Intergenerational Continuity of Substance Use

Abstract

Guided by rigorous methodology and a life-course perspective, the goal of this research is to address a gap in current knowledge on whether, when, and how strongly intergenerational continuity of substance use exists when examining age-equivalent and developmentally specific stages of the life course. Annual self-reported substance use measures were analyzed from a prospective, longitudinal, and nationally representative sample that originally consisted of 1,725 respondents and their families, who were then interviewed over a 27-year period from 1977 to 2004. Findings from multilevel random-intercept regression models provide support for intergenerational continuity when substance use occurs in emerging adulthood but not when limited to adolescence. Implications, limitations, and future research directions are discussed.

Keywords:

• alcohol use
• marijuana use
• drug use
• substance use
• intergenerational continuity
• intergenerational transmission
• adolescence
• emerging adulthood
• life course
• methodology

Notes

1 The notational system of G1, G2, and G3 are used to refer to each successive generation. As discussed in the method's section, G1 grandparents are not a focus of the current study.

2 Each annual substance use measure represents the number of times used, not the number of days used.

3 Substance use questions from Wave 10 (for G2 respondents) were different in format and timespan from other waves and therefore are not used in the present analysis (e.g., different substances, 6 months vs. 1 year, number of days used instead of number of times used).

4 In Wave 1 only, scale measures were converted to frequencies using regression transformations; see Elliott et al. (1989) Appendix A for details.

5 Substance use questions from Wave 11 (for G3 respondents only) were different in format and timespan from other waves and therefore are not used in the present analysis (e.g., different substances, 6 months vs. 1 year, and days of use instead of frequency of use).

6 One G3 respondent was 11 years of age and this case was retained to maximize the sample size.

7 The coefficient b = 1.55 is transformed into an odds ratio, OR = exp(1.55) = 4.71.

8 When a predictor and outcome variable are both logged, the regression coefficient is interpreted as “a d percent increase in X is associated with a 100 × (exp(b1 × (ln(1+(d/100))) –1) percent change in Y” and is sometimes referred to as elasticity (Gelman & Hill, 2007).