Does education affect smoking behaviors

Journal of Health Economics 26 (2007) 877–895

Does education affect smoking behaviors?Evidence using the Vietnam draft as an instrument

for college education

Damien de Walque ∗Development Research Group, The World Bank, 1818 H Street, NW, Washington, DC 20433, USA

Received 24 September 2006; received in revised form 11 December 2006; accepted 14 December 2006Available online 30 December 2006

Abstract

This paper tests the hypothesis that education improves health and increases people’s life expectancy.It does so by analyzing the effect of education on smoking behaviors. To account for the endogeneity ofsmoking, the analysis develops an instrumental variable approach which relies on the fact that during theVietnam War college attendance provided a strategy to avoid the draft. The results indicate that educationdoes affect smoking decisions: educated individuals are less likely to smoke, and among those who initiatedsmoking, they are more likely to have stopped.© 2007 Elsevier B.V. All rights reserved.

JEL classification: I 1; I 12; I 2

Keywords: Health; Education; Smoking; Vietnam war

1. Introduction

Smoking is the leading cause of premature adult mortality. Each year in the United States,tobacco use causes more than 440,000 deaths and results in more than 5.5 million years of potentiallife lost.1 The decision to smoke or not to smoke is a conscious choice that directly affects the healthstatus and ultimately the mortality of individuals. It therefore provides an interesting opportunityto investigate how education, by influencing behaviors, affects health outcomes.

∗ Tel.: +1 202 473 2517; fax: +1 202 614 0234.E-mail address: [email protected].

1 http://www.cdc.gov/tobacco/issue.htm (accessed on September 13, 2006).

0167-6296/$ – see front matter © 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.jhealeco.2006.12.005

mailto:[email protected]

http://www.cdc.gov/tobacco/issue.htm

dx.doi.org/10.1016/j.jhealeco.2006.12.005

878 D. de Walque / Journal of Health Economics 26 (2007) 877–895

The strong correlation between education category and health outcomes, even after controllingfor income, has been recognized as a robust empirical observation in the social sciences andeconomic literature (Deaton and Paxson, 2003; Fuchs, 1982; Grossman, 2006; Lleras-Muney,2005). This paper examines whether the effect of education on smoking behaviors and thereforeon improving health outcomes can be considered as causal.

The question of the causality of the effect of education is central in the economics literature onthe interaction between health and education. Explanations for the correlation between educationand health have been diverging. On the one hand, Grossman (1972) develops a model based oneducation entering as a factor in the health production function and emphasizes that educationimproves the access to health-related information and the processing of that information to makehealth-related decisions. In that model, the effect of education is causal. On the other hand, Fuchs(1982) and Farrell and Fuchs (1982) claim that the observed correlation between health andeducation is mainly due to unobservables, like the discount factor or the ability that causes thesame individuals both to study longer and to take greater care of their health.

The literature about the negative correlation between education and tobacco use has become akey element in the debate about whether education is causal in improving health.2 Several studieshave used instrumental variables techniques to assess causality. A recent paper by Currie andMoretti (2003) investigates the impact of maternal education on birth weight and on smokingbehavior during pregnancy using the dates of college openings as an instrument for maternaleducation. They find a positive effect of education on birth weight. They also obtain a negativeeffect of schooling on smoking during pregnancy, suggesting that this is the pathway leading tohigher birth weights. In both cases, they find instrumental variables which are higher than ordinaryleast squares.

A number of studies have used compulsory schooling laws as an instrument for education.Lleras-Muney (2005) uses changes in compulsory schooling ages in the United Stated in order toestimate the relationship between schooling and mortality using synthetic cohorts of the census.She also finds larger estimates when using the instrumental variable strategy. Adams (2002)obtains a similar result when he uses the same instrument to estimate the effect of education onself-rated health and functional ability. Arendt (2005) follows the same strategy using data andschool age reforms in Denmark. He also concludes that education has a positive impact on self-rated health. While Oreopoulos (2006) focuses mainly on the estimation of the effect of educationon earnings using variations in compulsory schooling as an instrument, he also presents resultsshowing a positive impact of education on health both in data from the United Kingdom and theUnited States.

This paper develops an instrumental variable approach that uses the fact that during the VietnamWar college attendance could serve a draft avoidance strategy.3 The results indicate that educationdoes affect smoking decisions: educated individuals are less likely to smoke, and among thosewho initiated smoking, they are more likely to have stopped.

In parallel research, Grimard and Parent (2006) develop a comparable strategy using a differentdata set. In their paper, they define their instrument differently, using a birth cohort dummyfor males rather than the actual values for the risk of induction. They obtain the result thateducation matters for smoking initiation but not for smoking cessation. They do not considercurrent smoking.

2 See Berger and Leigh (1989), Chaloupka and Warner (2000), Currie and Moretti (2003), Farrell and Fuchs (1982),Fuchs (1982), Grossman (1975), Kenkel (1991), Meara (2001), Sander (1995), Viscusi (1990, 1992).

3 This instrumental variable strategy was already developed in my dissertation (see de Walque, 2003).

D. de Walque / Journal of Health Economics 26 (2007) 877–895 879

Fig. 1. Percentage of individuals who are currently smoking by education level and gender. Smoking prevalence rates arecalculated for individuals born between 1937 and 1956 from the 1983, 1985, 1987, 1988, 1990, 1991, 1992, 1994 and1995 National Health Interview Surveys.

The paper is structured as follows. Section 2 presents descriptive statistics and the ordinaryleast square results. Section 3 contains the results of the instrumental variable strategy. Section 4concludes.

2. Data description and ordinary least square results

This paper uses nine smoking supplements from different years of the National Health Inter-view Survey between 1983 and 19954 to construct detailed smoking histories of each individual.In order to allow the comparability with the results of the instrumental variable approach pre-sented later, only individuals born between 1937 and 1956 are considered. To make sure thatthe education variable is not time sensitive, and given that most people will have reached theirdefinitive education level by age 25, I have sampled only individuals aged 25 and above.

Figs. 1 and 2 illustrate the strong correlation between educational achievement and smokingbehavior present in the sample. Fig. 1 shows that smoking prevalence, i.e. the proportion ofindividuals who were smoking at the time of the interview is declining strongly with educationlevels. While 55.1% of males with less than a high school degree were smoking at the time of theinterview, only 19.9% of males with a least a college degree were current smokers. Females wereoverall less likely to smoke, but the negative gradient between education and current smoking isalmost as strong for them. For both genders, the gradient is stronger at the college margin. Fig. 2reports similar statistics for the proportion of individuals who started and then stopped smoking.

4 The nine national surveys are from 1983, 1985, 1987, 1988, 1990, 1991, 1992, 1994 and 1995. (Department of Healthand Human Services, 1983, 1985, 1987, 1988, 1990, 1991, 1992, 1994 and 1995). The 1978, 1979 and 1980 surveyswould have included individuals under age 25 for whom the educational level might not have been definitive. The publicfile of the 1997, 1998, 1999 and 2000 surveys did not contain the veteran status of the individuals, a variable that is crucialfor the implementation of the instrumental variable approach.


Fig. 2. Percentage of individuals who started and then stopped smoking by education level and gender. Percentages ofindividuals who stopped smoking are calculated for individuals born between 1937 and 1956 from the 1983, 1985, 1987,1988, 1990, 1991, 1992, 1994 and 1995 National Health Interview Surveys.

Education increases greatly the likelihood that individuals quit smoking: while only 26.2% ofmales high school drop-outs who ever started smoking have subsequently stopped, 57.6% ofcollege graduates have done so. The positive gradient between education and smoking cessationis very similar among females.

Table 1 reports results from ordinary least square regressions where the dependent variable iseither an indicator for whether the individual is currently smoking (odd columns), or whether,among those who ever smoked, he did quit smoking (even columns). The regressions use a simplelinear probability model.

Because, as documented in Fig. 1, it is at the college margin that education seems to have thestrongest effect in decreasing smoking prevalence, the measure of education used in columns 1–6of Table 1 is the number of years of education above high school, so that individuals with nohigh school degree or just an high school degree are merged in the same category. In columns7 and 8, the measure of education is an indicator for whether or not the individual is a collegegraduate or more. Columns 1 and 2 control for gender, race, family income, age at interview,birth cohort and the date of the interview. The variable “trendmale” is included to account forthe fact that over time, the positive difference in smoking prevalence between males and femaleshas been decreasing (see de Walque, 2003). The measure of income is the logarithm of familyincome.5 The results suggest that 1 year of college education decreases smoking prevalenceby 4.0 percentage points and increases the probability of smoking cessation by 4.1 percentagepoints. These estimates, which are strongly significant, are consistent in magnitude with raw meancomparisons (see Figs. 1 and 2).

5 The National Health Interview Survey reports family income in $1000 brackets between $1000 and $20,000 and in$5000 brackets between $20,000 and $50,000. The variable is top-coded at $50,000. Since the data comes from surveysranging from 1983 to 1995, all reported incomes have been transformed in 1983 dollars using the inflation calculator ofthe Bureau of Labor Statistics.

D.de

Walque

/JournalofHealth

Econom

ics26

(2007)877–895

881Table 1Current smoking and smoking cessation for individuals born between 1937 and 1956: ordinary least square results

1 2 3 4 5 6 7 8Dependent variable Smoking now Stopped smoking Smoking now Stopped smoking Smoking now Stopped smoking Smoking now Stopped smoking

Years of education abovehigh school

−0.04*** 0.041*** −0.047*** 0.05*** −0.04*** 0.041*** – –[−10.53] [19.35] [−14.12] [19.87] [−11.08] [20.11] – –

College and above – – – – – – −0.173*** 0.178***

– – – – – – [−11.68] [18.56]

Male 0.102*** 0.048*** 0.073*** 0.063*** 0.079*** 0.051*** 0.077*** 0.052***

[14.56] [10.56] [12.08] [14.16] [11.16] [9.79] [10.89] [9.59]

Trendmale −0.001** −0.003*** −0.0002 −0.003*** −0.0003 −0.003*** −0.0003 −0.003***

[−2.14] [−7.44] [−0.53] [−9.41] [−0.84] [−7.62] [−0.69] [−7.68]

Log family income in1983$

−0.076*** 0.084*** – – −0.077*** 0.084*** −0.084*** 0.091***

[−27.94] [25.10] – – [−26.59] [25.37] [−22.52] [28.31]

Vietnam veteran – – 0.061*** −0.004 0.063*** −0.005 0.059*** −0.001– – [5.83] [−0.60] [5.94] [−0.73] [5.59] [−0.24]

Black −0.002 −0.051*** 0.036*** −0.095*** −0.0009 −0.051*** −0.0005 −0.052***

[−0.32] [−8.66] [5.65] [−16.69] [−0.13] [−8.75] [−0.06] [−9.05]

Other race −0.07** −0.014 −0.051 −0.039** −0.066** −0.015 −0.068** −0.01[−2.63] [−1.10] [−1.60] [−2.71] [−2.34] [−1.12] [−2.54] [−0.78]

R-square 0.068 0.09 0.055 0.074 0.069 0.09 0.065 0.08Observations 74810 40753 83069 45108 73952 40319 73957 40323

Analysis from smoking histories from the 1983, 1985, 1987, 1988, 1990, 1991, 1992, 1994 and 1995 National Health Interview Surveys for individuals age 25 and above atthe time of the survey. Individuals born between 1937 and 1956. Linear regression model, where the dependent variable is current smoking behavior (odd columns) or smokingcessation for individuals who ever smoked (even columns). Robust t-statistics, clustered by birth cohort and gender, in square brackets. Birth cohort, age at interview and surveydummies included in all specifications. Trendmale, a dummy for male multiplied by a yearly trend where born in 1937 = 0, accounts for the trend in the difference in smokingbehavior between males and females. See footnote 5 for details on the family income variable. Sample sizes vary across specifications as the Vietnam veteran status and theincome variable might have been missing.

*Denote statistical significance at the 10% confidence level.** Denote statistical significance at the 5% confidence level.

*** Denote statistical significance at the 1% confidence level.


Columns 3 and 4 further control for the Vietnam veteran status of the individuals, but with-out controlling for income: while the coefficient on education remain in the same range, theresults indicate that being a Vietnam veteran increases by 6.1 percentage points the probabilityof being a smoker, even after controlling for education, gender, race, age, birth cohort and yearof interview. This is a strong effect and will be discussed further, together with the instrumentalvariable approach. However, being a Vietnam veteran does not affect the probability of smokingcessation.

Columns 5 and 6 include controls for both income and actual Vietnam veteran status. Theresults indicate that, while income has a very significant negative effect on smoking prevalenceand a positive one on the likelihood to have stopped smoking,6 the negative effect of educationon smoking is very robust to the inclusion of income as a control. There is, in the educationvariable, something more than the ability to generate a higher income, which explains the smokingdecision.

Columns 7 and 8 uses the indicator for college graduate or more as an alternative measure ofeducation and yields consistent results, since the coefficient on the dummy for college graduateis roughly four-times larger than the coefficient on 1 year of college education, both for smokingprevalence and the probability of quitting.

The main conclusion from these ordinary least square regressions is that the effect of collegeeducation in reducing smoking prevalence and increasing the probability of smoking cessationis robust even after controlling for gender, race, age, birth cohort, date of interview and income.However, these results, if they lead to the conclusion that the effect of education goes over andabove the ability to generate an higher income, do not allow one to reject the hypothesis advancedby Fuchs (1982) that the fact that more educated individuals are less likely to smoke is due toselection on unobservables, like the discount factor.

Evidence from a plausibly exogenous variation in education would strengthen the argumentthat the effect of education in reducing smoking prevalence is causal. This motivates the imple-mentation of an instrumental variable approach that exploits the fact that college attendance hasbeen used as a draft avoidance strategy during the Vietnam War.

3. An instrumental variable approach: college attendance as a draft avoidance strategy

During the Vietnam War, some males enrolled in college to avoid the draft. Card and Lemieux(2000, 2001), by contrasting enrollment for males and females and the deviations from the trendin male/female ratio, estimate that this effect increased college graduation rates for males by 2percentage points. They construct a measure of the induction risk faced by each birth cohort ofmales: the average yearly number of inductions in Vietnam during the years when a particularbirth cohort was aged 19–22 divided by the size of the birth cohort. Based on their paper, I havereconstructed this measure.7 It is plotted in Figs. 3 and 4. Of course, for females, the inductionrisk is always zero.

6 Introducing the measure of current income in a regression of smoking cessation where, typically, the decision to quitsmoking has been taken in the past, might seem peculiar, unless current income is considered as a proxy for permanentincome. To maintain the parallelism with the regressions on current smoking behavior, this specification is still presented(columns 2 and 6), but a specification without income as a control (column 4) is also available.

7 The number of inductions per semester is from appendix Table 1 in Card and Lemieux (2000). The size of the cohortis taken from Department of Education (2000).


Fig. 3. College graduation rates by gender and birth cohort and risk of induction in Vietnam, by birth cohort. Collegegraduation rates are calculated for individuals born between 1937 and 1956 from the 1983, 1985, 1987, 1988, 1990, 1991,1992, 1994 and 1995 National Health Interview Surveys. The risk of induction in Vietnam is the average yearly numberof inductions in Vietnam during the years when that birth cohort was aged 19–22, divided by the size of that birth cohort,calculated from data in Card and Lemieux (2000) and Department of Education (2000).

Fig. 4. Risk of induction in Vietnam and risk of induction in Vietnam multiplied by the risk of being killed in action, bybirth cohort. The risk of induction in Vietnam is the average yearly number of inductions in Vietnam during the yearswhen that birth cohort was aged 19–22, divided by the size of that birth cohort, calculated from data in Card and Lemieux(2000) and Department of Education (2000). The risk of being killed in action is the ratio between the number of soldierskilled in action in Vietnam in a year and the level of troops during that year (see footnotes 11 and 12 for details and datasources).


3.1. Results

This measure of the risk of induction is used as an instrument for the effect of college educationon smoking prevalence in Table 2 , which contains the results from the first stage and instrumentalvariable regressions.8 The format is the same as in Table 1.

The first-stage regressions in the upper part of Table 2 indicate that the risk of induction isalways positively and significantly correlated with college education, indicating that males inbirth cohort that experienced a higher exposure to the risk of being drafted in the army duringthe Vietnam War were more likely to remain longer in college. Graphically, the effect of thedraft is illustrated in Fig. 3 in which the measure of the risk of induction is plotted together withmales and females college graduation rates, computed from the sample. The highest level of theinduction risk corresponds with the cohorts where the male college graduation rate is peaking,both in absolute terms and relative to the female college graduation rate. The instrument is ineffect defined at the birth cohort and gender level. In addition to the effect of the exposure tothe Vietnam draft, the instrument could thus pick birth-cohort-specific effects or secular trendsin differences between genders. In order to avoid this, all regressions include dummies for eachbirth cohort, a dummy for male and a trend accounting for the secular changes in males/femalesratios. The robust standard errors are also clustered by birth cohort and gender, as suggested byMoulton (1990).

The instrumental variable results are displayed in the lower part of Table 2. In columns 1 and2, no controls for Vietnam veteran status are included. The results of this instrumental variableregression (positive and significant effect of college education on smoking, but not significant onsmoking cessation) are different than the ordinary least square results (college education decreasessignificantly smoking prevalence and increases significantly smoking cessation). However, therisk of induction in the army, which is determined by the birth cohort, could also affect smokingdirectly and not only via education. In this case, the risk of induction would not be a validinstrument.

One potential channel through which the risk of induction can be directly correlated with smok-ing behavior is that the risk of induction is directly correlated with the probability of becoming aVietnam veteran. The ordinary least square results reported in Table 1 all show a strong positiveeffect of being a Vietnam veteran on smoking prevalence (but not on smoking cessation).9 With-out controlling for actually being a veteran, the instrument of the induction risk picks both theeffect of attending college (decreases smoking) and the probability of being a veteran (increasessmoking).

The National Health Interview Surveys, however, include the information on the actual veteranstatus of the individual, so the direct effect of being a Vietnam veteran on smoking behavior can beisolated. Columns 3–8 of Table 2, therefore control for actual Vietnam veteran status. Columns 3and 4 do not control for income.10 In column 3, the instrumental variable estimate of the effect of

8 Notice that this instrument is different than the Vietnam-Era Draft Lottery used by Angrist (1990) and Angrist andKrueger (1992). The draft lottery was in effect between 1970 and 1973, when the number of inductions was alreadydeclining.

9 Bedard and Deschenes (2006) also show that veterans are more likely to smoke.10 It can be argued that income should not be included in the first stage regressions, since current income is more likely

to be the consequence than the cause of educational achievement. Tables 2 and 3 both present results with and withoutincome included as a control variable in the regressions. Both specifications yield very similar results (compare columns3–4 with columns 5–6 in both tables).

D.de

Walque

/JournalofHealth

Econom

ics26

(2007)877–895

885

Table 2Instrumental variable estimation: instrument for years of education is the risk of induction in Vietnam

First-stage regression

Dependent variable Years of education above high school College and above

1 2 3 4 5 6 7 8

Induction risk 1.095*** 0.742*** 1.64*** 0.988*** 1.779*** 1.227*** 0.399*** 0.284***

[4.86] [2.64] [7.24] [3.52] [7.68] [4.25] [8.07] [4.67]


Instrumental variable regression

Dependent variable Smoking now Stopped smoking Smoking now Stopped smoking Smoking now Stopped smoking Smoking now Stopped smoking

Years of educationabove high school

0.076* 0.028 −0.06*** 0.052 −0.038* 0.051 – –[1.83] [0.61] [−3.10] [1.38] [−1.88] [1.62] – –

College and above – – – – – – −0.169* 0.219– – – – – – [−1.82] [1.57]

Male 0.058*** 0.051*** 0.081*** 0.063*** 0.078*** 0.049*** 0.077*** 0.05***

[3.64] [6.00] [6.98] [5.35] [6.83] [6.35] [6.89] [6.92]

Trendmale 0.00003 −0.003*** −0.0004 −0.003*** −0.0003 −0.003*** −0.0003 −0.003***

[0.05] [−7.80] [−0.85] [−8.94] [−0.64] [−7.64] [−0.55] [−8.06]


−0.178*** 0.093** – – −0.078*** 0.077*** −0.085*** 0.086***

[−4.71] [2.66] – – [−4.14] [3.22] [−5.21] [4.47]

Vietnam veteran – – 0.057*** −0.003 0.063*** −0.003 0.06*** 0.001– – [7.16] [−0.56] [7.57] [−0.49] [6.97] [0.12]

Black 0.024 −0.053*** 0.027* −0.094*** −0.0005 −0.05*** −0.0003 −0.05***

[1.24] [−6.44] [1.73] [−4.14] [−0.06] [−6.59] [−0.03] [−7.19]

886D

.deW

alque/JournalofH

ealthE

conomics

26(2007)

877–895

Table 2 (Continued )


Dependent variable Smoking now Stopped smoking Smoking now Stopped smoking Smoking now Stopped smoking Smoking now Stopped smoking

Other race −0.145*** −0.008 −0.044 −0.039* −0.067** −0.019 −0.069 −0.013[−4.85] [−0.28] [−1.41] [−1.77] [−2.29] [−0.84] [−2.45] [−0.69]

R-square (.) 0.088 0.052 0.074 0.068 0.089 0.065 0.084Observations 74810 40753 83069 45108 73952 40319 73957 40323

Analysis from smoking histories from the 1983, 1985, 1987, 1988, 1990, 1991, 1992, 1994 and 1995 National Health Interview Surveys for individuals age 25 and above atthe time of the survey. Individuals born between 1937 and 1956. Linear regression model, where the dependent variable is current smoking behavior (odd columns) or smokingcessation for individuals who ever smoked (even columns). Robust t-statistics, clustered by birth cohort and gender, in square brackets. Birth cohort, age at interview and surveydummies included in all specifications. Trendmale, a dummy for male multiplied by a yearly trend where born in 1937 = 0, accounts for the trend in the difference in smokingbehavior between males and females. See footnote 5 for details on the family income variable. The instrument for education is the induction risk in Vietnam faced by each birthcohort of males: it is the average yearly number of inductions in Vietnam during the years when that birth cohort was aged 19–22, divided by the size of that birth cohort. Seefootnote 7 for the data source for the instrument. Sample sizes vary across specifications as the Vietnam veteran status and the income variable might have been missing. (.)Denotes negative R-square.

* Denote statistical significance at the 10% confidence level.** Denote statistical significance at the 5% confidence level.



college education on smoking prevalence is negative and significant, and is somewhat larger thanthe ordinary least square estimate (−0.06 versus −0.047). The significance is lower in column4, when smoking cessation is analyzed. When controlling for family income, as in columns 5and 6, the instrumental variable estimates are very close to the ordinary least square estimatesand the actual mean differences in the raw data, but the result for smoking cessation is still onlyborderline significant. Comparable results are obtained in columns 7–8 with an indicator forcollege graduation as a measure of education.

Table 3 presents results using an alternative instrument: rather than only the risk of beingdrafted, I have created an instrument that also includes the risk of being killed in Vietnam. Theintuition behind this alternative instrument is that what might have caused individuals to enrollin college to avoid the draft is not so much the risk of being drafted, but the risk of being draftedand killed. I calculate the risk of being killed in action as the ratio between the number of soldierskilled in action in a year and the level of troops engaged in Vietnam during that year.11 Then, foreach year, I multiply the yearly induction rate with the probability to be killed in action.12

Fig. 4 compares the two risk measures used as instrument, the risk of induction and the riskof induction multiplied by the risk of being killed in action. Although these measures peak forthe same birth cohort, the slope is steeper when the probability of being killed in action is alsoconsidered. As reported in Table 3, when the risk of induction multiplied by the risk of beingkilled in action is used as an instrument (see columns 3–8, when actual Vietnam veteran status iscontrolled for), the results are qualitatively very similar to the results with the induction risk (seeTable 2) as instrument. However, the results are here always significant, even for the smokingcessation regressions. If anything, the point estimates of the coefficients are slightly higher thanthe ordinary least square results and the first instrumental variable strategy, especially in the caseof smoking cessation. It is especially important to establish the significance of the positive effectof college education on the decision to stop smoking because this is a “cleaner” margin to study therelationship between education and smoking. Smoking prevalence, the proportion of individualswho are currently smoking, is a stock determined by two flows: smoking initiation and smokingcessation. The median age of smoking initiation is 17, before most differences in educationalachievement take place, while the median age of smoking cessation is 37, when the schoolinglevels of most individuals can be seen as definitive. Interestingly, smoking cessation is also amargin where the actual Vietnam veteran status has no effect, as indicated by the results in theeven columns in Tables 1–3.

On the other hand, actual Vietnam veteran status has an important impact of the probabilityof being a current smoker, even after controlling for age, income, education, race and gender.The inclusion of this variable in the instrumental variable regressions appears to be crucial for itssuccess as indicated by the inconsistent results obtained when, as in columns 1 and 2, Vietnamveteran status is not controlled for.

11 The casualty data are taken from the website http://www.vietnamwall.org/casualty.html (accessed on May 11, 2004)and derived from the Combat Area Casualty File of 11/93 and the Adjutant General’s Center (TAGCEN) file of 1981,available from the National Archives. For the years 1961–1965, the number of killed in action (1864) has been evenlyspread over the 5 years, as no yearly details were available. The level of U.S. troops engaged in Vietnam for each year istaken from the website http://www.25thaviation.org/id430.htm (accessed on September 23, 2006).12 The final measure of the risk to be induced and killed for each birth cohort is calculated as the previous measure of

the risk of induction. I average the yearly number of inductions in Vietnam multiplied by the probability of being killedin action over the years when a particular birth cohort was aged 19–22, and divide this number by the size of the birthcohort.

http://www.vietnamwall.org/casualty.html

http://www.25thaviation.org/id430.htm

888D

.deW

alque/JournalofH

ealthE

conomics

26(2007)

877–895

Table 3Instrumental variable estimation: instrument for years of education is the risk of induction in Vietnam multiplied by the risk of being killed in action

First stage regression

Dependent variable Years of education above high school College and above

1 2 3 4 5 6 7 8

Induction risk* risk ofbeing killed in action

54.74*** 40.69*** 80.02*** 51.22*** 87.28*** 64.66*** 19.27*** 14.51***

[4.85] [2.87] [7.04] [3.62] [7.53] [4.45] [7.77] [4.73]



Dependent variable Smokingnow

Stoppedsmoking

Smokingnow

Stoppedsmoking

Smokingnow

Stoppedsmoking

Smokingnow

Stoppedsmoking

Years of education abovehigh school

0.069* 0.055 −0.059*** 0.068* −0.04** 0.066** – –[1.80] [1.29] [−3.11] [1.94] [−2.11] [2.24] – –

College and above – – – – – – −0.181* 0.297**

– – – – – – [−2.07] [2.24]

Male 0.061*** 0.046*** 0.081*** 0.056*** 0.079*** 0.044*** 0.078*** 0.045***

[3.91] [5.54] [6.98] [4.97] [7.25] [5.75] [7.23] [6.16]

Trendmale −0.00004 −0.003*** −0.0004 −0.003*** −0.0003 −0.003*** −0.0003 −0.003***

[−0.08] [−7.35] [−0.85] [−8.49] [−0.71] [−7.26] [−0.62] [−7.75]


−0.171*** 0.074** – – −0.077*** 0.077*** −0.083*** 0.076***

[−4.96] [2.31] – – [−4.31] [2.90] [−5.40] [4.16]

Vietnam veteran – – 0.058*** −0.001 0.062*** −0.003 0.059*** 0.006– – [7.28] [−0.16] [7.92] [−0.04] [7.43] [0.73]

Black 0.022 −0.049*** 0.027* −0.084*** −0.001 −0.047*** −0.0009 −0.048***

[1.20] [−5.32] [1.76] [−3.90] [−0.11] [−5.71] [−0.09] [−6.29]

D.de

Walque

/JournalofHealth

Econom

ics26

(2007)877–895

889

Other race −0.14*** −0.021 −0.045 −0.046 −0.066 −0.027 −0.067 −0.02[−4.79] [−0.74] [−1.42] [−2.01]* [−2.25]** [−1.15] [−2.40]** [−1.00]

R-square (.) 0.087 0.053 0.069 0.069 0.081 0.065 0.076Observations 74810 40753 83069 45108 73952 40319 73957 40323

Analysis from smoking histories from the 1983, 1985, 1987, 1988, 1990, 1991, 1992, 1994 and 1995 National Health Interview Surveys for individuals age 25 and above atthe time of the survey. Individuals born between 1937 and 1956. Linear regression model, where the dependent variable is current smoking behavior (odd columns) or smokingcessation for individuals who ever smoked (even columns). Robust t-statistics, clustered by birth cohort and gender, in square brackets. Birth cohort, age at interview and surveydummies included in all specifications. Trendmale, a dummy for male multiplied by a yearly trend where born in 1937 = 0, accounts for the trend in the difference in smokingbehavior between males and females. See footnote 5 for details on the family income variable. The instrument for education is the induction risk in Vietnam faced by each birthcohort of males multiplied by the risk of being killed in action. The risk of induction is the average yearly number of inductions in Vietnam during the years when that birthcohort was aged 19–22, divided by the size of that birth cohort (see footnote 7 for the data source). The risk of being killed in action is the ratio between the number of soldierskilled in action in Vietnam in a year and the level of troops during that year (see footnotes 11 and 12 for details and data sources). Sample sizes vary across specifications as theVietnam veteran status and the income variable might have been missing.




However, since the first stage of the instrumental variable strategy implies that it was possibleto avoid, or at least to delay, the draft by staying in college, it might be argued that Vietnamveteran status is endogenous. To address the issue of the potential endogeneity of veteran status,it is necessary to instrument both for education and veteran status. Since I have now two variablesto be instrumented (education and veteran status), I created a set of dummies13 for each value ofthe risk of induction rather than entering it as one variable.

The panel A of Table 4 presents results of regressions where I instrument for education only(columns 1 and 2), for veteran status only (columns 3 and 4) and both for education and veteranstatus (columns 5 and 6), using the set of risk of induction dummies as an instrument. Panel B ofTable 4 implements the same specifications but uses a set of dummies for the risk of inductionmultiplied by the risk of being killed in action, following the idea of Table 3.14 Both currentsmoking behavior and the quitting decision are considered as dependent variables. When currentsmoking is the dependent variable, I do find that the effect of college education on smokingis negative and significant and that the effect of veteran status is positive and significant. Withsmoking cessation as the dependent variable, the coefficient on education is always positive andsignificant (like in Table 3, but unlike Table 2 where it is positive but not significant), but like in theordinary least square results, there is no impact of Vietnam veteran status on smoking cessation.

It is important to find that the results of the analysis yield the same conclusions when onlyeducation is considered as endogenous, when only Vietnam veteran status is endogenous or whenboth variables are instrumented for. However, it should be noted that the magnitudes of the pointestimates vary depending on which variable is instrumented.

3.2. Interpretation of the instrumental variables estimates

The instrumental variable strategy implemented in this section uses the fact that college enroll-ment as a draft avoidance strategy during the Vietnam War era provides a plausibly exogenousvariation in educational attainment. During the Vietnam War, going to college was associatedwith an additional benefit: avoiding the draft and the disutility associated with it. This additionalbenefit induced some individuals that would otherwise not have enrolled in college to enroll.

It is however likely that the individuals induced to enroll in college because of the Vietnamdraft were not random among the group of individuals who, in the absence of the draft, would nothave gone to college. On the contrary, it is reasonable to expect that those males who decided toenroll because of the draft were the ones who were close (just under) to the margin of going tocollege. The possibility to avoid the draft if they were in college made them switch and decide toenroll in college. Typically, the “switchers” would have a lower discount rate than the individualswho did not enroll, despite the extra benefit of draft avoidance, but a higher discount rate than theones who would have gone to college even in the absence of the draft.

It then follows that, if there is heterogeneity in the returns to schooling, the instrumental variableestimate obtained is not a consistent estimate of the average effect of education on smoking in thepopulation (Card, 1995, 2001; Heckman et al., 1999). The instrumental variable estimate should

13 Notice that this set of dummies is not equivalent to the set of birth cohort dummies. First, all females are grouped inthe dummy with value zero. Also some birth cohorts share the value zero: there are 18 induction risk dummies for 20birth cohorts.14 Panels A and B of Table 4 yield very similar results. This is not a surprise given that once the induction risk and

the induction risk multiplied by the risk of being killed in action become a set of dummies they have almost the samedistribution (18 dummies in panel A and 17 dummies in panel B).

D.de

Walque

/JournalofHealth

Econom

ics26

(2007)877–895

891

Table 4

Education endogenous Vietnam veteran statusendogenous

Education and Vietnam veteranstatus both endogenous

1 2 3 4 5 6Dependent variable Smoking

nowStoppedsmoking

Smokingnow

Stoppedsmoking

Smokingnow

Stoppedsmoking

(A) Instrumental variable estimation: instrument for years of education and Vietnam veteran status is a set ofdummies for the risk of induction in VietnamYears of education above high school −0.061*** 0.099*** −0.040*** 0.041*** −0.095*** 0.109***

[−4.09] [5.56] [−11.20] [20.89] [−4.89] [6.61]

Male 0.089*** 0.035*** 0.078*** 0.042*** 0.088*** 0.037***

[8.10] [4.94] [8.58] [7.06] [7.81] [4.86]

Trendmale −0.0006 −0.003*** −0.0003 −0.003*** −0.0005 −0.003***

[−1.21] [−6.45] [−0.66] [−7.73] [−0.95] [−6.52]

Log family income in 1983$ −0.058*** 0.042*** −0.077*** 0.083*** −0.029* 0.035***

[−4.07] [3.09] [−25.93] [25.69] [−1.74] [2.85]

Vietnam veteran 0.056*** 0.006 0.065*** 0.018 0.095*** −0.005[6.02] [0.073] [4.22] [1.09] [5.46] [−0.36]

Black −0.005 −0.042*** −0.0009 −0.051*** −0.013* −0.041***

[−0.76] [−4.29] [−0.12] [−8.67] [−1.96] [−3.76]

Other race −0.053* −0.042* −0.066** −0.013 −0.029 −0.049**

[−1.85] [−1.96] [−2.31] [−1.04] [−0.85] [−2.15]

R-square 0.061 0.042 0.069 0.090 0.014 0.024Observations 73952 40319 73952 40319 73952 40319

892D

.deW

alque/JournalofH

ealthE

conomics

26(2007)

877–895

Table 4 (Continued )

Education endogenous Vietnam veteran statusendogenous

Education and Vietnam veteranstatus both endogenous

1 2 3 4 5 6Dependent variable Smoking

nowStoppedsmoking

Smokingnow

Stoppedsmoking

Smokingnow

Stoppedsmoking

(B) Instrumental variable estimation: instrument for years of education and Vietnam veteran status is a set ofdummies for the risk of induction in Vietnam multiplied by the risk of being killed in actionYears of education above −0.060*** 0.100*** −0.040*** 0.041*** −0.090*** 0.110***

High school [−4.07] [5.37] [−11.16] [20.89] [−4.69] [6.50]

Vietnam veteran 0.056*** 0.006 0.063*** 0.019 0.090*** −0.005[6.04] [0.077] [4.10] [1.13] [5.25] [−0.32]

R-square 0.062 0.040 0.069 0.090 0.024 0.021Observations 73952 40319 73952 40319 73952 40319

Analysis from smoking histories from the 1983, 1985, 1987, 1988, 1990, 1991, 1992, 1994 and 1995 National Health Interview Surveys for individuals age 25 and above atthe time of the survey. Individuals born between 1937 and 1956. Linear regression model, where the dependent variable is current smoking behavior (odd columns) or smokingcessation for individuals who ever smoked (even columns). Robust t-statistics, clustered by birth cohort and gender, in square brackets. Birth cohort, age at interview and surveydummies included in all specifications. Trendmale, a dummy for male multiplied by a yearly trend where born in 1937 = 0, accounts for the trend in the difference in smokingbehavior between males and females. See footnote 5 for details on the family income variable. In panel A, the instrument is a set of dummies for the values of the induction riskin Vietnam faced by each birth cohort of males: it is the average yearly number of inductions in Vietnam during the years when that birth cohort was aged 19–22, divided bythe size of that birth cohort (see footnote 7 for the data source). In panel B, the instrument for education is the induction risk in Vietnam faced by each birth cohort of malesmultiplied by the risk of being killed in action. The risk of induction is defined like in panel A. The risk of being killed in action is the ratio between the number of soldierskilled in action in Vietnam in a year and the level of troops during that year (see footnotes 11 and 12 for details and data sources).




rather be interpreted as the effect of education for the subgroup of males that has been inducedto enroll in college in order to avoid the Vietnam Draft. Imbens and Angrist (1994) define thisestimator as the local average treatment effect (LATE).

This qualification in the interpretation of the results is important, as it does not permit gener-alizing the estimates obtained from the instrumental variable estimates for the entire population.However, an estimation of the effect of education on smoking for those individuals who are atthe margin between going to college or not is interesting from a policy perspective, as this is thegroup that is likely to be induced to enroll by other interventions such as tuition subsidies.

4. Conclusions

The strong correlation between health and education is a well-known fact, but showing thatthis correlation comes from a causal impact of education on health has been more challenging.Theoretical explanations for the correlation can be classified into three broad categories. Oneexplanation stresses that education is an investment. Education will deliver a higher income, ahigher consumption level in the future, and raises the value of staying alive (Becker, 1993). Moreeducated individuals are healthier because their investment in the future gives them the rightincentives to protect their health. A complementary explanation, based on education entering as afactor in the health production function (Grossman, 1972), emphasizes that education improves theaccess to health-related information and the processing of that information to make health-relateddecisions. But, a third view (Farrell and Fuchs, 1982; Fuchs, 1982) claims that the observed corre-lation between health and education is mainly due to unobservables, like the discount factor or theability that causes the same individuals both to study longer and to take greater care of their health.

This paper uses the effect of education on smoking behaviors as an example where the corre-lation between education and health-related decisions is strong and well documented. To addressthe causality issue, this study isolates an exogenous variation in college education by exploitingthe fact that college attendance was used as a draft avoidance strategy during the Vietnam Warperiod. Using the risk of induction in Vietnam as an instrument for college education, and control-ling for actual Vietnam veteran status, instrumental variable estimates of the effect of educationon smoking behavior lead to the conclusion that education reduces significantly current smoking.For smoking cessation, the specification which uses the risk of induction multiplied by the risk ofbeing killed in action yield a positive and significant effect of college education on the probabilityto quit smoking. It is important to establish the significance of the positive effect of educationon the decision to stop smoking. Indeed, smoking cessation is more likely to occur when theschooling levels of most individuals can be seen as definitive. The results on current smoking andon smoking cessation are confirmed when Vietnam veteran status is considered as endogenous.

Rates of return to education, as traditionally calculated, only account for labor market earnings.If the effect of education on health and longevity is, as in the case of smoking reduction, causal, itmakes sense to attribute additional returns to education. I have proposed (de Walque, 2003, 2004)a method to estimate these extra returns in the context of the HIV/AIDS epidemic in Africa.Compared to the HIV/AIDS case, the potential addition to the returns to education is significantbut modest in the case of smoking, essentially because smoking kills at relatively late ages.

Acknowledgements

The author thanks Gary Becker, Mark Duggan, Michael Greenstone, Steven Levitt and TomasPhilipson for useful comments, suggestions and encouragements. I am grateful to Daniel Parent for


suggesting the Vietnam draft as an instrumental variable for college education. I also thank RaphaelDe Coninck, Michael Grossman, Ted Joyce, Don Kenkel, Darius Lakdawalla, Fabian Lange, ChrisRohlfs and participants of the Applications Workshop (May 5, 2003) at the University of Chicago,the NBER Health Economics seminar, the 24th Arne Ryde Symposium on the Economics ofSubstance Use at Lund University (Sweden), the Econometric Society World Congress (UniversityCollege London) 2005, the AEA Meetings (Boston) 2006 and Lehigh University for helpfulcomments and discussions. The findings, interpretations and conclusions expressed in this paperare those of the author and do not necessarily represent the views of the World Bank, its ExecutiveDirectors, or the governments they represent.

References

Adams, S.J., 2002. Educational attainment and health: evidence from a sample of older adults. Education Economics 10,97–109.

Angrist, J.D., 1990. Lifetime earnings and the Vietnam era draft lottery: evidence from Social Security AdministrativeRecords. American Economic Review 80 (3), 313–336.

Angrist, J.D., Krueger, A.B., 1992. Estimating the Payoff to Schooling Using the Vietnam-Era Draft Lottery. NBERWorking Paper 4067. Cambridge, MA.

Arendt, J.N., 2005. Does education cause better health? A panel data analysis using school reform for identification.Economics of Education Review 24, 149–160.

Becker, G.S., 1993. Human Capital: A Theoretical and Empirical Analysis with Special Reference to Education, third ed.The University of Chicago Press, Chicago, IL.

Bedard, K., Deschenes, O., 2006. The long-term impact of military service on health: evidence from World War II andKorea Veterans. American Economic Review 96 (1), 176–194.

Berger, M.C., Leigh, J.P., 1989. Schooling, self-selection and health. The Journal of Human Resources 24 (3), 433–455.Card, D., 1995. Earnings, schooling, and ability revisited. Research in Labor Economics 14, 23–48.Card, D., 2001. Estimating the return to schooling: progress on some persistent econometric problems. Econometrica 69

(5), 1127–1160.Card, D., Lemieux, T., 2000. Going to college to avoid the draft: the unintended legacy of the Vietnam war. Mimeo.Card, D., Lemieux, T., 2001. Draft Avoidance and College Attendance: The Unintended Legacy of the Vietnam War.

American Economic Review Papers and Proceedings, vol. 91, pp. 97–102.Chaloupka, F.J., Warner, K.E., 2000. The economics of smoking. In: Newhouse, J., Culyer, A. (Eds.), Handbook of Health

Economics 1A. North-Holland, Amsterdam.Currie, J., Moretti, E., 2003. Mother’s education and the intergenerational transmission of human capital: evidence from

college openings. Quarterly Journal of Economics 118 (4), 1495–1532.Deaton, A., Paxson, C., 2003. Mortality, income, and income inequality over time in Britain and the United States. In:

Wise, D.A. (Ed.), Perspectives on The Economics of Aging, 8. University of Chicago Press for NBER: Chicago, IL,Chicago, IL.

Department of Education, 2000. Digest of Education Statistics. UGSPO, Washington, DC.Department of Health and Human Services. National Health Interview Survey, 1983, 1985, 1987, 1988, 1990, 1991, 1992,

1994 and 1995. National Center for Health Statistics, Hyattsville, MD.de Walque, D., 2003. How Do Education and Information Affect Health Decisions? The Cases of HIV/AIDS and Smoking.

Ph.D. Dissertation. University of Chicago, Department of Economics.de Walque, D., 2004. How Does the Impact of an HIV/AIDS Information Campaign Vary with Educational Attainment?

Evidence from Rural Uganda. Policy Research Working Paper 3289. World Bank, Washington DC.Farrell, P., Fuchs, V., 1982. Schooling and health: the cigarette connection. Journal of Health Economics 1, 217–230.Fuchs, V., 1982. Time preference and health: an exploratory study. In: Fuchs, V. (Ed.), Economic Aspects of Health.

University of Chicago Press, Chicago, IL.Grimard, F., Parent, D., 2006. Education and Smoking: Were Vietnam Draft Avoiders Also More Likely to Avoid Smoking?

Departmental Working Papers 2006–05. McGill University, Department of Economics, Montreal.Grossman, M., 1972. On the concept of health capital and the demand for health. The Journal of Political Economy 80

(2), 223–255.Grossman, M., 1975. The correlation between health and smoking. In: Terlecky, N.E. (Ed.), Household Production and

Consumption. National Bureau of Economic Research, New York.


Grossman, M., 2006. Education and nonmarket outcomes. In: Hanushek, E., Welch, F. (Eds.), Handbook of the Economicsof Education. Elsevier Science, Amsterdam.

Heckman, J.J., Lalonde, R.J., Smith, J.A., 1999. The economics and econometrics of active labor market programs. In:Ashenfelter, O., Card, D. (Eds.), Handbook of Labor Economics, vol. III. Elsevier Science, Amsterdam.

Imbens, G., Angrist, J.D., 1994. Identification and estimation of local average treatment effects. Econometrica 62, 467–476.Kenkel, D.S., 1991. Health behavior, health knowledge, and schooling. The Journal of Political Economy 99 (2), 287–305.Lleras-Muney, A., 2005. The Relationship Between Education and Mortality in the U.S. Review of Economic Studies 72

(1), 189–221.Meara, E., 2001. Why is Health Related to Socioeconomic Status? The Case of Pregnancy and Low Birth Weight NBER

Working Paper 8231. Cambridge, MA.Moulton, B.R., 1990. An illustration of a pitfall in estimating the effects of aggregate variables on micro unit. The Review

of Economics and Statistics 72 (1), 334–338.Oreopoulos, P., 2006. Estimating average and local average treatment effects of education when compulsory schooling

laws really matter. American Economic Review 96 (1), 152–175.Sander, W., 1995. Schooling and quitting smoking. The Review of Economics and Statistics 77 (1), 191–199.Viscusi, W.K., 1990. Do smokers underestimate risks? The Journal of Political Economy 98 (6), 1253–1269.Viscusi, W.K., 1992. Smoking, Making the Risky Decision. Oxford University Press.

Does education affect smoking behaviors

Documents

Transcript of Does education affect smoking behaviors