Childhood self-control predicts smoking throughout life: Evidence from 21,000 cohort study participants.

Self-Control and Smoking

Why are some children more vulnerable than others to becoming tobacco users? One factor is environmental exposure: being raised by parents who smoke or in an environment where smoking is prevalent has been linked to early smoking initiation ( Bricker et al., 2006 ; Hiscock, Bauld, Amos, Fidler, & Munafo, 2012 ). Another factor is early personality. Enduring behavioral tendencies emerge early and have consequential effects on a range of adult outcomes. For example, the capacity to exert self-control over thoughts and actions from ages 3 to 11 predicts substance dependence at age 32 ( Moffitt et al., 2011 ).

Longitudinal studies have also linked childhood and adolescent conscientiousness to subsequent smoking (e.g., Pluess & Bartley, 2015 ). In fact, a substantial part of the health benefits of conscientiousness comes from not smoking ( Hampson, Edmonds, Goldberg, Dubanoski, & Hillier, 2015 ; Hampson, Goldberg, Vogt, & Dubanoski, 2007 ). Self-regulatory processes may underlie the development of conscientiousness ( Eisenberg, Duckworth, Spinrad, & Valiente, 2014 ) and explain why more conscientious individuals tend not to smoke and live longer, healthier lives ( Bogg & Roberts, 2004 ; Costa, Weiss, Duberstein, Friedman, & Siegler, 2014 ; Hampson et al., 2016 ; Turiano, Hill, Roberts, Spiro, & Mroczek, 2012 ).

Only a few longitudinal studies have shown that self-control problems place children at risk of subsequent smoking initiation in adolescence ( deBlois & Kubzansky, 2015 ; de Winter, Visser, Verhulst, Vollebergh, & Reijneveld, 2015; King, Fleming, Monahan, & Catalano, 2011 ; Lee, McClernon, Kollins, Prybol, & Fuemmeler, 2013 ; Moffitt et al., 2011 ) and, in turn, increase risk of smoking in young adulthood ( Lipkus, Barefoot, Williams, & Siegler, 1994 ; Welch & Poulton, 2009 ). The link between childhood self-control and smoking has not been examined using national data, nor has the potential protective role of self-regulatory skills in reducing the persistence of smoking across adulthood been identified. The paucity of data linking childhood self-control to adult smoking is surprising considering that recent reviews have implicated personality, behavioral, and neurobiological measures of impaired self-control in all stages of smoking, including initiation, maintenance, and relapse ( Bloom, Matsko, & Cimino, 2014 ; MacKillop et al., 2011 ).

Self-control, an important resource for resisting cravings and avoiding temptation, is likely to be vital to understanding who begins, continues, and gives up smoking. Children with low self-control are more susceptible to tobacco advertising and the influence of peers who smoke ( Audrain-McGovern et al., 2006 ; Piehler et al., 2012 ; Wills et al., 2010 ). Adolescents with better self-control are less likely to begin smoking as young adults ( Audrain-McGovern et al., 2009 ), and less impulsive smokers (presumably those with more self-control) are more successful in achieving their smoking cessation goals ( Ida, Goto, Takahashi, & Nishimura, 2011 ). Finally, interventions that increase self-control can reduce the risk of relapse among quitters ( Muraven, 2010 ).

Aims of the Present Study

Existing research strongly suggests a potential role for self-control in shaping smoking habits. However, this work has been limited by the use of non-nationally representative samples, short periods of follow-up, lack of multiwave data, and personality measures elicited after smoking initiation. We used two large cohort studies containing comprehensive data on tobacco use over a prolonged period spanning childhood to midlife.

We hypothesized that low self-control is a core reason why children take up smoking in adolescence and continue to smoke throughout life. Furthermore, we aimed to address four key limitations of the previous literature. First, we used childhood measures of self-control to predict later smoking and, where possible, eliminated from the sample any children who were already smokers at baseline. Other studies measuring smoking and self-control at the same time introduce potential ambiguities because self-control can change in response to prolonged exposure to nicotine, smoking deprivation, and cessation ( Bloom, Matsko, & Cimino, 2014 ; Ida et al., 2011 ; Sutin et al., 2013 ; Yamane et al., 2013 ). Second, we examined the potential confounding role of important contextual factors including low social class, parental smoking, and early individual differences including psychological distress and low cognitive ability, which increase the risk of tobacco use and covary with self-control ( Bricker et al., 2006 ; Hiscock et al., 2012 ; Lynskey & Fergusson, 1995 ). Third, we prospectively examined smoking status across multiple life stages, allowing us to test whether adolescent smoking was a pathway from early life self-control to adult smoking. Fourth, by examining changes in smoking status across adulthood, we could test whether self-control underlies the processes that shape population smoking levels: smoking initiation, relapse, and cessation.

National Child Development Study

We used the NCDS to extend our analysis to an older cohort. We also used the richer background data available in the NCDS to more stringently test the contribution of childhood self-control, measured at ages 7 and 11, to smoking behavior at ages 23, 33, 42, 46, 50, and 55, using a sample of 12,605.

Adult smoking

In both cohorts, participants indicated whether they “smoke cigarettes every day,” “smoke cigarettes occasionally but not every day,” “used to smoke cigarettes but don’t at all now,” or “never smoked cigarettes” at each wave across adulthood. We created a categorical variable at each wave classifying participants as “never smokers,” “ex-smokers,” or “smokers” (daily and occasional smokers combined). Our “smoker” definition followed the U.K. Office for National Statistics (ONS) smoking classification system, which combines daily and occasional smokers, allowing smoking rates in the sample to be compared with national statistics (see Figure 1 and Supplementary Materials, Section 3). Our second outcome, also reported at each wave, examined the number of cigarettes smoked per day by daily smokers (i.e., those who reported they “smoke cigarettes every day”). The questions used to elicit both smoking outcomes are described in the Supplementary Materials, Section 4. Participants provided smoking data in 72.4% of possible survey waves, and a set of weighted analyses (available on request) showed that accounting for selection bias and the association between baseline characteristics and missing data across survey waves did not affect the relationship between self-control and smoking status.

In the BCS, it was possible to identify those who met the ONS criteria for child smoking at baseline—defined as a child who smokes at least one cigarette per week on average. To maintain clarity regarding the direction of influence between self-control and smoking behavior, we therefore removed 91 participants who reported smoking 1 or more cigarettes per week at age 10, when self-control was elicited.

Adolescent smoking

In order to test whether adolescent smoking mediated the relationship between childhood self-control and adult smoking behavior, we examined the number of cigarettes smoked per week at age 16 (where 1 = Nonsmoker; 2 = 1 cigarette; 3 = 2–10 cigarettes; 4 = 11–20 cigarettes; 5 = 21–40 cigarettes; 6 = 41+ cigarettes) in both cohorts. The questions used to derive our adolescent smoking measure are detailed in the Supplementary Materials, Section 4.

Parental smoking

All analyses adjusted for parental smoking, which was measured via parent-report when the cohort member was aged 10 in the BCS and 16 in the NCDS. Maternal and paternal smoking habits were coded as 0 = Nonsmoker; 1 = 1–10 cigarettes per day; 2 = 11–20 per day; 3 = 21+ per day; 4 = missing data. The NCDS also included a category for parental pipe/cigar smoking. Where information on maternal smoking was unavailable at age 16 in the NCDS, we used maternal smoking levels prior to pregnancy. See Supplementary Materials, Section 4 for the individual parental smoking items used.

Covariates

The other childhood covariates were gender, general cognitive ability, psychological distress, and social class. In the BCS, cognitive ability was measured at age 10 using the British Ability Scales (BAS), which consist of two verbal and two nonverbal tests ( Elliott, Murray, & Pearson, 1978 ; Cronbach’s alpha = .93). In the NCDS, cognitive ability was measured at age 11 using 40 verbal and 40 nonverbal items ( Pigeon, 1964 ; Cronbach’s alpha = .94). Psychological distress was measured at age 10 in the BCS using 5 teacher-rated items from the Neuroticism/Anxiety subscale of the Child Developmental Behaviors scale (Cronbach’s alpha = .85). In the NCDS, distress was measured at ages 7 and 11 using a teacher-rated measure of psychological distress (see Egan, Daly, & Delaney, 2015 ) for further details, and individual distress items are included in Supplementary Materials, Section 2). Social class, elicited at birth and derived from the father’s occupation, was classified into five categories based on the Registrar General’s Social Classes: I = professional occupations; II = managerial and technical occupations; III = skilled occupations; IV = partly skilled occupations; V = unskilled occupations. Two additional categories were included to code for “Other” occupational categories (e.g., father unemployed/absent), and missing data.

In addition to these covariates, the comparatively richer background data available in the NCDS allowed us to include eight additional childhood variables that could have affected the association between early self-control and smoking. These variables were the cohort member’s race, family difficulties, household size, father’s age, and the presence of headaches/epilepsy, intellectual disability, psychiatric problems, and low birth weight. Details of the individual variables are described in the Supplementary Materials, Section 5.

Finally, we included three traits that are conceptually related to self-control in supplementary robustness tests. We included measures of child conduct problems and hyperactivity ( Lynskey & Fergusson, 1995 ) and assessed whether conscientiousness at age 16 diminished the contribution of childhood self-control to adult smoking independently of smoking behavior at age 16. Details of the measures used are included in Supplemental Materials, Section 6.

Early smoking initiation

We tested the indirect effect of childhood self-control on later life smoking by adding our measure of adolescent smoking to Model 1 and then using the khb command in Stata ( Karlson, Holm, & Breen, 2012 ) to calculate the mediation effect. The khb procedure is suitable for examining outcomes measured repeatedly over time and where direct cross-model comparisons cannot be made because the outcome variable is noncontinuous. In the current study khb performs the necessary decomposition to allow the indirect pathway from self-control through an ordinal mediator (adolescent smoking) to a categorical outcome (smoker/ex-/never adult smoker) to be estimated.