Tobacco Documents Online

. f i Journal of Health Economics 5(1986) 63-80. North-Holland RECElVIE-D I JUL 9 1986 / F Y TOBACCO TAXES AND HEALTH CARE COSTS Do Canadian Smokers Pay their Way? Greg L. STODDART and Roberta J. LABELLE* McMaster University, Hamilton, Ont., Canada L8N 3Z5 Morris L. BARER and Robert G. EVANS* Universit y of British Columbia, Vancouver, B.C, Canada V6T /WS Received March 1985, 6final version received September 1985 A. WHIST r- ( Through their health care utilization, smokers are generally perceived to be imposing a financial externality on non-smokers within health insurance systems. To investigate the empirical basis for this view, we estimated publicly financed health care expenditure attributable to smoking for the Canadian province of Ontario and compared it to tobacco taxes paid by Ontario smokers. Both initial estimates and the results of sensitivity analyses performed on key assumptions and parameters of the estimation methodology rejected the hypothesized existence of a financial externality arising from smokers' health care utilization. 1. Introduction The health consequences and the economic impact of cigarette smoking have been extensively documented. Smoking is a major cause of morbidity and mortality [USDHEW (1979), USDHHS (1982, 1984)] and generates significant costs, including both direct costs of health care and indirect costs of lost productivity [Luce and Schweitzer (1978)]. It causes property damage through fires, raises the cost of fire protection, and results in increased production costs of many goods and services through the need for extra ventilation and maintenance [Shillington (1977), Kristein (1977), Luce and Schweitzer (1978), Collishaw and Myers (1984)]. In addition, smoking imposes intangible costs of discomfort, pain and suffering on smokers themselves, their families and others. *We wish to thank Jane Fulton for her able research assistance. We are grateful to Fred Bass, David Feeny, Jonathan Lomas and Stephen Walter for constructive comments during the preparation of this manuscript. This research was supported by a grant from the Health Education and Promotion Branch of the Ontario Ministry of Health and by National Health Research Scholar Award no. 6610-1231-48. 0167fi296; 86;'S3.50 (i_~ 1986. Elsevier Science Publishers B.V. (North-Holland)

, 64 G.L Stoddart tt al. Tobacco taxes and health care costs Rigorous cost-benefit analysis of smoking is a difficult exercise.t The distribution of the costs and benefits of smoking, however, is as much a matter of public concern as their relative size. Smokers impose a physical externality on non-smokers, in the form of both discomfort at the time of exposure and subsequent health effects of `passive smoking'. The provision of separate areas for smokers and non-smokers in many service establishments, often in compliance with local by-laws resulting from public pressure by non- smokers, is one example of attempts to correct this externality. Increasingly, however, smokers are perceived to be imposing an additional, financial externality on non-smokers through the health care costs that smoking generates [Atkinson and Townsend (1977, p. 494), Wikler (1978, p. 318), Somers (1980, p. 1063)]. Concern about this financial externality often surfaces in a political context, where the line between 'encouraging individual responsibility' in lifestyle choices and 'victim blaming' can be easily blurred [Heckler (1984), Barer et al. (1984, ch. 1)]. It has prompted calls for special or increased taxes on cigarettes [Wilkinson et al. (1978), Heins (1978), Kristein and Grove (1978), Cady (1984)] and is often seen as one rationale for broader public policy initiatives against smoking [Wikler (1978), Ontario Council of Health (1984)]. Wikler (1978), Cullis (1978), and Harnes (1978) have identified the complexity of the externality issue. More recently, Leu and Schaub (1983) and Leu (1984, p. 115) have challenged the belief that smoking imposes a large cost burden on health service systems. Nevertheless, the impression persists that the health care costs attributable to smoking constitute a financial burden on non-smokers in excess of the revenues raised through excise taxes on tobacco. In this paper we present evidence relevant to the assessment of the financial externality associated with smoking. We estimate the publicly financed health care expenditure attributable to smoking for the Canadian province of Ontario during 1978 and compare this expenditure to tobacco tax revenue derived from Ontarians during the same year. Our results suggest that in the Canadian context of universal, publicly financed health 'The analysis attempted by Thompson and Forbes (1982) and subsequent comment by Woodfield (1984) demonstrate that cost-benefit analysis of smoking is not a straightforward task. From a social viewpoint, all direct, indirect and intangible costs must be weighed against consumer willingness-to-pay (actual expenditure plus consumer surplus) for tobacco products. Partial analyses of costs and benefits can be constructed from Luce and Schweitzer (1978) and Collishaw and Myers (1984). Luce and Schweitzer estimated the total cost of health care, lost production and fire losses attributable to or associated with smoking to be 527.5 billion for the United States in 1976, compared to expenditures of 515.7 billion by American smokers for all tobacco products. Collishaw and Myers estimated similar costs to be $5.2 billion for Canada in 1979, compared to consumer expenditures on tobacco of $3 billion. The interpretation of expenditure data as free and informed consumer choice is problematic, however, given the addictive nature of smoking and the very young ages at which decisions to smoke are frequently made. insura throul Hea specifi each c for trc sectiot genera specifi smoki utiliza insure smoki compi and tI the ep 7 we extern policy anc st: Epid smok illnm condi of infi such dose- consi: absen Trout As smok assoc w hicl ( w•ith 'Coi provid (1979), Furthe be fou ~ an earl ~ ~ + Gt N ~ ~

tercise. t The 'i as much a ;e a physical t the time of : provision of ;T.ablishments, ssure by non- in additional, n:. costs that rWikier (1978, al externality 'encouraging ning' can be tas prompted et al. (1978), often seen as king [Wikler dentified the k:haub (1983) ng imposes a te impression constitute a tised through ,tnent of the the publicly the Canadian •e to tobacco Our results tanced health xtt comment by i straightforward ueeighed against Abacoo products. eitur (1978) and health care, lost '.5 billion for the i smokers for all >a for Canada in interpretation of xever, given the ke are frequently G.L Stoddart et al.. Tobacco taxes and health cern costs 65 insurance, smokers do not impose a net financial externality on non-smokers through their health care utilization. Health care expenditure attributable to smoking is a function of the specific diseases or conditions associated with smoking, the proportion of each disease attributable to smoking, the quantities of health services utilized for treatment of each disease, and the unit costs of those health services. In section 2 we describe the illness burden of smoking, identifying diseases generally accepted as causally linked to smoking, and presenting age-sex specific estimates of the proportions of these diseases attributable to smoking. In sections 3 and 4 we combine the epidemiologic information with utilization and expenditure data to derive estimates of expenditures on insured hospital and physician services, respectively, that are attributable to smoking. We report revenues from tobacco taxes in section 5. The comparison of health care expenditure to tax revenue is made in section 6, and the results are subjected to a series of sensitivity analyses involving both the epidemiologic and the utilization and expenditure assumptions. In section 7 we identify factors contributing to the (mis)perception of financial externality and compare our results to those of other investigators. The policy implications of our results are discussed in section 8, and conclusions are stated in section 9. 2. Illness burden of smoking Epidemiologic studies have repeatedly demonstrated an association between smoking and increased risk of morbidity and mortality from a range of illnesses; however, the assessment that smoking 'causes' a particular condition must be made with caution because of the well known difficulties of inferring causation from correlation. Additional dimensions of the evidence such as the strength of the association, internal consistency, temporality, dose-response characteristics, biological plausibility, specificity and external consistency must be considered in arriving at assessments of causation in the absence of data from controlled experiments [Mausner and Bahn (1974), Trout (1981)]. As a starting point for estimation of health care expenditure attributable to smoking we identified, from epidemiologic reviews of evidence on the association between smoking and ill health, the four major conditions for which smoking has generally been accepted to be a causal agent.2 They are (with ICDA-8 codes in parentheses): (i) lung cancer (162, 163.0), including 2Comprehensive reviews of the association between smoking and ill health have been provided in the U.S. Public Health Service series of reports to the Surgeon General [USDHEW (1979), USDHHS (1982, 1984)]. See also Doll and Peto (1976), Tweed (1982), and Doll (1984). Further discussion of this and other epidemiologic evidence on the health effects of smoking can be found in Barer et al. (1984, ch. 4) which reports on the larger research project that generated an earlier version of the analysis and results in this paper.

66 G.L Stoddart et at., Tobacco taxes and health emt easts malignant neoplasms of the trachea, bronchus, lung and pleura, (ii) coronary heart disease (410, 411, 412, 414), including acute myocardial infarction, other acute and subacute forms of ischaemic heart disease, chronic ischaemic heart disease and asymptomatic ischaemic heart disease, (iii) bronchitis (490, 491), including unqualified bronchitis and chronic bronchitis, and (iv) emphysema (492). This initial assessment of conditions causally related to smoking is conservative. Some evidence exists linking smoking to mortality from other cancers, pulmonary disease, and certain ulcers [USDHHS (1982), Doll (Il984)). Smoking may also exacerbate the severity of conditions caused by other factors. Therefore smoking plays a larger role in the generation of health care expenditure than our four primary conditions indicate. The significance of an expansion in the set of smoking-related diseases was addressed through sensitivity analysis and is discussed in section 6. Baseline estimates of the proportion of each disease in the general population that is attributable to smoking were drawn from earlier work by Shillington (1977) on the economic consequences of cigarette smoking in Canada. Shillington reviewed data from studies of cause-specific mortality rates and morbidity rates among smokers and non-smokers to calculate the proportions attributable that are shown in table 1.3 Estimates derived from both morbidity and mortality data are shown. However, the lower quality and restricted availability of morbidity studies [Shillington (1977, p. 4)] rendered the use of morbidity-derived proportions less satisfactory for our purpose. Therefore, we produced range estimates of health care expenditure attributable to smoking by applying both the morbidity- and mortality-derived sets of proportions attributable to all utilization, both hospital and physician. In later sections we focus primarily upon the mortality-derived results because they generate an upper estimate of health care expenditure and thus contribute to a more robust test of the hypothesis that smokers do not `pay their way'. As is evident in table 1, the data necessary to calculate proportions attributable were unavailable for some age-sex cells. In other cases, such as coronary heart disease in females, the baseline estimates of Shillington 3Data sources, which included major prospective studies of smokers, are described in Shillington (1977, pp.4-7). Available sources did not always allow separate estimates for bronchitis and emphysema. Proportions attributable can be calculated from data on the number of smokers and non-smokers in a population and the morbidity or mortality rates of each group according to the formula for the etiologic fraction, ¢(RR -1)/(¢(RR-1)+ 1), where ¢ is the proportion of the population exposed to risk and RR is the relative risk of morbidity or mortality in those exposed, calculated as the rate of disease in exposed subjects divided by the rate of disease in unexposed subjects. Technically, the etiologic fraction measures the proportion of incident cases in a population that is attributable to an exposure; however, if the duration of illness is not significantly different in smokers and non-smokers it is also an acceptable measure of the proportion of prevalent cases attributable.

0 '+ -0 N ~' (td °, g' ~ o o w 9 CY A ~•C ..~.. (V .O-p y ~ ~ rp O0®o n ~ c~ ~ . - r y S QQ Table I Proportions of smoking-related diseases attributable to smoking, by age and sex.' Morbidity derived estimates Mortality derived estimates Lung cancer Coronary heart disease Bronchitis and emphysema Lung cancer Coronary heart disease Bronchitis Emphysema Male Male 15-24 b b 34.7 <35 0.0 0.0 0.0 0.0 25-34 b b 39.5 35-39 80.0 27.6 68.5 79,0 35-44 61.2 b 39.5 40-44 80.0 73.5 68.5 79.0 45-54 59.3 26.0 37.6 45-49 62.4 50.1 66.8 77.6 55-64 59.3 13.2 37.6 50-54 69.8 46.1 66.8 77.6 65 + 47.2 5.8 27.0 55-59 86.2 28.7 66.8 77.6 60-64 87.2 ' 20.1 66.8 80.1 65+ 69.8 8.5 57.0 76.0 Female 15-24 b 0.0 33.4 Female 15-34 0.0 0.0 0.0 25-34 b 0.0 37.1 35-39 41.5 0.0 0.0 35-44 42.8 0.0 37.1 40-44 41.5 0.0 0.0 45-54 37.2 0.0 31.9 45-49 36.0 23.8 54.9 55-64 37,2 0.0 31.9 50-54 36.0 23.8 54.9 65+ 16.6 0.0 13.6 55-59 21.9 17.9 54.9 60-64 21.9 17.9 54.9 65+ 8.6 4.0 40.6 'Source: Shillington (1977, pp. 4-7). bNo data available.

68 G.4 Stoddart a at., Tobacco taxes and health care costs probably understate actual proportions attributable since some effects of smoking on women have only recently become identifiable. To adjust for possible understatement of the proportions attributable in the baseline estimates, we conducted sensitivity analyses using an alternative assumption of significantly higher proportions. The application of the proportions attributable to the allocation of hospital and physician utilization is described in the following two sections. 3. Expenditure on hospital services Expenditure on hospital services attributable to smoking was calculated by disease, age, and sex category as the product of the number of hospital patient days, the proportion attributable to smoking, and the per diem cost of hospital care. The number of patient days utilized reflects both disease prevalence and prevailing clinical practice. Data on the age-sex distribution of patient days by ICDA disease category were obtained from the Health Division of Statistics Canada. The per diem hospital cost for each disease category was obtained by adjusting the 1978 per diecn cost of $165.06 for all Ontario hospitals [Ontario (1979)] with disease-specific weights derived from earlier work on Ontario hospital costs by Barer (1981). The resulting disease-specific per diems appear in table 2.` Table 2 presents estimates of hospital patient days attributable to smoking and hospital services expenditure attributable to smoking, by sex for each disease category. Estimates based on both morbidity- and mortality-derived proportions attributable are presented. 4. Expenditure on physician services In principle, expenditure on physician services attributable to smoking can be estimated, using methods analogous to those for hospital services, as the product of the utilization of physician services, the proportion attributable to smoking, and the average fee per service. However, Canadian ambulatory care utilization data are available only in considerably less detail than hospital utilization data. Information on both the number of physician `Barer estimated cost functions for Ontario hospitals with pooled cross-scction and time-series data. But the impact of the increasing technologic sophistication of hospitals (especially the spread of coronary care units) during the 1970s may not be fully reflected in the disease-specific per diems shown in table 2, if the care of smoking-related illness has increased in intensity more rapidly than that for all other illnesses taken together. Even markedly higher relative per diems for smolcing-related illnesses, however, do not alter the conclusions of this analysis, as demonstrated in section 6. It should also be noted that the per diems we employ are a ratio of total hospital expenditures to inpatient days. This implicitly assumes that outpatient and educational activity are attributable to smoking-related illness in the same proportions as inpatient. The growing emphasis on hospital-based ambulatory, or non-inpatient, care will affect our results only insofar as it may differ between smoking-related and other illnesses.

~0 .5 r^J .~ %D OU c. Xf ~• ~ A e`Z- ~ ~ N `C EE ~N c.y'n;3~c wo ~ 0 ,f.,2 0 o~A ~ W ~NIKi o ~ < 9 ~pr1'~ ao°a c`o• p rr o ~ e . A ~ O n» 0 -,, Table 2 Expenditure on hospital services attributable to smoking, Ontario, 1978, Patient days attributable to smoking Per di Cost attributable to smoking (S'000) Morbidity estimate Mortality estimate em cost (S) Morbidity estimate Mortality estimate Lung M 45,376 62,745 188.17 8,538 11,807 cancer F 7,751 5,580 188.17 1,459 1,050 Coronary M 42,677 85,677 146.90 6,269 12,586 heart disease F 0 20,480 146.90 0 3,009 Bronchitis M 15,335 33,381 122.15 1,873 4,077 & emphysema F 6,431 11,718 122.15 786 1,431 Total 117,570 219,581 18,925 33,960 GOTVC.RP ;:4C,Q%

70 G.L Stoddart et at.. Tobacco taxes and health care costs services utilized and expenditures for physician services in Ontario was unavailable on either an age-sex or an ICDA category basis. Alternative data sources and procedures were therefore employed. First, per capita expenditures on physician services by age, sex and ICDA disease category for the province of Saskatchewan in 1971 as reported by Shillington (1977) were converted to 1971 Ontario values, using inter- provincial expenditure indices available from Health and Welfare Canada [Canada (1979a)].s This conversion adjusted the Saskatchewan data for both the more intensive servicing patterns (higher real utilization per capita) and the higher level of physician fees that prevailed in Ontario. Next, the 1971 Ontario expenditure data were converted to 1978 values using indices of physician servicing intensity and fee levels, compiled by Barer and Evans (1984) and Wolfson, Evans and Lomas (1980), respectively. Census data [Canada (1979b)] were used to obtain provincial expenditures on physician services from the per capita data. Finally, the proportions attributable in table I were applied to the provincial expenditures on physician services by age, sex and disease category to obtain estimates of expenditure on physician services attributable to smoking. Estimates based on both morbidity- and mortality-derived proportions attributable are shown in table 3. Table 3 Expenditure on physician services attributable to smoking, Ontario, 1978 (S'000). Morbidity estimate Mortality estimate Lung M 625 890 cancer F 72 53 Coronary M 1,153 2,180 heart disease F 0 504 Bronchitis M 659 1,341 & emphysema F 153 218 Total ' 2,662 5,186 SThe Shillington data set was assembled on a`one-time' basis by integrating data from Statistics Canada, the Saskatchewan Medical Care Insurance Commission, and the Saskatchewan Cancer Commission [Shillington (1977, p. 8)]. Although it provides the best starting point for our estimate of expenditure on physician services, its adoption may result in some underestimation of expenditure on physician services if expenditures on lung cancer, coronary heart disease, bronchitis, and emphysema rose more rapidly over the 1971-78 period than expenditure on all diseases. In the absence of physician utilization or expenditure data by diagnostic category this is difficult to ascertain. Furthermore the use of Saskatchewan data carries with it the implicit assumption that relative physician service patterns were comparable across age, sex and disease categories in the two provinces. Given the magnitude by which tobacco tax revenue was found to exceed publicly financed health care expenditure, however, and given that expenditure on physician services constitutes approximately 10% of total health 5

G.L Stoddart et at., Tobacco taxes and health care costs 71 ntario was Alternative and ICDA eported by ising inter- ire Canada tta Cor both capita) and tv the 1971 ; indices of and Evans 'ensus data n physician ibutable in services by n physician bidity- and io, te tn$ data from on, and the vides the best tutay result in i lung cancer, 971-78 period diture data by richewan data re comparable ude by which ture, however, of total health 5. Revenue from tobacco taxes Tobacco excise taxes are levied at both the provincial and federal levels in Canada. Provincial tobacco tax revenue for Ontario for the fiscal year 1978/79 was $258,569,000 [Canada (1982)].6 Federal tobacco tax revenue, which was $710,579,000 nationally in fiscal 1978/79 [Canada (1981)], is not reported on the basis of revenue derived from the residents of each province.' Therefore the contribution of Ontario residents to federal tobacco tax revenue was estimated from data on the number of regular smokers in Ontario, average consumption of cigarettes, and federal excise tax and duty rates [Millar (1983), Canadian Tax Foundation (1979b)). The resulting estimate of 5226,446,000 should be viewed as conservative.8 6. Results and sensitivity analyses Publicly financed expenditures on hospital and physician services attributable to smoking for the four primary smoking-related diseases of lung cancer, coronary heart disease, bronchitis and emphysema are compared to tobaeco tax revenue in table 4. Using baseline estimates of proportions attributable, (both morbidity- and mortality-derived) we estimate that during 1978 between $20 million and $40 million approximately was spent by the province of Ontario for hospital and physician services attributable to care expenditure in this analysis, even a significant increase in these estimates would not alter the overall result. Expenditure on hospital services is the major determinant of total expenditure attributable to smoking, accounting for approximately 65%,. The remainder is accounted for by expenditures on othe>: health services or activities, introduced in section 6. bin 1979 the Ontario provincial tax rate was $0.012 per cigarette. Although the provincial component of the total tobacco tax rate varies across provinces, most provinces taxed at rates between $0.010 and 50.012 during the same year. Rates for the provinces of Newfoundland, which consistently taxes at higher than average rates, and Alberta, which consistently taxes at lower than average rates, were 50.016 and 50.0032, respectively [Canadian Tax Foundation (1979a)]. 'During fiscal 1978/79 the federal government levied an excise tax of 50.03 per five cigarettes and an excise duty of $5.00 per thousand cigarettes [Canadian Tax Foundation (1979b)], resulting in a total tax of $11.00 per thousand cigarettes. Although it is possible to obtain the portion of total federal tobacco tax revenue collected in each province, taxes are collected at the manufacturer or distributor level, not at the point of purchase. Consequently, these figures significantly overstate the contribution of residents of the provinces of Ontario and Quebec where most manufacturers are located. $Survey data generally underestimate average consumption, because of under-reporting by young smokers. An alternative method of estimating the contribution of Ontario residents to federal tobacco tax revenue would use the number of regular cigarette smokers in Ontario as a percentage of regular Canadian smokers, assuming that Ontarians' per capita consumption approximates the national average. This method resulted in an estimated contribution to federal tobacco tax revenue by Ontario residents of 5235,912,000. Excise taxes on cigars, pipe tobacco, and fine cut tobacco have been excluded from the analysis. A general manufacturers sales tax of 12% of selling price, levied by the federal government in fiscal 1978j/9, has also been exduded because reported revenue is not disaggregated by commodity.

72 G.L Stoddart et al., Tobacco taxes and heaith care costs Table 4 Publicly financed expenditures on hospital and physician services compared to tobacco tax revenue, Ontario, 1978 (S'000). Morbidity estimate Mortality estimate Expenditure on hospital services 18,925 33,960 Expenditure on physician services 2,662 5,186 Total expenditure 21,587 39,146 Provincial tobacco tax revenue 258,569 258,569 Contribution of Ontario residents to federal tobacco tax revenue 226,446 . 226,446 Total tax revenue 485,015 485,015 Expenditures on hospital and physician services as a% of tobacco tax revenue 4.5% 8 1'/ smoking. However, this expenditure amounted to only 4.5%-8.1 % of tobacco tax revenue derived from Ontarians. The sensitivity of these results to the estimation methods and values from which they were derived was tested by altering original values or assumptions in four areas, as shown in table 5. Both independent and interactive effects of the sensitivity factors were examined. All sensitivity analyses were performed on mortality-based figures, as these define the higher end of the range of our original estimates of health care expenditures. Estimates of per diem hospital costs were increased to allow for the possibility that the disease-specific adjustments derived from Ontario hospital cost analyses by Barer (1981) did not fully reflect increases in the relative technologic sophistication of treatment for smoking-related illnesses that may have occurred by 1978, especially for coronary care.9 Expenditures on hospital and physician services through Ontario's universal health insurance plan accounted for 68% of public sector health care expenditures in the province in 1978.11 For sensitivity analysis purposes, we assumed that the proportion of public expenditure on additional health 'Per diem values shown in table 2 for lung cancer and bronchitis and emphysema were increased by 10% to $207 and $134, respectively. The per diem cost for coronary heart disease was also set at S207 (to equal that for lung cancer), which represented a 40'/, increase. Because coronary heart disease accounts for almost half of the hospital patient days attributable to smoking in table 2, these alternative values are equivalent to the assumption of an across-the- board increase of 25% in the original per diem values. 1°i'he remainder consisted of expenditures on other institutional services, other health professional services, pharmaceuticals, public health and health promotion activities, research, and administration (Canada (1983)]. Although some of these expenditures were attributable to smoking, precise estimation is not possible because utilization and expenditure data for these categories of services and activities are not classified by disease.