• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of amjphAmerican Journal of Public Health Web SiteAmerican Public Health Association Web SiteSubmissionsSubscriptionsAbout Us
Am J Public Health. 2002 June; 92(6): 961–965.
PMCID: PMC1447495

Do Sex and Ethnic Differences in Smoking Initiation Mask Similarities in Cessation Behavior?


Objectives. This study compared success in smoking cessation by sex, ethnic status, and birth cohort.

Methods. African and European American respondents to the 1996 Current Population Survey (tobacco supplement) and the 1987 National Health Interview Survey (cancer control and cancer epidemiology supplements) constituted the study population. Elapsed time from smoking initiation to cessation was compared via nonparametric tests and survival analysis techniques.

Results. Findings showed that success in quitting was independent of ethnic status and sex and that population differences in smoking initiation age (assuming no differences in quitting behavior) could produce statistical associations between sex/ethnicity and smoking cessation.

Conclusions. Population differences in smoking initiation patterns can mask similarities in cessation rates. (Am J Public Health. 2002;92:961–965)

Cigarette smoking remains the most important contributor to preventable morbidity and mortality in the United States.1 Promotion of smoking cessation is a public health priority, but it continues to be the case that most individuals who successfully quit smoking do so on their own, without the aid of formal programs.2–7 Among individuals who have quit on their own, data from many epidemiological studies suggest that men are more likely to quit than are women,5,8,9 and Whites more likely to quit than Hispanics or African Americans.5,10–15 Results regarding sex and ethnic differences in cessation have not, however, been consistent.16,17 Specifically, when duration of smoking has been taken into account,16 no differences in quitting behavior have been found.

In this study, we examined ethnic and sex differences in successful quitting, taking duration of smoking into account. Elapsed time from smoking initiation to cessation (“time to quit”) was defined as individuals' reported age at quitting minus their reported age at initiation. This quantity approximates duration of smoking (it does not adjust for time off smoking due to quit attempts) and serves as an individual-level measure of successful quitting. In the framework of survival analysis, time to quit (or duration) is considered a “failure time,” a strategy that allows cessation to be viewed as a dynamic population process.

Conceptualizing cessation as a population process brings observational and measurement issues to attention. Specifically, it forces consideration of the possibility that the results of a process measurement taken at one point in time may differ from the results of the same measurement taken at a later time as a consequence of the time evolution of the process. We analyzed data sets collected in 1987 and 1996 in an effort to assess this possibility and its implications. We addressed the possibility of cohort differences in quitting behavior (generational trends) by constructing and comparing birth cohorts with respect to time-to-quit values.


Data collected in 2 national surveys, the Current Population Survey (CPS) of January 1996 and the National Health Interview Survey (NHIS) of 1987, were used in analyzing smoking cessation. Both the CPS and the NHIS are ongoing surveys targeting the civilian, noninstitutionalized population of the United States.

The NHIS is conducted by the National Center for Health Statistics and serves as the principal source of information on the health of the US population. In this survey, a multistage probability design is used to select sample households representative of the target population. Within selected households, information on all adult members (18 years or older) is collected by trained interviewers.18 In 1987, tobacco use and other cancerrelated information was assessed at the national level via supplementary cancer control and cancer epidemiology questionnaires.

The CPS is conducted by the Bureau of the Census and serves as the source of official government statistics on employment and unemployment. This survey also involves the use of a multistage probability design in selecting nationally representative households. Within households, information on all members 15 years or older is collected.19 In January 1996, a tobacco use survey was conducted as a supplement to the CPS.20 Proxy responses were permitted in this supplement (as shown in the first 3 CPS questions described subsequently), but these data were excluded from the analyses described here.

Information of interest in analyzing cessation was assessed via self-reports in both surveys. Variables used were derived from respondents' answers to the following questions: (1) Have you smoked 100 cigarettes in your entire life? (NHIS) or Has [household member in question] smoked 100 cigarettes in his/her entire life? (CPS); (2) Do you currently smoke cigarettes? (NHIS) or Does [household member in question] now smoke cigarettes every day, some days, or not at all? (CPS); (3) How old were you when you first started smoking cigarettes fairly regularly? (NHIS) or How old was [household member in question] when he/she first started smoking cigarettes fairly regularly? (CPS); and (4) How long ago did you stop smoking? (NHIS) or About how long has it been since you completely stopped smoking cigarettes? (CPS).

Individuals responding yes to question 1 were classified as ever smokers. Those responding no (NHIS) or not at all (CPS) to question 2 were classified as former smokers. Age at initiation was determined by response to question 3. Finally, for question 4, time to quit was defined for each individual as reported length of time from initiation to cessation of smoking. Among former smokers, current age (in days) and reported elapsed time since quitting were used in computing age at cessation.

Among CPS respondents, age at survey completion was used to estimate year of birth (1995 minus age); NHIS respondents reported their year of birth. Ten-year birth cohorts (1896–1905, 1906–1915, . . ., 1966–1975, 1976 and later), selected for compatibility with our previous investigations,21 were constructed according to birth year. Data for ethnic groups other than Black or White (CPS, n = 2221; NHIS, n = 1537) and data for which the time-to-quit variable was missing or unknown were not analyzed.

A successful quitter was defined as a former smoker who had maintained abstinence for 1 year or longer. Initially, we assessed sex and ethnic differences in successful quitting by comparing time-to-quit distributions using appropriate k-sample and 2-sample tests. These results did not, however, account for the complex sampling design used in the 2 surveys. Thus, another hypothesis-testing approach was taken to account for the complex design. The Kolmogorov–Smirnov statistic22,23 was used in pairwise comparisons of time-to-quit distributions. P values associated with these tests were adjusted, via estimation of design effect and effective sample size,24,25 to account for the survey design.

The adjustment procedure followed that of Kish.24 For each birth cohort and each combination of 2 subdomains—race/ethnicity and sex—a weighted Kolmogorov–Smirnov statistic and associated P value were calculated. Because survey design variables were not published in the CPS data, we calculated the variances (accounting for sampling design) for Kolmogorov–Smirnov statistics estimated in that data set by considering these statistics as differences in proportions and using the generalized variance functions supplied in the documentation.

In the case of the NHIS data, we used bootstrap resampling in estimating variance.26,27 Specifically, we produced 200 resamples by sampling with replacement from 62 pseudostrata; 3 (of 4) pseudo–primary sampling units were sampled independently in each pseudostratum. As a means of bias reduction,27 the weight associated with each selected record was multiplied by the factor 1.33 (4/3).

We then calculated design effects24 for the contrast by comparing the variance with the survey design taken into account and the simple random sampling variance. On the basis of these design effects, we calculated effective sample sizes and adjusted P values.23,28


Respondent characteristics are shown in Table 1 [triangle]. African Americans were oversampled in the 1987 NHIS. In those data, successful quitters were represented in all birth cohorts, although the sample size was small for the 1896 birth cohort (Table 1 [triangle]). Fewer African Americans overall, and no successful quitters belonging to the 1896 cohort, were sampled in the 1996 CPS.

—Numbers of Respondents, by Birth Cohort, Sex, Ethnic Status, and Reported Smoking Status

Figure 1 [triangle] displays male and female time-to-quit distributions by birth cohort. Distributions are presented as survival curves; probabilities of successful cessation are shown on the y-axis, and time-to-quit values are shown on the x-axis. Pointwise confidence bands are included. The similarity of male and female cessation patterns is evident in all birth cohorts, with confidence bands indicating that the distributions are statistically indistinguishable. A similar conclusion held for the comparison of African American and European American cohorts (data not shown). In both cases, confidence bands were calculated with an assumption of simple random sampling. That is, the complex sampling design of the survey data was not accounted for in these results. As a remedy, hypothesis-testing procedures (as described earlier) were used, and results (available on request) confirmed that the compared distributions were statistically indistinguishable.

—Male and female time-to-quit distributions, by birth cohort: Current Population Survey, 1996.

The quit ratio is a commonly used measure of smoking cessation in populations. We examined the proposition that 2 populations having identical time-to-quit distributions can display significantly different quit ratios if the pattern of initiation is different in the populations. If members of one group systematically initiate smoking at earlier ages than members of a comparison group, and both groups are equally likely to quit after a given duration of smoking, then the early-initiating group would be expected to contain a larger proportion of quitters at most observation times.

A Monte Carlo simulation experiment was performed to assess this proposed mechanism quantitatively. Mean prevalence rates of current smoking and quit ratios were generated for 2 populations under the assumption of identical time-to-quit distributions but differing initiation patterns. We compared quit ratios by computing a relative ratio representing the quit ratio of one group divided by the ratio of the other group.

The experiment revealed that lower mean age at initiation for one population results in relative ratios above 1.0 that decrease toward 1.0 as observation time increases; larger differences in mean age at initiation result in larger relative ratios. Variability in initiation age has an effect independent of mean age; other factors being equal, initiation occurring over shorter periods of time in one population increases the relative ratio. The simulated prevalence of smoking is initially larger in early-initiating groups but becomes smaller in comparison as cessation proceeds. The results of this experimental simulation confirm that statistical associations of sex and ethnic status with quit ratios on the order reported in the literature could be produced by differences in mean age at initiation of 1 or 2 years.


The analyses described here lead to 2 main conclusions. First, our findings contradict previous research results suggesting that success in quitting cigarette smoking is a function of ethnic status or sex and that women and African Americans lag behind in terms of smoking cessation. The validity of these findings is supported by 2 considerations: (a) the data examined are nationally representative; and (b) results were consistent over 2 separate observations (1987 and 1996) and for all cohorts examined, whose experiences span the history of mass cigarette use in the United States. Second, given that ethnic and sex differences in initiation age have been demonstrated for African American and European American populations,29–32 use of the quit ratio to compare cessation rates in these populations is problematic. We suspect that any measure of cessation that does not account for an individual's duration of smoking may involve similar limitations.

Two possibilities may threaten the validity of these conclusions. First, comparison of time-to-quit distributions could lead to error if reported initiation age, smoking status, or age at cessation were systematically misclassified. In the literature of which we are aware,33–35 evidence does not support the presence of sex or ethnic bias in reports of these variables. A second possibility is that time to quit may be misrepresented if quit attempts or time away from smoking resulting from such attempts differ systematically between populations (e.g., if number of quit attempts or time away from smoking during attempts is systematically greater in one population). Data relating to quit attempts among former smokers were not collected in the CPS, but such data were collected in the cancer control supplement of the 1987 NHIS. Analyses of these data indicate no significant male–female or Black–White differences in either number of quit attempts or duration of most recent (or only) quit attempt.

Implications of these conclusions for smoking cessation surveillance and research are clear. With respect to tobacco use surveillance, use of the quit ratio appears to be inadequate. It is important to measure and monitor both initiation and cessation if present status and trends in smoking are to be correctly interpreted. Similarly, with respect to smoking cessation research, measures of quitting that account for duration of smoking may be most appropriate, both in analyzing factors associated with cessation and in evaluating interventions. Additional research on the relation between number of quit attempts and duration of attempts might be helpful in connecting clinical trial results to long-term successful cessation.


This research was supported in full by the National Cancer Institute (grant RO3-CA83337).


Both authors participated in the initial conception and design of the study. G. A. McGrady was primarily responsible for the data analysis. Both authors contributed to interpretation of data and to drafting and revision of the article.

Peer Reviewed


1. Centers for Disease Control and Prevention. Smoking-attributable mortality and years of potential life lost—United States, 1984. MMWR Morb Mortal Wkly Rep. 1997;46:444–451. [PubMed]
2. Cancer Facts and Figures. New York, NY: American Cancer Society; 1986.
3. Fiore MC, Novotny TE, Pierce JP, et al. Methods used to quit smoking in the United States: do cessation programs help? JAMA. 1990;263:2760–2765. [PubMed]
4. Lando HA, Pechacek TF, Fruetel J. The Minnesota Heart Health Program community Quit and Win contests. Am J Health Promot. 1994;9:85–87. [PubMed]
5. Coambs RB, Lis S, Kozlowski LT. Age interacts with heaviness of smoking in predicting success in smoking cessation. Am J Epidemiol. 1992;135:240–246. [PubMed]
6. Pederson LL. Smoking behavior of Canadians. In: Canada's Health Promotion Survey 1990: Technical Report. Ottawa, Ontario, Canada: Health and Welfare Canada; 1993:91–101.
7. Prochaska JO, DiClemente CC, Velicer WF, et al. Predicting change in smoking status for self-changers. Addict Behav. 1985;10:395–406. [PubMed]
8. Hymowitz N, Cummings M, Hyland A, et al. Predictors of smoking cessation in a cohort of adult smokers followed for five years. Tob Control. 1997;6(suppl 2):S57–S62. [PMC free article] [PubMed]
9. Derby CA, Lasater TM, Vass K, et al. Characteristics of smokers who attempt to quit and those who recently succeeded. Am J Prev Med. 1994;10:327–334. [PubMed]
10. Gilpin E, Pierce JP, Goodman J, et al. Reasons smokers give for stopping smoking: do they relate to success in stopping? Tob Control. 1992;1:256–263.
11. Tunstall CD, Ginsberg D, Hall SM. Quitting smoking. Int J Addict. 1985;10:1089–1112. [PubMed]
12. Haire-Joshu D, Morgan G, Fisher EB. Determinants of cigarette smoking. Clin Chest Med. 1991;12:711–725. [PubMed]
13. Effectiveness of smoking-control strategies: editorial note from the Centers for Disease Control. JAMA. 1992;268:1645–1646. [PubMed]
14. Carmody TP. Affect regulation, nicotine addiction, and smoking cessation. J Psychoactive Drugs. 1992;24:111–122. [PubMed]
15. Centers for Disease Control and Prevention. Cigarette smoking among adults—United States, 1993. MMWR Morb Mortal Wkly Rep. 1994;43:925–930. [PubMed]
16. Freund KM, D'Agostino RB, Belanger AJ, et al. Predictors of smoking cessation: the Framingham Study. Am J Epidemiol. 1992;135:957–964. [PubMed]
17. Fiore MC, Novotny TE, Pierce JP, et al. Trends in cigarette smoking in the United States: the changing influence of gender and race. JAMA. 1989;261:49–55. [PubMed]
18. Sample Design and Estimation Procedures for the National Health Interview Survey, 1985–1994. Hyattsville, Md: National Center for Health Statistics; 1989. DHHS publication PHS 89-1384.
19. The Current Population Survey Design and Methodology. Washington, DC: US Bureau of the Census; 1985. Technical paper 40.
20. Current Population Survey, January 1996: Tobacco Use Supplement [machine-readable data file]. Washington, DC: US Bureau of the Census; 1998.
21. McGrady GA, Ahluwalia JS, Pederson LL. Smoking initiation and cessation in African Americans attending an inner-city walk-in clinic. Am J Prev Med. 1998;14:130–137. [PubMed]
22. Siegal S. Nonparametric Statistics for the Behavioral Sciences. New York, NY: McGraw-Hill Book Co; 1956:127–136.
23. Kendall M, Stuart A. The Advanced Theory of Statistics: Inference and Relationship. Vol. 2. 4th ed. London, England: Charles Griffen & Co Ltd; 1979.
24. Kish L. Survey Sampling. New York, NY: John Wiley & Sons Inc; 1965.
25. Skinner CJ, Hold D, Smith TMF. Analysis of Complex Surveys. New York, NY: John Wiley & Sons Inc; 1989.
26. Efron B, Tibshirani RJ. An Introduction to the Bootstrap. New York, NY: Chapman & Hall; 1993.
27. Korn EG, Graubard BI. Analysis of Health Surveys. New York, NY: John Wiley & Sons Inc; 1999.
28. Press WH, Teukolsky SA, Vetterling WT, Flannery BP. Numerical Recipes in C: The Art of Scientific Computing. 2nd ed. New York, NY: Cambridge University Press; 1992.
29. Escobedo LG, Anda RF, Smith PF, Remington PL, Mast EE. Sociodemographic characteristics of cigarette smoking initiation in the United States: implications for smoking prevention policy. JAMA. 1990;264:1550–1555. [PubMed]
30. Gilpin EA, Lee L, Evans N, Pierce JP. Smoking initiation rates in adults and minors: United States, 1944–1988. Am J Epidemiol. 1994;140:535–543. [PubMed]
31. Centers for Disease Control. Differences in the age of smoking initiation between blacks and whites—United States. MMWR Morb Mortal Wkly Rep. 1991;40:754–757. [PubMed]
32. Faulkner DL, Escobedo LG, Zhu B, Chrismon JH, Merritt RK. Race and the incidence of cigarette smoking among adolescents in the United States. J Natl Cancer Inst. 1996;88:1158–1160. [PubMed]
33. Krall EA, Valadian I, Dwyer JT, Gardener J. Accuracy of recalled smoking data. Am J Public Health. 1989;79:200–206. [PMC free article] [PubMed]
34. Means B, Habina K, Swan GE, Jack L. Cognitive research on response error in survey questions in smoking. Vital Health Stat 6. 1992;No. 5.
35. Clark P, Gautam SP, Hlaing WM, Gerson LW. Response error in self-reported current smoking frequency by black and white established smokers. Ann Epidemiol. 1996;6:483–489. [PubMed]

Articles from American Journal of Public Health are provided here courtesy of American Public Health Association
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...