Tobacco Documents Online

WORKING PAPERSERIES : THE SOCIAL SECURITY COST OF SMOKIN6 John B. Shoven Jeffrey 0. Sundberg John P. Bunker Working Paper No. 2234 NATIONAL BUREAU OF ECONOMIC RESEARCH, INC.

Th* National Bureau oi Econoinic Resea" is a peivate, noe-profit, nan-pardsaA orgaaizaNon eusaald in 4aaedire aaalysis of the Aamiean etonocny. 11ds pap.r has not undergone the rm+ewacco~+ded otficfai NM publicadoas; in patdatiar, it has not been wbaritted for approval by ahe Board of Dkecto.am lt is Intended to malce raait: oi NeQt rexarcfi ava8abfe to other economists in pcdiaiinary fora to encourage dstmsioa and suggestions for rcviaion befoee final publication.

NBER WORKING PAPER SERIES THE SOCIAL SECURITY COST OF SMOKING John B. Shoven Jeffrey 0. Sundberg John P. Bunker Working Paper No. 2234 NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 May 1987 Paper presented at the National Bureau of Economic Research Conference on the Economics of Aging, March 19-22, 1987, New Orleans, LA. This work originated when John Shoven and John Bunker were Fellows at the Center for Advanced'Study in the Behavioral Sciences in 1984-85. The research reported here is part of the NBER's research program in Taxation. Any opinions expressed are those of the authors and not those of the National Bureau of Economic Research: 0

. NBER Working Paper ai2234 May 1987 The Social Security Cost of Smoking ABSTRACT Our paper is an ezaaination of the Social Security cost of smoking from.an individual point of viev. It is well latown that smokers have a shorter life expectancy than nonsmokers. This means that by smoking they are giving up potential Social Secutity bensfits. We estimate this caat and consider the effects on the system as a whole. Qe use mortality ratios, which relate the annual death.probabilities of smokers and nonsmokers, and the percentage of smokers in each age group to break down the life tables for men and women born in 1920 into the approximate life tables for smokers and nonsmokers. ue then calculate expected Social Security taxea and benefits for each group, using median earnings as a base. We find that smoking costs men about $20,000 and women about $10,000 in expected net benefits. The implication of this for the system as a whole is that the prevalence of smoking has a direct effect on the financial viability of the system; every decrease in the number of smokers in sociery increases the system's liability. Changes in smoking behavior should be•recognized as affecting the system. John B. Shoven Department of Economics Stanford University Encina Hall, 4th Floor Stanford, CA 94305 (415) 723-3712 Jeffrey 0. Sundberg John P. Bunker Department of Economics Stanford University Stanford University School of Medicine Stanford, CA 94305 HELP Building Room #7 (415) 725-0959 Stanford, CA 94305 (415) 723-6426

1. INTRODIICTION Smoking•in the United States is associated with enormous costs to society. The Congressional Office of Technology Assessment has estimated the armua1 cost of medical cars for smoking related ilZaeas at $15 to $30 billion, and that smoking related illness is responsible for an additional $49 to $70 billion in lost productivity. There are also substantial costs to the iadividual who smokes in tezms of lost wages aver a life time., primarily affecting those who die of smoking related disease while still active wage earners in the work force. Costs to the individual also include approximately $500 to $1,000 per anm"" for pack and two pack-a-day smokers to purchase cigarettes. An additional cost to the individual is the loss of Social Security benefits as a result of smoking-induced loss of life expectancy. '!he-data presented herein estimate the masgtitude of this loss for single and married men and women born in 1920 and 1923, respectively. While most of the previous literature on the costs of smoking and the benefits of quitting has overlooked the implications of smoking behavior on pension plans (see, for ezample, Oster, Colditz, and Kelly (1984)), this is by no means universal. Cori et al. (1983) estimated that the savings realized by Ford Motor Company if the health of their employees improved (in terms of less expensive medical insurance, disability insurance, and life insurance) would be much smaller than the additional pension costs due to their increased longevity. Atlcinson and Townsend (1971) noted that the fiaancial benefits the British National Health Service would enjoy if there was a forty percent reduction in smoking in Britain would be more than offset by the increased cost of retirement pensions. In this paper, we eT'm{*+~ the Social Security consequences of smoking from the individual or household perspective. From that vantage point, Social Security can be thought of as a prepaid life annuity. Contributions or taxes N 0 N W ~ N 0 0 ~ 1

are collectad during one's work life which entitle aae to an iadezed life aanuity b.ginaiag at age 63. The a:muity can be cosasenced at age 62 vith a roughly fair actuarial adjustm.nt, and can be started at an ago beyond 63 with somewhat higher benefits reflecting the shorter expected raw{mng lifetime. In general, the system is not actuarially fair (favoring sove cohorts relative to oc.hers, those with low incomes or short covered c=reers relative to others, and aarrieds, especially oae.earaer couples, relative to singles). Oar point is that the system is imfair in a way v.ry relevant to the dscision of whether or not to smoka. Social Security does not have separate benefit strvctures for smokers and nonsmokers even though ssok.rs have a much lower chance of reachiag retirement age and a shorter expected length of retirement conditional on reaching that age. The Office of Teci~aology Assesssent (1985) estimatad that 273,000 p.opie died in the United States in 1982 of saoking related disease. Of those, 44 percent, or 121,000 died before they reached their 65th birthday. They may have never collected anything from Social Security. If they were sarriad, their spoase may collect vidov's benefits, but it is clear that their premature deaths greatly reduce their raturn on their participation in Social Security. Smoking also affects the liedicare portion of the Social Security system. While we concentrate oa OASI, it is probably worth noting that the health insurance component is siailarly affected. Many estimates of the effect of smoking an the total desand for health care services in the caantr9 find that it is small in the long run. Smokers certainly experience more health problems per year of life, but this is offset by the fact that they live fe.rer years. Qith a lower incidence of smoking, there would be more elderly additional health care services. The reduced dsmand caused by who require the improved health status of the former ssoicers is offset by extra care needed by the ~ 0 N W ~ C 0 ~ 2

additional elderly. Thers aight be some initial reduction in the deaaad for health care if smoking was reduced. The improveaents in health status would presumably occur before the ags structure was significantly zltared. However, in the long ztm the two effscts offset each other. Despits the fact that total health car. daoaad may be little affected by smoking, Iiedicare's finances are almost certainly affected. 'rhe reason is that it is a prepaid health insurance aumuity for those over 65. Medicare does not bear the higher health costs of nonelderly smokers, but benefits financially from the fewer numbers of elderly due to smoking. The other side of the coin is that smokers, as with their retirement benefits, pay while they work for old age health insurance which they are less likely to collect, or to collect for a shorter period than nonsmokers. Wright (1986) estisatss that each person who quits smoking increases the deficit faced by the HI component of Social Security for just these reasom . Our study is the QASI analog of Wright's HI research. We assemble separate life tables for smokers and nonsmokers and then estimate the Social Security tazes, benefits, and transfers for members of the 1920 birth cohort. 74 do this for those who earn median wages for their age and cohort and for those who earn 60 percent of the median, in each case begina3ag at age 20. The results can be previewed by saying that re find the expected loss in nst Social Security b.nefits accompanying smoking to be very large relative to the other costs of smoking. The loss exceeds the lifetime costs of cigarettes, is large relative to the estimates of the medical costs and lost wages diu to excess morbidity and mortality, and is perhaps tsn times greater than the corresponding Medicare figures of Wright. The next section of the paper briefly reviews what is kaowa about tfie effect of smoking on mortalir.y. It discusses disease specific effects and also our technique of using mortality ratios to yield approximate separate life 3

tables for smokers and nonsmokers. The third section of the paper describes our simulatioa procadure for calculsting the Social Security costs of smoking. Zt presents separate results for siagle iadividusls, one-earaer and ttiro-earaer couples because of their separate t=.atmsat by Social Security. Qs conclude the paper with an interpretation of rhat our fiad3ags imply about the private and social incentives to quit ssokin;. 2. EFFECT OF SISOS=G ON !SG'gT1T_rrv There can be no statistical doubt that smoking is associated with increas.d mortality hazard rates. The overall fiadina of the 1979 Surgeou G•aeral's report an the snbject was that the mortality of alI sale cigarette smokers is about 170 percent of that of male nonsmokers. For two-pack-a-day saok.rs, the average mortality ratio is 200 percent. For particular diseases the relative haxard is eve n areater. For azaaple, two separate studies find smokers are between 9 and 15 tim.s more Likely than nansmokers to die of lung cancer (Lubin et a1. (1984) and Covell and Hirst (1980)). The risk of dying of arterosclerotic and degenerative heart disease and ayocardial iasufficiearies has been estimated at 2.7 tiaes as great for smokers as nonsmokers (Covell and Hirst (1980)). There is further evidence that there is a significant interacicioa between smoking and other eaviroam.ntal factors such as exposure to asbestos. The finding is that rhile smoking is a major cause of lung caacer, smoking coabiaed with other assaults (snch as iadcstrial exposure) greatly iacr:ases the mortality hazards (Schneiders.a and L.via (1974). Our development of separate life tables for ssokers•aod nonsmokers uoLlizes the fiadiags.of E. C. Hasstond (1966) regarding the effect of smoking on N 0 N W ~ mortality. Qoriciag for ttu American Cancer Sociecy, he conducted a comprenensiveCA four-year study trackiag a population of over one million subjects. He determines the death rates and the prevalence of certain causes of death for 4

smokers and nonsmokers of many different characteristics. The tachnique vas to ezamine death certificates for the cause of death and to request information from the attanding doctor whenever cancer was mentioned on the death certificate. Hamsond's results are a very detal.led set of mortality ratiosl for diffarent types of smokers and for several different causes of death. In 1959 and 1960, Hamoond enrolled aver one million volunteers from t9renty-five different states to provide data on mortality. Subjects were classified by sex, age, type of tobacco smoked (cigarette, cigar, pipe, or none), age at which subject began smoicing, daily amount of smoking, and degree of smoke inhalation. Each subject was contacted an=tally for four years to track the number and tiaing of fatalities in each group. Death certificatas vere received for aver 97% of reported deaths to provide better information as to causes of death. Using the accumulatad data, Hammond combined subjects with similar characteristics into five-year and tsn-year age cohorts, and divided the number of deaths in each cohort during the study period by the nsaber of "person- years• experienced in each cohort. This provided cohort death rates over the period for groups of similar age and sex, and varying smoking habits. This allowed Hammond to calculate'mortality ratios for different groups. A sample of his findings is shown in Table 1. The separate mortality tables that ws have produced are contained in the Appendix to this paper. The basic life tables used are the cohort life tables for men and vomen born in 1920, as estimated by the Social Security Administration. The mortality hazards are shown in column 8 of the appendix ~ table for men and .roasn. If we let Q'.(a) represent the one-year death 0 probability for males as a fnnction of age (similarly QA(a) for females), ~ ~ N CA ~-

Tabl. I Current Number Per Day 1-9 10-19 20-39 40+. 1-9 10-19 20-39 40+ Mortality Sstios for Smokers as Dat.raia.d by B. C. Sasmoad Ags 35-44 43-54 55-64 65-74 75-84 M+sa rith History of Only Cigarstta Smoking * 1.84 1.33 1.50 1.36 1.36 2.26 1.92 1.65 1.55 1.91 2.41 2.05 1.71 1.26 2.59 2.76 2.26 1.81 * Vos.n rith History of Only Cigarette Smoking 0.90 0.95 0.99 1.09 1.07 0.97 1.22 1.31 1.18 1.21 1.35 1.54 1.46 1.51 * * 1.96 * * * * Sigaifias a vary low rn:mbsr of sxp.ctad daat3ss (small sampla or low death rats). SOQBCE: Hammoad, p.133. 6