Skip to main content
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2011 Aug 28.
Published in final edited form as: J Health Soc Behav. 2008 Dec;49(4):484–498.

Racial Convergence in Cigarette Use from Adolescence to the Mid-Thirties*

FRED C PAMPEL 1
PMCID: PMC3163096  NIHMSID: NIHMS317390  PMID: 19181051

Abstract

Cigarette smoking by whites and African Americans shows puzzling age differences: An African American advantage during the teen years no longer appears in mid-adulthood. This study uses two data sets to examine whether the life-course change is real–and not due to misleading comparisons across different cohorts–and then whether the racial convergence is consistent with resource or stress arguments emphasizing, respectively, cessation among whites or late initiation among African Americans. First, multilevel growth models using data from the National Youth Survey–a prospective, longitudinal study of a randomly selected national sample of teens followed from ages 12 to 18 in 1977 to ages 26 to 34 in 1992–reveal that the racial convergence in smoking prevalence over the life course among this cohort is due primarily to greater white cessation. Second, consecutive cross-sectional samples of the National Health Survey replicate the broad patterns found in the NYS and show that convergence in smoking trajectories by race has strengthened over time. Together, the results most favor a resource explanation of the different life-course patterns of smoking among whites and African Americans.


Smoking among non-Hispanic whites and African Americans differs strikingly across ages. On one hand, African Americans smoke considerably less than whites during the teen years (Ellickson et al. 2004). According to the 2004 Monitoring the Future (MTF) survey (Johnston et al. 2005:633–634), 27.6 percent white, but only 10.7 percent of African American, high school seniors smoked–a 16.9 percentage point gap. This racial difference in teen smoking has widened considerably over the last several decades. In 1978, the prevalence of smoking among white and African American high school seniors equaled, respectively, 37.6 and 32.7 percent–a gap of 5.9 percentage points. Cigarette use has become one component of health behavior that gives African American teens a distinct advantage over white teens.

On the other hand, adult African Americans smoke at rates similar to, or slightly higher than, those of adult whites. For example, figures computed from the 2005 National Health Interview Survey for persons ages 40–59 show current smoking of 23.8 percent among non-Hispanic whites and 25.9 percent among African Americans (National Center for Health Statistics 2007). The slightly higher figure for African Americans stems in part from the influence of socioeconomic status (SES). Over the past several decades, cigarette use has become concentrated among lower education, occupational prestige, and income groups (Barbeau, Krieger, and Soobader 2004; Honjo et al. 2006). Since the social and economic disadvantage of African Americans should raise their smoking prevalence, adjusting for education reduces the gap reported in the NHIS data to near zero (also see Kiefe et al. 2001). However, the near-identical smoking levels among adults still contrast markedly from the African American advantage among teens, a fact noted in the literature (Edelan, Tucker, and Ellickson 2007; Gardiner 2001; Geronimus, Neidert, and Bound 1993; Moon-Howard 2003; Trinidad et al. 2004).

Less clear is how this racial convergence in smoking might have emerged. In general terms, the convergence involves changes in smoking that can be viewed from a life-course perspective (Bachman et al. 2002; Graham et al. 2006; Lynch, Kaplan, and Salonen 1997). Racial patterns of smoking reflect varying pathways or trajectories through early adolescence, young adulthood, and later adulthood. A life-course perspective suggests that these pathways reflect different social experiences during the transitions from adolescence to early and middle adulthood (Alwin and Wray 2005; Yu 2006).

Describing better the trajectories in smoking seems crucial to understanding racial disparities in health. Given the well-known standing of cigarette use as the largest source of premature mortality in high-income nations (Department of Health and Human Services (DHHS) 2004), efforts to reduce smoking among African Americans can do much to improve their health and longevity and to counter racial disparities in mortality (Fagan et al. 2004; Lauderdale 2001). The NIH Strategic Plan to Reduce and Ultimately Eliminate Health Disparities (NIH 2000) thus gives prominent attention to tobacco use. Knowing more about the life-course pathways that eliminate a potential African American advantage in smoking can help identify key strategies for change (Kandel et al. 2004).

SOURCES OF RACIAL CONVERGENCE IN SMOKING

Racial convergence in smoking across ages can result from three patterns of life-course stability and change (Geronimus et al. 1993).1 These patterns are explained by theories focusing on resources, stress, and cohort differences.

First, whites may be more likely to quit smoking in young adulthood (King et al. 2004). Because early quitting moderates the harm of the habit on later health, anti-tobacco strategies encourage young smokers to stop (DHHS 2000). And studies do indeed demonstrate a decline in smoking during the twenties as smokers take on new responsibilities and recognize their youthful mistake in starting in the first place (Bachman et al. 1997; Chen and Kandel 1995). However, lower SES groups and racial minorities have more trouble quitting than higher SES groups and whites (Barbeau et al. 2004; DHHS 1998; Honjo et al. 2006). Higher rates of successful cessation among whites thus may reduce the race gap in smoking as age increases.

The key theoretical mechanism underlying this argument relates to resources. The desire to quit smoking is common among all groups, but high-SES smokers succeed more than others do in realizing this goal (Barbeau et al. 2004; King et al. 2004). Whites have more financial resources than minority groups to use for counseling, nicotine replacement therapy, assistance from physicians, and prescription drugs (Honjo et al. 2006); they have greater educational skills and knowledge to use in devising solutions to nicotine addiction (Mirowsky and Ross 2003); and they have more varied occupational experiences to draw on in overcoming obstacles (Sorensen et al. 2004). In addition, forms of social capital, such as friendship, neighborhood, and acquaintance networks that offer support, encouragement, and approval for quitting (Chen, White, and Pandina 2001), may help white smokers more than African American smokers. These resources may not operate during the teen years, when unconventionality and a search for independence dominate peer groups (Turbin, Jessor, and Costa 2000). During adulthood, however, they may become central to greater cessation among whites and to convergence across races in smoking.

Second, African Americans may start smoking at older ages (Geronimus et al. 1993; Moon-Howard 2003; Trinidad et al. 2004). During the teen years, high levels of parental disapproval of smoking, involvement in religious activities, and attraction to sports appear to limit smoking among African Americans (Carmona et al. 2004; Ellickson et al. 2004; Ellickson, Perlman, and Klein 2003). However, during the transition to adulthood, when African Americans leave high school and their neighborhoods for college and the labor market, they may face heightened levels of racism and discrimination (Wallace et al. 1995; Williams 2005). Under difficult social and personal circumstances, minority groups may come to use cigarettes and nicotine as a form of self-medication and coping (Bennett et al. 2005; DHHS 1998). Although smoking typically begins during the teen years, experiences with racism may promote smoking initiation at later ages among African Americans (Gardiner 2001; Trinidad et al. 2004). As a result, African American smoking may steadily increase during the twenties to reach the levels of whites.

The key theoretical mechanism underlying this argument relates to stress. Studies demonstrate a relationship between stress and smoking that helps explain high usage rates among low SES groups (Dugan, Lloyd, and Lucas 1999; Udry, Li, and Hendrickson-Smith 2003). With disadvantaged groups facing a struggle in daily life to meet demands they often cannot satisfy, smoking and nicotine may offer some relief from chronic stress to those unable to purchase more expensive forms of pleasure or therapy (Graham 1995; Wilkinson 1996). Deprived people under stress might also reason that, given the other sources of premature mortality they face, smoking represents a relatively minor additional threat (Lawlor et al. 2003). Thus stress leads to higher smoking, and the added stress experienced by African Americans in young adulthood may increase their smoking relative to that of whites.

Third, the apparent racial convergence in age-based patterns of smoking may come from cohort differences. Since 88 percent of those who have ever smoked a cigarette did so for the first time by age 18 (DHHS 1994:67), cohort experiences at young ages greatly influence adult smoking (Jacobson et al. 2001; Preston and Wang 2006). Thus, African American and white adults today may have similar rates of smoking because teen African Americans and whites had similar rates several decades ago. This claim matches reports from the Monitoring the Future surveys that race differences in smoking among high school seniors were small in the late 1970s (Johnston et al. 2005). Conversely, the widening racial gap among high school seniors that occurred more recently may show up among adults in the future once the younger cohorts more fully replace older cohorts. In suggesting that the racial gap in smoking during the teen years persists largely unchanged through adulthood, this argument downplays the importance of race differences in cessation and late initiation, and emphasizes the importance of teen initiation. It suggests that the ostensible racial convergence in smoking is an artifact of confounding age with cohort and highlights the need to compare age patterns of smoking within as well as across cohorts.

In summary, a narrower race gap in smoking at adult ages may stem from (1) whites’ greater access to resources that help them quit, (2) greater strains faced by African Americans in young adulthood that lead to late initiation, or (3) persistence of cohort-based patterns established as teens. Data are not available to test directly for the mechanisms specified by each theory (i.e., problem-solving resources, stressful perceptions of discrimination, and the social contexts of cohorts during their teen years). However, describing age trajectories by race and gender can offer insights into the underlying processes. The preponderant influence of resources should show up in relatively stable smoking during the post-teen years among African Americans but falling smoking among whites. The preponderance of differential stress should show in smoking that levels off in early adulthood among whites but continues to rise among African Americans. The preponderance of cohort influence should show first in largely parallel within-cohort changes in smoking among whites and African Americans and second in changing differences in the race gap across cohorts. Of course, all three forces may operate simultaneously, and, given the overlap between resources and stress, support may emerge for each of the arguments. But presenting each hypothesis as being conceptually distinct helps organize the arguments and facilitate the tests to follow.

Few studies have examined the longitudinal age patterns of change by race. Instead, studies compare age patterns of smoking by race with cross-sectional data (Moon-Howard 2003), examine trajectories for age of initiation but not age of cessation (Edelen, Tucker, and Ellickson 2007; Trinidad et al. 2004), or contrast racial patterns in cessation but not in age of initiation (King et al. 2004; Weden and Kimbro 2007). Dozens of longitudinal studies examine smoking trajectories but without making racial comparisons (e.g., Bachman et al. 1997; Chen and Kandel 1995). Of those that do examine racial differences, most cover only a short time period. For example, studies of smoking using the National Longitudinal Survey of Adolescent Health have followed youth for up to seven years (Hu, Davies, and Kandel 2006; Kandel et al. 2004), while other studies have focused on changes from ages 13 to 23 (Ellickson et al. 2004) or from 18 to 30 (Bachman et al. 1997, 2002; Kiefe et al. 2001). Coming closest to the goals of this study, Geronimus et al. (1993) use retrospective data on initiation and cessation from a 1987 survey to chart changes from ages 15 to 35 for women. Yet studies rarely have longitudinal data that cover changes in smoking from the early teen years through periods of late initiation and cessation during the twenties and into the thirties, and also use other data to examine age patterns across multiple cohorts.

METHODS

Data

This study uses two data sources to examine trajectories of smoking among whites and African Americans from the teen years through early and middle adulthood. The first data set, the National Youth Survey (NYS), is a prospective longitudinal study of a randomly selected national sample of teens ages 12–18 in 1977 to ages 26–34 in 1992. Although designed to study life-course changes in criminal offending, it contains a few items on cigarette use, includes measures of family background and educational attainment, and covers a wide age span. The second data set, the National Health Interview Surveys (NHIS), consists of consecutive cross-sectional surveys of persons age 18 and older that were done from 1976 to 2006. Although not longitudinal, the NHIS data offer an opportunity for comparisons across as well as within cohorts, and they avoid problems of panel attrition with new random samples each year.

The NYS began in 1976 with a survey of 1,725 boys and girls ages 11–17 (born from 1959 to 1965). The sample comes from a multistage cluster design that selected 7,998 households and all youths ages 11–17 in these households who were mentally and physically able to complete the interviews. Of 2,360 eligible youths, 1,725 (73 percent) agreed to participate, and both a parent and the respondent completed surveys. The sample appears representative of the population. No significant differences in age, sex, or race–ethnicity existed among those refusing and agreeing to participate, and the sample proportions matched Census data for the age group (Elliott, Huizinga, and Ageton 1985).

Additional surveys of respondents were done in 1977, 1978, 1979, 1980, 1983, 1986, 1989, and 1992, by which time the cohort had reached ages 26–34. The surveys were conducted primarily through face-to-face interviews. Attrition reached 13 percent over the first five waves, and 21 percent by wave nine. Even with the loss of respondents–an inevitable result of any longitudinal study covering such a long time spancovering such a long timeinformation remains for about 1,300 cases. Although no significant differences appear in demographic characteristics of continuing participants and dropouts (Menard et al. 1994), the analyses to follow adjust for changes in the sample across waves.

However, items on cigarette smoking are not available for all waves and cases. The survey for wave 1 did not include questions on substance use. Starting with wave two in 1977 at ages 12–18, questions on use of tobacco and other substances were asked to a random sample of half the original respondents. Starting with wave four in 1979 at ages 14–20, all respondents answered questions on use of tobacco. Since those not asked about smoking until wave four are randomly selected, they do not differ in ways other than chance from those asked earlier. This randomness means cases can be added to the analysis of patterns of smoking at later ages and waves without biasing the results. And even for those not asked about smoking until later ages, measures of SES characteristics obtained from earlier waves and ages, including some based on responses from parents, can be used to predict later smoking.

The NYS sample is 53 percent male, 47 percent female, 79 percent non-Hispanic white, 15 percent African American, and 6 percent other race or ethnic groups. Given the small numbers and disparate backgrounds of the other category, this analysis is limited to whites and African Americans. For African Americans, the number of respondents with data on tobacco use equals 252. Combining persons and waves, the total number of cases for African Americans reaches 1,478. For whites, 1,310 respondents have data, and the total number of person-wave cases equals 8,049.

The NHIS is a continuous survey of the U.S. civilian, non-institutionalized population that uses a multistage probability sampling design to select representative samples of households and interview household members age 18 and older. It has asked questions on smoking in 24 surveys over the 31 years from 1976 to 2006 (National Center for Health Statistics 2007). It does not include younger teens, and the selection of a new sample each year means one cannot compare the same individuals at different ages. However, the high quality of the surveys for the study of cigarette use and the newly chosen random sample of respondents each year avoid some of the problems of the NYS. Given the later starting age of the NHIS, the analysis follows the cohorts to age 37, an age that is older than for the NYS but young enough to include more recent birth cohorts.

Variables

The NYS measure of cigarette use comes from a question on the number of cigarettes used per day, with those smoking zero cigarettes coded as nonsmokers and those smoking one or more cigarettes a day coded as smokers. Note that the measure may miss irregular smokers. Note also that comparisons across age on this measure come not from direct questions on age of initiation and cessation but from the proportion of respondents at each age reporting that they smoke some cigarettes. Net increases in smoking reflect an excess of new smokers over quitters, and net decreases reflect an excess of quitters over new smokers.

The NHIS measures smoking differently than the NYS. Ever smokers are defined as persons who report smoking at least 100 cigarettes in their lifetime. Among ever smokers, current and former smokers are distinguished by their answers to the question, “Do you smoke now?” The 1978 figures on current smoking for those age 18 equal 34.6 for whites and 32.2 for African Americans, leaving a gap of 2.4 percentage points. The percentages of white and African American NYS respondents in 1978 at age 18 who said they smoked one or more cigarettes a day equal 33.1 and 28.0–a gap that is larger than for the NHIS. The difference may stem from the forms of questions (cigarettes smoked versus current and former smoking), the definition of smoking (“daily” versus “daily or occasional”), and the nature of the samples (following the same sample versus new random samples each year). In any case, the discrepancies stress the need to examine age patterns in smoking by race using both the NYS and the NHIS.

As for other variables, the first wave of the NYS has information from a parent on family background. Two broad measures reflect the SES of the respondent’s family. One comes from questions on the years of schooling completed by the mother and father, which are averaged into a single measure of parents’ education ranging from 1 (some grade school) to 7 (post-graduate degree). Respondents missing data for one parent are assigned the value of the other parent, and those missing data for both parents, 2.4 percent of the cases, are assigned the mean of the race-group to which they belong. The occupation of the household’s principal wage earner is coded into seven categories from low prestige to high prestige (those with no occupation listed, 5.7 percent of the cases, are assigned the mean of the respondent’s race-sex group). These measures are available only for the time of the first wave (i.e., 1976, when the respondents were ages 11–17) rather than for all waves.

All waves of the NYS contain measures of the respondent’s schooling. One measure indicates whether a respondent attends (equal to 1) or does not attend (equal to 0) an academic school at the time of each interview. Another measure equals current grade of those in school or the last completed grade of those not in school (ranging from 5 to 17, with 17 indicating graduate or professional school). Unlike the measures of parents’ education and occupation, these measures change with age and wave of the survey, and they control for changes over the life course in educational success. The NHIS data include a control for the respondents’ years of completed education.

Estimation

The NYS analysis uses multilevel growth-curve models to describe smoking trajectories over the life course (Raudenbush and Bryk 2002), but with adjustment for selection bias due to attrition (Willson, Shuey, and Elder 2007). In the multilevel models, ages for each individual serve as level one units, and individuals serve as level two units. Smoking is treated as a function of an age quadratic and time-varying control variables within individuals, and the growth-curve coefficients for age and age-squared are then treated as linear functions of race, sex, and time-invariant controls across individuals. The models allow for use of subjects with incomplete data at level one, a key need given the loss of cases through attrition and the late start of half of the NYS sample in answering smoking questions.

The models include respondent’s age, school attendance, and current or completed years of schooling as level one time-varying determinants and include race, sex, parents’ education, and occupation of main wage earner at wave 1 as level two time-invariant determinants. At level two, race and sex take the form of dummy variables (African American males, white females, and African American females compared to the omitted reference group of white males) that affect the levels and age patterns of smoking of individuals. Also, level two contains a propensity measure to adjust for attrition bias. A logistic regression model computes for each wave and each respondent the predicted probability of being included in the survey based on predictors available in wave 1.2 For each respondent used in the multilevel analysis, the propensity score measures the average predicted probability of inclusion across waves and is included as a level two control for attrition (Willson et al. 2007). This control does not greatly change the results, as race and gender are not closely related to selection, but it does adjust for a common problem in longitudinal analyses by correcting for correlation of the error term with covariates.

Deleting missing data leaves 1,560 individuals and 9,496 person-wave cases available for analysis. The intraclass correlation coefficient of .602 reveals that more than half of the variation occurs between individuals, and the remainder occurs over time and within individuals. Estimates come from HLM 6 (Raudenbush et al. 2004).

The NHIS modeling proceeds differently. With the consecutive cross-sectional design, pseudo-growth curves can be estimated with logistic regression models of current smoking by using age, age-squared, race-sex dummy measures, cohort, and education as independent variables. The effects of age and age-squared are allowed to vary not only by race and sex but also by cohort (and, implicitly, period). In both its additive and its interactive forms, the influence of cohort takes a linear form, which reflects the steady decline in smoking across cohorts and smoothes changes in the age pattern of smoking across cohorts. Since the model includes 24 additive and interactive terms, the results are presented in the form of predicted age curves by race, sex, and cohort. With 181,064 whites and 37,945 African Americans, the NHIS contains enough cases for race comparisons across both ages and cohorts.

RESULTS

National Youth Survey

Descriptive statistics illustrate in rough terms that levels of smoking among whites and African Americans appear to converge over the life-course. Table 1 presents figures on the percentage of current smokers (and the number of cases) for five age groups: 12–15, 16–19, 20–23,24–27, and 28–34. For males, the percentage of smokers among whites exceeds that for African Americans until ages 24–34. The gap begins at 3 percentage points at the youngest ages, rises to 6 percentage points at ages 20–23, and then reverses directions at older ages. For ages 28–34, smoking is higher by 7 percentage points among African Americans. However, the differences by race among males do not reach statistical significance. For females, the percentage of white smokers exceeds that for African Americans at all ages, but the gap declines from 22 percentage points at ages 20–23 to 12 percentage points at ages 28–34. The percentage of smoking among African American females differs statistically from that among white females for ages under 28 but not for ages 28–34.

TABLE 1.

Percent Current Smokers by Race, Sex, and Age Group for NYS and NHIS Data

NYS 12–15 16–19 20–23 24–27 28–34
White Males % 18 33 39 39 35
N (573) (1185) (837) (765) (728)
AA Males % 15 28 33 46 42
N (117) (275) (174) (162) (138)
White Females % 27* 40* 45* 41* 37
N (542) (1115) (814) (767) (723)
AA Females % 9* 21* 23* 20* 25
N (95) (177) (128) (111) (101)

NHIS 18–21 22–25 26–29 30–33 34–37

White Males % 30* 36* 35 34 33
N (12873) (15542) (17364) (18429) (18607)
AA Males % 21* 30* 32 34 36
N (2411) (2485) (2694) (2819) (2886)
White Females % 29* 33* 31* 30 29
N (15065) (18284) (20855) (22024) (22030)
AA Females % 15* 21* 26* 27 28
N (3732) (4859) (5277) (5586) (5205)
*

p < .01 for t-test of difference in percentages between white and African American (AA) males and between white and African American (AA) females.

Table 2 presents more formal models of age-based patterns of change. The multilevel models begin in simple form with age and age-squared determining smoking at level one and the race-sex variables determining the intercept or level of smoking at the mean age. This additive model assumes that race and sex affect the average level but not the age patterns of smoking. Note that the signs for the coefficients of age (positive) and age-squared (negative) are consistent with a rise, leveling off, and decline in smoking from the teen years to the thirties.3

TABLE 2.

Coefficients and Z-Ratios for Multilevel Logistic Regression Estimates of Current Smoking, NYS

Independent Age Additive + SES Controls Age Interactive + SES Controls
Variables b Z b Z b Z b Z
Fixed Effects
Intercept −.632* −10.92 −.625* −10.94 −.613* −10.37 −.603* −10.40
 ×AA Males −.345* −2.56 −.471* −3.32 −.302* −2.09 −.430* −2.93
 ×White Females .232* 2.78 .246* 2.99 .193* 2.25 .201* 2.40
 ×AA Females −.660* −4.40 −.813* −5.22 −.741* −4.92 −.891* −5.77
Age 3.390* 13.42 4.293* 12.04 3.726* 9.85 4.662* 10.20
 ×AA Males −.014 −.02 −.346 −.39
 ×White Females −.318 −.57 −.314 −.56
 ×AA Females −3.467* −3.86 −3.687* −4.12
Age 2 −.692* −13.93 −.862* −12.41 −.750* −9.51 −.919* −10.23
 ×AA Males .078 .42 .143 .77
 ×White Females .019 .16 .017 .14
 ×AA Females .741* 3.74 .781* 3.95
Parents’ Education −.015 −.34 −.016 −.34
Parents’ Occupation −.083* −2.89 −.083* −2.87
In School −.160* −3.94 −.162* −3.97
Years School −.077* −5.17 −.074* −4.70
Propensity Score −6.763* −9.09 −5.579* −6.89 −6.717* −9.00 −5.559* −6.83
Deviance 24374 24324 24368 24318

Random Effects Variance
Comp.
Chi-
Square
Variance
Comp.
Chi-
Square
Variance
Comp.
Chi-
Square
Variance
Comp.
Chi-
Square

Intercept 4.60 3740* 4.26 3563* 4.69 3704* 4.34 3526*
Age 28.01 648 22.26 641 28.78 648 22.64 641
Age 2 .97 603 .76 599 1.00 603 .78 598

p < .05

The coefficients for the race-sex dummy variables show that, on average, smoking prevalence is higher among white females and lower among African American males and females than among white males. Note that a dummy variable for sex and a dummy variable for race would not capture the high smoking of white females and the low smoking of African American females. The model deviance is significantly lower with the race-sex interaction than with additive race and sex terms.

Controlling for SES tends to widen race differences in smoking. The next column in Table 2 controls for education and occupation of parents, school attendance, and years of schooling.4 These SES-related variables reduce smoking and modestly change the age, race, and sex patterns of smoking. The controls increase race differences as the lower SES of African Americans otherwise tends to mask their lower propensity to smoke.

Graphs depict the meaning of these coefficients. Figure 1a plots the age-based probability of smoking predicted by model 1 for the four race-sex groups (without the SES controls). With the age trajectories constrained to be equal in the logits, the graph of probabilities shows a rise and decline in smoking for all four groups. White females have a higher curve than white males and African American males, and African American females stand out as having a particularly low trajectory. Figure 1b presents the curves implied by the model with controls for SES (which take their mean values in computing the predicted probabilities). The controls lead to greater separation among the groups but otherwise reveal the same ranking and similar age trajectories as the model without the controls.

FIGURE 1.

FIGURE 1

a. NYS Age Patterns of Smoking: Additive No Controls

b. NYS Age Patterns of Smoking: Additive SES Controls

c. NYS Age Patterns of Smoking: Interactive No Controls

d. NYS Age Patterns of Smoking: Interactive SES Controls

Allowing the age trajectories to interact with race and sex (i.e., including product terms of race and sex times age and age-squared) reveals mixed evidence of life-course differences. In columns 5–6 of Table 2, race and sex affect the age and age-squared coefficients. The interaction coefficients for African American men and white women do not reach statistical significance, and overall the deviance falls only slightly. For African American women, however, the significant negative coefficient for the age term and the positive coefficient for the age-squared term imply that their curve rises and falls less than for white men. Controlling for SES in the last columns of Table 2 does not greatly change the age patterns for race and sex.

The extent of the race differences in trajectories shows in Figures 1c (without SES controls) and 1d (with SES controls). The probabilities of smoking among white males and females in Figure 1c both rise quickly to a peak around age 25 and decline thereafter. In contrast, the trajectory for African American males peaks later and declines less. The trajectory for African American females, although much lower than for all other groups, continues to rise through the thirties. As a result, African American male smoking converges with white male smoking, and African American female smoking converges with white female smoking. The curves in Figure 1d that control for SES reveal much the same pattern. Overall, life-course patterns are most distinct for African American females.

A simple exercise apportions the change in race differences to the decline in white smokers and the rise in African American smokers. The exercise compares changes in the race gap for two periods: (1) a period of rising smoking from the late teens to the peak in the mid-twenties when a faster rate of initiation among African Americans can reduce the race gap, and (2) a period of declining smoking from the mid-twenties to the mid-thirties, when a faster rate of cessation among whites can reduce the gap. The former period reflects the influence of later initiation among African Americans, and the latter period the influence of earlier cessation among whites.

For males, the decomposition compares the predicted percent smokers at age 17 when the African American advantage peaks; at age 26 when smoking prevalence averaged across both white and African American males peaks; and at age 34 when the sample reaches its oldest age (a table of results is available on request). The change from ages 17 to 26, a period of rising smoking, involves a decline in the African American advantage by 4.9 percentage points, or 37.7 percent of the total decline in the race gap. This change reflects a higher rate of initiation among African Americans during young adulthood. The change in the gap from ages 26 to 34, a period of declining smoking, equals 8.1 percentage points and accounts for 62.3 percent of the total change in the gap. This change largely involves greater cessation among whites. Thus, greater white cessation at older ages accounts for more of the racial convergence in smoking prevalence than does the faster rate of increase among African American males at younger ages. With controls for SES, 26.4 percent of the change in the gap occurs during the early period and 73.6 percent in the later period, again showing the importance of greater white cessation at older ages.

The decomposition for females gives even more weight to cessation among whites. African American females move from an advantage of 24.3 percentage points at age 22 to a disadvantage of 1.7 percentage points at age 34–a change of 22.6 percentage points. About 98.2 percent of this change (96.2 percent with SES controls) comes from the greater decline among white females at older ages. The slow but steady increase in smoking through the twenties among African American females contrasts with the steep decline among white females.

Although the descriptive patterns and decomposition demonstrate racial convergence, the NYS results lack a strong statistical basis. In part because of the small number of African Americans in the sample, consistently significant effects do not emerge. The larger sample available from the NHIS data can extend these findings.

National Health Interview Surveys

Table 1 lists smoking prevalence by age group for the NHIS data. At ages 18–21, white male smoking exceeds African American male smoking by 9 percentage points (p < .01), but by ages 34–37 African American male smoking exceeds white smoking by 3 percentage points (not significant). Similarly, the 14 percentage-point advantage of African American females at ages 18–21 (p < .01) declines to an insignificant 2 percentage point difference by ages 34–37. These figures present averages across multiple cohorts and do not isolate changes within cohorts but, like the NYS data, demonstrate convergence across race in the age patterns of smoking.

Logistic regression models using the NHIS allow comparisons across cohorts in the race-specific age trajectories of smoking. Likelihood ratio tests show that models allowing age patterns of smoking to vary by race and sex significantly improve on models restricting age patterns to be the same across race and sex; further, likelihood ratio tests show that models allowing interactions between age, sex, and race to vary by cohort improve significantly on cohort-invariant models. However, the many interaction terms make interpretation of coefficients difficult, and the huge number of cases limits the value of significance tests. Graphing age patterns of smoking by race for multiple cohorts helps interpret the results.

To illustrate the strategy, Figure 2a presents the predicted age trajectories in smoking by race and sex when including additive but not interactive cohort controls (i.e., age trajectories are averaged across cohorts). Despite some differences, the NHIS results display a pattern of convergence similar to the NYS results. African American males at age 18 have a predicted proportion of smokers that is 10.1 percentage points lower than that of white males (t = 16.8, p < .001), but by age 37 the African American prevalence is only .6 percentage point lower (t = .94, p < .343). African American female smoking begins lower than that of all other groups, but the gap with smoking by white females declines from 15.5 percentage points at age 18 to 4.6 percentage points at age 37 (both differences are significant).

FIGURE 2.

FIGURE 2

a. NHIS Age Patterns of Smoking: All Cohort Average

b. NHIS Age Patterns of Smoking: 1951 Cohort

c. NHIS Age Patterns of Smoking: 1965 Cohort

d. NHIS Age Patterns of Smoking: 1979 Cohort

Perhaps more importantly, the race and sex differences in age trajectories change across cohorts. Figure 2b depicts the predicted age patterns for an older cohort born in 1951 (at the 10th percentile in the cohort distribution) from the full interactive model. Smoking of white and African American males moves in tandem, rising from age 18 to a peak at ages 30–31, declining afterward, and showing little convergence. In contrast, white and African American females reach parity by age 27, but then the gap increases again. For older cohorts, then, little evidence of convergence emerges.

Figure 2c shows the pattern for the 1965 cohort–at the 50th percentile and the last birth year of the NYS cohort. For the years in which this cohort reaches young adulthood, convergence has become apparent. The gap in male smoking by race falls from 7.0 percentage points at age 18 to 1.6 percentage points at age 37. For females, the gap falls from 12.5 percentage points to 7.5.

Figure 2d graphs the patterns for a newer cohort born in 1979 (at the 90th percentile of the cohort distribution). The graph reveals further evidence of a wide gap at younger ages that narrows at older ages. Among males, an African American advantage of 9.1 percentage points at age 18 reaches 12.7 percentage points at age 24 before falling to 2.8 percentage points at age 37. For females, a race gap persists over all ages but falls during young adulthood: The gap rises from 13.1 percentage points at age 18 to 18.6 percentage points at age 26 and then falls to 9.0 percentage points at age 37.

An exercise of decomposition similar to that used with the NYS leads to much the same conclusions for the NHIS. Focusing on the 1965 and 1979 cohorts, which reveals clear evidence of convergence, the decomposition shows that the primary source of change comes after smoking prevalence peaks, when race differences in cessation bring white smoking down to the lower levels for African American smoking. For males, 74.6 percent of the narrowing in the 1965 cohort and 93.9 percent of the narrowing in the 1979 cohort come at these older ages, showing that white cessation influences the gap more than late initiation among African Americans. For females, the percentages equal 51.0 for the 1965 cohort and 94.8 for the 1979 cohort. Both the NYS and the NHIS data demonstrate the importance for racial convergence of white cessation at older ages.

CONCLUSION

Use of the longitudinal NYS data over a 16-year period from ages 12 to 34 demonstrates varied patterns of cigarette use by whites and African Americans during adolescence and young adulthood. African Americans, particularly women, adopt cigarettes more slowly during the teen years but exhibit late initiation and low cessation during the twenties. As a result, the African American advantage in the teen years largely disappears by the early thirties. The convergence stems for the most part from greater rates of white cessation during young adulthood. Replication with consecutive cross-sectional data from the NHIS shows some differences from the NYS in levels of smoking (particularly for African American and white women) but roughly similar patterns of change. African Americans begin at age 18 with lower smoking but by the mid-thirties reach levels close to those of whites, a pattern that strengthens for newer cohorts.

Although Ellickson et al. (2004) find little evidence of convergence from ages 13 to 23 because of lower rates of transition from experimental to regular smoking among African American youth, the findings here, which extend the ages studied up to 34 or 37, show evidence of convergence. This result matches that of Geronimus, Neider, and Bound (1993), who reach the following conclusion about smoking among women ages 15–45 using retrospective data on smoking initiation and cessation from a 1987 survey:

White women initiate smoking at younger ages than Black women, but among smokers, White women are more likely than Blacks to quit. The result is that adolescent and young adult Black women are less likely to smoke than their White counterparts, but this health advantage disappears rapidly. (p. 1261)

The results demonstrate the value of a life-course approach to smoking. Although whites and African Americans end up with similar smoking prevalence during the late twenties or early thirties, they get to that point in different ways. African American and white males exhibit similar increases in smoking through the mid-twenties, but African American males show a smaller and later decline in the late twenties. African American females show little decline through their thirties and exhibit eventual convergence with the smoking prevalence of white females. Controls for SES characteristics of the parents and educational attainment of the youth tend to expand rather than eliminate the racial gap over the life course.

Going beyond within-cohort results, the NHIS results for recent cohorts more clearly illustrate the pattern of racial convergence in smoking. New cohorts show a large gap between whites and African Americans at younger ages but also show that the gap declines with age. These findings reflect a major change across cohorts. Where whites and African Americans once exhibited similar age trajectories of smoking, a steep decline in smoking has occurred among young African Americans. In more recent cohorts, however, this African American advantage narrows considerably through adulthood.

These findings have implications for theories of health behavior and inequality. Among disadvantaged groups, inequality in health generally begins with strain and lack of resources during childhood and adolescence (Alwin and Wray 2005). The continued accumulation of problems over the life course then creates more serious health problems in late adulthood and old age. For smoking, however, the African American disadvantage emerges more slowly, at least among more recent cohorts. An advantage exists during the teen years–one that has been growing for several decades–but to a large extent it is lost in the next decade of the life course. The loss comes from the greater cessation of whites and the fewer resources for quitting among African Americans. Although to a lesser extent, it also comes from the continued initiation among African Americans and the greater stress they experience relative to whites during their twenties. The findings overall most favor theories of smoking based on lack of resources.

However, resources and stress involve more than measured SES characteristics. The persistence of race differences in smoking with controls, particularly for younger ages and more recent cohorts, suggests that something more is involved in health disparities than SES differences. Health lifestyles (Cockerham 2000) may develop within race as well as SES groups and reflect cultural as well as socioeconomic sources of divergence. Seen as a key component of a health lifestyle, tobacco use may vary by race, as young whites and blacks develop different norms for use of tobacco and other substances. These race-specific norms appear to change across cohorts as well as converge within cohorts. Such patterns of change, particularly the development of anti-smoking norms among African American youth, have received little attention and deserve more study. Cultural influences on smoking appear important in understanding puzzling changes across race groups.

In terms of policy, the results are consistent with conclusions that general strategies to prevent initiation and promote cessation apply to both whites and African Americans (Kandel et al. 2004). However, they suggest a difference in emphasis. Prevention of teen initiation is relatively more important to whites, while prevention of initiation at older ages is relatively more important to African Americans (Moon-Howard 2003). In addition, the lower rate of cessation at older ages among African Americans should be a key concern for developing policies to reduce disparities. Still further, the greater cohort-based decline of teen smoking among African American youth compared to white youth has policy implications. Recent legislative changes have made the purchase of cigarettes more costly for lower-income groups, including African Americans. However, DeCicca et al. (2008) find that prices have little influence on youth initiation, while anti-tobacco sentiment has an important influence. Such sentiment appears to have grown particularly among African American youth and may be key in promoting declines among other race groups.

Despite some advantages of the NYS for studying long-term patterns of change in smoking, the data set has limitations. The simple measure of current smoking based on cigarettes smoked per day lacks the detail available from other surveys that are designed specifically to study cigarette use and include items on ages of adoption and cessation. A more complete history of starting and stopping for each respondent, and perhaps even measures of biological nicotine markers (cotinine), would allow for additional analyses and insights. In addition, the availability of only about 250 African Americans for the NYS analysis and the need to separate males and females weaken the statistical power of the tests for race differences. The NHIS contains a larger number of cases and more statistical power, but the composition of the cohorts changes over time with population change and new random samples for each survey. Neither data set is ideal, but together they provide consistent evidence of convergence.

Acknowledgments

This research was funded by grant R21 HD51146 from the NICHD-funded University of Colorado Population Center.

Footnotes

1

Studies have largely discounted claims that error in self-reported cigarette use accounts for race differences (Wallace et al. 1995; DHHS 1998).

2

The propensity score comes from eight separate binomial logistic regressions, one for each wave, that treats inclusion in the wave as the dependent variable, and variables gathered at wave one for the full sample as the independent variables. All 1,621 white and black respondents are used in each logistic regression. The wave one variables include measures obtained from parents (a scale of child trouble in dealing with others, a scale of neighborhood problems, parents’ education, and parents’ occupation) and measures obtained from the child (high school GPA, involvement in athletics, involvement in community activities, working at a job, and parents’ involvement at school). Race has little influence on attrition, but school GPA, holding a job, and participating in athletics or community activities generally reduce attrition, and each model has a chi-square that significantly improves on the intercept model. A single propensity score for each respondent then comes from the average of the wave-specific predicted probabilities.

3

The variance component for the intercept is significant, but those for age and age-squared are not. At the same time, the deviance declines significantly when random effects for age and age-squared are included, suggesting the possible value of modeling the race and sex determinants of variation in age trajectories.

4

Computing the pseudo-variance explained is made difficult by use of both the logit link function and multilevel models, and such measures can give inconsistent results (Singer and Willett 2003). For example, a decrease in the pseudo-variance explained can occur with the addition of more variables to a model. In rough terms, the simple model explains low amounts of the pseudo-variance, about 3 percent at level one and 4 percent at level two. Rather than focusing on variance explained, however, the models aim to describe the modest but substantively meaningful influence of race, sex, and age on patterns of smoking.

REFERENCES

  1. Alwin Duane F., Wray Linda A. A Life-Span Developmental Perspective on Social Status and Health. Journals of Gerontology: Series B. 2005;60B(special issue 11):7–14. doi: 10.1093/geronb/60.special_issue_2.s7. [DOI] [PubMed] [Google Scholar]
  2. Bachman Jerald G., O’Malley Patrick M., Schulenberg John E., Johnston Lloyd D., Bryant Alison L., Merline Alicia C. The Decline of Substance Use in Young Adulthood: Changes in Social Activities, Roles, and Beliefs. Lawrence Erlbaum; Mahwah, NJ: 2002. [Google Scholar]
  3. Bachman Jerald G., Wadsworth Katherine N., O’Malley Patrick M., Johnston Lloyd D., Schulenberg John E. Smoking, Drinking, and Drug Use in Young Adulthood. Lawrence Erlbaum; Mahwah, NJ: 1997. [Google Scholar]
  4. Barbeau Elizabeth M., Krieger Nancy, Soobader Mah-Jabeen. Working Class Matters: Socioeconomic Disadvantage, Race/Ethnicity, Gender, and Smoking in NHIS 2000. American Journal of Public Health. 2004;94:269–78. doi: 10.2105/ajph.94.2.269. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bennett Gary G., Wolin Kathleen Yaus, Robinson Elwood L., Fowler Sherrye, Edwards Christopher L. Perceived Racial/Ethnic Harassment and Tobacco Use among African American Young Adults. American Journal of Public Health. 2005;95:238–40. doi: 10.2105/AJPH.2004.037812. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Carmona R, Gfroerer J, Caraballo R, Lee SL, Husten C, Pechacek T, Robinson RG, Lee C. Prevalence of Cigarette Use among 14 Racial/Ethnic Populations–United States, 1999–2001. Morbidity and Mortality Weekly Report. 2004;53:49–52. [PubMed] [Google Scholar]
  7. Chen Kevin, Kandel Denise B. The Natural History of Drug Use from Adolescence to the Mid-Thirties in a General Population Sample. American Journal of Public Health. 1995;85:41–7. doi: 10.2105/ajph.85.1.41. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Chen Ping-Hsin, White Helene Raskin, Pandina Robert J. Predictors of Smoking Cessation from Adolescence into Young Adulthood. Addictive Behaviors. 2001;26:517–29. doi: 10.1016/s0306-4603(00)00142-8. [DOI] [PubMed] [Google Scholar]
  9. Cockerham William C., Bird Chloe E., Conrad Peter, Fremont Allen M. Handbook of Medical Sociology. 5th ed. Prentice Hall; Upper Saddle River, NJ: 2000. The Sociology of Health Behavior and Health Lifestyles; pp. 159–172. [Google Scholar]
  10. DeCicca Philip, Kenkel Donald, Mathios Alan, Shin Yoong-Jong, Lim Jae-Young. Youth Smoking, Cigarette Prices, and Anti-Smoking Sentiment. Health Economics. 2008;17:737–49. doi: 10.1002/hec.1293. [DOI] [PubMed] [Google Scholar]
  11. Department of Health and Human Services (DHHS) Preventing Tobacco Use among Young People: A Report of the Surgeon General. Department of Health and Human Services, U.S. Public Health Service, Centers for Disease Control and Prevention; Atlanta, GA: 1994. [Google Scholar]
  12. Department of Health and Human Services (DHHS) Tobacco Use among U.S. Racial/Ethnic Minority Groups: A Report of the Surgeon General. Department of Health and Human Services, U.S. Public Health Service, Centers for Disease Control and Prevention; Atlanta, GA: 1998. [Google Scholar]
  13. Department of Health and Human Services (DHHS) Reducing Tobacco Use: A Report of the Surgeon General. Department of Health and Human Services, U.S. Public Health Service, Centers for Disease Control and Prevention; Atlanta, GA: 2000. [Google Scholar]
  14. Department of Health and Human Services (DHHS) The Health Consequences of Smoking: A Report of the Surgeon General. Department of Health and Human Services, U.S. Public Health Service, Centers for Disease Control and Prevention; Atlanta, GA: 2004. [Google Scholar]
  15. Dugan Shaun, Lloyd Barbara, Lucas Kevin. Stress and Coping as Determinants of Adolescent Smoking Behavior. Journal of Applied Social Psychology. 1999;29:870–88. [Google Scholar]
  16. Edelen Maria Orlando, Tucker Joan S., Ellickson Phyllis L. A Discrete Time Hazards Model of Smoking Initiation among West Coast Youth from Age 5 to 23. Preventive Medicine. 2007;44:52–54. doi: 10.1016/j.ypmed.2006.09.004. [DOI] [PubMed] [Google Scholar]
  17. Ellickson Phyllis L., Orlando Maria, Tucker Joan S., Klein David J. From Adolescence to Young Adulthood: Racial/Ethnic Disparities in Smoking. American Journal of Public Health. 2004;94:293–99. doi: 10.2105/ajph.94.2.293. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Ellickson Phyllis L., Perlman Michal, Klein David J. Explaining Racial/Ethnic Differences in Smoking during the Transition to Adulthood. Addictive Behaviors. 2003;28:915–31. doi: 10.1016/s0306-4603(01)00285-4. [DOI] [PubMed] [Google Scholar]
  19. Elliott Delbert S., Huizinga David, Ageton Suzanne S. Explaining Delinquency and Drug Use. Sage; Beverly Hills, CA: 1985. [Google Scholar]
  20. Fagan Pebbles, King Gary, Lawrence Deirdre, Petrucci Sallie Anne, Robinson Robert G., Banks David, Marable Sharon, Grana Rachel. Eliminating Tobacco-Related Health Disparities: Directions for Future Research. American Journal of Public Health. 2004;94:211–17. doi: 10.2105/ajph.94.2.211. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Gardiner Phillip S. African American Teen Cigarette Smoking: A Review. In: National Cancer Institute, editor. Changing Adolescent Smoking Prevalence: Where It Is and Why. National Cancer Institute; Bethesda, MD: 2001. pp. 213–25. [Google Scholar]
  22. Geronimus Arline T., Neidert Lisa J., Bound John. Age Patterns of Smoking in U.S. Black and White Women of Childbearing Age. American Journal of Public Health. 1993;83:1258–64. doi: 10.2105/ajph.83.9.1258. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Graham Hilary. Cigarette Smoking: A Light on Gender and Class Inequality in Britain? Journal of Social Policy. 1995;24:509–27. [Google Scholar]
  24. Graham Hilary, Francis Brian, Inskip Hazel M., Harman Juliet. Socioeconomic Life-course Influences on Women’s Smoking Status in Early Adulthood. Journal of Epidemiology and Community Health. 2006;60:228–33. doi: 10.1136/jech.2005.039784. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Honjo Kaori, Tsutsumi Akizumi, Kawachi Ichiro, Kawakami Norito. What Accounts for the Relationship between Social Class and Smoking Cessation? Results of a Path Analysis. Social Science and Medicine. 2006;62:317–28. doi: 10.1016/j.socscimed.2005.06.011. [DOI] [PubMed] [Google Scholar]
  26. Hu Mei-Chen, Davies Mark, Kandel Denise B. Epidemiology and Correlates of Daily Smoking and Nicotine Dependence among Young Adults in the United States. American Journal of Public Health. 2006;96:299–308. doi: 10.2105/AJPH.2004.057232. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Jacobson Peter D., Lantz Paula M., Warner Kenneth E., Wasserman Jeffrey, Pollack Harold A., Ahlstrom Alexis K. Combating Teen Smoking: Research and Policy Strategies. University of Michigan Press; Ann Arbor: 2001. [Google Scholar]
  28. Johnston Lloyd D., O’Malley Patrick M., Bachman Jerald G., Schulenberg John E. Monitoring the Future National Survey Results on Drug Use 1975–2004. Vol. 1, Secondary School Students. National Institute on Drug Abuse; Bethesda, MD: 2005. NIH Publication No. 05-5727. Retrieved September 15, 2006 ( http://monitoringthefuture.org/pubs/monographs/voll_2004.pdf) [Google Scholar]
  29. Kandel Denise B., Kiros Gebre-Egziabher, Schaffran Christine, Hu Mei-Chen. Racial/Ethnic Differences in Cigarette Smoking Initiation and Progression to Daily Smoking: A Multilevel Analysis. American Journal of Public Health. 2004;94:128–35. doi: 10.2105/ajph.94.1.128. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Kiefe Catarina I., Williams O. Dale, Lewis Cora E., Allison Jeroan J., Sekar Padmini, Wagenknecht Lynne E. Ten-Year Changes in Smoking among Young Adults: Are Racial Differences Explained by Socioeconomic Factors in the Cardia Study? American Journal of Public Health. 2001;91:213–18. doi: 10.2105/ajph.91.2.213. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. King Gary, Polednak Anthony, Bendel Robert B., Vilsaint My C., Nahata Sunny B. Disparities in Smoking Cessation between African Americans and Whites: 1990–2000. American Journal of Public Health. 2004;94:1965–71. doi: 10.2105/ajph.94.11.1965. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Lauderdale Diane S. Education and Survival: Birth Cohort, Period, and Age Effects. Demography. 2001;38:551–61. doi: 10.1353/dem.2001.0035. [DOI] [PubMed] [Google Scholar]
  33. Lawlor Debbie A., Frankel Stephen, Shaw Mary, Ebrahim Shah, Smith George Davey. Smoking and Ill Health: Does Lay Epidemiology Explain the Failure of Smoking Cessation Programs among Deprived Populations? American Journal of Public Health. 2003;93:266–70. doi: 10.2105/ajph.93.2.266. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Lynch JW, Kaplan GA, Salonen JT. Why Do Poor People Behave Poorly? Variation in Adult Health Behaviours and Psychological Characteristics by Stages of the Socioeconomic Lifecourse. Social Science and Medicine. 1997;44:809–19. doi: 10.1016/s0277-9536(96)00191-8. [DOI] [PubMed] [Google Scholar]
  35. Menard Scott, Mihalic Sharon W., Morse Barbara J., Burton B, Huizinga David, Elliott Delbert S. Crime, Drug Use, Mental Health and Victimization: Adult Epidemiology and Life Course Developmental Patterns. Institute of Behavioral Science; Boulder, CO: 1994. [Google Scholar]
  36. Mirowsky John, Ross Catherine E. Education, Social Status, and Health. Aldine de Gruyter; New York: 2003. [Google Scholar]
  37. Moon-Howard Joyce. African-American Women and Smoking: Starting Later. American Journal of Public Health. 2003;93:418–20. doi: 10.2105/ajph.93.3.418. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. National Center for Health Statistics . National Health Interview Survey (NHIS) 2007. Retrieved October 10, 2007 ( http://www.cdc.gov/nchs/about/major/nhis/quest:data_related_1997_forward.htm) [Google Scholar]
  39. National Institutes of Health . NIH Strategic Research Plan to Reduce and Ultimately Eliminate Health Disparities. 2000. Retrieved September 30, 2006 ( http://www.nih.gov/about/hd/strategicplan.pdf) [Google Scholar]
  40. Preston Samuel H., Wang Haidong. Sex Mortality Differences in the United States: The Role of Cohort Smoking Patterns. Demography. 2006;43:631–46. doi: 10.1353/dem.2006.0037. [DOI] [PubMed] [Google Scholar]
  41. Raudenbush Stephen W., Bryk Anthony S. Hierarchical Linear Models: Applications and Data Analysis Methods. 2nd ed. Sage; Thousand Oaks, CA: 2002. [Google Scholar]
  42. Raudenbush Stephen W., Bryk Anthony, Cheong Yuk Fai, Congdon Richard. HLM 6: Hierarchical Linear and Nonlinear Modeling. Scientific Software International; Lincolnwood, IL: 2004. [Google Scholar]
  43. Singer Judith D., Willett John B. Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. Oxford University Press; New York: 2003. [Google Scholar]
  44. Sorensen Glorian, Barbeau Elizabeth, Hunt Mary Kay, Emmons Karen. Reducing Social Disparities in Tobacco Use: A Social-Contextual Model for Reducing Tobacco Use among Blue-Collar Workers. American Journal of Public Health. 2004;94:230–39. doi: 10.2105/ajph.94.2.230. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Trinidad Dennis R., Gilpin Elizabeth A., Lee Lora, Pierce John P. Has There Been a Delay in the Age of Regular Smoking Onset among African Americans? Annals of Behavioral Medicine. 2004;28:152–57. doi: 10.1207/s15324796abm2803_2. [DOI] [PubMed] [Google Scholar]
  46. Turbin Mark S., Jessor Richard, Costa Frances M. Adolescent Cigarette Smoking: Health-Related Behavior or Normative Transgression? Prevention Science. 2000;1:115–24. doi: 10.1023/a:1010094221568. [DOI] [PubMed] [Google Scholar]
  47. Udry J. Richard, Li Rose Maria, Hendrickson-Smith Janet. Health and Behavior Risks of Adolescents with Mixed-Race Identity. American Journal of Public Health. 2003;93:1865–70. doi: 10.2105/ajph.93.11.1865. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Wallace John M., Jr., Bachman Jerald G., O’Malley Patrick M., Johnston Lloyd D., Botvin Gilbert J., Schinke Steven, Orlandi Mario A. Drug Abuse Prevention with Multiethnic Youth. Sage; Thousand Oaks, CA: 1995. Racial/Ethnic Differences in Adolescent Drug Use: Exploring Possible Explanations; pp. 55–80. [Google Scholar]
  49. Weden Margaret, Kimbro Rachel Tolbert. Racial and Ethnic Differences in the Timing of First Marriage and Smoking Cessation. Journal of Marriage and Family. 2007;69:878–87. [Google Scholar]
  50. Wilkinson Richard. Unhealthy Societies: The Affliiations of Inequality. Routledge; London: 1996. [Google Scholar]
  51. Williams David R. Journals of Gerontology: Series B. special issue II. 6O. 2005. The Health of U.S. Racial and Ethnic Populations; pp. 53–62. [DOI] [PubMed] [Google Scholar]
  52. Willson Andrea E., Shuey Kim M., Elder Glen H., Jr. Cumulative Advantage Processes as Mechanisms of Inequality in Life Course Health. American Journal of Sociology. 2007;112:1886–1924. [Google Scholar]
  53. Yu Stella. The Life-Course Approach to Health. American Journal of Public Health. 2006;96:768. [Google Scholar]

RESOURCES