Economics of Smoking

Effects Of Other Cigarette Demand Determinants

Educational Attainment

Two important stylized facts are pertinent for describing the relationship between educational attainment and smoking. First, more highly educated persons tend to be healthier than others and, second, on average more highly educated persons smoke less. Although the associations are indisputable, there is controversy about causation and, if there is causality, the magnitude of the effect. Moreover, the relative importance is not clear of the various pathways through which educational attainment influences smoking.

Although the correlation between education, health behaviors, and health is well established, the notion that additional schooling affects health behaviors and causes health improvements is not. Reverse causality from health in general, and health behaviors in particular, to years of schooling completed, and/or the presence of third factors correlated with schooling and smoking but omitted from the analysis may cause educational attainment, health behaviors, and health to vary in the same direction. In high-income countries, but not necessarily in other countries, there is little reason to expect reverse causation from health to years of schooling because generally the temporal lag between the completion of formal education and the time at which health declines is several decades.

Possible omitted third variables is a more likely source of bias. Among these variables are: native ability, time preference, including at the time schooling choices are made, genetic factors, poor health in early life which may be positively correlated with poor health in adulthood (and which may also affect educational attainment in early, low income in early life which similarly may affect educational attainment and, independently, health in later life (Jayachandran and Lleras-Muney, 2009).

One’s ability to make valid causal inferences depends on the quality of the identification strategy. Various strategies have been used. For example, de Walque (2010) constructed panels based on smoking histories, finding that among women, college education has a negative influence on the probability of smoking and more educated persons’ smoking responded more quickly to diffusion of information on smoking. He offered explanations for differences by gender. Other identification strategies were based on availability of high school and college openings, draft avoidance during the Vietnam War and abolition of secondary school fees. In general, studies find causal effects, but there are exceptions. Most empirical evidence on effects of educational attainment comes from the USA and other high-income countries. However, empirical evidence from other countries supports the finding of a negative effect of educational attainment on smoking.


Peer effects have different potential roles depending on the stage of the life cycle. For adolescents, peers may encourage smoking. In the past, in adult life, smoking may have been seen as useful in promoting business or social interactions. Also, spouses may influence a person’s smoking patterns. Isolating peer effects is difficult in particular because choice of peers is endogenous. People who enjoy being around smokers are likely to associate with smokers and conversely for persons who suffer from being around smokers. The conventional wisdom is that peer effects are important in influencing adolescents to start smoking.

Powell et al. (2005) analyzed peer effects on smoking. Peer effects were measured as behavior of all students in a school less the student in question. The instrumental variables were characteristics of other students in the school. They found that peer effects have an important role in influencing individual adolescents to smoke. In particular, moving from a school in which no students smoke to one in which a quarter of the students smoke increases the probability that a youth smokes by 14.5% points. Although the specific estimates vary, virtually all studies in the USA and elsewhere, have found positive effects of peer group smoking.

One type of peer effect reflects interactions of siblings and spouses within a household, where the expectation is that the effect would be strong. The relationship between decisions about smoking of spouses may reflect several underlying influences: correlation due to matching in the marriage market; bargaining within marriage; and social learning.

Health Shocks

Surviving a major health shock, such as a heart attack, may affect continued smoking for at least two reasons. First, the shock may reveal inherited susceptibility, i.e., new information about the personal effects of a life style including smoking. Second, the person may seek to forestall further health damage to self. However, the health shock may serve to increase unhealthy behaviors if it leads to thinking that they will gain little or no benefit from cessation given that they do not have much time to live. The onset of smoking-related health shocks other than cancer, nonsmoking related health shocks, onset of activities of daily living limitations, and onset of fair/poor health all increase the probability of smoking cessation. Smoking-related health shocks may generate new information on the effects of smoking to the individual. But the nonsmoking-related health shocks must be affecting smoking though another mechanism.

Health shocks to spouses as well as their smoking behavior may influence an individual’s smoking decision. Among never smokers, but not for current and former smokers, spousal smoking has a negative effect on a person’s longevity expectations. Spousal smoking-related health shocks also reduce longevity expectations of never smokers for reasons that are not understood because they have no effect on such expectations for current smokers. This might occur through three channels: Consumption externalities – one spouse’s welfare affects the other spouse’s welfare; altruism – one spouse reduces smoking in response to the other spouse’s bad health; and learning about risks of smoking from the health experience of the other spouse. There is some evidence suggestive that consumption externalities are at work.


Even though there is a substantial amount of noneconomic literature on the effect of stress on smoking, the economic literature on the topic is quite limited, and only some results support an effect of stress on smoking. Job-related stress seems to increase smoking at the extensive margin, but the relationship is not generally statistically significant at the intensive margin. Death of a parent within the previous 2 years increases the probability of continuing to smoke. Without fixed effects, being separated, divorced, widowed in the past 2 years leads to continued smoking, but with fixed effects included, these relationships disappear.

Sex Work in Developing Countries