Statistical Issues in the Estimation of the Causal Effects of Smoking Due to the Conduct of the Tobacco Industry

  • Donald B. Rubin
Part of the Statistics for Social Science and Public Policy book series (SSBS)


A major legal issue for the past several years has been the tobacco industry’s liability for health-care expenditures incurred because of its alleged misconduct beginning in the mid-1950s. Quantifying answers to such causal questions is a statistical enterprise, which has been especially active in the last quarter century.

This chapter summarizes my formulation of a statistically valid approach for estimating the potential damages in the tobacco litigation. Six distinct statistical tasks are outlined, although no specific estimates are produced. These six tasks are: formulation of mathematical statistical framework; assembly of data to estimate health-care-expenditure relative risks of smoking in the actual world; design of the statistical analyses to estimate these expenditure relative risks—a problem closely related to causal inference in observational studies; assembly and analysis of appropriate data to estimate the prevalence of different types of smoking behaviors and other health-expenditure-related factors in the relevant population—a problem of survey inference; assembly and analysis of appropriate data to estimate the dollar pots of health-care expenditures of various types in the relevant population—another problem of survey inference; assembly and analysis of information concerning the prevalence of smoking and other health-expenditure-related factors in a counterfactual world without the alleged misconduct of the tobacco industry—a problem involving explicit assumptions justified by actual-world experimental and observational data. This sixth task is the critical step where the alleged misconduct, and thus causal inferences, enter the equation; the second through fifth tasks involve the careful assembly and analysis of actual-world data. The outputs from the last five tasks (2 – 6) are input into an equation, which is derived in the first task, to give an estimate of the causal effect of the alleged misconduct on health-care expenditures.

The plausibility and validity of the results depend critically on the use of detailed information on health-care expenditures, smoking behavior, and covariates (i.e., background characteristics and nonsmoking health-care-expenditure-related factors). The reason is that this detail is used to justify the key assumption in the mathematical statistical formulation in the first task. The need for this level of detailed information places extra demands on the data-based efforts in the last five tasks.

The formulation presented here distinguishes issues of fact about the actual world, involving the health-care-expenditure relative risks of smoking and the prevalence of smoking behaviors and other health-care-expenditure-related factors, from issues using actual-world facts to conjecture about the counterfactual world. The results show that, under the key assumption, counterfactual-world estimation enters the equation only through the differences between actual-and counterfactual-world prevalences of smoking and other health-expenditure-related behaviors in subpopulations defined by background charactertics.


Propensity Score Causal Effect Smoking Behavior Actual World Statistical Issue 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. [1]
    Agency for Health Care Policy and Research (1992), National Medical Expenditure Survey, Calendar Year 1987, Center for General Health Services Research, AHCPR, Rockville, MD: Public Health Service.Google Scholar
  2. [2]
    Anderson, S., Auquier, A., Hauck, W.W., Oakes, D., Vandaele, W., and Weisberg, H.I. (1980), Statistical Methods for Comparative Studies, New York: John Wiley.MATHCrossRefGoogle Scholar
  3. [3]
    Barwick, J. and Katikineni, D. (1992), Task NMS.3180 Report HHS Events Editing,Volume I, Division of Intramural Research, Agency for Health Care Policy and Research, Rockville, MD.Google Scholar
  4. [4]
    Best, J.A., Brown, K.S., Cameron, R., Manske, S.M., and Santi, S. (1995), Gender and predisposing attributes as predictors of smoking onset: implications for theory and practice, Journal of Health Education, 26, 2, S59 - S60.Google Scholar
  5. [5]
    Chase, G.R., Kotin, P., Crump, K., Mitchell, R.S. (1985), Evaluation for compensation of asbestos-exposed individuals. II. Apportionment of risk for lung cancer and mesothelioma, Journal of Occupational Medicine, 27, 189–255.CrossRefGoogle Scholar
  6. [6]
    Clogg, C.C., Rubin, D.B., Schenker, N., Schultz, B., and Weidman, L. (1991), Multiple imputation of industry and occupation codes in Census public-use samples using Bayesian Logistic Regression, The Journal of the American Statistical Association, 86, 413, 68–78.CrossRefGoogle Scholar
  7. [7]
    Cochran, W.G. (1957), Analysis of covariance: its nature and uses, Biometrics, 13, 261–281.MathSciNetCrossRefGoogle Scholar
  8. [8]
    Cochran, W.G. (1965), The planning of observational studies of human populations, The Journal of the Royal Statistical Society A, 128, 234–265.CrossRefGoogle Scholar
  9. [9]
    Cochran, W.G. and Rubin, D.B. (1973), Controlling bias in observational studies: a review, Sankya -A, 35, 4, 417–446.MATHGoogle Scholar
  10. [10]
    COMMIT Research Group (1995a), Community intervention trial for smoking cessation (COMMIT): I. Cohort results from a four-year community intervention, American Journal of Public Health, 85, 183–192.CrossRefGoogle Scholar
  11. [11]
    COMMIT Research Group (1995b), Community intervention trial for smoking cessation (COMMIT): II. Changes in adult cigarette smoking prevalence, American Journal of Public Health,85, 193–200.CrossRefGoogle Scholar
  12. [12]
    Dehejia, R. and Wahba, S. (1999), Causal effects in non-experimental studies: reevaluating the evaluation of training programs, The Journal of the American Statistical Association 94, 448, 1053–1062.CrossRefGoogle Scholar
  13. [13]
    Ebrahim, S. and Smith, G.D. (1997), Systematic review of randomised controlled trials of multiple risk factor interventions for preventing coronary heart disease, British Medical Journal, 314, 1666–1674.CrossRefGoogle Scholar
  14. [14]
    Efron, B. and Tibshirani, R.J. (1993), An Introduction to the Bootstrap,New York: Chapman and Hall.MATHGoogle Scholar
  15. [15]
    Engle, H.A. (1999), Verdict form for Phase I, Howard A. Engle,M.D.,et al.,Plaintiffs,v R.J. Reynolds Tobacco Company et al.,Defendants,General Jurisdiction Division Case No 94–08273/c/a-22, Circuit Court of the Eleventh Judicial Circuit in and for Dade County, Florida.Google Scholar
  16. [16]
    Ezzati-Rice, T., Johnson, W., Khare, M., Little, R.J.A., Rubin, D.B., and Schafer, J. (1995), A simulation study to evaluate the performance of model-based multiple imputations in NCHS Health Examination Surveys, Bureau of the Census Eleventh Annual Research Conference, 257–266.Google Scholar
  17. [17]
    Federal Judicial Center (1994), Reference Manual on Scientific Evidence.Google Scholar
  18. [18]
    Fisher, F. (1999), Preliminary expert report of Franklin M. Fisher, July 1, 1999. Blue Cross/Blue Shield of New Jersey et al.,Plaintiffs,vs. Philip Morris, al.Defendants,No. 98 Civ. 3287 (JBW).Google Scholar
  19. [19]
    Flay, B.R. (1985), Psychosocial approaches to smoking prevention: a review of findings, Health Psychology, 1985, 4, (5), 450.CrossRefGoogle Scholar
  20. [20]
    Gastwirth, J.L. (1988), Statistical Reasoning in Law and Public Policy,Vol.,2, New York: Academic Press, Inc., 807–810.MATHGoogle Scholar
  21. [21]
    Greenland, S. and Finkle, W.D. (1995), A critical look at methods for handling missing covariates in epidemiologic regression analyses, American Journal of Epidemiology, 1995, 142, 1255–1264.Google Scholar
  22. [22]
    Harris, J.E. (1997), Expert report, November 3, 1997: Health-care spending attributable to cigarette smoking and to cigarette manufacturers’ anti-competitive conduct: State of Washington Medicaid Program, 1970–2001, State of Washington,Plaintiff,u American Tobacco,et al.,Defendants, Superior Court of Washington in and for King County, Washington.Google Scholar
  23. [23]
    Harris, J.E. (1999), Final expert report, November 29,1999: Defendant’s anticompetitive anddeceptive onduct was a significant contributing factor in the development of smokin-related diseases among beneficiaries of Blue Cross/Blue Shield plans and increased health-care spending by those plans. Blue Cross and Blue Shield of New Jersey, Plaintiff,v,Philip Morris Incorporated,et al.,Defendants, No. 98 Civ. 3287 (JBW).Google Scholar
  24. [24]
    Harrison, G.W. (1998), Expert report, April 27, 1998: Health care expenditures attributable to smoking in Oklahoma, The State of Oklahoma,ex rel.,et al.,Plaintiffs,vs. Reynolds Tobacco Co.,et al.,Defendants, Case No. CJ-96–1499-L, District Court of Cleveland County, Oklahoma.Google Scholar
  25. [25]
    Kennickell, A.B. (1991), Imputation of the 1989 Survey of Consumer Finances: stochastic relaxation and multiple imputation, Proceedings of the Survey Research Methods Section of the American Statistical Association, 1–10.Google Scholar
  26. [26]
    Klausner, R. (1997), Evolution of tobacco control studies at the National Cancer Institute, Tobacco Control, 1997, 6 (Suppl 2), S1–S2.CrossRefGoogle Scholar
  27. [27]
    Lauer, M.S. (1998), Expert report: Smoking and coronary heart disease, undated; and Supplemental report: Smoking and stroke, October 22, 1998; Iron Workers Local Union No. 17 Insurance Fund and Its Trustees, et al., Plaintiffs, v Philip Morris Incorporated, et al., Defendants, Civil Action No. 1:97 CV1422, District Court, Northeastern District of Ohio.Google Scholar
  28. [28]
    Little, R.J.A. and Rubin, D.B. (1987), Statistical Analysis with Missing Data, New York: John Wiley and Sons. Translated into Russian in 1991: Finansy and Statistika Publishers: Moscow, Andrei Nikiforev, translator.Google Scholar
  29. [29]
    Massey, J.T., Moore, T.F., Parsons, V.L., and Tadros, W. (1989), Design and estimation for the National Health Interview Survey, 1985–94, Vital Health Statistics, 2, 110, National Center for Health Statistics, DHHS Publication No. (PHS) 89–1384.Google Scholar
  30. [30]
    Nelson, D.E., Emont, S.L., Brackbill, R.M., Cameron, L.L., Peddicord, J., and Fiore, M.C. (1994), Cigarette smoking prevalence by occupation in the United States, Journal of Occupational Medicine, 36, 5, 516–525.Google Scholar
  31. [31]
    Nicholson, W.J. (1999), Expert report, August 30, 1999: Evaluation of cigarette smoking interactions in the development of asbestos-related disease, Falise et al., Plaintiffs v The American Tobacco Company et al., Defendants, United States District Court Eastern District of New York, (CV97–76640).Google Scholar
  32. [32]
    O’Hara, P., Connet, J.E., Lee, W.W., Nides, M., Murray, R., and Wise, R. (1998), Early and late weight gain following smoking cessation in the Lung Health Study, American Journal of Epidemiology, 148, 821–832.CrossRefGoogle Scholar
  33. [33]
    Ogden, D.W. (1999), News conference, September 22, 1999, FDCH transcripts,…newsanalysis/transcripts&NSinitiaLfrm
  34. [34]
    Paulin, G. and Raghunathan, T.E. (1998), Evaluation of multiple imputation inferences, Proceedings of the American Statistical Association 1998 Survey Methods Research Section.Google Scholar
  35. [35]
    Reinisch, J.M., Sanders, S.A., Mortensen, E.L., and Rubin, D.B. (1995), In utero exposure to phenobarbital and intelligence deficits in adult men, The Journal of the American Medical Association, 274, 19, 1518–1525.CrossRefGoogle Scholar
  36. [36]
    Rabinovitz, F.F. (1999), Expert Report, September 1, 1999: Estimation value of the tobacco industry’s share of the indemnity and expenses of the Manville personal injury settlement trust, Robert A. Falise, et al., Plaintiffs v The American Tobacco Company, et al., Defendants,United States District Court, Eastern District of New York, (CV97–76640).Google Scholar
  37. [37]
    Rosenbaum, P.R. and Rubin, D.B. (1983), The central role of the propensity score in observational studies for causal effects, Biometrika, 70, 41–55.MathSciNetMATHCrossRefGoogle Scholar
  38. [38]
    Rosenbaum, P.R. and Rubin, D.B. (1985), Constructing a control group using multivariate matched sampling incorporating the propensity score, The American Statistician, 39, 33–38.Google Scholar
  39. [39]
    Rothman, K.J. and Greenland, S. (1998), Modern epidemiology, 2nd ed., Lippincott-Raven, Philadelphia.Google Scholar
  40. [40]
    Rubin, D.B. (1973), The use of matched sampling and regression adjustment to remove bias in observational studies, Biometrics, 29, 1, 184–203.Google Scholar
  41. [41]
    Rubin, D.B. (1979), Using multivariate matched sampling and regression adjustment to control bias in observational studies, The Journal of the American Statistical Association,74, 366, 318–328.MATHCrossRefGoogle Scholar
  42. [42]
    Rubin, D.B. (1984), William G. Cochran’s contributions to the design, analysis, and evaluation of observational studies, W.G. Cochran’s Impact on Statistics, P.S.R.S. Rao and J. Sedransk (eds.), New York: John Wiley, 37–69.Google Scholar
  43. [43]
    Rubin, D.B. (1987), Multiple Imputation forNonresponse in Surveys, New York: John Wiley.CrossRefGoogle Scholar
  44. [44]
    Rubin, D.B. (1996), Multiple imputation after 18+ years, Journal of the American Statistical Association, 91, 473–489 (with discussion, 507–515, and rejoinder, 515517).Google Scholar
  45. [45]
    Rubin, D.B. (1998), What does it mean to estimate the causal effects of “smoking”?, Proceedings of the Section on Statistics in Epidemiology of the American Statistical Association, 18–27.Google Scholar
  46. [46]
    Rubin, D.B. (2000), Estimating the causal effects of smoking, to appear in Statistics in Medicine.Google Scholar
  47. [47]
    Rubin, D.B. and Thomas, N. (2000), Characterizing the effect of matching using linear propensity score methods with normal covariates, Biometrika, 79, 4, 797–809.MathSciNetCrossRefGoogle Scholar
  48. [48]
    Rubin, D.B. and Thomas, N. (2000), Combining propensity score matching with additional adjustments for prognostic covariates, to appear in The Journal of the American Statistical Association, 95, 573–585.CrossRefGoogle Scholar
  49. [49]
    Rybak, D.C. and Phelps, D. (1998), Smoked: The Inside Story of the Minnesota Tobacco Trial, MSP Books: Minneapolis, MN, 394–395.Google Scholar
  50. [50]
    Samet, J.M. (1998), Trial testimony at 3482 and Deposition testimony at 369–370, State of Minnesota and Blue Cross and Blue Shield of Minnesota,Plaintiffs, u Philip Morris, Inc. et al., Defendants, Court File No. C1–94–8565, District Court of Ramsey County, Minnesota.Google Scholar
  51. [51]
    Schafer, J. (1997), Analysis of Incomplete Multivariate Data, London: Chapman and Hall. CrossRefGoogle Scholar
  52. [52]
    Simpson, E.H. (1951), The interpretation of interaction in contingency tables, The Journal of the Royal Statistical Society B, 13, 238–241.MATHGoogle Scholar
  53. [53]
    Statistical Solutions, Ltd. (1999), SOLAS 2.0 for Missing Data Analysis,
  54. [54]
    Stellman, S.D., and Garfinkel, L. (1986), Smoking habits and tar levels in a new American Cancer Society prospective study of 1 2 million men and women, Journal of the National Cancer Institute, 76, 1057–1063.Google Scholar
  55. [55]
    Strecher, V.J. (1999), Computer-tailored smoking cessation materials: a review and discussion, Patient Education and Counseling, 36, 107–117.CrossRefGoogle Scholar
  56. [56]
    U.S. Department of Health and Human Services, Public Health Service, Office of the Surgeon General, Publication No. 1103. (1964), Smoking and Health: Report of the Advisory Committee to the Surgeon General of the Public Health Service, U.S. Government Printing Office, Washington, DC.Google Scholar
  57. [57]
    U.S. Department of Health and Human Services, Public Health Service, Office of the Surgeon General, (1994), Preventing Tobacco Use Among Young People: A Report of the Surgeon General, Centers for Disease Control and Prevention, Government Printing Office, Washington, DC.Google Scholar
  58. [58]
    U.S. Department of Health, Education, and Welfare, Public Health Service, Office of the Surgeon General, Publication No. 79–50066, (1979), Smoking and Health: A Report of the Surgeon General, U.S. Government Printing Office, Washington, DC.Google Scholar
  59. [59]
    U.S. Department of Transportation, NHTSA (1998), Multiple imputation of missing blood alcohol content (BAC) in FARS, Research Note: National Highway Traffic Safety Administration, Washington, DC.Google Scholar
  60. [60]
    Velicer, W.F., Prochaska, J.O., Bellis, J.M., DiClemente, C.C., Rossi, J.S., Fava, J.L., and Steiger, J.H. (1993), An expert system intervention for smoking cessation, Addictive Behaviors, 18, 269–290.CrossRefGoogle Scholar
  61. [61]
    Viscusi, W.K. (1991), Age variations in risk perceptions and smoking decisions, The Review of Economics and Statistics, 73,4,577–588.CrossRefGoogle Scholar
  62. [62]
    Warner, K.E., Chalouka, F.J., Cook, P.J., Manning, W.G., Newhouse, J.P., Novotny, T.E., Schelling, T.C., and Townsend, J. (1995), Criteria for determining an optimal cigarette tax: the economists’s perspective, Tobacco Control, 4, 380–386.CrossRefGoogle Scholar
  63. [63]
    Winkleby, M.A., Feldman, H.A., and Murray, D.M. (1997), Joint analysis of three U.S. community intervention trials for reduction of cardiovascular disease risk, Journal of Clinical Epidemiology, 50, 6, 645–658.CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media New York 2000

Authors and Affiliations

  • Donald B. Rubin

There are no affiliations available

Personalised recommendations