wu et a1.1985 29 None Garfinkel et al. 1985 134 >/i' = 2.1 for women whose Lee et al. 1986 47 husbands smoked more than 20 cigarettes daily None Akiba et al. 1986 103 None Pershagen (in print) 67 None SoueCE: US Department of Health and Human Services (USDHHS, 1986). trol studies provided some information on tobacco exposure at work, but the data were limited and inconclusive. For all these reasons, we assume in this study that smoking is not a cause of death among non-smokers.7 Concerning the relationship of 'passive' smoking to other cancers, the Surgeon General writes: There are, at present, insuff cient data to ade- quately evaluate the role of involuntary smok- ing in adult cancers other than primary carcin- oma of the lung. With regard to cardiovascular disease, the Surgeon General writes: More detailed characterizations of exposure to ETS [Environmental Tobacco Smoke] and specific types of cardiovascular disease as- sociated with this exposure are needed before an effect of involuntary smoking on the etiology of cardiovascular disease can be established. Smokers Know that Smoking is Detriment- al to Their Health Finally, we assume that smokers are aware that smoking may be harmful to their health. Such an assumption is especially reasonable in Canada, where the tobacco industry is required to display very promi- nent warnings on all cigarette packages. Indeed, such warnings must occupy an area equal to 20 per cent of the principal panels of the cigarette pack. They must be 'legible and prominently displayed in con- trasting colours,' and the message is unam- biguous: Smoking reduces life expectancy. Smoking is the major cause of lung cancer. Smoking is a major cause of heart disease. Smoking during pregnancy can harm the baby. Smokers' Burden on Society 303

If the numerous anti-smoking cam- paigns are added to these warnings, it would be simply unrealistic to assume that smokers are not aware that smoking may be hazardous to their health. Yet, one could argue that people still un- derestimate the risks or do not pay enough attention to these warnings. But this view would not be consistent with the findings of W.K Viscusi (1990): This paper uses a national survey of 3,119 in- dividuals to exaTMine the effect of lung cancer risk perceptions on smoking activity. Both smokers and nonsmokers greatly overestimate the lung cancer risk of cigarette smoking, and the extent of the overestimation is much greater than the extent of underestimation. One could still argue that in the 1940s and 1950s when smoking was portrayed as glamorous and good, people began smoking but that now they find it very difficult to quit. This is correct as a statement of fact but irrelevant in the context of this analysis. The real point is whether there is or there was a market failure giving rise to external costs as opposed to private costs such as loss of income due to premature death at- tributed to tobacco. Even in the 1940s nobody hid the truth from smokers because nobody knew that tobacco was so harmful to the health. Indeed, the problem at hand is not a prob- lem of information but rather of uncer- tainty. Moreover, it is not because many see smoking as addictive that an external cost results automatically. Indeed, using a rational addiction model, G.S. Becker and F S. Murphy (1988) show that a consumer may be perfectly aware that smoking is harmful and addictive, and, in spite of that, decide to smoke to maximize his/her utility over time $ In other words, if decisions are taken in full knowledge of the consequences as they exist at the time the decision is made, this becomes a purely private and rational deci- sion involving no more than private utility 304 Andri Raynauld and Jean-Pierre Vidal and costs. III External Costs To determine whether smokers are a bur- den on society, one must first establish the importance of external costs related to smoking. In application of our first assump- tion above, supplementary medical and hospital costs are involved and they are deemed to be external given the public na- ture of the health care system in Canada. We also include costs associated with fire losses imputed to negligent smokers. Various other costs are mentioned in the literature but not included herein, such as: 1/ time lost on the job due to smoking; 2/ ab- senteeism without loss of pay; 3/ additional cleaning and ventilation costs in areas used by smokers; 4/ the price non-smokers would pay to be always smoke-free; and 5/ the price relatives and friends would pay to re- duce the risk of a smoker's death. The first two categories of costs are held to be negligible when they are compared with all the other reasons or pretexts to waste time on the job or stay away from work. Moreover, in a majority of cases, the alleged external cost is very much a private cost; people are not paid when they do not work or they lose either money or leisure from taking 'sick days,' which are inter- changeable with additional wages or days off. In any case, if smoking had been signifi- cantly related to costs for the employer, or to productivity, then smokers' wage rates would certainly have reflected it. If they did not, one cannot easily dismiss the idea that smoking was an excuse that might easily be substituted for another one, like having a coffee or a soft drink. Furthermore, in a study involving 33,032 individuals, C.E. Bonilla (1989) found that smoking had no bearing on absenteeism. The same applies to cleaning and venti- lation costs. If a systematic association had been established involving smokers, mar- ket forces would have led to lower salaries and smokers would have internalized the 2501357040

Table 3- Proportion of cigarette smokers among Canadians (percentages) 1986 35.40 33.60 18.60 31.40 25.30 11.30 Note: The proportion of smokers in the overall population is 22.12%. It is found by applying the proportions of Table 3 to the total population by age group and sex, found in Statistics Canada (1988), summing them up and taking the percentage. The 0 to 15 age group is assumed not to contain regular smokers. sot:xcE: Health and Welfare Canada (1988). costs. Time lost in the work place as well as cleaning and ventilation costs are not real external costs. On the other hand, the price non- smokers would pay to live in a smoke free environment is a genuine externality. _ v-wever, smokers would undoubtedly be r..iling to pay to smoke wherever they like and not be blamed for it. This externality, in the opposite direction, could very well cancel out or even more than offset the ex- ternal cost borne by non-smokers. This is especially true now, since new laws and in- ternal regulations forbid smoking in many public and work places. Finally, even if those close to smokers were prepared to pay to reduce the risk of death, this cost cannot be considered purely external, since friends and relatives share their concerns with the smoker, who is very likely to take their views into account. For those reasons, we have not included these considerations in our evaluation. On the other hand, if we follow the logic ' of the situation and demonstrate that s kers risk premature death, we must ac- cept the fact that smokers stop requiring the health services they would have re- quired had they not smoked and lived longer. As we will see in section V (Methodological Issues), several authors support this logic. Therefore, to arrive at a net value of ex- ternalities, one has to calculate the reduc- tion of health costs due to premature death and balance it against the supplementary costs attributable to smoking. Additional Hospital Care Costs The computation of supplementary hospi- talization costs is based on the proportions of cases attributable to smoking, for each disease, each sex, and each age-group, pro- portions that are called etiologic fractions W and are defined as: 8(~i-i) ~- , 8(¢-1) + 1 where 8 is the proportion of the population by age and sex exposed to the risk factor and V, is the relative risk coefficient 9 B is given above in Table 3, and V, has already been given in Table 1. The computation of costs was then car- ried out as follows: (1) The number of hospital care days, throughout Canada in 1986, was estab- lished for each relevant disease or cate- gory of disease, each sex, and each age- group_10 (2) These numbers of days were then mul- tiplied by the corresponding etiologic fractions, by age, and by sex, to obtain the number of hospital care days attribu- table to smoking, that is, the number of hospital care days that presumably would have been saved had no one smoked. (3) The sum over diseases, sex, and age- groups multiplied by $352.22, which is taken to be the average cost of one hospi- talization day, gave $546.5 million.ll Although the calculations are very simple, they are so numerous that it is im- possible to give a detailed account of the more than 7,000 pages of tables that could Smokers' Burden on Society 305

Table 4 Additional hospital care costs: summary of intermediary results for men 1986 International classification of diseases - 9th revision (ICD - 9) Malignant neoplasms lip, oral cav. and phar. (140-149) esophagus (150) stomach (151) pancreas (157) larynx (161) trachea, bronchus and lung (162) bladder (188) kidney and other urinary organs n.o.s. (189) Ischaemic heart diseases (410-414) Chronic obstructive pulm. dis. and allied cond. (490-496) Overall Total number of hospital days Proportion attributable to smoking (%) Hospital days attributable to smoking Imputed hospital cost (million dollars) 55,739 75 41,805 14.7 29,228 57 16,636 5.9 64,285 16 10,011 3.5 48,041 33 15,788 5.6 33,859 76 25,626 9.0 343,995 75 258,605 91.1 92,897 29 26,994 9.5 41,147 29 11,784 4.2 1,077,802 15 162,672 57.3 651,273 72 468,015 164.8 2,438,266 1,037,937 365.6 Notes: Due to rounding, totals may differ slightly. The proportion of days attributed to smoking is a weighted average of the etiologic fractions defined above as the proportion of cases attributable to smoking by age, sex, and disease. The etiologic fractions themselves are some- what higher than the average values. For instance, 89% of all days spent for men aged 25 to 44 and suffering from a chronic obstructive pulmonary disease were attributed to smoking. This might be surprising when compared with the average value of 72%, but it results from the fact that no cases were attributed to smoking under the age of 15. Cancer of the cervix uteri has been deleted from this table but it has been taken into consideration for women. be accounted for in this whole study. However, Table 4 has been constructed as a heuristic device to show the kind of opera- tions involved. Taking the first line as an example, the total number of hospitaliza- tion days spent for malignant neoplasms of the lip, oral cavity, and pharynx was 55,739 in 1986 for men. Since the proportion at- tributable to smoking was established at 75 per cent, the number of hospitalization days attributed to smoking was 41,805. At $352.22 per day, the corresponding cost amounts to $14.7 million. Additional Medical Services Costs In this case, we had to extrapolate from statistics relating to Saskatchewan to ar- rive at an overall picture for Canada, be- 306 Andre Raynauld and Jean-Pierre Vidal cause this is the only provincial authority gathering the required data on the specific illnesses. With reference to Saskatchewan, during 1986-87, $4.7 million was spent for neo- plasms; $4.7 million for diseases of the circulatory system; and an additional $5 million for chronic obstructive pulmonary diseases.l2 Since Saskatchewan's overall health services cost $184.8 million in 1985, while the cost for all Canada was $6.3 billion, we arrived at a multiplication factor of 33.82, which when applied to the previous num- bers gave Canadian extrapolations of $159.1 million for neoplasms; $160.2 million for diseases of the circulatory system; and $167.6 million for chronic obstructive pul- 2501357042

monary diseases. Finally, assuming that the cost of medi- cal services attributable to smoking was proportional to the number of hospital days due to the same cause, we obtained a cost for medical services attributable to tobacco of approximately $68 million: $22 million for neoplasms; $5 million for diseases of the circulatory system; and $41 million for dis- eases of the respiratory system. The Cost of Smokers' Negligence In table 5 of the Federal Fire Commis- sioner's annual report (Labour Canada, 1987), property losses from fires blamed on smoking are listed at $52.2 million for 1986. To these property losses we added $2 nillion for the nation's reduction in well- being resulting from the loss of wooded land. Our reasoning was as follows: 1/ We knew the fight against forest fires costs $102.3 million in 1986, and that fires blamed on 'recreation' account for 2 per cent of devastated areas (see Petawawa National Forestry Institute, 1989). 2/ So, assuming that the government would keep spending money to protect the en- vironment until the value of the lost en- vironment would be worth no more than the money spent to protect it, we esti- mated that the value of the wooded land lost represented at most 2 per cent of $102.3 million. Obviously, this is an over- estimation because the category 'recrea- tion' is much broader than the `neg- ligence on behalf of a smoker' 13 The Reduction of Future Hospital Care Costs Consistent with the extreme assumptions adopted for the hazards of smoking, we im- puted 21,841 deaths to tobacco use by apply- ing the etiologic fractions to the statistically recorded deaths in Canada by cause, age, and sex. Subsequently, a model repre- senting the anticipated life and death of each of these deceased individuals has been constructed, assuming that all of them died earlier than they would have had they not smoked 14 The fundamental premise leading us to infer a reduction in future cost is that al- though smokers might have lived longer had they never smoked, they would nevertheless have been subject to the same risks as any non-smoker. At one time or another, they would have been ill, and, eventually, most would have spent some time in the hospital. Consequently, in our 'life-and-death' model, each prematurely deceased in- dividual is classified according to the year of his or her unavoidable demise (the mo- ment when he or she would have died any- way), the cause of that demise, and the age that he or she would have reached at that time. From this information, we computed the following reduction in hospital care costs: L For each year, from 1986 to 2071, the number of deaths by ICD-9 category and by age was multiplied by the average length of a hospital stay (in days) rele- vant to each category and age. Summing up over categories and ages, a number of hospitalization days saved was then ob- tained for each year from 1986 to 2071. (There would be no survivors alive after 2071. Since no one died from smoking before being 15 years old, nobody would live beyond the age of 100 years: 2071- 1986 + 15). 2/ Again, for each year from 1986 to 2071, the previously obtained numbers of hospitalization days were multiplied by the average cost of each day, established at $352.22 for 1986-87, and assumed to remain at that level in constant dollars. 3/ Finally, the present value was taken at an annual compounded real discount rate of 3 per cent.15 From the above calculations, we esti- mated the reduction in hospital care cost at $133.4 million. It is worth indicating that this amount is probably a significant underestimation, since our calculations do not reflect hospi- tal stays not directly related to the ultimate death of the indiuiduals concerned.ls Smokers' Burden on Society 307 I

The Reduction of Future Medical Services Costs From our 'life-and-death' model, we also obtained the number of deceased smokers who might have survived in each of the years from 1986 to 2071. It was then possible to estimate the value of medical services saved because of these premature deaths. Our methodology can be described as follows: 1/ Had the dead smokers never smoked and lived longer, we assumed that, at any specific age, they would have required, on average, the same kind of medical at- tention as anyone else at that age; con- sequently, their average medical costs would have been the same as those of other citizens.17 2/ We then applied the average cost of medi- cal services, broken down into five-year age spans, to the number of deceased smokers who, presumably, would have survived. As above, the present value was taken at a 3 per cent discount rate. The reduction in future medical services was established at $123.5 million. The Reduction in Residential Care Facilities for the Aged Finally, we considered the reduced expen- diture on residential care facilities for the aged. 1/ We assumed that the number of old age beneficiaries of residential care, ex- pressed as a percentage of the total popu- lation by age group, would remain con- stant after 1986,18 as well as the average cost horne by society at $35.17 per day (in 1986 dollars) is 2/ Subsequently, we applied the proportion of old age beneficiaries to the number of deceased smokers for each year from 1986 to 2071, taking into account age and sex, to calculate how many of them would have been beneficiaries had they not succumbed in 1986. 3/ Finally, by straightforward multiplica- tion of the above figures by $35.17, we ar- rived at the annual amounts saved from 1986 to 2071. 308 Andre Raynauld and Jean-Pierre Vidal Taking the present value, we arrived at a reduced cost for residential care facilities for the aged of $204.9 million. The Net Cost of Externalities Based on our extreme assumptions about the risks run by smokers, we estimate the net external cost of smoking in Canada for 1986 at $207 million. This figure includes supplementary costs presumably incurred because some people smoke: $546.5 million in hospitali- zation; $68.1 million in medical services; $52.2 million in property losses from fire; and $2 million for the destruction of part of the nation's wooded land, for a total of $668.8 million. Reductions in future costs include: $133.4 million in hospitalization; $123.5 million in medical services; and $204.9 mil- lion in residential care facilities for the aged. Substracting this $461.8 million re- duction from the $668.8 million, we arrive at a net external cost of $207 million. IV Transfers The next question is how much of this net external cost of $207 million is paid by the smokers themselves and how much is paid by non-smokers. Put in these terms, it be- comes obvious that the question is only a small part of a much broader issue. To determine whether smokers overall constitute a financial burden on non- smokers or on society as a whole, it is not enough to determine if taxes paid by smokers are sufficient to cover the expenses they are responsible for. In addition we have to estimate the relevant taxes smokers would have paid had they lived longer and also, we have to include the net savings re- lated to pension benefits, which are reduced by the early deaths of smokers. The Transfers from Externalities In the preceding section we established the net external cost of smoking at approxi- mately $207 million. Since we estimate that 69.15 per cent of taxpayers are non- 2501357044

smokers20 $143 million will be financed by non-smokers. This amounts to a transfer of $143 million from non-smokers to smokers. However, there are people of working age among the deceased smokers who would have lived longer, had they not smoked. Then, we must estimate the taxes they would have paid. The following de- scribes our methodology: ll According to Health and Welfare Canada (Table 6), in 1986, the costs of hospitali- zation and medical services were $17.5 billion and about $7 billion respectively. 2/ According to Statistics Canada (1989b:119), the cost of residential care facilities for the aged was approximately $3 billion, which means that applying our 65 per cent rule, $2 billion was paid for from public sources. • According to our calculations, there were 12 million Canadian workers in 1986. 4/ Thus, assuming that the costs were fully recovered by the government from the taxpayers, each taxpayer contributed $2,252.70. 5/ Assuming that this contribution would remain at that level in constant dollars, we multiplied it by the number of deceased taxpayers for each year after 1986. Taking the present value as we did before, we estimated the contribution of deceased smokers, had they not smoked, at $146 million. Of this amount, $100.9 million (69.15%) represents a reduction of the fi- nancial advantage to non-smokers. The net transfer relative to external cost, from non- --nokers to smokers, is therefore approxi- -lately $244 million (143 + 100.9). The Transfers through Tobacco Taxes Smokers paid $4.1 billion in tobacco taxes, over and above the taxes paid by non- smokers in 1986. However, the transfer going from smokers to non-smokers is not equal to that amount because smokers benefit as well as non-smokers from the public services they finance. More pre- cisely, since non-smokers represent 77.88 per cent of the population, they benefit from smokers' taxes by approximately $3.2 billion. To obtain this total of $4.1 billion, we simply added the federal and provincial consumption taxes on tobacco. Federal taxes have been established at $1,948.3 mil- lion: $277.3 million for the federal sales tax;21 $1,107.5 million for the excise tax; $552.6 million for the excise duty;22 and $11 million for the tariff duty.23 Provincial sales taxes have been established at $2,119.7 million. This estimate applies to calendar year 1986, and comes from Statis- tics Canada (1986) 24 The Transfers through Pension Plans We assumed at the outset that smoking was hazardous to health and that, as a con- sequence, smokers had a shorter life expec- tancy than non-smokers. Consistent with that assumption, we find that if smokers die earlier than non-smokers with similar pen- sion benefits, there must be a transfer in favour of non-smokers and to the detriment of smokers. We estimate this transfer at $1.4 billion. In 1986, total contributions to A ension plans amounted to $29.5 billion. Since there were 11.8 million workers, the aver- age contribution was about $2,504.13. On average, for the same year, the Old Age Security and Supplement regimes paid out $4,604.71 to 107 per cent of the 65 and over age group (because spouses' allow- ances are also paid to those less than 65 years old, as are pensions to Canadians abroad); Canada and Quebec Pension Plans $2,865.73 to 96 per cent of those 65 and over; and private pension plans $7,261.27 to 45 per cent of those 65 and over. From these observations, we undertook the following calculations: 1/ From the calculations we referred to pre- viously, we know how many deceased smokers would have worked, had they not smoked, for each year after 1986. To determine how much they would have contributed to financing pension funds, we multiplied that number by the aver- age contribution of $2,504.13. Smokers' Burden on Society 309

Table 5 Total net transfers from smokers to non-smokers 1986 In millions of dollars Net external costs -244.0 Additional taxes paid 3,168.2 Pension plans 1,417.4 Total net transfers 4,341.5 Note: Due to rounding, total may differ. 2/ From the 'life-and-death' model, we also know, for each year after 1986, how many deceased smokers would have been alive and over 65 years old had they not smoked. To determine how much they would have received from each type of pension plan, it is sufficient to mul- tiply their number, first, by the propor- tion of those 65 and over receiving money from that particular plan and, second, by the average amount of bene- fit paid out. Taking the present value of the differ- ences between additional contributions and benefit payments for each relevant year after 1986, we arrive at $2 billion. This is the amount that would have been necessary in 1986 to guarantee that the deceased smokers would benefit from pension plans exactly as non-smokers. Since smokers would participate in gathering this sum, only part of it is a transfer to non-smokers. Using once again the proportion of Canadian taxpayers who did not smoke in 1986 (69.15%), a transfer of $1.4 billion was obtained.26 The Net Transfer If we sum the transfers over the three cate- gories above, we get Table 5. From our assumptions and the above analysis, it emerges that any notion that smokers are a burden on society in general or on the non-smoking population in partic- ular is without foundation. Actually, for 1986 in Canada, non-smokers enjoyed a standard of living $4.3 billion higher than it would have been if there had been no 310 Andre Raynauld and Jean-Pierre Vidal smokers at all, according to our assump- tions and methodology. V Methodological Issues Our results differ very considerably from the conclusions reached by several authors, including those who examined the Canadian situation. These major differences come mainly from the methodology used and involve a number of considerations that may or may not be held relevant to the analysis of the financial burden of smokers on society. There are three major issues: 1/ the re- duction of future costs associated with the assumed early deaths of smokers; 2/ the no- tion of financial transfers between smokers and non-smokers; and 3/ finally, the most important one, the inclusion or exclusion of personal income losses in the balance sheet. The Reduction of Future Costs Although E.R. Shillington (1977) and N. Collishaw and G. Myers (1984) ignored the reduction of future healths costs, we main- tain that these savings must be deducted from the supplementary costs associated with smoking. It is a matter of very simple logic to argue that if smokers die earlier than non-smokers, between the time of their early death and the time they would have died had they not smoked they will not use the services typically used by non- smokers. In any case, in approaching the problem in this way one may claim the sup- port of several authors, such as A.B. Atkin- son and J.L. Townsend (1977), W.F. Forbes and M.E. Thompson (1983), G.T. Watts (1983), R.E. Leu and T. Schaub (1983, 1984), and A. Markandya and D.W. Pearce (1989)?7 In dollar terms, this item is not crucial, but it is not negligible either. As already in- dicated above, it reduces tobacco-related health cost by $462 million. Transfers Financial transfers between smokers and non-smokers are not considered in the ref- 2501357046

erence studies cited above. In our view, such transfers, which determine who pays and who benefits between smokers and non-smokers respectively, are an essential component in any assessment of the burden that smokers may or may not impose on society. There are two steps in such an analysis. First, one must incoporate taxes paid by smokers, because on efficiency grounds alone and consistent with the Pigovian ap- proach, such taxes are a compensation for the external costs generated. An optimal tax in this context is one that just equals this external cost at the margin. Up to this point we follow A.B. Atkinson and 'r.W. Meade (1974), as well as A. Markandya and D.W. Pearce (1989). However, we make a second, additional, step in this study. As the calculations above indicate, total external costs and taxes are adjusted by the relevant proportions of smokers to derive transfers in the strict sense of the word. We take a broad view of transfers as it should be. In addition to the taxes on tobacco consumption, our transfers include estimates of the net flow of pension con- tributions and benefits between smokers and non-smokers. The bottom line on over- all transfers is given in Table 5. Own Income Losses Personal income losses are the main em- phasis in most of the literature on the 'economic consequences' (cost) of smoking. For example, personal income loss repre- sents 86 per cent of the 'economic con- sequences' in the study by Shillington (1977); 77 per cent in Collishaw and Myers (1984); and 55 per cent in USDHHS (1990). It seems that all these were inspired more or less directly by D.P. Rice (1966). The concept of 'economic consequences' was renamed 'economic cost' by Leu (1983), and Leu and Schaub (1984), who ration- alised it as the sum of the 'monetary private costs' to smokers plus the 'monetary exter- nal costs' of smoking. The 'economic cost' of smoking would then be the sum of its pri- vate and external components - that is, its social cost - but without considering intan- gible costs such as pain and suffering. It is this concept of 'economic cost' that has been applied to show what a burden smokers are to society. The language used in the Report to Congress, USDHHS (1990), is unambiguous when the authors refer to 'the cost or value to society' (p.38) or when they remark that 'smoking causes large numbers of deaths and a very large dollar cost to society' (p.40). Ultimately, it is suggested that taxes on tobacco should compensate for this huge burden. As we remarked in the introduc- tion, there are numerous published opin- ions to this effect. The fundamental issue to consider in this case is whether own-income losses as- sociated with the premature death of smokers are a private cost or an external cost. Our thesis is that such income losses are a private cost and should not be in- cluded in an analysis of the financial bur- den of smokers on society. The first argument in support of this thesis has been included earlier as one of our three basic assumptions in this study, namely that smokers behave rationally and are well-informed of any possible health hazard from smoking. It might still be argued, however, that even though smokers have knowingly ac- cepted the risk of sickness or early death, they nevertheless impose a loss on society. This evokes the image of smokers dying in their forties and thereby depriving society of another two decades of productive work. Such reasoning may seem persuasive, but it is almost invariably in error. In fact, society tends to reward individu- als strictly in accordance with the contribu- tion it receives from each. It therefore neither gains nor loses with the increase or decrease of its overall numbers. When another worker is introduced into the economy, production rises, leading to the growth of national product. But this does not imply that society as a whole is any more rich or prosperous. This reasoning is not new or original. Smokers' Burden on Society 311

Indeed, there is a strong consensus among economists on this point. In the words of E.J. Mishan (1971a): the loss of potential future earnings, can be rationalized only if the criterion adopted in any economic reorganization turns on the value of its contribution to GNP, or, more accurately, to net national product. But although financial journalists manage to convey the contrary im- pression, maximizing GNP is not an acceptable goal of economic policy. Mishan's judgment highlights the simple truth that the real victim of a death is the dead person concerned, not society as a whole. The three methodological issues just raised are largely responsible for the huge differences between our empirical results and those of our reference studies. The myth and reality of the financial burden of smokers do not rest on the minute details of computations but on the relevant factors to be considered in the analysis. The next section will confirm this conclusion. VI Sensitivity of Results to Relative Risk Coefficients Remembering that our results regarding health costs are based on epidemiological studies which differ widely in their find- ings, a legitimate question arises about the sensitivity of our conclusions to the relative risk coefficients used. All calculations have therefore been done again using 14 differ- ent hypotheses covering a very wide range of possible assumptions regarding relative risk coefficients. They range from smoking is not dangerous at all (,y = 0), to smoking is ten times more dangerous than in the worse scenario envisaged in the reports we used (y= 10). As expected, gross external costs in- crease with the danger smoking poses to health. When -y equals 0, tobacco use is harmless to health, and the only external cost is fire at $54 million. When Y equals 1, relative risk coefficients being identical to 312 Andre Raynauld and Jean-Pierre Vidal those assumed in section II, we get the same result as in section III, of $668.8 million. Af- terwards, costs tend to stabilize toward an upper bound which cannot be exceeded: ob- viously, when y- -, all the etiologic frac- tions tend to 100 per cent, at which point no further costs are left to be imputed.28 On the other hand, when smoking is harmless, there is no reduction in future costs; however, as it becomes increasingly dangerous, it is accompanied by a reduction in future costs. Two characteristics are worth noting: first, the curve describing the reduction in future costs increases at a lower rate than the additional cost curve does; and second, the rate of increase of the reduction cost curve decreases more slowly than the additional cost curve does. These characteristics result in a very in- teresting net external cost curve, shown on Graph 1, which represents the difference between gross external costs and reduction of future costs. It will be recalled that in our reference scenario, where y equals 1, smokers im- posed a net external cost of $207 million on Canadians in 1986. As is apparent from Graph 1, this net external cost appears to be in the neighbourhood of a maximum: whatever the relative risk coefficients one chooses, the net external cost is never sig- nificantly higher than $207 million. Reduc- ing the risk brings a movement to the left of ` y equals 1' and leads to a decline in net external cost, but increasing the risk, al- though it brings a movement to the right, also leads to a decline in net external costs because the reduction in future costs grows at a faster rate than the increase of addi- tional costs in present value (as of 1986). Actually, if it were possible for relative risk coefficients to be as high as when y equals 10, we would have to conclude that there would be a net external benefit of $725,000 (which appears close to zero in Graph 1). The existence of this maximum of ap- proximately $207 million, allows us to ob- serve that, whatever the risk smoking poses to health, the present level of taxation is un- 2501357048

 1. NET COST 0r- EXTERNALITIES justifiable on grounds of efficiency. In other words, since the net external cost function is not an increasing monotonic function of the risk involved in tobacco use, there is no level of danger at which the present level of taxation would be justified. As for the transfer between smokers and non-smokers, whatever the danger tobacco might pose to health, non-smokers always enjoy a much higher standard of living be- cause of smokers. Given our assumptions, iokers are never a burden to non- s.nokers: at worst, non-smokers receive ap- proximately $3 billion from smokers, and at best $7 billion. Without doubt, the net global transfer flow is always to the advan- tage of non-smokers and to the detriment of smokers. At the beginning of this paper, we noted that many authors evaluated the 'economic consequences' of smoking as one huge sum, which has subsequently been interpreted as a burden smokers would be imposing on others. New taxes have been proposed and justified on these grounds. We have shown instead that net addi- tional external costs borne by non-smokers worked out to $244 million for Canada in 1986. However, smokers are responsible for a much larger transfer flow in the other direction. In the pension area alone, non- smokers benefit from a transfer of $1.4 bil- lion mainly because smokers tend to die before non-smokers do if we use risk coeffi- cients established by the medical profes- sion. Finally, the massive tax burden borne by smokers alone means that they account for a further transfer of close to $3.2 billion to the benefit of non-smokers. Overall, as Table 5 indicates, smokers make a net over- all contribution of $4.3 billion to the bene- fit of non-smokers. Whatever the degree of risk or danger at- tributed to tobacco, the validity and direc- tion of these conclusions remain un- changed. Smokers' Burden on Society 313 _,:h,.

Notes 1 See Forbes and Thompson (1978), Wilkinson et aL (1978), Heins (1978), Kristein and Grove (1978), Cady (1983), Warner (1983), Cady (1986) and Hof- fenberg (1988). Studies on the 'economic con- sequences' of smoking such as those of Shillington (1977), Collishaw and Myers (1984) and the US Department of Health and Human Services (USDHHS,1990), have been interpreted as imply- ing a financial burden on the part of smokers. Strictly speaking, this implication is incorrect. See also Rice (1966), Stoddart (1987), Choi and Nethercott (1988), Freour et al. (1976), Luce and Schweitzer (1978), Shultz (1985), Hinds (1986), Gorsky, Schwartz and Dennis (1990), Gray et al. (1988). A closely related literature claims to follow 'cost-bene6t' analysis. A typical example is found in Forbes aqd Thompson (1982), and a criticism in Woodfield (1984). 2 We used the Reports of the years 1982, 1983 and 1984. For a Canadian study, taking our relative risk coefficients from an American source may ap- pear to be a second best, especially if one believes, as we do, that geographical variables may be im- portant. However, some Canadian studies are in- cluded in these reports. 3 Although we did not review the medical literature systematically, we did come across some very dis- turbing findings for the 'conventional wisdom' about cigarette smoking and health. These lead us to believe: first, that the results of epidemiological studies might very well be strongly biased by the omission of some crucial variables from the statis- tical analyses, such as personality (especially risk aversion) which is responsible for a larger set of detrimental behaviours as well as beneficial ones, like exercising; and second, that adequate care had not always been taken in building the samples used in epidemiological studies. See Seltzer (1989 and 1980). Sterling and Weinkam (1987), and Tol- lison (1986). 4 We assume implicitly that the morbidity risk coefficients are the same as the mortality risk coefficients, because the statistics required to dis- tinguish the two are inadequate. This assumption is not favourable to our conclusions. 5 Prospective studies are based on a random sample of individuals who are monitored over a number of years to determine how many among them will die of a lung cancer and were subject to smoke. Retrospective studies by contrast examine, ex- post, people who have died of a lung cancer and seek to determine how many among them were subjected to smoke. A retrospective study does not give a relative risk coefficient (), strictly speak- ing, but an approximation of it 6 Our emphasis in the text. 7 We do not deny that for those who are already vic- tims of chronic obstructive lung disease, smoke could be the cause of some discomfort. We do not deny either that some people may complain of minor eye irritations, and the like, if they stay long enough in a room filled with thick smoke. How- ever, what we do hypothesize is that it is far from established that someone can contract a disease requiring significant health care only because of environmental tobacco smoke. Later, we will ex- plain why we do not evaluate the sums non- smokers would be willing to pay to avoid such inconveniencies. 8 See also Chaloupka (1990), and Becker, Grossman and Murphy (1990). 9 For further details on epidemiologic methods see Miettinen (1972,1974), and Walter (1975; 1976). 10 Preliminary data for 1985-1986, from the Cana- dian Health Information Center. These data will be revised and published in the catalogue 82-206 (Statistics Canada, forthcoming b). 11 The average cost per day is taken from prelimi- nary data from the Canadian Health Information Center. Once revised, it will be published in table 117 of the catalogue 83-233 (Statistics Canada, forthcoming a) for the year 1986-1987. 12 See Saskatchewan Cancer Foundation (1987:25), and Saskatchewan Health, Medical Care In- surance Branch (1989). 13 Marginal cost and benefit are increments. Usually we assume these increments to be small. How- ever, in the present circumstances, we have no choice but to take the total cost as if it were an in- crement over 0. 14 Appendixes are available from the authors upon request. Appendix 1 gives the number of deaths according to cause and sex. It also describes the model which simulates what would have been the life and death of smokers had they not smoked and lived longer. Appendix 3 discusses the discrepancy between our 21,841 estimate of the number of deaths due to smoking and that of 35,131 from Col- lishaw, Tostowaryk and Wigle (1988). 15 From 1946 to 1990, the ex post real rate of return on 10-year Canadian bonds was never higher than 3%, except once in 1989. Indeed, it has been almost nil or even negative for approximately two de- cades, from the end of the 40s to the end of the 50s, and from the mid-'70s to the mid-'80s. Thus, over the long run, it is reasonable to believe that the maximum real rate of return on risk-free assets is not superior to 3%. Obviously, the discount rate could be inferior, but then, our conclusions would be even more favourable to smokers. 16 Additional medical services costs were found to be $68 million compared to $546.5 million for addi- tional hospital care costs which gives a ratio of one to eight. We will find in the following section that the reduction of future medical services costs is $123.5 million. Applying the ratio.of one to eight 314 Andre Raynauld and Jean-Pierre Vidal 2501357050

would lead us to think that the reduction of future hospital care costs could be as high as $988 mil- lion. 17 We took as a reference the average costs as calcu- lated by Regie de 1'Assurance Maladie du Quebec (1987). We made two adjustments: one to recon- cile the average payments on a fee-for-service basis with overall medical services, and one to rec- oncile the per capita medical service cost in Que- bec with that in Canada. 18 During 1986, the total number of old age re- cipients of residential care was established at 155,381. Assuming linear progression, this figure is derived from an interpolation carried out for June 1, between the March 1, 1986, figure of 154,502 (see Statistics Canada, 1989a) and the 158,019 figure of March 1, 1987, (see Statistics Canada,1989b:119). The breakdown of this figure by age-group was performed according to table 6 in (Statistics Canada, 1989a). 19 We first found that public sources account for 65% of the residential care facilities income by sub- stracting private financial sources (co-insurance, $539 million; supplements for rooms, $26 million; miscellaneous income, $46 million) from a total in- come of $L73 billion (Table 9 of Statistics Canada, 1989a). We then took 65% of the average daily cost of $54.40 (Statistics Canada, 1989b:119), to obtain $35.17. 20 In an appendix, which is available from the authors upon request, we explain how we obtained the number of taxpayers among the smokers who would have survived between 1986 and 2071 had they not smoked. 21 This estimate is for calendar year 1986, and has been supplied by the Department of Finance, Government of Canada. 22 The estimates of the federal excise tax as well as the federal excise duty apply to fiscal year 1986- 1987 and come from the Public Accounts (Canada, 1986). 3 The estimate of the tariff duty applies to the cal- endar year and comes from Statistics Canada (1986). 24 The figure published originally was $1,987.7 mil- lion. This figure was increased to include revenues from the general sales tax on tobacco in New- foundland and Ontario. 25 The breakdown is as follows: $13,234,590 for Old Age Security and Supplement; $7,394,088 for Canada and Quebec Pension Plans; and, $8,832,962 for private pension plans. 26 Although this approach based on averages is quite rough, it still requires more than 250 tables, which shows that a more precise estimation would have required a study of its own. Detailed calculations would certainly have to take into account the changes in the number and age distribution of beneficiaries, of spouse's allowances and of pen- sions to survivors. As well, it would be necessary to calculate the change in the amounts paid out because a member of a couple dies. But, since the average income between men and women differs, it would also be necessary to take into account that the amount offered by some programs depends on general income and on other benefits already being paid out. 27 Leu and Schaub (1983), have been criticized by Forbes and Thompson (1985). A reply may be found in Leu and Schaub (1985). 28 The form of the curve does not depend on the transformation function, which is linear, but on the etiologic fraction function (see section III). More precisely, the higher the Euclidian norm uik N, the higher will be the first derivative of the cost function (which is positive) and the higher will be the absolute value of the second derivative (which is negative). References Atkinson, A.B. and T.W. Meade (1974) 'Methods and preliminary findings in assessing the economic and health services consequences of smoking, with particular reference to lung cancer,' Journal of the Royal Statistical Society, Part 3:297-312. Atkinson, A.B. and J.L. Townsend (1977) 'Economic aspects of reduced smoking,' Lan- cet, 3:492-4. Bartlett, L. (1988) `Smoking: we cannot afford the cost,' Canadian Medical Association Journal, 138:644-5. Becker, G.S., M. Grossman and K.M. Murphy (1990) 'An Empirical Analysis of Cigarette Addiction.' Center for the Study of the Economy and the State Working Paper 61. University of Chicago, Chicago. Becker, G.S. and KM. Murphy (1988) 'A Theory of Rational Addiction,' Journal of Political Economy, 96:675-700. Bonilla, C.E. (1989) Determinants of Employee Absenteeism. Washington: US Chamber of Commerce, National Chamber Foundation. Cady, B. (1986) 'Payment by Nonsmokers for Smoking Related Illness,' Journal of the American Medical Association, 256:10:1291. - (1983) 'Cost of Smoking,' New England Journal of Medicine, 308(18):1105. Canada (1986) Public Accounts 1986-1987 (Ot- tawa). Canadian Tax Foundation. (1986-87) The National Finances (Toronto). Chaloupka, F.J. (1990) 'Rational Addictive Be- havior and Cigarette Smoking.' National Smokers' Burden on Society 315

Bureau of Economic Research Working Paper 3268. Cambridge, MA. Choi, B.C.K and J.R. Nethercott (1988) 'The Economic Impact of Smoking in Canada,' In- ternational Journal of Health Planning and Management, 3:197-205. Collishaw, N.E. and G. Myers (1984) 'Dollar Estimates of the Consequences of Tobacco Use in Canada, 1979,' Canadian Journal of Public Health, 75:192-9. Collishaw, N.E., W. Tostowaryk and D.T. Wigle (1988) 'Mortality Attributable to Tobacco Use in Canada,' Canadian Journal of Public Health, 79:166-9. Forbes, W.F. and M.E. Thompson (1978) 'Ciga- rette Smoking: Medical Costs vs Tax Re- ceipts,' Journal of the American Medical Association, 240:9:828. - (1982) 'Cogts and `benefits' of cigarette smoking in Canada,' Canadian Medical As- sociation Journal, 127:831-2. - (1983) 'Estimating the Health Care Costs of Smokers,' Canadian Journal of Public Health, 74:3:183-90. - (1985) 'Reasons for the Disagreements on the Impact of Smoking on Medical Care Ex- penditures: A Proposal for a Uniform Ap- proach,' Social Science and Medicine, 2L•7:771-3. Freour, P. et al. (1976) 'Le coat du tabagisme en France,' Bulletin de l'Academie Nationale de Mbdecine,160:6:583-91. Gorsky, RD., E. Schwartz and D. Dennis (1990) 'The morbidity, mortality, and economic costs of cigarette smoking in New Ham- pshire,' Journal of Community Health, 15:3:175-83. Gray, A.J. et al. (1988) 'The cost of cigarette smoking in New Zealand,' New Zealand Medical Journa1,101:844:204-7. Health and Welfare Canada (HWC) (1988) The Smoking Behauiour of Canadians -1986 (Ot- tawa: Supply and Services Canada). - (forthcoming) Table 6 from National Health Expenditures in Canada 1975-1987. Heins, M: (1978) 'Health Costs of Alcohol and Tobacco: Who Pays?' New England Journal of Medicine, 298. Hinds, M.W. (1986) 'Medical Care Costs Attribu- table to Cigarette Smoking in Kentucky,' Southern Medical Journal, 79:6:665-8. Hoffenberg, R (1988) 'Penalising Smokers and Drinkers,' Lancet,1:8586:649. Kristein, M.M. and DA Grove (1978) 'Who Pays Health Costs of Alcohol and Tobacco?' New 316 Andrd Raynauld and Jean-Pierre Vidal England Journal of Medicine, 299:606-7. Labour Canada (LC) (1987) Report of the Domin- ion Fire Commissioner, Annual report 1986- Losses caused by fires in Canada (Ottawa). Leu, RE. (1983) 'What Can Economists Con- tribute?' In M. Grant M. Plant and A. Willi- ams (eds.), Economics and alcohol (London: Croom Helm). - and T. Schaub (1983) 'Does Smoking In- crease Medical Care Expenditure?' Social Science and Medicine, 17:1907-14. - (1984) 'Economic Aspects of Smoking,' Ef- fectiue Health Care, 2:3:111-23. -(1985) 'More on the Impact of Smoking on Medical Care Expenditures,' Social Science and Medicine, 21:7:825-7. Luce, B.R and S.O. Schweitzer (1978) 'Smoking and Alcohol Abuse: A Comparison of their Economic Consequences,' New England Journal of Medicine, 298:10:569-7L Markandya, A. and D.W. Pearce (1989) 'The so- cial costs of tobacco smoking,' British Jour- nal of Addiction, 84:1139-50. Miettinen, O.S. (1972) 'Components of the crude risk ratio,' American Journal of Epidemi- ology, 96:2:168-72. - (1974) 'Proportion of disease caused or pre- vented by a given exposure, trait or interven- tion,' American Journal of Epidemiology, 99:5:325-32. Mishan, E.J. (1971a) Cost-Benefit Analysis (Lon- don: Allen & Unwin). - (1971b) 'Evaluation of life and limb: a theoretical approach,' Journal of Political Economy, 79:4:687-705. -(1981) 'The value of trying to value a life,' Journal of Public Economics,15:L-133-37. Petawawa National Forestry Institute (PNFI) (1989) Unpublished tables (Chalk River. For- estry Canada). Regie de 1'Assurance Maladie du Quebec (RAMQ) (1987) Statistiques annuelles 1986 (Quebec: Service des communications de la RAMQ). Rice, D.P. (1966) Estimating the Cost of Illness. Health Economic Series No 6. (Washington: US Department of Health, Education and Welfare, Public Health Service). Saskatchewan Cancer Foundation (1987) An- nual Report 1986-1987. Regina. Saskatchewan Health, Medical Care Insurance Branch (1989) Totalisation spdciale: cout des services midicaux suivant les codes de mala- dies410a414et490a496. Seltzer, C.C. (1980) 'Smoking and Coronary 2501357052

Heart Disease: What Are We to Believe?' American Heart Journal, 100:275-80. - (1989) 'Framingham Study Data and "Es- tablished Wisdom" About Cigarette Smoking and Coronary Heart Disease,' Journal of Clinical Epidemiology, 42:8:743-50. Shillington, E.R. (1977) Selected Economic Con- sequences of Cigarette Smoking (Ottawa: National Ministry of Health and Welfare). Shultz, J.M. (1985) 'Perspectives on the Economic Magnitude of Cigarette Smoking,' New York State Journal of Medicine, 85:7:302-6. Statistics Canada (1986) Imports, merchandise trade -1986. Cat. 65-23 (Ottawa). -(1988) Age, sex and marital status. Cat. 93- 101 (Ottawa). -(1989a) Residential care facilities for the aged. Cat. 83-237 (Ottawa). . - (1989b) Health Reports. Cat. 82-003 Quar- terly (Ottawa). -(forthcoming a) Table 117 of Cat. 83-233. -(forthcoming b) Table 1 of Invoice 41017 (Formerly Hosp. Morb.) Cat. 82-206. Sterling, T.D. and J.J. Weinkam (1987) 'Errors in Estimates of Smoking-Related Deaths Derived from Nonsmoker Mortality,' Risk Analysis, 7:4:463-75. Stoddart, G.L. et al. (1986) 'Tobacco Taxes and Health Care Costs: Do Canadian Smokers Pay Their Way?' Journal of Health Econom- ics, 5:63-80. Sullivan, P. (1989) 'Make it too costly for people to -smoke, finance minister advised,' Canadian Medical Association Journal, 140:5:546. Tollison, R.D. (1986) Smoking and Society, Toward a More Balanced Assessment (Toronto: Lexington Books). TJS_ Department of Health and Human Services (USDHHS) (1982) The Health Consequences o f Smoking: Cancer. A Report of the Surgeon General (Washington: Superintendent of Documents, US Government Printing Of- fice). - (1983) The Health Consequences of Smok- ing: Cardiovascular Disease. A Report of the Surgeon General (Washington: Superinten- dent of Documents, US Government Print- ing Office). - (1984) The Health Consequences of Smok- ing: Chronic Obstructive Lung Disease. A Re- port of the Surgeon General (Washington: Superintendent of Documents, US Goverm ment Printing Office). - (1986) The Health Consequences of Inuo1- untruy Smoking. A Report of the Surgeon General (Washington: Superintendent of Documents, US Government Printing Of- fice). -(1990) Smoking and Health, A National Status Report. A Report to Congress. Pub- lication No. (CDC) 87-8396 (Washington: DHHS). Viscusi, W.K. (1990) 'Do Smokers Underesti mate Risks?' Journal of Political Economy, 98:6:1253-69. Walters, S.D. (1976) 'The estimation and inter- pretation of attributable risk in health re- search,' Biometrics, 32:828-49. - (1975) 'The distribution of Levin's measure of attributable risk,' Biometrika, 62:2:371- 75. Warner, K.E. (1983) 'The Economics of Smok- ing: Dollars and Sense,' New York State Jour- nal of Medicine, 83:13:1273-4. Watts, G.T. (1983) 'Cost of smoking,' New Eng- land Journal of Medicine, 309:15:929. Wilkinson, R. et al. (1978) 'Health Costs of Alco- hol and Tobacco: Who Pays?' New England Journal of Medicine, 298:22:1262-3. Woodfield, A.E. (1984) 'Costs and "benefits' of cigarette smoking in Canada: comment,' Canadian Medical Association Journal, 130(2):118-20. Smokers' Burden on Society 317