Tobacco Documents Online

The Taxes of Sin Do Smokers and Drinkers Pay Their Way? ' Wiilard G. Manning, PhD; Emmett B. Keeler, PhD; Joseph P. Newhouse, PhD; ,Elizabeth M. Sioss, PhD; Jeffrey Wasserman, PhD We estimate the lifetime, discounted costs that smokers and drinkers Impose on others through collectively financed health insurance, pensions, disability insur- ance, group, iife insurance, fires, motor-vehicie accidents, and the criminal Justice system. Aithough nonsmokers subsidize smokers' medical care and group life insurance, smokers subsidize nonsmokers' pensions and nursing home payments. On balance, smokers probably pay their way at the current level of excise taxes on cigarettes; but one may, nonetheless, wish to raise those taxes to reduce the number of adolescent smokers. In contrast, drinkers do not pay their way: current excise taxes on alcohol cover only about hatf the costs imposed on others. (J.9MA.1989,2s1:1604-1so9) POOR health habits, such as smoking and heavy drinking, carry costs not only for smokers and heavy drinkers, but for everyone else as well. Concern about these costs- has prompted not only health-promotion efforts, but also pro- posals to increase both federal and state excise taxes on cigarettes and alcohol. Pbr such taxes to be at an economically efficient level, they must at least cover the costs to others that arise from smok- ing and heavy drinking. We term the costs to others external costs, in con- trast to those borne by the smoker or heavy drinker, which we term internal costs. Some external costs are obvious, for example, the damage caused by drunk driving and passive smoking; others are more subtle, for example, the higher medical costs of smokers that are S nanced by health insurance premiums and payroll taxes. SWh premiums and payroll taxes are'the same for smokers and nonsmokers (unlike individual life insurance premiums). As a result, non- smokers may susidize smoking. Our purpose in this article ia to quan- tify external costa. Earlier estimates of the costs p~gmroking and drinking~' (Of- ~ Fiom The lJniveRitya M~idrquL Ann Arbor (or Man. nNq): The w1NO Vorporatlon, suua Mw" CaGr (urs Manwft. KAel86 Newhou». sloa. and wasserman); C,e oivt.ton oa tieaNN Policy Rmearct, and Educmion, Harv.rd UNuenlry, c;ambridqe, Mass (or NewtxousA): and syneM.amuccsraw4iii. sara earcam caar (orweasmmen} The opini«r and c«,cknton.xpreseed h&vin ane aoWyame af8,e sulhorsr,d drou/d notbe carwuued as mpmswdv m® poaicis «o*0ora or n,e RAND Co rpo«atbn or.riyaa«icr a tt» Us cRwemm«x or any aft individuals named herein. Reae" mciumts w ft oep.ran«rt a N"un sef vices Manepement Lnd Pbft The Univsrsky d Miohl- yan. 1420 Washinqton Fleiphts, Ann Arbor. M! 4910® (Or Mannkp). fice of Zlechnology Assessment, unpub- lished data, 1985) are not suitable for analysis of taxes because they do not always distinguish between internal and external costs, nor do they calculate the lifetime costs of poor health habits. METHODS Ext®rnal Costs and Their Estimation We illustrate our conceptual frame- work in terms of smoking, but the same principles apply to our analyses of drinking. Table 1 illustrates the division be- tween the internal and external costs of smoking. In the case of alcohol abuse, we also consider the costs of motor-vehi- cle accidents and criminal justice. One goal of an economically efficient tax on smoking or tobacco is to have the smoker bear the costs that he imposes on others when deciding whether or how much to smoke. Costs imposed on other family members, however, are dif6cult to c)assify as internal or exter- nal because it is not clear whether those coda would, in any event, be taken into account by the smoker. If they would be, then they are internal costs. Al though our base-case estimates classify such costs as internal, we show the ef- fect of treating certsin costs borne by otherfsmily members as external. A simple example that considers only medical costs may clarify the division between internal and external cost. Suppose a worker has a group health insurance policy that pays 75% of his medical bMs, and suppose that amoking a pack of cigarettes per day raises medi cai. bi7ls by;Ci000. The amount the work er pays, $1500 (fl.2b X 6000 =1500), is a component of internal costs. Because the smoker does not pay higher premi- ums that reflect his or her higher costs, the remainder of the cost, $4500, is a component of external costs. Zb estimate external costs, we should not contrast the medical and other ex- penses of smokers to nonsmokers, be- cause nonsmokers differ from smokers in other ways that affect the various components of cost such as medical ex- penses. For example, according to the 1983 National Health Interview Survey (NHIS), those who never smoke are 1.5 times more likely than current smokers to have more than a high school educa- tion. Rather, we contrast smokers to a hypothetical group of "nonsmoking smokers," people who are like smokers in age, sex, education, drinking habits, and several other ways described here- in, except that they have never smoked.' To test how sensitive our esti- mates are to differences between smok- ing and not smoking, however, we also contrast medical and other costs of smokers to those of actual nonsmokers. Our methods estimate lifetime costs by tracking expenditures for two hypo- thetical cohorts of men and women from age 20 years to death. One cohort smokes; the other does not. We develop life tables for each cohort showing the probability of surviving to each age from age 20 years. These tables come from applying estimates of the relative risk of smoking to the 1980 life tables of the US population.` Reiative risk was estimated by applying the 1984 Centers for Disease Control health risk apprais- al program' to the ever smokers in our sample twice-once with their actual smoking status and once with their smoking status changed to `hever smoked." In judging any policy that has long- term effects, it is important to discount future costs, thereby making costs that occur at different times commensurate. A dollar received today is worth more thana dollarreceived 15 years from now (even without inflation). A current dol- lar can be invested and earn interest so that at the end of 15 years it wil1 be worth more than $2 (at M Because the proper rate of discount is controversial, we have computed results for rates that span the range between 0% and 10%. The expected net external costs per pack are the sum of the immediate costs , 1604 JAMA, March 17, 1989-Yd 261, No. 11 Sin Tax-Marriftet ai TIMN 325066

k 'r ta a) _ t per pack and the cumulating lifetime costs per pack. We assume that the costs of flres, motor-vehicle accidents, and criminal justice are immediate; ie, each cigarette or ounce of ethanol has a certain probability of causing such costs in the immediate period after purchase, but once the cigarette is smoked or the alcohol consumed, the probability drops quickly to zero. For such costs, we di- vide estimated national annual costs by the annual packs (or excess ounces). The cumulative net lifetime external costs are given by the following: 96 1 a"0 x p(Aax)t x C(H)t t-20 96 - 4 at-20 xP(AWH)t X C(NH)t. t-zo where 8 indicates the annual discount factor (1l[1 +r]) if r is the discount rate; P(AIH), the probability of surviving from age 20 years to at least age t years, conditional on smoking; C(H), the annu- al costs minus taxes and premiums for smokers of age t; P(AINH), the proba- bility of surviving from age 20 years to at least age t years,'conditional on not smoking; and C(NH), the annual costs minus taxes and premiums for smokers of age t years if they had never smoked. The external costs come from collec- tively financed programs, including health insurance, pensions, sick leave, disability insurance, and group life in- surance. These programs are financed by taxes and premiums that do not dif- ferentiate between smokers and non- smokers. Because smokers have shorter life expectancies, they will pay less of the taxes and premiums that fi- nance these programa. Tb simplify the calculation of how much smokers and nonsmokers pay annually to finance these programs, we assume that each pays the same proportionpf earnings, where the proportion•isijust enough to finance these programa. The discount- ed, expected lifetime costs per pack are calculated by divi"_ the lifetime costs by the expected„ number of packs smoked in a lifetime In estima '*.tE~e external costs of smoking and S~m g, we relied on self- reported cons ption. Because people underreport their consumption, we have corrected for the difference be- tween actual and reported use. The re- ported number of packs per day was multiplied by 1.5, and reported alcohol consumption was multiplied by 2.5.'-' Our figures for pension income have been corrected for a 21% rate of underreporting.' Our estimates are based on data from a number of sources. The primary source for those under age 60 years is The RAND CorporationiB Health In- JAMA, March 17, 1989-W1261, No. 11 Table 1.-Costs of Smoidtq 1Ype InttmN Ezt.rn.1 Fremahxe death Smoker and tarrwy- Coworkers and aa,ees• Pak+ and sulterlnp Smoker and famky~ Coworkers and aU+ers• Medical costs Copaymstris kuturana mOrbursemenb Sick kave,, Uncovered sick losst , Covered sick bsst OisabiWy F«eqon. k;am. not r.W.c.d by disablYty Yisuranoe DtaabiYty insurance Group We iriaurance Negligible Death bensfN Pension Defined-oor0lbution plans Social Security and deflned-bereffl plans Wages Foregone disposable income Taxes on earnings Otfwr costs Property bea due to ftra paid by person insured propstty ias due to ftres Tobacco prcducss GOar.tta PtxctWa . . 4 'Pramature tttortaYty and suffettrtp among fart* members and coworkers k caused by passive smkinq. VS/e daasNy casis borne by ofhsr famlty members as internal costs, tsy otwered, we mean a,bJea to some kind of rnurat,ce or tncanw ror~laeemeM plan #Exdse uxss on cigarettes couid be considaed neQatiw external costs. M they.re so defined, the object of attr exercise would be b determine R external costs were zero, rather fhan equal to the current sxdss tax. Table 2.-External Costs per Pa& of Gpat:?ttes' Dlscotut: Raes External Costs 0% 6% 10% Costs per pedc $ Medical carat o.3s 0.26 0.18 Sick leave 0.01 0.01 0.01 Group Yfe insurance 0.11 0.05 0.02 Nurskq home -0.26 -0.03 0.00 Re!ltementpeasiorfx -1.82 -024 -0.02 Fhs 0.02 0.02 0.02 Taxes on eaminps b finartc. above proprams, $ -0.65 -0.09 -0.02 Total net ooets per pedc, t§ -0.91 0.15 024 LHe exped.nc.y at.pe 20 y per pedc, mm -137 -28 -6 'The rsmbsr of packs ®f cigarettes arm totrecled for underreporting. Costs (in 19b8 dollars) per pack ara calculated by divldtq by the discounted nurnber of pscks smoked. tlndudes all rwt mawrttitx weM, and denW car.. #Induees dtsab+Nty ins+,ranc.. gThe sum of costs mtrws taxes on eaminps, ep, costs at 5% equals 0.15 $0.26+0.01 +0.05-0.03-024+0.02 -(-0.o4 surance Experiment (HIE), because of its detailed information regarding hab- its and the medical reasons for the utili- zation of medical are.'a" Because per- sons aged 62 years or older at the time of enrollment were excluded from the HIE sample of 5809 persons, we used data regarding persons greater than age 59 years from a 1983 supplement to the NHIS. It included information re- garding health habits, health care use, andworklosainasampleof22418per- sons. In addition, we compared the 1983 NHIS results for nonelderly persons with those from the HIE. We have in- flated all cost data to 1986 dollars using the consumer price index. We estimated differences in spending for medical care servicesbetween those with and without each habit. Such dif- ferences, of course, may or may not be caused by the habit. We addressed this ambiguity in two ways. First, we con- trolled for the confounding characteris- tics described in the next section. Sec- ond, although our base-case estimates include all medical services except ma- ternity services and well care, we exam- ined their sensitivity to considering only costs that arise from diagnoses TIMN 325067 thought to be directly related to smok- ing and-excessive diinking; such as can- cer of the lung and cirrhosis of the liver. In addition to medical expense, we estimated the difference in days lost from work between persons with and without each habit, controlling for the confounding variables described herein. The collectively financed cost of days lost from work was computed by multi- plying the daily wage by 0.38, the em- ployers' average share of the cost of work loss through covered sick leave. - When estimating the cost of drinking, we controlled for smoking status, and conversely. Had we not done so, we would have attributed some of the costs of smoking to drinking if smokers tend to drink heavily. We classified persons as former cigarette smokers, current cigarette smokers, current pipe or cigar smokers, and never smokers based on their responses to a smoking history questionnaire filled out at the time of enrollment in the study. We classified persons as abstainers, former drinkers, and current drinkers based on respons- es to the same questionnaire. We col lapsed information regarding the cur- rent drinkers' consumption of beer, Sin Tax-Mannitq et ai 16N

5 -20r- / Percent Discount Rate -30 -40 ~ -50 .4 -60 ~ V - 70 - 80 -90 -100 Extemai costs of poor health habits at altemative discount rates. Table 3.-Sensitivity of Extemal Costs (in 1986 Dollars) per Pack to Assumptions at 5% Discount Rate 10 Fxtamal Costs Baa. Casa• All Data From National NeaftN Interview Survey Comparison With Nawr Smoker l.ower Bountft Total Costs* Costs per padc, $ Medical care 0.26 0.26 0.30 0.15 0.36 Sick leave 0.01 0.05 0.04 0.01 0.03 Group Gte insuraroe 0.05 0.05 0.06 0.05 0.05 Nuraing home -0.03 -0.03 -0.02 -0.03 -0.03 Retirement pension§ -0.24 -0.24 -0.20 -0.38 -0.24 Fires 0.02 0.02 0.02 0.02 0.02 Taxea on eamings per pack. t -0.09 -0.09 -0.091 -0.06 -0.931 Tota/ net costs per pack. tf 0.15 0.20 028/ -0.15 4'• •Enect of a,anyir,a «,rert and tom,er smokers to neW arrrokars, w;th other a,araaer+mcs hNd aon,r,rx. tNamow definition of medical atteas, with no effects of smoking on early reur.m.nt. tinerueea:,temal costs. gtndudes asability inaixuro.. than Value smoksnm per paac b~ecauae of htq/w earnirq _ iat~but~~ imp actually tt~iatsnie~ h cerft ~ .amnp ~sratsa ix causapy related to smaanp, and we have asaumed tt»y aw na teamrqe, na taxes an.am:qa. #Sum of costs mirrn taxae an earrwqs. ••t.osa of it. and pain aod;uFFerinp by smoker and tanrity not icftxled: see t.xt wine, and spirita into a single variable- monthly cons>#mption of ethanol in ounces. Within the category of current drinkers, heatiy drinkers include those who repqFte,an average of two or more drinks daily'(five or more actual drinks daily, with= allowance for underreport- ing). Becaus4 light drinking may not be harmful, we calculate the cost per ounce in excess of two reported drinks per day." Thus, the drinking analogue of •nonsmoking smokers are "controlled" heavy drinkers; ie, we estimate the ef- fect of hypothetically reducing the con- sumption of those with more than two reported drinks per day to two reported drinks per day. Our base-case analysis also controlled for health insurance coverage, age, sex, race, education, the use of seat belts, family income, exercise, self-assessed measures of physical, mental, and gen- eral health, and family size. We included education and seat belt use to measure attitudes that may differ between those with varying health habits-attitudes that may affect work loss and use of medical services independently of smoking and drinking. Pensions and Other Costs In addition to the costs of inedical care and work loss, we calculated the other components of cost shown in Table 1, Data regarding pension and disability payments by age,. sex, and educatioq status come from the Current Fbpuz~~ tion Survey. That survey is also N source of earnings data, which we use tc calculate taxes to finance the progrants. Our estimate of annual property togo from 5res1hat are associated with ciga_ rette smoking is $340 million (in 19% dollars).' B6cause of fire insurance, we have assumed these costs are entirely external, but our estimates are not sen. sitive to this assumption. Our estimates of certain annual ex- ternal costs of alcohol abuse are as fol- lows: property damage from motor-ve- hicle accidents, $3.6 billion, and from fires, $507 million; criminal justice, $3.1 billion; and social programs, $54 md- lion.u It is extremely difficult, and to some distasteful, to place a dollar value on the innocent lives lost due to fires, passive smoking, or drunk driving. Neverthe- less, it is often necessary, implicitly or explicitly, to place a value on lives lost when judging the merits of alternative policies, for example, policies leading to air pollution control or increased auto- mobile safety. For this analysis, we in- clude an explicit value for the lost lives to avoid the systematic undercounting of the costs to society that would occur if we included only the differences in use of medical care, sick leave, etc. Zb define a value for innocent lives lost because of fires, passive smoking, I and drunk driving, we used a method based on the willingness to pay for a small change in the probability of sur- viving.'T This yields a value of $1.66 mil- lion per life (around $10 per hour, using years of life expectancy discounted at 5%), considerably more than the value of lost earnings. We believe earnings are an inappropriate measure of the val- ue of life, in part, because they attribute a relatively low value to those who are out of the labor force." RESULTS Smoking External Costs per Pack of Ciga- rettes.-If costs are not discounted, each pack of cigarettes increases medi- cal costs by $0.38, but saves $1.82 in public and private pensions due to a 137- minute reduction in life expectancy. Overall, there is a net savings of $0.91 per pack in undiscounted costs (Table 2). Results change markedly if costs are discounted at 5%, largely because pen- sion costs change from -$1.82 (at 0%) to -$0.24 (at 5%) per pack. Pensions are received late in life, so discounting dramatically decreases the differential between smokers and nonsmoking smokers. Using a 5% discount rate, the 1606 JAMA. March 17, 1989-V01261, No. 11 TIMN 325068 SinTax-Manningetal

total external costs per pack are $0,16, fnd they rise to $0.24 per pack at a 10% discount rate. The main reason these results are so much lower than, for ex- ample, the estimate from the Office of Technology Assessment of $2.1,5 per pack (unpublished data, 1985) is our eacclusion of changes in lifetime earn- ings from smoking, which are internal costs. . Sensitivity of Costa to Assump- tions.-Clearly, the magnitude of any subsidy from nonsmokers to smokers is sensitive to the discount rate, especially below 5% (Figure). Table 3 shows the effect of varying other assumptions. For comparison, the first column re- peats the results from Table 2 for a 596 discount rate. To test how sensitive the results are to the data source selected, we used NHIS data for young as well as old persons (Table 3, column 2). Medical costs per pack do not change, but cov- ered sick leave costs rise to $0.05 per pack, and the total net costs rise from $0.15 to $0.20 per pack. To test how sensitive the results are to different assumptions about how smoking affects health, we contrast smokers with actual never smokers, rather than nonsmoking smokers (Table 3, column 3). The results are relatively insensitive to this modification also; ex- ternal costs rise to $0.28 per pack. This figure probably overstates the true costs because it treats all the differences between smokers and never smokers, except wages, as causally related to smoking, whereas smokers may have different patterns of medical use and retirement for reasons unrelated to smoking. As another test, we restricted medical costs to those arising from diag- noses thought to be related to poor health habits; medical costs fell $0.11 (Table 3, column 4). The estimates de- scribed herein assumed that a cohort of nonsmoking smokers would retire in a manner similar to people- who never smoked. However, wc.aiso computed effects on taxes and pensions, assuming that the pattern of netirement among nonsmoking smokerr;Vwould be the same as among smokar4~, ie, quitting would not affect age -o( retirement (Table 3, column 4). Comb,ining these assump- tions leads t~ a lower boundary of -$0.15 (at a 5~'n dfscount rate) on costs peT pack. r ~ Finally, the ltst column in Table 3 gives total costs; that is, it includes the portion of costs that are financed by the person: It does not, however, include the costs of premature mortality and suffering, which is why a question mark appears in the lower right corner of the table. Other Costs of Smoking.-Our esti- mates of the costs of smoking in Table 2 do not include the adverse effects of Table 4.-External Costs of Heavy Drinken per Excess Ouncu' Olseountflate External Costs 0% ax 10% IN.Ckal and Pen6*n oosts per excess ounce. S Medical carei 0.26 0.10 0.05 Sick Msve _ f 0.06 0,05 0.04 Group Nfe nauranp 0.02 0.02 0.02 N,rsi<,4 hon,e -0.01 t t Reur.rwx Oneiong -0.04 0.03 0.02 Taxes an eamirqs, S -0.35 -0.06 -0.02 Net rnediea/ and pension costs per excess ounce, S 0.63 0.26 0.15 Mo1or•vehicls accidents and criminal Juatlca costs per excess ounee, S Lives of nondrinkers 0.58 0.58 0.58 NI other coetsN 0.35 0.35 0.35 Total net costs per excess ounce, $1 1.56 1.18 1.08 Life expectancy at age 20 y per excess auna. min -20 -8 -4 'Coets (in 1988 dollars) per excess ounce are calculated by dividing by the d'iseounted number of exeess ounces. tExcludes matemdy, weN, and dental car., and medical care coets b others caused by drunk drtvkq. #Indicates Npure +s less trun 0.005. glnduasa dasab+Nty r,euranc.. liThe 50.35tient 8qun Includes certain internal costs, such as the property damage in motoavehide acddents paid by the alcoholic driver in deductibles or otlW copayments and higher premiums but excludes the external costs associated with the effects of alcoholism on spouses and children (eg, their use of insured mental health services) and ttwse associated with the increased risk of alcoholism for these dependents. 1Sum of coats minus taxes on earnirqs. passive smoking on those outside the smoker's family. Passive smoking causes an estimated 2400 lung cancer deaths per year, and it has also been linked to reduced lung function among children of smokers, a higher incidence of respiratory problems for children and others, as well as the displeasure of con- suming unwanted cigarette smoke.' Most of these costs are within the family and are internal or external costs de- pending on the extent to which the smoker considers the welfare of others in his family when he smokes. The fig- ures in Table 2 assume that such costs are internal. I£ however, we treat the costs of the 2400 deaths as entirely ex- ternal and use an estimate of willingness to pay for lower mortality of $1.66 mil- lion per life,,' external costa per pack would rise $0.14. Because deaths in smoking-related firea are also almost entirely within the family, we have treated the costs as internal and did not include them in our estimates. However, if we were to treat the costs of such deaths as external, some 1600 people in 1984 (J. Hall, oral communication, Aug 13, 1987), we would increase the external costs of cig arettes by $0.09 per pack of cigarettes.. The smoker loses 28 discounted min- utes of life expectancy (at a 5% discount rate) for each pack smoked (Table 2), which accounts for $0.93 of discounted wages (many of the lost minutes occur when not working). Using our estimat ed willingness to pay for lower mortality of $10 per hour, the 28 minutes is worth approximately $5. Although we consid- er the $5 an internal cost, it may none- theless be relevant to an economically efficient tax, a point we will come to later. Heavy Drinking External Medical and Pension Costs per Excess Ounce of Alcohol.- Using undiscounted values, each excess ounce of alcohol, ie, those consumed in excess of two reported drinks per day, has external medical and pension costs of $0.63 and causes a loss of 20 minutes of life expectancy (Table 4, column 1). At a 5% discount rate, external medical and pension costs per excess ounce fall to $0.26. In contrast to smoking, heavy drinking increases all categories of costs (at a 5~'n discount rate), even pensions, because the large effects of early retire- ment, which triggers pension and dis- ability payments, outweigh the shorter life of drinkers. At a 10% rate of dis- count, medical and pension costs fall to $0.15 per excess ounce. Before discussing the other costs of drinldng shown in Table 4, we describe the sensitivity of our estimates of medi- cal and pension costs to different as- sumptions (Table 5). Fbr convenience, the first column of Table 5 repeats the results from Table 4 for a 5% discount rate. Medical and pension costs are not sensitive to the source of data (Table 5, column 2), nor do they change much if we compare heavy drinkers with actual abstainers and light drinkers rather than hypothetical controlled drinkers (cutting back to two reported drinks per day among those consuming more than that amount) (Table 5, column 3), nor do they change when drinking is not treat- ed as a cause of disability retirement (Table 5, column 4). Restricting medical costs to those arising from diagnoses thought to be related to poor health habits makes vir- tually no difference to our estimates (Table 5, column 4), implying that the JAMA, March 17,1M-Vol M1, No.11 TIMN 325069 . SinTax-Manningetal 1607

Table S.-Sens,tiwty ot Medical and Pension Costs (in 1988 OW1ars) per Excess Ounce of Ethanol 10 Assumptions, 5% Discount Rats Costs t3ass Cass' An Oata From National FNatth Intsrvisw Survey Abstatn.n and Light Drinksrs Lower t3ounQ} Total Costs; Medical and pension casu. s Medical cars per excess ounee§ 0.10 0.11 0.07 0.11 0.16 Sick b.vs ! 0.05 1 0.10 0.05 0.13 Group Afs insurance 0.02 0.01 0.04 0.02 0.02 'NursUq homs I I -0.01 ( 1 Retirement pensbnl 0.03 0.05 -0.15 -0.05 0.03 Taxss on sanynqs, i -0_.06 -0.06 -0.14I -0.03 -0.64•• Net medical and peruion eosb per excess ounca. Stt 0.26 0.23 0.20 0.15 1 LI(e expectancy at aqs 20 y per excess ounce, min -8 -7 -19 -6 -8 'Effect of chargirq heavy drinker to convo86d drk*sr, with other eharactaristies held oonstant tNarrow definition o/ medical sNeets, with no aHsct on early retUerrNnt. $Irdudas iMemd costs. gExdudes matearwty4 wslt, and dental can. plndicates figure Is less than 0.005. 11ncludes dsatslity insuxanos. #We have used the earnings of abafainars and YQht drinkers to compute taxes. Theas earnings arA considerably higher than for drinkers, even after controlling for educatkm. To the extent that these samirps differences are not caused by drUYcinq, we should use drinksrs' eaminqs; in that oase, the -0.1411pure woukd be -0.03. '•Eamirqs, not taxes on saminpsm t}Sum of costs minus taxes on eaminqs. medical costs shown in the first column are largely due to differences in medical use that are related to habits. In con- trast, the external costs of smoking are sensitive to the definition of relevant medical costs, suggesting that the broader definition of smoking effects may overstate medical costs and total external costs. Other External Costs.-Although our estimates include the additional probability that a drinker will be killed in a traffic accident, they do not account for the deaths of innocent bystanders and nondrinking passengers in such ac- cidents. The Department of Transpor- tation estimates that about 7400 of the 22 400 people who died in alcohol-relat- ed traffic accidents in 1985 were not drinking.' Based on a willingness to pay for a human life of $1.66 million and the estimated volume of drinking from the 1983 NHIS, the v4lue of the 7400 lost lives is $0.58 per exe_ess ounce of ethanol (Table 4, bottom). This figure is low because it does not include medical, dis- ability, and suffering costs of surviving nondrinldngv~ctims of alcohol-related accidents. Un'the other hand, the figure is high to tha extent that not all drink- ing-relatedk,accidents are caused by alcohol. • , In additinn, there are annually $7.2 billion of other costs described previ- ously herein, principally costs of the criminal justice system and property damage in alcohol-related motor-vehi- cle accidents. These costs add another $0.35 per excess ounce. Senaitivity of Results Although $0.15 per pack of cigarettes and $1.19 per excess ounce of alcohol are our best estimates of the external eco- nomic costs of smoking and heavy drink- ing, the values are sensitive to four fac- tors: discount rate, value assigned to lives lost in drunk driving-related acci- dents, amount of underreporting, and treatment of persons who die of causes related to passive smoking and fires. Discount Rate.-The sensitivity to the discount rate is more pronounced with smoking, where the estimated ex- ternal costs would be almost $0.20 lower per pack if we used a 3% rather than a 5% discount. The sensitivity of drinking costs to discounting is much less. For smoking, consumption starts early, but deaths come much later than in the case of drinking. The shorter the time be- tween consumption and death, the less sensitive the estimates are to discounting. Dollar Value of Life.-Because the assumed vaiue of life is on the low end of estimated values, our estimates of drinking costs are conservative. Underreporting.-Assuming that the reported level of consumption were closer to the actual level of consumption would raise our estimates of the exter- nal cost, because we would inflate the level of reported packs and ounces by a smaller factor when computing costs per pack and ounce. For example, had we assumed respondents reported 60% of their actual alcohol consumption, we would only have multiplied reported ounces by 1.67 (100/60) rather than 2.5 (100/40) to estimate actual ounces, and the- estimated cost per excess ounce would be 50% (2.5J1.67=1.5) higher. In the case of alcohol, our cost estimate is conservative because the 40% figure we used is at the low end of the estimates found in the literature.' Within-Family Costs.-We ignored 1608 JAMA, March 17, 1989-bbl 261. No. 11 costs of $0.23 per pack associated witb deaths caused by passive smoking and fires because we assumed they were i>s the famiiy and taken into account by the smoker. Defining these costs as exter. nal would more thaq double our estimat. ed external cost of`smoking. Our estimates are relatively insensi. tive to other assumptions..8ecause the external costs of drinking are dominab. ed by costs associated with"drunk driv- ing, such costs are relatively insensitive to discounting (Figure). The choice of data used to estimate effects (HIE va NHIS) has little effect on the results. Our estimates of the external costs of alcohol were made per excess ounce, but excise taxes apply per ounce, not per excess ounce. Forty percent of total consumption represents ounces in ex- cess of two reported drinks per day (five actual drinks per day, given our esti- mate of underreporting). Zb convert our figures per excess ounce to figures per ounce, one should multiply them by 0.4, reducing the estimated cost of $1.19 per excess ounce to $0.48 per ounce. Our estimate of the external cost of smoking, $0.15 per pack, is well below the current average (state plus federal) excise and sales taxes of $0.37 per pack ($0.32 of the $0.37 are excise taxes).30 However, the $0.37 tax rate approxi- mately equals the estimated external cost of $0.38 if we were to treat all lives lost to passive smoking and fires as ex- ternal costs. By contrast, our estimate of the external cost of alcohol, $0.48 per ounce, is well above the current average (state plus federal) excise and sales tax- es of $0.23 per ounce.n (The average excise tax is taken across distilled spir- its, wine, and beer, where the excise taxes are $0.25, $0.03, and $0.09 per ounce of ethanol, respectively.) Thus, smokers probably pay enough taxes to cover the net costs they impose on oth- ers, but heavy drinkers do not. We noted in the introduction that eco- nomically efficient excise taxes should at least cover external costs. By this criterion, taxes on alcohol are too low; whether cigarette taxes are high enough depends on one's appraisal of three other arguments for taxation of cigarettes and alcohol. (Each of these arguments would further strengthen the case for increasing alcohol taxes. ) The first argument takes cognizance of the regret expressed by most smok- ers and their attempts to quit. Smoking tends to start in adolescence or early adulthood, at a time when individuals are not well informed and may not ap- preciate the consequences of their aa tions.s Cigarettes (and alcohol) are ad- dictive, so itis more difficult to quit than to avoid starting the habit. Because over 85% of smokers begin smoking be- fore age 20 yearsa and some evidence TIMN 325070 Sin Tax-Manning et aa

~~ suggests that the proportion of those under 20 years of age who smoke is sen- sitive to taxes," higher taxes may de- )y the crease the number of individuals who 'xter. become addicted. ima6, Some may see,this argument as pater- nalistic, but it is not, if judged by the sensi, tastes of the individual attempting to e the quit; those tastes arguably determine unat= the economically efficient tax. If the loss driv. in life expectancy of 28 minutes per pack sitive is relevant to economic efficiency be- ce of cause of later regret, an economically E va efficient tax would be on the order of $5 s. _per pack, the estimated value of the 28 sta of minutes. , but A second and related reason to tax per cigarettes is that many adults do not total appreciate the risks. Despite the warn- • ex- ing labels on cigarettes, 2096 to 2596 of (five adult smokers say they do not know the esti- risks of smoking.° A higher tax would vert deter initiation of smoking, thus com- um pensating for any undervalued risk. a by A third reason to tax addictive com- 1.19 modities is that such taxes are likely to lead to a relatively small change in be- t of havior among those already addicted. low Suppose, for example, there were no ral) external costs, no ignorance, and no re- ack gret associated with smoking. From the g).'D point of view of raising revenue, it may )xi- still be wise to tax cigarettes because it 'nal is preferable to tax items for which be- vea havior does not change; there is less ex- induced inefficiency." This argument ate could also justify higher cigarette taxes Per than at present. ige Despite the uncertainties surround- ax- ing our estimates, in the case of alcohol, tge the difference between the actual tax ,ir- and external costs is so large that, in our ise view, a strong case can be made for an )er increase in federal alcohol taxes. The is, tax increase should occur at the federal to level, not the state level, to prevent th- bootlegging acrass state lines. The case is especially strong for raising taxes on :o- beer and wine, which, is toted previ- ild ously herein, are much lower (per ounce be of ethanol) than taxes on distilled spir- ?rl its. Strategies such as banning advertis- ;h ing or promoting.ndgative advertising of may be complgaaeatary.$ of 'Ib the degreettiat external costs of ;e alcohol abuse; stem from people who In drink in bars ~nd yestaurants and then drive home while intoxicated, there is a ~ case for an additional tax on alcohol sold •- by the drink We have not tried to ascer- g tain what proportion of external costs y stem from alcohol consumed in bars and $ restaurants relative to that consumed in ~ homes. Ideally, society would tax drunk driv- ers to force them to pay the external 1 costs of drunk driving rather than tax = alcohol. 'Ib some extent, society does so with fines, suspension of driving li- '- censes, jail sentences, and civil liability. However, the present legal system does not make, nor could it reasonably make, drunk drivers bear fully the external costs of their actions, especially in those cases where there is a loss of innocent lives." For example, liability insurance partially shields drunk drivers. We close by considering two argu- ments against higher excise taxes. First, tobacco and alcohol taxes consti- tute a larger proportion of the income of the poor than of the well-to-do."-" How- ever, alcohol and tobacco taxes each supply only 1% of federal revenues. As a result, rather small changes in the indi- vidual income tax structure could readi- ly compensate for the effect of increased excise taxes on the distribution of in- come, if that were deemed desirable. Drinkers and smokers would still pay more, but low income individuals, as a group, need not pay more. Second, light drinkers may argue that they impose few or no external costs, but would unfairly pay a higher tax burden. There are two responses. First, suppose that a given amount of revenue to finance government expen- diture must be raised from various tax- es, including excise taxes on alcohol. As a group, persons whose consumption of alcohol is below the population average of 1.7 reported drinks (over four actual drinks) per day will benefit from shift- ing more of the tax burden to alcohol taxes and away from other taxes (eg, payroll taxes). In fact, of adults who drink, three fourths drink less than this amount. Second, to the degree that higher taxes deter alcohol abuse, the resulting decrease in external costs will offset increases in the tax burden of light drinkers. Because excise taxes must be propor- tional to consumption and because the external costs of smoking and drinking are not proportional to consumption, there will not be, in practice, a tax that does not leave someone subsidizang someone else. The task of determining how such subsidies will flow fails to our political institutions. We hope our esti- mates contribute to more informed decisions. Thia work was eupported by grant R01-HS-0b278 from the National Center for Health Services Re- search and lbchnology Assessment. We thank Thomas Vogt, MD, Robert Leu, PhD, and Bernard Friedman, PhD, for suggestions and guidance; Robert Amler, MD, for help with the Health Risk Appraisal model; Kenneth Warne.r, PhD, Charles Phelps, PhD, James Kahan, PhD, Bridger Mitchell, PhD, and Jim Smith, PhD, for careful reviews; Bernadette Benjamin and Janet Hanley, MS, for programming and data manage- ment; Joyce Peterson, PhD, for editorial aseia- tance; and Stephen Marcus, PhD, and Selwyn Waingrow for their support, comments, and advice. Rd.r.nes. 1. US Dept of lfanaportation: Drunk Driving Fbets National Highway 1raf5c Safety Admfnis- trstioa,1986. 2. Luce BR, Schweitzer SO: Smoking and alcohol abuse: A comparison of their economic con.r quences. N Engl I Mod 1978;19&b696b71. 3. Leu RE: Anti-smoking publicity, taxation, and the demand for ciprettes. I Health Econ 1984; 3:101-116. 4. Vital Statiatiu ojtAs United States,19d0. Hy- att.vil1e, Md, National Center for Health Statis- tia,1984. 5. CDC Xeaitk Risk Appraisal Urer Manual. At- lanta, Centus for Disease Contro1,1984. 6. Pechman JA: Fadans! Thx Ibtiey, ed 3. Wash- ington, DC, Brookinga Inatitution,19T7. 7. Warner KE: Possible increaaes in the underre- porting of cigarette consumption. J Am Stat Aisoe 1978;73:314318. 8. F%U Special Report to the US Congress on Al- cohol and HiattAFitim the SeerstaryajflealtAand Human Services. US Dept of Health and Human Servicee,1983. 9. Money income d househokb, fum7les, and per- aona in the United Statee:1984, in CYrnnt Fbprla- tiox Reporla, Consumer Income, Sarie+ P.60. US Depto[Commerce, BureauoftheCenw,19®6, vol 151, pp 166-170 10. Newhouse JP: A design for a health insurance ezperiment. Inquiry 1974;11:5-27. 11. Brook RH, Ware JE, Davies-Avery A, et aL• Overview of adult health atat,ua measures 8elded in RAND's Health Insurance Study. Mod Car. 1979;17(euppl):1-131. 12. Price DN: Cash beneHts for ahort.term sick- ness: Thirty-five years of data, 1948-1983. Soc Se- cur Bu111986;49:b38. 13. Marmot MG, Ro®e G, Shipley MJ, et aL• Alcohol and mortality: A u-shaped curve. Lancst 1981; 1:580-G83. 14. Dyer AR, Stamler J, Paul o, et aL• Alcohol consumption and 17-year mortality in the Chicago Western Electric Company atudy. Prev Mod 1980;9:78-90. 15. Klataky AL, Friedman GD, Siegelaub AB: Al- cohot and mortality: A ten-year %aisevPerman- ente experience. Ann Intern Mod 1981;96:139-146. 16. Harwood HJ, Napolitano DM, Kristiansen PL, et aL• Economic Costs to Society of Alcohol and DrugAbuse and Mental ltlneaa Reeearch'lfiangle Park, NC, Reaeaech'lfiangie Inatidtte.1984. 17. Shepard DS, 7ectih•^ } RJ: Survival versus eonaumption. Management Sci 1984;30:423-4.49. 18. Howard RA: Life and death decia9on analyai., in Prnccediwpt Second Lawranea Symportum on Sy,tems and Dwision Analysis. Berkeley, Uni- vercty of CaliforniaPseaa,1978. 19. TJu Ileala Consequences of Ixrooluatarlr Smoking: A Report to the SxrpaoR Ceeemi. US Deptdaealth and Himun Servicn,1996. 20. 1ke 7hx Burd.n on lbbaaco. Washington, DC, 'lbb.cco Inatitute,M6. 21. Publfe Revenues Fmm Alcohol Beusragea ~~ 1DC~Diet~ed Spirits Council of the U 22. Warner BE: Selliap Smatt: Ciyandts Advsr- titiap ancl Public EealtJl. Washington, DC, Ameri- can Public Health Msociatian,1986. 2s. Lewit EM, Coate D, Grossman 1(:'17u effects ~E~ on teenage ®oim~s. J 24. Wasserman J: Ezeise Thxa, Repriatios, and the Demandfor Cigarettes, publication P-7498-RG. Santa Monica, Caliiy The RAND Corpoiation, 1988. 25. Ramsey F: A contribution to the theory of tax- ation. Eco%J 1927;37:47-6L 26. ManII Sl, Sinning L.4, Hicl®aaJH: Culpabil- ity and accountability of hospitalized injured alco- hol impaired drivera: A prospective study. JAMA 1984;262r1880-1883. 27. Toder EJ: Ieaues in the taxation of cigarettes, in The Cigarette Eaeise 1bs. Cambridge, Mass, Harvard University Pcese,198b, pp 66-87. 28. Rock SM: Measurement of tax progreeri.it,y: Application. Publie Finance Q 1983;11:109-180. 29. Harria JE: Increasing the federal ezciee tax on agarettee. J Hea1fA Eaox 19sT,1:117-120. Sin Tax-Manninp et al 1609 ' I LAM& March 17, 'M-Vd 26',-"°. " ~ TIMN 325071