Casting doubt on the causal link between intelligence and age at first intercourse: A cross-generational sibling comparison design using the NLSY *

• We tested the causal link between AFI and intelligence using inter-generational data.
• We explicitly parsed between- and within-family variance.
• Using between-family design, we found positive relationships between child AFI and intelligence.
• Using a within-family design, the relationship between intelligence and AFI vanishes for maternal and child intelligence.
• We find no evidence for intelligence being a direct causal influence on AFI.
• Although “smart teens don’t have sex (or kiss much either)” (Halpern et al. 2000) at a descriptive level,
• their reasons for delaying these activities do not appear to be caused by their smartness.

Halpern et al. (2000) published a study based on early Add Health data with the provocative title “Smart Teens Don’t Have Sex (or Kiss Much Either).” Several following papers reported the same result, a positive correlation between the intelligence of adolescent girls and age at first intercourse (AFI). However, the causal mechanism has not been carefully investigated. Harden and Mendle (2011) used Add Health data within a biometrical design and found that the relationship between intelligence and AFI was fully accounted for by shared environmental differences, suggesting at least the location of the causal mechanism — the part of the household environment shared by siblings that influences both child intelligence and AFI.In this study, we use an intergenerational sibling comparison design to investigate the causal link between intelligence and AFI, using the National Longitudinal Survey of Youth 1979 and the NLSY-Children/Young Adult data. We measured maternal IQ using the AFQT, child IQ using PPVT, PIAT, and Digit Span, and AFI, using respondent self-report. Our analytic method used Kenny's (2001) reciprocal standard dyad model. This model supported analyses treating the data as only between-family data (as in most past studies), and also allowed us to include both between- and within-family comparisons. These analyses included two forms, first a comparison of offspring of mothers in relation to maternal IQ, then a comparison of offspring themselves in relation to offspring IQ.When we evaluated the relationship between maternal/child intelligence and AFI, using a between-family design, we replicated earlier results; smart teens do appear to delay sex. In the within-family analyses, the relationship between intelligence and AFI vanishes for both maternal intelligence and child intelligence. The finding is robust across gender and age. These results suggest that the cause of the intelligence-AFI link is not intelligence per se, but rather differences between families (parental education, SES, etc.) that correlate with family-level (but not individual-level) intelligence.