All in the family British Style: Does Family Smoking Cause British Youth to Smoke?

Smoking causes and complicates symptoms of chronic diseases that impose costs on smokers and society. In response, governments enact tobacco control policies; many of which specifically aim to limit smoking initiation (e.g. cigarette taxes, marketing restrictions, and restrictions on youth access). While many of these policies directly affect behavior, other policies may indirectly reduce smoking initiation. Here we study how smoking initiation varies with parental smoking behavior - an example of such an indirect mechanism that researchers commonly assert to be a determinant of youth initiation probabilities. We also analyze separately the correlation between smoking behavior of a parent and child of the same gender to test another hypothesis commonly found in the literature - that a parent of the same sex more strongly influences youth behavior because that parent serves as a role model.

Researchers frequently present evidence showing that smoking prevalence rates are higher among children of smoking parents than among children of non-smoking parents. Few studies attempt to identify whether a child smokes because her parents smoke. Here we do with instrumental variable (IV) methods and innovative data.

We draw data from the British Household Panel Study (BHPS). Those data track lifetime smoking behavior so we observe whether a youth smokes in every year of life. Because the BHPS is a household-based panel that follows household members into adulthood, we observe lifetime smoking behavior of parents and children. Our instrument exploits two stylized facts. First, smoking initiation usually takes place between the age of 14 and 18. Second, smoking is an addictive behavior, and thus smoking status in teenage years predicts smoking status in adulthood. Using those stylized facts, we predict whether or not a parent smoke using the average cigarette price each parent faced when he/she was a teenager. Because the average British woman bears her first child when she is 23-25 years old, most parents start to smoke a decade or more before their children face their decision. We argue that this temporal distance creates exogenous variation, as cigarette prices parents and children face as teenagers differ.

In our naive results, when parental smoking is treated as exogenous, we find a strong positive correlation between the probability that a child smokes when one or both parents smoke. However, the positive relationship largely disappears when we endogenize the parent's decision to smoke. We conclude that the inter-generational transmission of smoking behavior is mainly due to unobserved effects at the family level, rather than to a true causal effect of parental smoking. When parents are considered separately, OLS estimates find strong correlations between the smoking status of daughters and mothers, and sons and fathers. However, these gender-specific correlations disappear in the IV models. The results again suggest that the correlation between the smoking behavior of mothers and daughters and of fathers and sons is due to gender-specific factors common to mothers-daughters and fathers-sons, rather than true causation.