Does staying in school (and not working) prevent teen smoking and drinking?

https://doi.org/10.1016/j.jhealeco.2012.05.004Get rights and content

Abstract

Previous work suggests but cannot prove that education improves health behaviors. We exploit a randomized intervention that increased schooling (and reduced working) among male students in the Dominican Republic, by providing information on the returns to schooling. We find that treated youths were much less likely to smoke at age 18 and had delayed onset of daily or regular drinking. The effects appear to be due to changes in peer networks and disposable income. We find no evidence of a direct impact of schooling on rates of time preference, attitudes towards risk or perceptions that drinking or smoking are harmful to health, though our measures of these factors are more limited.

Introduction

Schooling is a strong predictor of health, in both developed and developing countries. These associations are large: for example in the year 2000, one more year of schooling was associated with approximately one more year of life expectancy in the U.S. More educated individuals are also less likely to smoke or drink excessively, and in general have better health-related behaviors (see Cutler and Lleras-Muney, 2008 for a review). However, there is considerable debate about whether these associations reflect causal effects. The correlation may instead be driven by omitted variables bias; for example, high discount rates can influence both schooling and health-related behaviors, since both require forgoing utility today in return for future benefits (higher wages or better health). Alternatively, causality may be reversed; for example, students who drink regularly may not perform well in school and therefore drop out, be held back or expelled. A number of recent studies have addressed these concerns using instrumental variables and/or natural experiments, such as changes in compulsory schooling laws, but they have yielded mixed evidence on the health-education relationship.1 However, these studies generally suffer from two primary concerns. First, they rely on difficult to test identifying assumptions. Second, most do not explore the mechanisms that explain why schooling affects health behaviors.

In this paper, we take advantage of a unique panel data set and a randomized intervention in the Dominican Republic to overcome these two challenges. The intervention provided students with information on the returns to schooling. While the standard model of human capital suggests that education responds to the returns to schooling, it is the returns perceived by decision-makers that matter, not the returns measured by economists (Manski, 1993). And there are many reasons to believe that students, particularly in developing countries, may not be well informed of the true returns.2 For example, youths in rural communities or small towns where few adults have any education will have little information from which to infer the returns, including the returns in the urban sector. If students in particular underestimate the returns to schooling, the provision of information on the true returns alone may affect schooling. Consistent with this hypothesis, Jensen (2010) finds that 8th grade male students in the Dominican Republic significantly underestimate the returns to schooling. And students at randomly selected schools who were provided with information on the measured returns completed on average 0.20 years more schooling over the next 4 years than those who were not given this information. To the extent that this intervention affects drinking and smoking only through the impact on schooling (a point we discuss below), the random assignment in this experiment provides an exogenous shock to schooling, uncorrelated with omitted variables and with a clear direction of causality, with which to identify the relationship between schooling and alcohol and tobacco use.

The survey also collected data on a number of potential determinants of drinking and smoking that will allow us to understand the mechanisms through which schooling affects these behaviors. The mechanisms we explore are the behavior of peer networks, discretionary income, rates of time preference, attitudes towards risk, and perceptions of the health consequences of these behaviors. Although not an exhaustive list of mechanisms, these are some of the most commonly cited in the literature (Cutler and Lleras-Muney, 2008). We briefly explain how schooling would affect behaviors though each. First, youths who drop out of school and enter the labor market will have more income at their disposal, which makes it easier to afford alcohol and tobacco. Second, schooling may change the youths’ peer sets. Youths who stay in school will spend a significant fraction of their time with peers who are also in school, and thus for example of a similar age. By contrast, those who drop out may spend more time with older people, such as in the workplace. Third, schooling may directly affect knowledge of the health risks of drinking and smoking. For example, high schools may have required “health” classes that provide information on the health risks. Health knowledge could also be indirectly affected, if more schooling results in exposure to more and/or different sources of health information. Fourth, schooling may affect an individual's rate of time preference or attitude towards risk (Fuchs, 1982), because for example schooling instills particular values, and it requires discipline and patience. In light of this discussion, we note that it is not likely to be possible or even meaningful to talk about the pure effects of schooling on health behaviors, even though most of the literature does not account for this distinction; leaving school in some but not all cases means entering the labor force, and it may be that it is what happens in the workplace, rather than what happens in school, that affects whether a teen drinks or smokes. Thus, our analysis will focus on the combined effects of work and school.

Our study focuses on smoking and drinking (especially daily or regular drinking). These behaviors are two of the most important risk factors in explaining early mortality, accounting for about 14% of deaths worldwide (WHO, 2010). Additionally, drinking and smoking are significant policy concerns because of the externalities associated with their consumption, such as second-hand smoke. Excessive drinking is also associated with increases in deaths from accidents (such as motor vehicle injuries) and crime, especially among adolescents. Our data focuses on the period of adolescence, the period during which both drinking and smoking typically start; most individuals have already tried alcohol by their early teens,3 and most adult smokers begin smoking before the age of 18.4 Thus, this age range is a particularly important one to study. And there are several other reasons why smoking and excessive drinking among teens is of particular concern. First, the health consequences of these behaviors are a function of exposure, so even just delaying initiation will also delay the onset of the adverse health consequences, and thus increase life expectancy. Second, delaying initiation of smoking reduces the likelihood of ever smoking (Gruber and Zinman, 2001, Auld, 2005).5 Finally, recent evidence from the medical and biology literature suggests that brain development, which is not complete during the teen years, is sensitive to alcohol and nicotine; thus drinking and smoking among teens may have more severe long-term impacts, even compared to such behaviors among adults.6 To the extent that adolescents aren’t fully informed of these costs or are unable to make fully rational decisions,7 preventing adolescents from engaging in these behaviors is a worthwhile public policy goal.

Our results show that in addition to increasing schooling and decreasing work, youths who received the treatment were significantly less likely to smoke 4 years later, and experienced daily or weekly drinking at a later age. These changes appear to be due to the effects of school and work on exposure to peers that drink and smoke as well as the amount of disposable income the youth has. The changes do not appear to be driven by any direct impact of school or work on rates of time preference, attitudes towards risk or perceptions of the adverse health consequences of these behaviors; however, our measures of these factors are limited and imperfect, so we cannot conclusively rule out such effects.

The remainder of this paper proceeds as follows. In Section 2, we discuss the data and experimental design. Section 3 presents the results and Section 4 discusses the limitations of our study and concludes.

Section snippets

Survey information

The sample was drawn in two stages. First, from the 30 largest cities and towns (representing about two-thirds of the population), we chose 150 sampling clusters at random, with the number of clusters chosen in each city or town approximately proportional to its share of the combined population of the 30 cities/towns. For each of the 150 clusters, we selected the school where students from that cluster attend 8th grade, the final year of primary school. From each school we selected 15 boys8

Empirical strategy

In order to explore the impacts of the intervention on drinking and smoking, we estimate regressions of the form,Yi=β0+β1×Treatmenti+Xiα+εiwhere Y is the outcome of interest for individual i, and Treatment is an indicator equal to one if the individual received the treatment. Standard errors are adjusted for clustering at the school level – the level of randomization. We present regressions with and without additional controls (father's education, school performance and log of family income).

Discussion and conclusion

We find that an intervention providing information on the market returns to schooling, which increased schooling and decreased work, lead to a reduced incidence of smoking at age 18 and a later onset of daily drinking. Our results confirm that at least part of the gradient between education and health-related behaviors among teens is indeed causal. These behavioral changes are likely to carry important private and social gains both directly in terms of health, as well as indirectly via

References (59)

  • M.C. Auld

    Causal effects of early initiation on adolescent smoking patterns

    Canadian Journal of Economics

    (2005)
  • G.S. Becker et al.

    The endogenous determination of time preference

    Quarterly Journal of Economics

    (1997)
  • U. Benzion et al.

    Discount rates inferred from decisions: an experimental study

    Management Science

    (1989)
  • H.P. Binswanger

    Attitudes toward risk: experimental measurement in rural India

    American Journal of Agricultural Economics

    (1980)
  • Bureau of Labor Statistics, U.S. Department of Labor, 2000. Report on the Youth Labor Force....
  • J. Carpenter et al.

    Behavioral development economics: lessons from field labs in the developing world

    Journal of Development Studies

    (2008)
  • F.J. Chaloupka et al.

    Price, tobacco control policies and youth smoking, NBER Working Paper 5740

    (1996)
  • C.F. Chabris et al.

    Individual laboratory-measured discount rates predict field behavior

    Journal of Risk and Uncertainty

    (2008)
  • S. Chan et al.

    What you don’t know can’t help you: pension knowledge and retirement decision-making

    Review of Economics and Statistics

    (2008)
  • Clark, D., Royer, H., 2009. The effect of education on adult mortality and health: evidence from England, mimeo,...
  • M. Coller et al.

    Eliciting Individual discount rates

    Experimental Economics

    (1999)
  • D. Cutler et al.

    Education and health: evaluating theories and evidence

  • A. Delvande et al.

    Measuring subjective expectations in developing countries: a critical review and new evidence

    Journal of Development Economics

    (2010)
  • O. Deschesnes

    Estimating the effect of family background on the returns to schooling

    Journal of Business and Economic Statistics

    (2007)
  • J. Dominitz et al.

    Eliciting student expectations of the returns to schooling

    Journal of Human Resources

    (1996)
  • C.M. Dormitzer et al.

    The Pacardo Research Project: Youthful Drug Involvement in Central America and the Dominican Republic

    Pan American Journal of Public Health

    (2004)
  • A.M. Dozier et al.

    Tobacco use in the Dominican Republic: understanding the culture first

    Tobacco Control

    (2006)
  • E. Duflo et al.

    The role of information and social interactions in retirement plan decisions: evidence from a randomized experiment

    Quarterly Journal of Economics

    (2003)
  • P. Dupas

    Do teenagers respond to HIV risk information? Evidence from a field experiment in Kenya

    American Economic Journal

    (2011)
  • Cited by (36)

    • Does better education mitigate risky health behavior? A mendelian randomization study

      2022, Economics and Human Biology
      Citation Excerpt :

      If the association between education and health behaviors reflects omitted variables or reverse causality, this implies that investments in education would not reduce poor health by altering or preventing unhealthy behaviors. Empirical studies have typically found a negative association between education and smoking status, but the results based on exogenous changes in schooling laws have been inconclusive, indicating either a negative (Arendt, 2005; Jensen and Lleras-Muney, 2012; Jürges et al., 2011) or zero effect (e.g., Braakmann, 2011; Clark and Royer, 2013; Dilmaghani, 2021; Kemptner et al., 2011; Li and Powdthavee, 2015). Similarly, higher education has been related to higher alcohol consumption (e.g., Cutler and Lleras-Muney, 2010), but this relationship may not be causal (Braakmann, 2011; Clark and Royer, 2013; Fletcher, 2015).

    • Unhealthy consumption behaviors and their intergenerational persistence: The role of education

      2020, China Economic Review
      Citation Excerpt :

      However, a large body of literature has documented mixed results regarding the link between education and health behavior. Many studies reveal that education has a negative effect on the probability of smoking or binge drinking (Cowell, 2006; Jensen & Lleras-Muney, 2012; Kemptner, Jurges, & Reinhold, 2011; Nayga Jr., 1999), suggesting that more educated people are likely to have better health status and are less likely to engage in unhealthy behaviors, while other studies cannot provide strong evidence that education leads to healthier behavior (Reinhold & Jurges, 2010; Xie & Mo, 2014; Zhong, 2015). Given the existing literature, however, the mechanism underlying the connection between education and unhealthy consumption behaviors is still not fully understood.

    • Educational gradient of health in rural China

      2019, Social Science Journal
      Citation Excerpt :

      Another strand of health economics research is to specify the multiple indicators of health. Although self-reported health is on the research agenda (e.g., Silles, 2009), the majority of current studies focus on the objective variables, such as health behavior (Jürges, Reinhold, & Salm, 2011; Li & Powdthavee, 2015), smoking (Jensen & Lleras-Muney, 2012; Maralani, 2013), mortality (Gathmann, Jürges, & Reinhold, 2015), hospitalization (Arendt, 2008), and so forth. What is also of interests for health economists is how parental education is linked to their children’s health conditions (Chou et al., 2010; Weitzman, 2017).

    • The Impact of Education on Health Outcomes and Behaviors in a Middle-Income, Low-Education Country

      2018, Economics and Human Biology
      Citation Excerpt :

      Behrman et al. (2015) relied on within-monozygotic twins fixed effects and showed that education improved mental health and reduced smoking in China. Exploiting a randomized controlled experiment in the Dominican-Republic, Jensen and Lleras-Muney (2012) showed that informing male students about returns to education lead them to stay in school longer, and enhanced pro-health behaviors (i.e., decreased smoking, delayed onset of drinking) by the age of 18.4 To fill the gap in the literature regarding the impact of education on health in developing countries, we use data from the Republic of Turkey.

    View all citing articles on Scopus

    We would like to thank Eric Driggs, Jason Fiumara, Zachary Jefferson, Magali Junowicz, Yesilernis Peña, Louisa Ramirez, Rosalina Gómez, Alexandra Schlegel and Paul Wassenich for valuable research assistance. Assistance and financial support from the Fundación Global Democracia y Desarrollo (FUNGLODE) and President Leonel Fernández is gratefully acknowledged. This work greatly benefitted from comments from Sherry Glied, Seema Jayachandran, Eduardo Ramos, seminar participants at UC Santa Barbara and an anonymous referee.

    View full text