Education and the Hispanic Mortality Advantage: the Role of Smoking

Andrew Fenelon ***WORKING PAPER. PLEASE DO NOT CITE WITHOUT AUTHOR’S PERMISSION***

Population Studies Center, University of Pennsylvania 239 McNeil Building, 3718 Locust Walk, Philadelphia, PA 19104, [email protected]

Introduction Nearly three decades of research into the so-called “Hispanic/Latino Paradox” has failed to generate a convincing explanation for this pervasive epidemiological phenomenon (Markides and Eschbach 2005). Despite having lower levels of material wealth and fewer years of education than non-Hispanic whites in the United States, most Hispanic groups exhibit lower death rates at most adult ages (Markides and Coreil 1986; Sorlie et al. 1993; Hummer et al. 2000; Abraido-Lanza et al. 1999; Elo, Turra, et al. 2004). The National Center for Health Statistics released the first ever Hispanic life tables in 2010, showing a substantial survival advantage relative to non-Hispanic whites (Arias 2010). While a mortality advantage has been observed among many different Hispanic subgroups (Hummer et al. 1999, 2000), the specific reasons for the advantage remain unclear in the literature. Some recent research suggests that health behaviors are essential to understanding the overperformance of Hispanics with respect to health and longevity. Singh and Siahpush (2002) show that Hispanics have substantially lower mortality than non-Hispanic whites from cancers and respiratory diseases, which are commonly associated with smoking. They also find lower reported smoking prevalence among Hispanics. Abraido-Lanza et al. (2005) likewise find that Hispanics have a generally better behavioral profile than non-Hispanic whites. Blue and Fenelon (2011) directly calculate the burden of cigarette smoking among Hispanics and non-Hispanics whites using three indirect estimation techniques, they find that a majority of the longevity advantage of Hispanics compared with non-Hispanic whites can be attributed to lower mortality from smoking among Hispanics. However, the Hispanic mortality advantage is not a characteristic of all Hispanics. First, foreign-born Hispanics exhibit a large advantage, while US-born Hispanics show a health and mortality profile more similar to non-Hispanic whites (Palloni and Arias 2004). Second, Hispanics have the largest advantage at the lowest levels of education (Turra and

Goldman 2007). College educated Hispanics and non-Hispanic whites have very similar health performance. Ostensibly this is because whites have a rather steep educational gradient in mortality while Hispanics, especially immigrants have a relatively flat gradient (Goldman et al. 2006). The proximate reasons for this ethnic difference in gradient steepness are not yet identified. Since the size of the Hispanic mortality advantage varies widely by education, any explanation should be able to account for this educational pattern. In this paper, I examine the contribution of smoking to the educational pattern in Hispanic and non-Hispanic white mortality. I use a large, nationally-representative survey to calculate smoking attributable-risk for three population subgroups in the United States: non-Hispanic whites, MexicanAmericans1, and all other Hispanics2. Evidence for the Hispanic Mortality Advantage Early findings regarding the health and mortality experience of Hispanics demonstrated that Mexican-Americans in the Southwest exhibited adult death rates similar to non-Hispanic whites, despite substantial socioeconomic disadvantage (Becker et al. 1988; Markides and Coreil 1986). These studies evaluated the Hispanic population using Spanish surname or other indirect methods of classification, and suffered from a lack of generalizability (Sullivan et al. 1984). The first analysis to use a large, population-based sample was by Sorlie et al. (1993) who investigated the survivorship of Hispanic subgroups using the National Longitudinal Mortality Study (NLMS) 1979-1987. They found significantly lower mortality among Mexican-Americans, Puerto Ricans, Cubans, and other Hispanics compared with non-Hispanics, especially at older ages, with the largest advantage for cardiovascular disease and cancers. Since then, evidence of the Hispanic mortality and

1

The terms Mexican, Mexican-American, or Chicano will hereafter be used interchangeably for those of Mexican origin residing in the United States. 2 Other Hispanics include all those professing Hispanic origin but not identifying as being of “Mexican” origin. Other Hispanics include individuals of Cuban, Puerto Rican, Caribbean, Central American, South American, and Spanish origin.

health advantage has come from a variety of data sources, each valuable for examining different aspects of the phenomenon. The major sources of data on Hispanic mortality are U.S. vital statistics and nationally-representative surveys. Vital statistics use Hispanic ethnicity coded on death certificates and census estimates of the Hispanic population. Differences in the identification of Hispanic ethnicity on death certificates (which is reported by a physician) and the census (which is self-reported) have the potential to underestimate Hispanic mortality; Elo et al. (2004) find that data from the Social Security Administration (NUMIDENT), which does not suffer from ethnicity mismatch, imply a smaller advantage than that given by vital statistics. Nationally-representative surveys with prospective mortality follow-up partially solve this issue, since ethnicity is self-reported and respondents are matched to records in the National Death Index. The use of large-sample surveys also allows the researcher to examine the Hispanic advantage across a variety of other covariates, including socioeconomic variables and health behaviors (e.g. Rogers et al. 2005). Although the size of the observed Hispanic advantage varies somewhat depending on the data source used, it has been a remarkably consistent finding for the past two decades. Hypotheses explaining the Hispanic paradox typically fall into two broad categories: selective migration and cultural effects. Since the majority of U.S. Hispanics are foreign born, any examination of the Hispanic mortality experience must consider who migrates and who does not. Selective migration can refer to both in-migration of the healthy (healthy migrant effect) and out-migration of the unhealthy (“salmon bias”). The former concerns the greater human capital and health resources that may be necessary to undertake an international move, such that we observe a highly select group of individuals from sending countries, which may offset the negative effects of the poor socioeconomic profile (AbraidoLanza et al. 1999). The latter suggests that foreign-born individuals in the United States may

return to their countries of origin when they become ill, both leaving behind a relatively healthy stock of foreign-born individuals in the United States and causing many deaths to be unobserved in American vital registration (Palloni and Arias 2004; Palloni and Ewbank 2004). Aspects of Hispanic culture in the U.S. may also be protective for the health of Hispanics relative to non-Hispanic whites (Abraido-Lanza et al. 2005). Hispanics may benefit from stronger social support and community networks. There is evidence that Mexican immigrants have better health when they live in neighborhoods with greater concentrations of Mexicans (Eschbach et al. 2004; LeClere, Rogers, and K. D. Peters 1997). Hispanic culture may also promote attitudes about health that promote well-being or may reduce unhealthy behaviors such as cigarette smoking or drug use (Markides and Eschbach 2005; AbraidoLanza et al. 2005). A recent study indicates that first generation Hispanic immigrants in California eat more fruits and vegetables, and drink less soda than native-born whites (Allen et al., 2007). Recent immigrants are also less likely than whites to be overweight and obese, though this effect deteriorates with greater acculturation (Goel et al. 2004; Kaplan et al. 2004). Cigarette smoking is a particularly promising behavior to explain the health advantage of Hispanics. This is because it is (a) responsible for a large number of deaths in the United States (Peto et al. 1994; Preston, Glei, and Wilmoth 2010b) and (b) much more prevalent among whites than among Hispanics (CDC 2009). The mortality burden of smoking among whites is very large, and is likely to be much smaller in the Hispanic population. Cigarette Smoking Among Hispanics and non-Hispanics Current survey data show that Hispanics are less likely to be current smokers and smoke fewer cigarettes on average than whites (Barbeau, Krieger, and Soobader 2004; Caraballo and Lee 2004; Sudano and Baker 2006; Bethel and Schenker 2005). According to the 2008 National Health Interview Survey, 20.7 percent of Hispanic men were current

smokers compared with 23.5 percent of white men, while 10.7 of Hispanic women and 20.6 percent of white women were smokers (CDC 2009). Among Mexican-Americans, smoking appears to be extremely light, with a substantial fraction of Mexican smokers smoking fewer than five cigarettes per day (Caraballo et al. 2001, 1998; Trinidad et al. 2009). One study found that among Mexican-Americans who identify as current smokers, more than onequarter had serum cotinine concentrations lower than the amount necessary to be considered smokers (Caraballo et al. 2001). Smoking prevalence alone does not capture the total burden of smoking in a population. Smoking begins in adolescence or early adulthood, but the negative health effects are observed only after decades of the behavior. Since detailed cohort smoking histories are unavailable for most populations, researchers often must turn to indirect measures of the impact of smoking. The lung cancer death rate has been shown to be an accurate indicator of the damage from smoking in a population, since it simultaneously reflects smoking duration and intensity (Peto et al. 1992; Preston, Glei, and Wilmoth 2010a). Smoking is also the primary cause of lung cancer in developed countries, responsible for more than 90% of male lung cancer deaths and more than 70% of female deaths (Ezzati and Lopez 2003). Along with lower reported smoking prevalence, Hispanics exhibit lower prevalence of chronic diseases associated with smoking, including lung cancer (O’Brien et al. 2003). The Role of Education The relationship between SES and health in the U.S. is well documented and persists regardless of the health or SES measures used (Case, Lubotsky, and Paxson 2002; Elo and Preston 1996; J. P. Smith 2004). This relationship often mediates race/ethnic differentials in health and mortality. For Hispanics, however, the relationship between SES and mortality is different. Unlike whites, Hispanics have a relatively weak socioeconomic gradient in health and mortality, and this pattern is especially pronounced among Mexican immigrants

(Goldman et al. 2006; Turra and Goldman 2007). Among Hispanics, women tend to exhibit the weakest gradient. At age 70, Mexican women with less than 12 years of education have mortality rates only 5% higher than those with 13 or more. The corresponding excess is 60% among white women (Goldman et al. 2006). As a result, the Hispanic advantage is much larger at lower levels of education and relatively small or non-existent among collegeeducated individuals. For example, at ages 50 and above, Mexican-American women experience mortality rates 15-25% lower than whites at low levels of education (<12 years), and either similar to or higher than whites at the highest educational levels (Turra and Goldman 2007). Hispanics also exhibit relatively weak gradients in chronic disease incidence and a number of related risk factors including smoking, physical activity, and heavy drinking (Crimmins, Hayward, and Seeman 2004; Winkleby and Cubbin 2004; A. Buttenheim et al. 2010). These distinct patterns may reflect the specific characteristics of migrant origin locations. For example, rural areas in Mexico (presumably the predominant origin of MexicoU.S. migrants) exhibit relatively little variation in health and health behaviors by level of education or income (A. Buttenheim et al. 2010; K. V. Smith and Goldman 2007). It is also possible that the proximate factors responsible for the SES-health relationship – such as resource disparities, stress, health behaviors and health-related knowledge, and access to care – are less closely tied to status among Hispanics (Kimbro et al. 2008; Markides and Eschbach 2005). It may also reflect the effects of acculturation among Hispanic immigrants. As foreign-born individuals assimilate into U.S. society, some negative effects of the immigration process may weaken the relationship between high SES and good health for those with more education. Each measure of SES reflects a distinct aspect of individual status and has independent effects on health. The use of income, wealth, and/or occupation presents

challenges in exploring health dynamics given the possibility of selection and reverse causation (Liberatos, Link, and Kelsey 1988; J. P. Smith 2004). These measures also may change over an individual’s life course, often in response to health events. As a single measure of SES, education has the advantage of being determined (usually) in early adulthood and remaining fixed throughout an individual’s life (Preston and Taubman 1994). Education is also relatively easy to measure and is readily available in most surveys. This paper examines the impact of smoking on education gradients in mortality among Hispanics and whites. I compare the mortality experience of Hispanics to that of whites in three education categories: less than high school, high school graduate, and college or more. Methods Data Data come from the National Health Interview Survey (NHIS) smoking supplements between 1987 and 2004.3 Respondents in the NHIS are linked to the National Death Index (NDI) to ascertain mortality status through the multiple cause-of-death (MCD) up to the end of 2006. NHIS samples are obtained through the Integrated Health Interview Series (IHIS 2010) at the University of Minnesota and mortality follow-up linkage is provided by the National Center for Health Statistics. The relevant sample contains smoking status and mortality follow-up information on 281,567 individuals aged 35 or more at the interview over 16 survey years. 57,467 deaths, including 4,914 from lung cancer, are observed by the end of 2006. Individuals are weighted using supplement-specific annual person weights for survey years 1987 – 1995 and using eligibility-adjusted mortality sample adult weights for 1997 – 2004.4

3

Information on individual smoking supplements is available from the National Center for Health Statistics http://www.cdc.gov/nchs/nhis/tobacco/tobacco_questions.htm 4 No supplement providing information on tobacco use is available for 1989 or 1996.

Method The impact of smoking on mortality at the individual level depends on the intensity of current and past cigarette use and the lifetime duration of smoking. Smoking attributable risk, the number of deaths that would not occur in the absence of smoking, can be calculated based on the relative mortality risk of smokers relative to non-smokers. To estimate these values, I use loglinear hazard regression models predicting all-cause mortality from smoking behavior ln !! !

= !! + !! !! + !! !! + !! !! + !

(1)

!! are 5-year age groups – 35-39, 40-44,…,80-84, 85+. Individuals who enter the NHIS below age 35 are not considered in the current sample until they reach age 35. !! refers to the smoking status of individual i (see below), !! is an indicator of whether the individual is obese (body mass index 40+). Models are run separately by education and sex for three groups: non-Hispanic whites, Mexican-Americans, and other Hispanics. Variables Smoking status is ascertained in the NHIS through a series of questions. Respondents are asked if they have smoked more than 100 cigarettes in their entire lives. If they have not, they are classified as “never smokers”. All others are considered ever smokers. Respondents are then asked if they smoke cigarettes currently, every day or most days. Ever smokers who no longer smoke currently are classified as “former smokers”. Current smokers also report the number of cigarettes they usually smoke per day. Smoking status is measured using a sixcategory variable accounting for current smoking behavior and daily cigarette consumption. The categories are: lifelong never smoker, former smoker, current very light smoker (fewer than 5 cigarettes per day), current light smoker (5-9 per day), current medium smoker (10-19 per day), and current heavy smoker (20+ per day). The very light category is included partly

to capture the extremely low cigarette consumption of Mexican-Americans (Caraballo et al. 2001). Education is measured using a categorical variable for the number of years completed. The categories are: 11 or fewer years of education (“low education”), 12 years (“medium education”, and 13 years or more (“high education”). Models also control for obesity, defined as having a body mass index (BMI) of 40 or greater. This is intended to capture confounding in the relationship between smoking and mortality (Rogers et al. 2005). The amount of confounding is expected to be relatively small (M. J. Thun, Apicella, and Henley 2000). Estimating Smoking Attributable-Risk The number of deaths attributable to cigarette smoking in each ethnic group is calculated using a conventional attributable-risk approach (Rockhill, Newman, and Weinberg 1998). This method predicts the number of deaths that would not have occurred if smokers experienced no excess mortality relative to non-smokers. Adjusted for confounding factors, the mortality experience of the lifelong never smokers represents how the mortality of the population would look in the absence of smoking. The excess is the fraction that is attributable to cigarette smoking (A) ! !!

!!!

=

− !!!∗ !!!

!!!

is the predicted death rate for a specific sex-ethnicity subgroup between ages x and

x+n. !!!∗ is the estimated age-specific death rate among never-smokers in the same subgroup. Contribution of Smoking to the Hispanic Advantage by Education The contribution of smoking to the Hispanic advantage depends on the size of the disparity in both the observed data as well as in the hypothetical scenario in which smokingattributable mortality is absent. The age-specific death rates in each scenario imply sex- and

ethnicity-specific adult life expectancies. The relative change in the Hispanic-non-Hispanicwhite life expectancy gap after the removal of smoking represents the proportion of the advantage that is attributable to smoking. Formally, the contribution is calculated as ∆! = !!

!!

− !! !!

−

!!

∗

− !! ∗

− !!

and !! are life expectancies at age 35 for the Hispanic subgroup and whites,

respectively. The * superscript refers to the equivalent measures calculated using death rates estimated in the absence of smoking ( !!!∗ ). ∆! can be interpreted as proportional decrease in the Hispanic life expectancy advantage with the removal of the effect of cigarette smoking. The value is computed separately for each educational category and for all categories combined. Results Table 1 presents life expectancy at age 35 by Hispanic status, sex, and education. Both Mexican-Americans and other Hispanics outlive whites at each level of education, though the size of the advantage is largest at low education. Mexican men have an advantage of 4.3 years and other Hispanic men an advantage of 4.2 years in this education group. In comparison, their advantages are only 0.4 years at high education. While high education white men outlive low education white men by 6.4 years, Mexican and other Hispanic men show corresponding educational gaps of 2.4 and 2.6 years, respectively. The pattern for women is very similar. The total education-adjusted advantages are 2.1 – 2.6 years for Hispanic men and women. These differences are very similar to those found using vital statistics (Arias 2010; Markides and Eschbach 2005). Descriptive analysis confirms that Hispanics have less education than non-Hispanics whites but are substantially less likely to smoke cigarettes. Table 2 shows demographic and behavioral characteristics by sex and Hispanic subgroup. 51% and 44% of non-Hispanic

white men and women, respectively, have college education or more, compared with only 24% and 21% of Mexican men and women, and 39% and 34% of other Hispanic men and women. The majority of Mexican-Americans have less than a high school education. Whites are more likely to smoke than Mexican-Americans and other Hispanics, though the difference is much larger among women. 24% of white women were current smokers, compared with only 14% of Mexican women and 18% of other Hispanic women. White men and women also report substantially higher cigarette consumption than Hispanics. White men who smoke consume an average of 23 cigarettes per day, and women 18 per day. By contrast, Mexican men smoke only 11 and women only 9 per day. Other Hispanics consume an intermediate number. These results imply that the burden of smoking will be greater for whites than for Hispanics. Figure 1 presents the number of years of life expectancy at age 35 lost to smoking by education for each group. These come from subtracting observed life expectancy from life expectancy calculated after removing smoking-related deaths. As expected, white men and women exhibit strong educational gradients in the impact of smoking, while Mexicans and other Hispanics do not. White men with low education lose more than 5 years to smoking compared to 3.5 years among those with high education. White women with low education lose nearly 4 years while those with high education lose only 2.5 years. Mexicans and other Hispanics lose fewer years to smoking, and have weaker or non-existent educational gradients in smoking. Mexican men lose fewer than 2 years and other Hispanic men around 3 years. Smoking reduces life expectancy by 1.2 years among Mexican women with low education and 0.8 among those with high education. Other Hispanic women lose about 2 years at all levels of education. The differences in the strength of the gradient between whites and Hispanic groups suggest that the contribution of smoking to the Hispanic advantage should be largest at low education.

Figure 2 shows the overall contribution of smoking the life expectancy advantage of Mexican-Americans and other Hispanics. The black bar represents the portion of the difference that is explained by smoking. The grey bar represents other factors, or the size of the advantage in the absence of smoking. Smoking explains 1.7 years of the difference between Mexican and white women and 2 years between Mexican and white men. This represents 66% of the difference for women and 96% for men. It is also responsible for around 1 year of the advantage of other Hispanic men and women, which is 40% of the difference. When we calculate the contribution of smoking by education by education, several interesting patterns emerge (Figure 3). The first is that the Hispanic life expectancy advantage is greatest at low education; Each Hispanic subgroup outlives whites by more than 4 years in this educational category. The advantage falls to one year or less at high education. The number of years explained by smoking follows a similar pattern by education. Smoking explains 2-3 years of the advantage of each Hispanic subgroup at low education, but as few as 0.5 years at high education. As a result, the educational pattern of the contribution of smoking partially produces the overall educational pattern of the Hispanic advantage. Among Mexican men and women, 26% of the difference in the size of their advantage between low education and high education results from a lower in the contribution of smoking at high education. Among other Hispanics it is 35%. Discussion The consistent observation of lower mortality among Hispanics in the United States, particularly the foreign-born, has perplexed social scientists for several decades (Palloni and Arias 2004). The “Hispanic paradox” remains a powerful example of a situation in which low socioeconomic status (SES) does not translate into higher mortality, and as such, it represents an especially valuable empirical finding. In searching for the reasons why Hispanics do suffer

poor health as a function of their lower SES, we approach a more comprehensive framework for understanding the SES-mortality relationship. Previous studies have suggested that health behaviors might contribute to the health advantage of Hispanics, particularly Mexican-Americans (Rogers et al. 1996; Morales et al. 2002; Abraido-Lanza et al. 2005; Singh and Siahpush 2002, 2001; Markides and Eschbach 2005). Blue and Fenelon (2011) quantified the effect of smoking using indirect methods and found that a majority of the Hispanic advantage could be explained by smoking-related mortality. This uses individual-level data on smoking behavior and mortality outcomes and produces estimates of the impact of smoking that are very similar. The main contribution of this paper is identifying smoking as a major producer of the educational pattern of the Hispanic advantage. The larger advantage among less educated Hispanics derives largely from heavy smoking among less educated whites. Among the college educated, both the life expectancy advantage and contribution of smoking are substantially reduced. The educational pattern of the Hispanic mortality advantage comes from the combination of a weak education gradient among Hispanics and a strong one among whites (Turra and Goldman 2007). A relatively weak education gradient in health and mortality has been a particularly consistent finding in the literature on Hispanic health, and likely reflects the characteristics of migrant origin destinations or the migration process itself (A. Buttenheim et al. 2010). There are several possible explanations, including selective migration, acculturation effects, or behavioral factors. Selective migration may explain the pattern if only the healthiest individuals with less education migrate (Abraido-Lanza et al. 1999). In this case, low education immigrants would overperform with respect to health relative to expectations given their level of education. Research finds only weak evidence for health selection among Mexican immigrants, either in terms of chronic conditions or health

behaviors (Rubalcava et al. 2008). Migrants and non-migrants in Mexico have similar selfrated health, although migrants appear less likely to be overweight. It is also possible that Hispanic immigrants in the US suffer similar health effects of the acculturation process. Upon entry into the United States, Hispanic immigrants may integrate into lower status groups by virtue of racial discrimination or residential segregation. This “segmented assimilation” may present similar difficulties and stresses to all Hispanics regardless of socioeconomic background (Rumbaut 1997). This may partially explain why greater duration of residence in the US is associated with increases in unhealthy behaviors such as smoking and poor diet (Bethel and Schenker 2005; Goel et al. 2004). Yet segmented assimilation is unlikely to explain a large portion of Hispanics’ weak education gradient since a similar finding has been observed among adults in Mexico as well (K. V. Smith and Goldman 2007). Another explanation for the weak health gradient among Hispanics is similarly weak educational gradients in influential health behaviors. This paper provides direct evidence that the weak educational gradient in smoking is partially responsible for the weak educational gradient in mortality among Mexicans and other Hispanics. This is consistent with existing research demonstrating the role of smoking in socioeconomic differences in adult mortality (Jha et al. 2006; Denney et al. 2010). This is also consistent with evidence indicating small differences in smoking behavior, obesity, and alcohol use by socioeconomic status in Mexico (K. V. Smith and Goldman 2007). Some evidence even suggests that better educated and wealthier individuals in Mexico are more likely smoke, presumably due to the high cost of cigarettes (A. M. Buttenheim et al. 2010; Thrasher et al. 2009). If these patterns of behavior are imported to the US when immigrants arrive, they would help to weaken educational gradients of Hispanics.

The principal limitation of this study is the inability to account for migration effects. Although evidence for the impact of migrant selectivity or salmon bias is mixed (Turra and Elo 2008; Abraido-Lanza et al. 1999; Palloni and Morenoff 2001), it has the potential to bias the conclusions of this paper. Especially if foreign-born Hispanic smokers are more likely to return to their origin countries when they become ill, the estimates of the impact of smoking may be overstated. We can be reassured that burden of smoking is actually lower among Hispanics, since baseline smoking data (that is not subject to salmon biases) indicate a lower prevalence and daily cigarette consumption for these groups. Still, future research should attempt to calculate the contribution of health-selective return migration to differences in smoking-related mortality between Hispanics and non-Hispanic whites. Conclusion As an explanation for the Hispanic mortality advantage, cigarette smoking is consistent with both selective migration and cultural effects. Smoking may be an important dimension on which migrant selection operates. For various reasons, smokers may be less likely to move to the United States, though whether this is true is not yet known. Alternatively, cigarette smoking may be a less culturally acceptable behavior in the origin countries of many Hispanic immigrants. For instance, smoking prevalence may be low in rural Mexico leading to low prevalence among recent Mexican immigrants in the US. Future research is required to specify the factors responsible for superior health behaviors among Hispanic individuals that help to produce their health advantage. This information will be integral to understanding why most Hispanic groups do not experience poor health as a result of low socioeconomic position.

References Abraido-Lanza, A. F., M. T. Chao, and K. R. Florez. 2005. “Do healthy behaviors decline with greater acculturation? Implications for the Latino mortality paradox.” Social Science & Medicine 61(6):1243-1255. Abraido-Lanza, A. F., B. P. Dohrenwend, D. S. Ng-Mak, and J. B. Turner. 1999. “The Latino mortality paradox: A test of the ‘salmon bias’ and healthy migrant hypotheses.” American Journal of Public Health 89(10):1543-1548. Arias, E. 2010. “United States Life Tables by Hispanic Origin.” Vital Health Statistics 152(2). Barbeau, E. M., N. Krieger, and M. J. Soobader. 2004. “Working class matters: Socioeconomic disadvantage, race/ethnicity, gender, and smoking in NHIS 2000.” American Journal of Public Health 94(2):269-278. Becker, T. M., C. Wiggins, C. R. Key, and J. M. Samet. 1988. “ISCHEMIC HEARTDISEASE MORTALITY IN HISPANICS, AMERICAN-INDIANS, AND NONHISPANIC WHITES IN NEW-MEXICO, 1958-1982.” Circulation 78(2):302-309. Bethel, J. W., and M. B. Schenker. 2005. “Accidturation and smoking patterns among Hispanics - A review.” American Journal of Preventive Medicine 29(2):143-148. Blue, Laura, and Andrew Fenelon. 2011. “Explaining low mortality among US immigrants relative to native-born Americans: the role of smoking.” International Journal of Epidemiology 40(3):786 -793. Retrieved September 8, 2011. Buttenheim, A.M., N. Goldman, A. R. Pebley, R. Wong, and C. Chung. 2010. “Do Mexican immigrants ‘import’ social gradients in health to the US?” Social Science & Medicine 71(7):1268-1276. Buttenheim, A.M., R. Wong, N. Goldman, and A.R. Pebley. 2010. “Does social status predict adult smoking and obesity? Results from the 2000 Mexican National Health Survey.” Global Public Health 5:413-426. Retrieved September 14, 2011. Caraballo, R. S., and C. W. Lee. 2004. “Tobacco use among Mexicans and their descendents in the United States.” Salud Publica De Mexico 46(3):241-250. Caraballo, R. S., G. A. Giovino, T. F. Pechacek, and P. D. Mowery. 2001. “Factors associated with discrepancies between self-reports on cigarette smoking and measured serum cotinine levels among persons aged 17 years or older - Third National Health and Nutrition Examination Survey, 1988-1994.” American Journal of Epidemiology 153(8):807-814. Caraballo, R. S. et al. 1998. “Racial and ethnic differences in serum cotinine levels of cigarette smokers - Third National Health and Nutrition Examination Survey, 19881991.” Jama-Journal of the American Medical Association 280(2):135-139. Case, A., D. Lubotsky, and C. Paxson. 2002. “Economic status and health in childhood: The origins of the gradient.” American Economic Review 92(5):1308-1334.

CDC. 2009. “Cigarette smoking among adults and trends in smoking cessation - United States, 2008.” MMWR Morb Mortal Wkly Rep 58(44):1227-32. Crimmins, E., M. D. Hayward, and T. Seeman. 2004. “Race/Ethnicity, Socioeconomic Status, and Health.” in Critical Perspectives on Racial and Ethnic Differences in Health in Late Life, edited by N. B. Anderson, R.A. Bulatao, and B. Cohen. Washington DC: National Academy Press. Denney, J. T., R. G. Rogers, R. A. Hummer, and F. C. Pampel. 2010. “Education inequality in mortality: The age and gender specific mediating effects of cigarette smoking.” Social Science Research 39(4):662-673. Elo, I. T., and S. H. Preston. 1996. “Educational differentials in mortality: United States, 1979-85.” Social Science & Medicine 42(1):47-57. Elo, I. T., C. M. Turra, B. Kestenbaum, and B. R. Ferguson. 2004. “Mortality among elderly Hispanics in the United States: Past evidence and new results.” Demography 41(1):109-128. Eschbach, K., G. V. Ostir, K. V. Patel, K. S. Markides, and J. S. Goodwin. 2004. “Neighborhood context and mortality among older Mexican Americans: Is there a barrio advantage?” American Journal of Public Health 94(10):1807-1812. Ezzati, M., and A. D. Lopez. 2003. “Measuring the accumulated hazards of smoking: global and regional estimates for 2000.” Tobacco Control 12(1):79-85. Goel, M. S., E. P. McCarthy, R. S. Phillips, and C. C. Wee. 2004. “Obesity among US immigrant subgroups by duration of residence.” Jama-Journal of the American Medical Association 292(23):2860-2867. Goldman, N., R. T. Kimbro, C. M. Turra, and A. R. Pebley. 2006. “Socioeconomic gradients in health for white and Mexican-Origin populations.” American Journal of Public Health 96(12):2186-2193. Hummer, R. A., R. G. Rogers, S. H. Amir, D. Forbes, and W. P. Frisbie. 2000. “Adult mortality differentials among Hispanic subgroups and non-Hispanic whites.” Social Science Quarterly 81(1):459-476. Hummer, R. A., R. G. Rogers, C. B. Nam, and F. B. LeClere. 1999. “Race/ethnicity, nativity, and US adult mortality.” Social Science Quarterly 80(1):136-153. IHIS. 2010. “Integrated Health Interview Series: Version 2.0.” Jha, P. et al. 2006. “Social inequalities in male mortality, and in male mortality from smoking: indirect estimation from national death rates in England and Wales, Poland, and North America.” Lancet 368(9533):367-370. Kaplan, M. S, N. Huguet, J. T Newsom, and B. H McFarland. 2004. “The association between length of residence and obesity among Hispanic immigrants.” American Journal of Preventive Medicine 27(4):323–326.

Kimbro, R. T., S. Bzostek, N. Goldman, and G. Rodriguez. 2008. “Race, ethnicity, and the education gradient in health.” Health Affairs 27(2):361-372. LeClere, F. B., R. G. Rogers, and K. D. Peters. 1997. “Ethnicity and mortality in the United States: Individual and community correlates.” Social Forces 76(1):169-198. Liberatos, P., B. G. Link, and J. L. Kelsey. 1988. “THE MEASUREMENT OF SOCIALCLASS IN EPIDEMIOLOGY.” Epidemiologic Reviews 10:87-121. Markides, K. S., and J. Coreil. 1986. “The Health of Hispanics in the Southwestern UnitedStates - An Epidemiologic Paradox.” Public Health Reports 101(3):253-265. Markides, K. S., and K. Eschbach. 2005. “Aging, migration, and mortality: Current status of research on the Hispanic paradox.” Journals of Gerontology Series B-Psychological Sciences and Social Sciences 60:68-75. Morales, L. S., M. Lara, R. S. Kington, R. O. Valdez, and J. J. Escarce. 2002. “Socioeconomic, cultural, and behavioral factors affecting Hispanic health outcomes.” Journal of Health Care for the Poor and Underserved 13(4):477-503. O’Brien, K. et al. 2003. “Cancer statistics for Hispanics, 2003.” Ca-a Cancer Journal for Clinicians 53(4):208-226. Palloni, A., and E. Arias. 2004. “Paradox lost: Explaining the Hispanic adult mortality advantage.” Demography 41(3):385-415. Palloni, A., and D. C Ewbank. 2004. “Selection processes in the study of racial and ethnic differentials in adult health and mortality.” Critical perspectives on racial and ethnic differences in health in late life. Palloni, A., and J. D. Morenoff. 2001. “Interpreting the paradoxical in the Hispanic paradox Demographic and epidemiologic approaches.” Pp. 140-174 in Population Health and Aging - Strengthening the Dialogue between Epidemiology and Demography, Annals of the New York Academy of Sciences, edited by M. Weinstein, A. I. Hermalin, and M. A. Stoto. Retrieved (://000173961000010). Peto, R., A. D. Lopez, J. Boreham, M. Thun, and C. Heath. 1992. “Mortality From Smoking in Developed Countries - Indirect Estimation From National Vital-Statistics.” Lancet 339(8804):1268-1278. Peto, R., A. D. Lopez, J. Boreham, M. Thun, and C. Heath. 1994. Mortality from Smoking in Developed Countries 1950-2000. Oxford: Oxford University Press. Preston, S. H., and P. Taubman. 1994. “Socioeconomic Differences in Adult Mortality and Health Status.” in Demography of Aging, edited by L. Martin and S. H. Preston. Washington DC: National Academy Press. Preston, S. H., D. A. Glei, and J. R. Wilmoth. 2010a. “A new method for estimating smoking-attributable mortality in high-income countries.” International Journal of Epidemiology 39(2):430-438.

Preston, S. H., D. A. Glei, and J. R. Wilmoth. 2010b. “Contribution of Smoking to International Differences in Life Expectancy.” in International Differences in Mortality at Older Ages: Dimensions and Sources, edited by E. Crimmins, S. H. Preston, and B. Cohen. Washington, D.C.: National Academy Press. Rockhill, B., B. Newman, and C. Weinberg. 1998. “Use and misuse of population attributable fractions.” American Journal of Public Health 88(1):15-19. Rogers, R. G., R. A. Hummer, P. M. Krueger, and F. C. Pampel. 2005. “Mortality attributable to cigarette smoking in the United States.” Population and Development Review 31(2):259-292. Rogers, R. G., R. A. Hummer, C. B. Nam, and K. Peters. 1996. “Demographic, socioeconomic, and behavioral factors affecting ethnic mortality by cause.” Social Forces 74(4):1419-1438. Rubalcava, L. N., G. M. Teruel, D. Thomas, and N. Goldman. 2008. “The healthy migrant effect: New findings from the Mexican family life survey.” American Journal of Public Health 98(1):78-84. Rumbaut, R. G. 1997. “Paradoxes (and orthodoxies) of assimilation.” Sociological Perspectives 40(3):483-511. Singh, G. K., and M. Siahpush. 2001. “All-cause and cause-specific mortality of immigrants and native born in the United States.” American Journal of Public Health 91(3):392399. Singh, G. K., and M. Siahpush. 2002. “Ethnic-immigrant differentials in health behaviors, morbidity, and cause-specific mortality in the United States: An analysis of two national data bases.” Human Biology 74(1):83-109. Smith, J. P. 2004. “Unraveling the SES-health connection.” Population and Development Review 30:108-132. Smith, K. V., and N. Goldman. 2007. “Socioeconomic differences in health among older adults in Mexico.” Social Science & Medicine 65:1372-1385. Sorlie, P. D., E. Backlund, N. J. Johnson, and E. Rogot. 1993. “MORTALITY BY HISPANIC STATUS IN THE UNITED-STATES.” Jama-Journal of the American Medical Association 270(20):2464-2468. Sudano, J. J., and D. W. Baker. 2006. “Explaining US racial/ethnic disparities in health declines and mortality in late middle age: The roles of socioeconomic status, health behaviors, and health insurance.” Social Science & Medicine 62(4):909-922. Sullivan, T. A., F. P. Gillespie, M. Hout, and R. G. Rogers. 1984. “ALTERNATIVE ESTIMATES OF MEXICAN-AMERICAN MORTALITY IN TEXAS, 1980.” Social Science Quarterly 65(2):609-617. Thrasher, James F. et al. 2009. “Does the availability of single cigarettes promote or inhibit cigarette consumption? Perceptions, prevalence and correlates of single cigarette use

among adult Mexican smokers.” Tobacco Control 18(6):431-437. Retrieved April 20, 2011. Thun, M. J., L. F. Apicella, and S. J. Henley. 2000. “Smoking vs other risk factors as the cause of smoking-attributable deaths - Confounding in the courtroom.” Jama-Journal of the American Medical Association 284(6):706-712. Thun, M. et al. 1997. “Trends in tobacco smoking and mortality from cigarette use in cancer prevention studies I (1959 through 1965) and II (1982 through 1988).” Pp. 305-382 in Changes in cigarette-related disease risks and their implications for prevention and control, smoking and tobacco control. Smoking and tobacco control monograph no. 8, NIH publication no. 97-4213, edited by D. Burns, L. Garfinkel, and J. M. Samet. Bethesda, MD.: Cancer Control and Population Sciences, National Cancer Institute, U.S. National Institutes of Health. Trinidad, D. R. et al. 2009. “Intermittent and light daily smoking across racial/ethnic groups in the United States.” Nicotine & Tobacco Research 11(2):203-210. Turra, C. M., and I. T. Elo. 2008. “The impact of salmon bias on the Hispanic mortality advantage: New evidence from social security data.” Population Research and Policy Review 27(5):515-530. Turra, C. M., and N. Goldman. 2007. “Socioeconomic differences in mortality among US adults: Insights into the Hispanic paradox.” Journals of Gerontology Series BPsychological Sciences and Social Sciences 62(3):S184-S192. Winkleby, M. A., and C. Cubbin. 2004. “Racial/Ethnic Disparities in Health Behaviors: A Challenge to Current Assumptions.” in Critical Perspectives on Racial and Ethnic Differences in Health in Late Life, edited by N. B. Anderson, R.A. Bulatao, and B. Cohen. Washington, DC: National Academy Press.

Table 1: Life expectancy at age 35 by education for Mexican-Americans, other Hispanics and non-Hispanic whites Men

Women

non-Hispanic white

MexicanAmerican

other Hispanic

non-Hispanic white

MexicanAmerican

other Hispanic

Low Education

38.3

42.6 (4.3)

42.5 (4.2)

42.9

47.8 (4.9)

48.2 (5.3)

Medium Education

41.4

44.8 (3.3)

45.1 (3.7)

47.3

49.1 (1.9)

48.5 (1.2)

High Education

44.7

45.0 (0.4)

45.1 (0.4)

49.2

50.5 (1.2)

50.0 (0.7)

Total

41.9

44.0 (2.1)

44.2 (2.3)

46.9

49.5 (2.6)

49.3 (2.4)

Difference High-Low 6.4 2.4 2.6 6.3 2.7 1.8 Notes: Numbers refer to life expectancy in years at age 35 estimated from hazard regression by sex and ethnicity. Numbers in parentheses refer to the education-specific advantage of each Hispanic subgroup relative to non-Hispanic whites. Low education=less than high school; medium education=high school diploma; high education=college or more. Difference High-low is the life expectancy gap between high education and low education in each subgroup.

Table 2: Demographic and behavioral characteristics of sample by sex and Hispanic ethnicity: 1987 - 2004 Men

Mean Age Education (percent) Low Medium High Body Mass Index (BMI) Not Obese (BMI<30) Obese (BMI≥30) Smoking Status Never smoker Former smoker Current (0-4 per day) Current (5-9) Current (10-19) Current (≥20) Current smoker1 Mean cigarettes per day2 Observations Number of Deaths

NonHispanic White 54.6

MexicanAmerican 49.5

14.7 33.9 51.4

Women Other Hispanic 51.1

NonHispanic White 57.4

MexicanAmerican 50.7

Other Hispanic 52.1

50.5 25.7 23.9

31.1 29.9 39.0

16.6 39.3 44.2

52.7 26.7 20.6

36.4 29.5 34.1

84.2 15.8

77.7 22.3

83.4 16.6

84.8 15.2

73.9 26.1

81.3 18.7

38.1 34.6 1.7 1.4 5.0 19.2 27.3 22.5

48.0 26.4 8.5 4.0 6.4 6.8 25.6 10.6

47.6 26.0 5.6 3.1 7.0 10.7 26.4 14.4

53.6 22.8 1.9 1.9 6.3 13.5 23.6 18.1

72.0 13.7 5.2 3.0 3.4 2.8 14.3 9.1

66.9 15.4 3.9 3.2 5.1 5.5 17.7 9.8

106,991 23,715

7,790 825

6,121 770

142,170 30,445

9,538 800

8,957 912

Notes: N=281,567. Supplement specific weights are used for 1987 - 1995. Adult sample weights are used for 1997-2004. Low education=less than high school; medium education=high school diploma; high education=college or more. 1 Refers to the prevalence of current smoking irrespective of consumption. 2 Among current smokers

Figure 1: Years of life expectancy lost to smoking by education for non-Hispanic whites, Mexican Americans, and other Hispanics: age 35+ (a) Men 6.0 Low ed.

Medium ed.

High ed.

5.0 4.0 yrs lost to 3.0 smoking 2.0 1.0 0.0 White

Mexican

Other Hispanic

(b) Women 4.5 4.0

Low ed.

Medium ed.

High ed.

3.5 3.0 yrs 2.5 lost to smoking 2.0 1.5 1.0 0.5 0.0 White

Mexican

Other Hispanic

Figure 2: Overall contribution of smoking of the Hispanic advantage in life expectancy at age 35 3.0

Mexican-Americans

Other Hispanics

2.5 2.0 Life Exp. 1.5 Advantage (years)

Other Factors Smoking

1.0 0.5 0.0 Women

Men

Women

Men

65.8%

95.9%

37.4%

42.7%

Figure 3: Contribution of smoking to the Hispanic life expectancy advantage at age 35 by education (a) Men 5

Other Hispanic

4 3

Other Factors

Life 2 Exp. Advantage (years) 1

Smoking

0 -1 -2

Low

Medium

Low

High

Medium

High

(b) Women 6 Mexican

Other Hispanic

5 4 Life Exp. 3 Advantage (years) 2

Other Factors Smoking

1 0 -1

Low

Medium

High

Low

Medium

High