The effects of alcohol use on high school absenteeism.

Austin, Wesley A.

I. Introduction

In many health-related and social science fields, there has long been concern about the various harmful effects of alcohol use. A significant consequence of alcohol use is the potential reduction in human capital accumulation by drinkers. This issue is particularly acute during adolescence and early adulthood, in which decisions regarding high school completion and college attendance are first considered, and academic performance realizations that affect longer-term educational and economic outcomes are initially observed. Excessive drinking has been associated with this age group despite prevention programs and its illegality until the age of 21. For instance, data from the recent National Survey on Drug Use and Health (NSDUH) found approximately 18 percent of youths (15--18 years old) and approximately 43 percent of young adults (18--25 years old) engaged in binge drinking (i.e. the consumption of five or more alcoholic beverages in one sitting, in the past month).

Several reasons might lead heavy drinking to impair human capital formation. Intoxication could interfere with school enrollment and attendance, and time spent in activities where drinking occurs could substitute away from time allocated to studying. This could hurt academic performance in the short term, which might diminish the ability or incentive to continue schooling over the longer term. Risks stemming from intoxication, such as injury from accidents or fights, pregnancy and disease from unsafe sex, conflicts with parents or law enforcement, and a tarnished reputation with school authorities might also limit the capability of a student to attend school (Cook and Moore, 1993). Alternatively, social interactions associated with drinking might improve academic achievement by providing a means of relieving stress (Williams et al., 2003).

Why is the potential impact of alcohol use on school absenteeism relevant for the discipline of economics? Substantively, human capital obtained from schooling bears directly on earning potential and strong relationships to individual health have been established. Moreover, variables such as, years of completed schooling, and dropout, are commonly examined education outcomes among broader literatures on human capital, given that they are easily measured and have a salient impact on future wages. This analysis augments the literature by investigating effects on absenteeism, i.e., classes missed due to "skipping" and illness or injury, which has not been widely studied despite potential effects on the educational outcomes mentioned above. Alcohol use, by raising absenteeism, can have confounding effects on outcomes such as grades and school completion.

As noted in the literature overview, some evidence has established a negative relationship between the regularity and intensity of drinking and human capital measures, such as school completion, and grades. But distinguishing whether these relationships are causal, such that increased alcohol consumption directly increases, for example, skipping classes, or merely correlational, with changes in other confounding variables simultaneously leading to drinking and lower class attendance, is worthy of investigation.

For economists and policy makers, obtaining an accurate estimate of the magnitude of the causal effect that alcohol use has on educational outcomes should be a primary econometric concern. This task is a natural one to tackle using econometric techniques such as instrumental variables (IV) regression.

II. Literature Overview

The relationship between alcohol use and human capital accumulation has been addressed by economists, but research on the topic had been fairly limited, with measures of drinking and educational outcomes, as well as conclusions, varying across studies. Early research produces evidence of a negative relationship between alcohol consumption and educational achievement, but either makes no attempt to econometrically deal with the potential endogeneity of drinking in education equations, or does so in a way that has since been criticized as unsatisfactory. Instrumental variables regression has been used in some previous studies but (as mentioned below) some of the studies have been faulted as not properly identifying the drinking/education relationship. So it is unclear whether or not this negative correlation indeed represents declines in educational outcomes that are caused by drinking.

Absenteeism from school by adolescents may not be a recent phenomenon; however, its growth could be on the rise. Causes of adolescent absenteeism may include: medical conditions; psychological problems; environmental risk factors such as homelessness, poverty, teen pregnancy, school violence, school climate, parental involvement; and family and community factors such as parental divorce and unsafe neighborhoods (Kearney, 2008). Of particular significance is the fact that one of the strongest correlates of recent absenteeism by 8th and 10th graders is drug use (Henry, 2007). By extension, the results of Bhatt (2011) could impact education as he finds that binge drinking influences transfers of money from parent to child; this, in turn, could affect decisions regarding drinking and schooling.

Cook and Moore (1993), analyze National Longitudinal Survey of Youth (NLSY) data on the 753 members of the two youngest cohorts (those ages 14 or 15 in 1979) who were enrolled in 12th grade as of the 1982 interview and estimate IV models in which the effect of current alcohol use on postsecondary schooling is identified by state excise taxes on beer and an indicator for whether the student could legally drink based on the state's MLDA (minimum legal drinking age). Results from three separate specifications show that heavy drinking in 12th grade decreases subsequent schooling. Students that live in states with higher beer taxes and minimum alcohol purchasing ages continue further in school and are more prone to graduate from college.

Dee and Evans (2003) call into question the causal effect interpretation of these results. They argue that the use of cross-state alcohol policy variation to identify the effects of drinking on other outcomes is potentially problematic because such variation might be correlated with unobservable attributes that affect both alcohol use and educational attainment.

Wolaver (2002) uses generalized method of moments to estimate a three-equation IV model in which alcohol consumption of college students, their study hours and academic performance are simultaneously determined. Utilizing measures of the ease of obtaining alcohol, parents' drinking behaviors, family attitudes about drinking and religiosity and peer studying and drinking as instruments, results indicate that heavy drinking reduces the probability of an A average by 12 to 18 percentage points. However, overidentification tests suggest that the exclusion restrictions as a whole were invalid.

Two studies concerning the impact of drinking on the school-related behaviors have appeared in the economics literature and merit attention. Roebuck et al. (2004) examines the likelihood of quitting school and truancy among a sample of 12-18 year olds who had not yet completed high school. A probit regression shows that those who consumed any alcohol over the previous year are 0.6 percent more likely to not be enrolled in school, but a negative binomial regression found no relationship between days truant and any past year drinking among those enrolled. Although measures of illegal drug use were included in the models, the potential endogeneity of drinking was not incorporated.

Markowitz (2001) estimates the effects of the number of days the respondent drank and binge drank over the prior 30 days on high school age students who fought and carried weapons. Analysis of these behavior variables has connecting implications for missed school days in that disciplinary sanctions for such actions may include suspension or expulsion from school. Results from the study, using a 2SLS procedure in which state level price measures, the beer tax and an indicator of whether marijuana is decriminalized serve as instruments, show that the probability of having been in a physical fight during the previous year rose by about six percentage points with each day of drinking and 11 points with each binge drinking day. However, the IV methodology is subject to the same criticism as that of Cook and Moore (1993), as state fixed effects were not included and thus cross-state price variation may have contributed to identification. Indeed, when census division indicators are added, both drinking measures become negative and insignificant in the fighting equations, but significantly positive in the gun carrying equation, while the F-statistics for the joint significance of the instruments fell from around four to below two and insignificant.

In addition to being plagued by weaknesses in instrumental variables procedures, the literature has focused mostly on college students. The focus of this article, however, is entirely on high school students.

III. Data

The National Survey on Drug Use and Health (NSDUH), sponsored by the Substance Abuse and Mental Health Services Administration (SAMHSA), is administered annually to approximately 55,000 civilian, non-institutionalized individuals age 12 and over, chosen so that the application of sample weights produces a nationally representative sample, with approximately equal numbers of respondents from the 12-17, 18-25 and 26 and over age groups. Using merged data from the 2007 and 2008 NSDUH, absenteeism is examined among a sample of 15,718 currently enrolled high school students, ages 14-17. Data from the NSDUH allow for both breadth and depth of coverage on the topic. Breadth comes from the ability to study aspects of educational outcomes using data from an elaborate questionnaire covering a wide array of youth experiences. Depth is provided by numerous variables on demographics, family income, family composition and relocation, among others.

An equally important facet of the NSDUH data is that they are conducive for utilization in IV regression methodology where the causal effect of alcohol use on human capital can be estimated. Abundant information is collected on experiences related to substance consumption, including measures of peer use of alcohol and drugs, ease of substance availability and the perceived risks involved in substance use. An assortment of variables are observed, therefore, that have the potential to serve as instruments for the proposed model, in the sense that they are likely to be highly correlated with alcohol use but would not have any obvious reason to be otherwise directly associated with educational outcomes.

A potentially problematic facet of the data is non-random measurement error emanating from the self-reported nature of responses. Although IV will eliminate bias from random measurement error, it cannot salvage data suffering from systematic measurement error. However, studies on the quality of self-reported academic variables and drinking data suggest that such reporting bias should be minimal. Cassady (2001) finds that self-reported GPA values are "remarkably similar to official records" and therefore are "highly reliable" and "sufficiently adequate for research use." Similarly, Grant et al. (1988), Midanik (1988), and Reinisch et al. (1991), conclude that youth drinking self-reports are reliable, based on the consistency of responses to alcohol use questions from repeated interviews. Harrison and Hughes (1997) find that survey methods not requiring subjects to verbally answer questions, as in the NSDUH, increase the accuracy of substance use self-reports.

IV. Research Method and Empirical Specification

In determining causation, the primary methodological question is whether drinking is properly specified as an exogenous variable with respect to educational variables or should instead be treated as endogenous. Consider the following equations, in which drinking (D) is a function of exogenous factors and an educational variable such as school absenteeism (A) is a function of some (but not all) of the same exogenous determinants as well as D,

D = [[alpha].sub.0] + Z[[alpha].sub.1] + X[[alpha].sub.2] + [omega], (1)

A = [[beta].sub.0] + [[beta].sub.1]D + X[[beta].sub.2] + [epsilon]. (2)

In the above equations, which apply to individual NSDUH respondents (with the corresponding observation-level subscript suppressed), vectors X and Z represent sets of exogenous variables that affect both drinking and absenteeism (X), and drinking but not absenteeism (Z), [omega] and [epsilon] are error terms that encompass all factors influencing the corresponding dependent variable that are not explicitly controlled for on the right hand side of the equations, and the [alpha]'s and [beta]'s are parameters to be estimated. Econometrically, alcohol use is exogenous in equation 2 if it is uncorrelated with the error term [epsilon]. This condition holds, by definition, if none of the unobserved schooling determinants are related to drinking. If so, there is no need to estimate equation 1; a single equation regression method such as OLS will produce consistent estimates of the causal effect of drinking, [[beta].sub.1]

However, two sources of endogeneity could possibly lead to a nonzero correlation between alcohol use (D) and the error term in (2). One is unobserved heterogeneity, which occurs should any of the unmeasured educational variable (e.g. absenteeism) determinants that are subsumed in the error term [epsilon] be correlated with alcohol use; then the resulting estimate of [[beta].sub.1] in (2) would suffer from omitted variable bias, which cannot be eliminated directly because the omitted variables are not recorded in the data. Disruptive events such as parental separation or divorce might simultaneously be responsible for greater alcohol consumption and increased school absenteeism on the part of students.

Such events are not observed and thus are not held constant in the regression, and the positive correlation between drinking and absenteeism that they induce becomes embedded into the alcohol use coefficient, which is thus biased as an estimate of the causal drinking effect. Conversely, unmeasured ability or socioeconomic background could create a negative bias in the estimated drinking effect if higher ability students are better able to function normally after alcohol consumption, or students who have more money to spend on alcohol (and therefore drink more) also enjoy greater academic success and are hence less likely to be absent from school. If such events could be incorporated, it stands to reason that coefficient estimates of [[beta].sub.1] would most likely be smaller in magnitude. However, the issue is primarily one of simple nonzero correlation emanating from unobserved heterogeneity, not magnitude per se.

The other potential source of endogeneity is reverse causation. If alcohol use and educational variables like absenteeism are simultaneously determined, the outcome will not only be a function of drinking, as specified in equation 2, but also will be a contributing factor to the decision regarding whether and how much alcohol to consume. In terms of equation 2, shocks to the error term [epsilon] that, by definition, influence educational outcomes will ultimately extend to drinking through the feedback effect of educational outcomes on alcohol consumption, thus creating a correlation between alcohol use and [epsilon] that renders the estimate of the causal drinking effect [[beta].sub.1] inconsistent. To investigate the possibility that alcohol use is endogenous as an explanatory factor for absenteeism, this analysis uses the method of instrumental variables (IV).

To use IV, there must be at least one, preferably two or more, variables (i.e. instruments or IVs) that affect alcohol use but have no direct impact on school absenteeism (A). In the case of exactly one instrument Z, the IV method works by estimating the causal drinking effect [[beta].sub.1] as the ratio of the sample correlation between the instrument and absenteeism to the sample correlation between the instrument and alcohol use.

Consider equation (3) below:

[[beta].sub.1] = c[??]rr[Z,A]/c[??]rr[Z,D]

where the "^" symbolizes that the quantity is estimated from the data and the correlations are estimated while holding constant the vector X of explanatory factors. Because the instrument is exogenous and related to absenteeism only through drinking, the sample correlation between the instrument and absenteeism is purely a product of that between drinking and absenteeism. Thus, the sample correlation between the instrument and absenteeism merely needs to be standardized with the sample correlation between the instrument and drinking in order to be used as an estimate for the causal effect of drinking on school absenteeism.

Equation 3 makes transparent the two important conditions that the instrument vector Z must satisfy in order for IV to produce consistent estimates of the causal drinking effect [[beta].sub.1]: First, the instruments must be highly correlated with alcohol use but not correlated with school absenteeism through any other mechanism besides drinking. If the correlation between the instruments and drinking is not statistically significant, the denominator in (3) is statistically equal to zero, thus rendering the expression for [[beta].sub.1] indeterminate. The strength of this correlation is judged from the F-statistic for the joint significance of [[alpha].sub.1] in equation 1. Minimally, [[alpha].sub.1] should be significant at the 1 percent level.

Second, if a direct correlation between the instruments and school absenteeism exists outside of the pathway from the instruments to drinking to absenteeism, the numerator in (3) includes variation that is not part of the relationship between drinking and absenteeism, and consequently the expression is no longer a consistent estimate of the causal effect of drinking. The reason multiple instruments are preferred is this overidentifies equation 2, which allows for specification tests to determine the empirical validity of excluding the instrument set Z from (2).

One other methodological point merits attention. Although IV estimates are consistent if the instrument strength and exogeneity conditions outlined above are satisfied, they are inefficient relative to OLS if it turns out that alcohol use is truly exogenous with respect to absenteeism, in which case the OLS estimates can be interpreted as causal effects. Thus, it is desirable to econometrically test the null hypothesis that drinking is exogenous in the absenteeism equation. This is done using a Hausman (1978) test, which proffers that, if drinking and the error term are uncorrelated, IV and OLS estimates should differ only by sampling error. If the null hypothesis of exogeneity is rejected, OLS estimates are inconsistent and conclusions about causal effects should be based on IV estimates; failure to reject the null means that OLS estimates may be interpreted as causal effects are preferable because of their smaller standard errors.

1. School Absenteeism

Using equations (1) and (2) above, the analysis is conducted utilizing a sample of currently enrolled high-school age students (14-17 years old). With a response range from 0 to 30 days, all respondents are asked to report the number of school days, over the past month, that were missed due to skipping and because of illness or injury. The latter merits attention in that drinking may be the reason for reporting illness or injury.

2. Drinking Variables

Among the varied alcohol use measures utilized are: the number of days the respondent drank in the past year (which is coded as '0' for nondrinkers and those that consumed no drinks in the previous year), and the number of drinks consumed in the previous month (which is coded as '0' for nondrinkers and those that consumed no drinks in the previous month). Binge drinking is defined as consuming five or more drinks on the same occasion on at least one day in the past 30 days. Although the timing of the number of drinks and binge drinking variables is not an ideal match for the absenteeism variables, in the sense that past month drinking cannot literally affect behavior that preceded the past month, this work will follow that of previous studies in assuming that previous month drinking patterns proxy those occurring in the recent period prior to the previous month.

The impact on absenteeism from alcohol abuse or dependence in the past year is also examined. This is accomplished by an indicator in the NSDUH of whether respondents exhibited symptoms of alcohol abuse or dependence in the past year. This is retrospectively coded by SAMHSA based on responses to questions corresponding to criteria outlined in the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders, the clinical standard for establishing drug abuse and dependence.

3. Explanatory Variables

Several variables from the NSDUH data are considered explanatory in the model: family income is measured in four categories: less than $20,000; $20,000-$49,999; $50,000-$74,999; $75,000 or more, with less than $20,000 as the omitted category in the regressions. Population density is represented by indicators for two categories: an MSA with one million persons or greater and an MSA of less than one million persons, with non-MSA areas as the omitted category. For race, indicators are specified for Caucasians, African Americans, Native Americans, Asians, non-white Hispanics and multiracial. A binary measure of gender is included and family size is measured using two variables: the number of members if the household has one to five members and an indicator for those with over five.

Indicators for the last grade completed are 9th, 10th or 11th grade, with 12th as the omitted category. And factors are included to indicate a presence of the mother or father in the household as well as indicators for whether or not parents assisted the student with homework always, sometimes or seldom in the past 12 months, with never as the omitted category. A variable regarding how many times the respondent moved in the past year is included. Age indicators are not incorporated as age 16 is used as an instrument. By proxy however, age is included as an explanatory variable--the last grade in which the student was enrolled is included in the X vector as described above. For example, those in the 10th and 11th grades tend to be 16 and 17 years old respectively.

4. Instrumental Variables

Several NSDUH variables conceivably influence drinking without having direct effects on school absenteeism and are thus candidates to serve as instrumental variables. The specific variables utilized are: student is 16 years of age; peer use of alcohol; and perceived risk of bodily harm from marijuana use.

Age 16 could be connected to drinking as students of this age obtain drivers' licenses (and possibly fake IDs), potentially rebel against parents and attend more social functions where alcohol is consumed. Being 16 years old is not presumed to be directly connected to school absenteeism except through the aforementioned "party" behavior. Also, age 16 is suited to be an IV as it is possibly correlated to increased experimentation with substance use (including alcohol) during the teenage years. In addition, there is no a priori reason to expect that simply being age 16 suddenly leads to greater school absenteeism; it is most likely caused by the desire to socialize with friends or it is the result of negative health effects that result from "partying." Also, the exogenous nature of one's age makes it potentially attractive as an instrumental variable vis-a-vis the criteria described above.

For risk associated with marijuana use, a binary measure indicates if the respondent feels there are great/ moderate risks or slight/no risks of harm, physically or otherwise, from consuming marijuana once or twice a week. While risk aversion could arguably be linked to educational outcomes in that more risk averse individuals are more likely to accumulate human capital (i.e. attend and graduate high school, maintain high grades) as possible "insurance" against having a low standard of living, in this instance the domain of risk involves only substance use. In the model, risk aversion involves concerns about the direct consequences of bodily harm from consuming an illegal substance. Therefore it is assumed to have no direct connection to educational variables (i.e. absenteeism). Some evidence shows that the risk associated with one variable is indeed different from risk associated with another--risk aversion is not evenly spread among all aspects of human decisions (Heath and Tversky, 1991). Therefore, the risk associated with substance use is different from, and quite possibly not correlated with, the risk associated with failure to obtain human capital; this risk however could be correlated to drinking in that those who associate greater risk with marijuana use are also less likely to consume another substance: alcohol.

Peer use information reflects respondents' perceptions regarding whether none, a few, most, or all students in the same grade at his or her school consume alcohol. For the peer use variable, a binary measure is defined to designate if the respondent feels that most or all schoolmates consume alcohol. Potential endogeneity of the peer variable, stemming from a possible connection between one's own behavior and perceptions about the behavior of others, should be mitigated in that the relevant data cover all classmates rather than simply friends, who are presumably chosen by the respondent. Norton et al. (1998), Gaviria and Raphael (2001) and Kremer and Levy (2003) each find evidence that increased rates of drinking among peers raised the propensity to consume alcohol.

V. Empirical Findings

The causal effect drinking has on the days missed due to skipping and illness or injury is estimated using the respective instrumental variables listed above. The main results of the IV analysis are also compared with parameter estimates obtained using OLS methodology. While the discussion that follows concentrates on the effects of alcohol consumption and results from specification tests, appendix 1 (for binge drinking) shows the IV coefficients and marginal effect standard errors of all explanatory variables on both absenteeism measures.

Table 1 presents select summary statistics. The mean number of days drinks were consumed in the past year is about 13 while the mean number of drinks consumed in the past month is 3.4. Mean alcohol abuse/ dependence is 0.06. Mean days missed due to skipping is 0.34 and days missed because of injury/ illness is about one day. About 90 percent of respondents in both samples live in an MSA, roughly equally split between MSAs with populations greater than and less than one million. For the sample, fathers are less likely to be present in the household than are mothers and the proportion of parents that help with homework is also very high.

1. First Stage Regression Results

Table 2 presents the probit results for the drinking measures on the instruments. Being age 16 increases the number of drinks consumed in the past year by approximately five. The number of drinks consumed in the past month rises by roughly one, while the probability of binge drinking in the last 30 days rises by about 0.03 percentage points and the likelihood of being categorized as abusive/dependent on alcohol increases by 0.07.

For respondents who report most or all their schoolmates use alcohol, the number of days drinking occurred in the past year rises by almost nine days, the number of drinks consumed in the past month rises by roughly 2.5, the probability of binge drinking in the last 30 days rises by 0.06 points, and the likelihood of being categorized as abusive/ dependent on alcohol increases by 0.04.

Risk of bodily harm from marijuana use has the strongest overall effect. Of those who perceive that there is moderate to great risk of harm from marijuana use, the number of days drinking occurred in the past year is lowered by about 10 days. The number of drinks consumed in the past month is reduced by about three, while the likelihood of binge drinking in the last 30 days falls by 0.07 percentage points. The likelihood of being categorized as abusive/ dependent on alcohol falls by 0.06 points. The F-statistics and associated p-values indicate that the instruments are jointly significant for all the drinking measures.

2. The Effects of Drinking on Absenteeism

The findings in Table 3 show that drinking among high school students leads to increases in absenteeism. An additional day increase in the number of past year drinking days elevates days skipped by 0.012 and days missed because of illness/ injury by approximately 0.01, relative to refraining from drinking. For each additional drink increase in the number of drinks the respondent consumed in the past month, days missed because of skipping rise by 0.04 and days missed due to illness increase by about 0.05. Respondents that drank in the prior month experience a 12 percent increase in days skipped and an approximate four percent increase in sick days, above the mean values for the sample.

Binge drinking and abuse/ dependence on alcohol further increase absenteeism. For students who engaged in binge drinking in the previous month, the number of days missed due to skipping is elevated by almost two days. For those classified as abusive/dependent with respect to alcohol, the number of days missed due to skipping escalates to approximately 2.5 days per month compared to those not abusive/ dependent. School days missed due to illness or injury rise by 1.5 days for binge drinkers and about 2.5 days for those who are abusive/ dependent with respect to alcohol. Particularly noteworthy is that racial category is not statistically significant in determining skipping school. However Asians miss fewer days due to illness compared to Caucasians, while other racial categories are not statistically significant. Gender is not significant for absenteeism due to skipping, but it is for absenteeism due to illness (females miss more days due to illness).

For all drinking indicators, the overidentification tests have associated p-values that offer evidence in support of the hypothesis of instrument exogeneity at the 10 percent level. For both absenteeism variables, the Hausman tests afford the same general result for all drinking measures: statistically significant differences between IV and OLS estimates are present; hence, drinking and the error term are most likely correlated and this offers evidence that drinking may be considered endogenous in the absenteeism equation.

Overall, there is a strong indication that drinking, possibly by raising the opportunity cost of high school attendance and impairing cognitive functioning, increases absences in high school. As drinking impairs the learning process, this may lead to greater student frustration which engenders even greater absenteeism. And, considering the additional resources the student devotes toward drinking if the student binge drinks or is abusive/ dependent on alcohol, there is compelling evidence that absenteeism is largely and positively impacted by this heavy alcohol use.

3. Instrumental Variable Robustness and Absenteeism

To determine if there is any sensitivity in the main results attributable to changes in the instrument set, regressions are performed with varying pairs of instruments with results presented in Table 4. The instrument that is omitted from the IV combination is utilized as an explanatory variable and its coefficient and standard error is reported.

All drinking measures have positive effects on both absenteeism factors and the effects are noticeably similar to those in the main regression where all three instruments are employed. For both absenteeism outcomes, the overidentification test results support exogeneity for all IV pairs. In addition, Hausman tests show there are statistically significant differences between IV and OLS estimates in all specifications, again providing evidence that drinking may be considered endogenous. And the additional instrument not used to identify drinking is never significant as an explanatory variable in the absenteeism equation.

4. General Comment on Instrument Robustness and OLS

Overall, the robustness analyses offer some evidence to support the hypothesis that the instruments are exogeneous. Overidentification tests on differing pairs of instruments demonstrate that exogeniety is generally robust to any instrument set employed. And the effect of alcohol use on defiant behaviors is remarkably similar regardless of the utilized instrument set.

Throughout the analyses, OLS parameter estimates, relative to IV, consistently underestimate the magnitude of the positive effects in the main specifications for absenteeism. This could be ascribed to the prospect that higher ability (i.e. higher achieving) students perform better academically even when they drink and these higher achievers are more likely to have lower absenteeism rates. In addition, higher income students (who spend more on alcohol and therefore drink more) may command more resources that can be channeled toward education and this in turn could serve to subdue absenteeism. It could also be that higher achieving students simply "like" school more, which in turn lowers the incidence absences. High achievers may have a higher opportunity cost with respect to misconduct, so absenteeism is naturally less for these students, and the simple exhilaration of academic success could engender better school attendance. The results shown in Appendix 2 emphasize the potential effects of reverse causation. As a specification check of the relationship between absenteeism and drinking, a test of drinking as a function of absenteeism is performed. The table summarizes the effects of days absent due to skipping/sickness on drinking in the past month at age 16, utilizing the same explanatory variables. Reverse causation running from absenteeism to drinking is present (i.e. the effect of days missed due to skipping and sickness on the drinking outcome is significant).

It should be noted that when smaller samples (only age 16) are utilized, reverse causation may mitigate the impact of variables in the model. Therefore, reverse causation may serve as at least a partial explanation as to why OLS estimates are smaller than IV estimates.

Measurement error could be another explanation for the IV/ OLS differences. However, IV estimates are consistent even when alcohol use is measured with random error.

As an additional "check" for IV accuracy, Appendix 3 offers a stepwise regression for the binge drinking and abuse/ dependence measures, which, as noted above, have the most negative impact on school attendance. In this procedure, observed variables are sequentially added into the model. The results show that the binge drinking and abuse/ dependence coefficient estimates are not materially affected by the stepwise addition of other explanatory variables. The magnitude of both coefficient estimates is very similar to estimates generated absent the stepwise procedure. Generally, this suggests the primary coefficient estimate of interest in the model--the effect of drinking on absenteeism--is not susceptible to weakening as a result of adding explanatory variables.

VI. Concluding Remarks

This study contributes to the literature by examining the effects of youth drinking on days missed due to skipping school and due to illness or injury, while accounting for unobserved endogeneity. The economics literature has established a negative link between drinking and educational variables, but many of these studies do not account for the possibility that the negative correlation between these factors may be the result of unobserved variables that cause simultaneous increases in drinking and unfavorable outcomes in education variables. For studies that have incorporated unobserved endogeneity, instrumental variable procedures have been problematic. While the literature has investigated outcomes such as school completion and grades, absenteeism, though not strictly an educational outcome but an influence on them, has not been widely addressed. The results of this study buttress those of other papers which find a negative relationship between substance use and educational variables.

The results of this study complement one other major study in this area: the findings of Roebuck et al. (2004), where substance use (specifically, marijuana) contributed to dropout and truancy. The findings presented above suggest alcohol use effects truancy as well.

Evidence is offered that school absenteeism in high school is raised when students use alcohol more frequently and intensely. Binge drinking and abuse of alcohol have the most detrimental impact on school attendance. Throughout the analysis, extensive empirical testing generally confirm instrument exogeneity and thus demonstrate that adolescent alcohol consumption may be treated as endogenous. Also, OLS regressions tend to underestimate the effects of alcohol use on both absenteeism factors.

Although there is no direct analysis of the effectiveness of laws and other programs designed to curtail youth drinking, the conclusions in this paper support the premise that reducing adolescent alcohol use enhances human capital accumulation. Minimum legal drinking ages, high school anti-substance use programs and other policies aimed at lowering youth drinking may well be justified on human capital grounds. A further examination of the effectiveness of public policies that purport to reduce youth drinking would also prove valuable. Though the instrumental variables generally prove to be effective and useful, further research should include continued exploration for reliable instruments to ensure that the relationship between drinking and academic variables is properly identified. Also, this study is limited in that it does not account for state fixed effects and by the fact that, given the instruments selected, age indicators could not be used in the model.

APPENDIX 1.
All IV Estimates on Absenteeism for Binge Drinking (n=15,718)

 Days missed due to skipping

 IV coefficient
Explanatory variables (Marginal Effect SE)

Binge drinking 1.785 (0.229)
Mother in household 0.061 (0.040)
Father in household -0.054 (0.032)
Female gender 0.003 (0.021)
Last grade completed (9th grade) 0.074 (0.036)
Last grade completed (10th grade) 0.034 (0.043)
Last grade completed (11th grade) -0.044 (0.043)
Race (African American) 0.024 (0.074)
Race (Native American) 0.198 (0.182)
Race (Asian) 0.069 (0.112)
Race (non-white Hispanic) 0.003 (0.072)
Number in family 0.028 (0.015)
Number in family (>5) 0.150 (0.068)
Family income ($20,000-$49,999) -0.149 (0.046)
Family income ($50,000-$74,999) -0.252 (0.046)
Family income ($75,000 or more) -0.304 (0.047)
MSA segment with 1+ million persons 0.056 (0.037)
MSA segment of less than 1 million 0.058 (0.035)
number of times moved (past 12 months) 0.077 (0.021)
Parents help with homework (always) -0.223 (0.055)
Parents help with homework (sometimes) -0.200 (0.056)
Parents help with homework (seldom) -0.128 (0.062)

 Days missed clue to illness

 IV coefficient
Explanatory variables (Marginal Effect SE)

Binge drinking 1.954 (0.392)
Mother in household -0.016 (0.079)
Father in household -0.160 (0.050)
Female gender 0.145 (0.035)
Last grade completed (9th grade) 0.147 (0.056)
Last grade completed (10th grade) 0.037 (0.064)
Last grade completed (11th grade) -0.069 (0.062)
Race (African American) 0.068 (0.063)
Race (Native American) 0.312 (0.230)
Race (Asian) -0.364 (0.078)
Race (non-white Hispanic) -0.080 (0.052)
Number in family 0.022 (0.024)
Number in family (>5) 0.240 (0.111)
Family income ($20,000-$49,999) -0.323 (0.069)
Family income ($50,000-$74,999) -0.520 (0.072)
Family income ($75,000 or more) -0.608 (0.072)
MSA segment with 1+ million persons 0.111 (0.061)
MSA segment of less than 1 million 0.176 (0.059)
number of times moved (past 12 months) 0.270 (0.036)
Parents help with homework (always) -0.060 (0.070)
Parents help with homework (sometimes) -0.027 (0.073)
Parents help with homework (seldom) -0.026 (0.083)

Standard errors are in parentheses.

APPENDIX 2.
OLS Descriptive Statistics-Effects of Absenteeism/ Other Variables
on Past Month Drinking at Age 16

 (Marginal
Variable OLS coefficient Effect SE)

Number of days missed due to skipping 1.165 * 0.218
Variable
Mother in household 1.202 1.126
Father in household 1.284 0.867
Female -2.290 * 0.640
Last grade completed (9th grade) -2.025 2.121
Last grade completed (10th grade) -3.287 *** 1.764
Last grade completed (11th grade) -2.723 1.693
Race (African American) 4.024 ** 1.884
Race (Native American) 1.198 3.282
Race (Asian) 0.869 5.585
Race (non-white Hispanic) 3.394 *** 1.868
Number in family -0.859 ** 0.409
Number in family (>5) -4.461 ** 1.853
Family income ($20,000-$49,999) -0.285 1.014
Family income ($50,000-$74,999) -1.070 1.188
Family income ($75,000 or more) -1.549 1.159
MSA segment with 1+ million persons 0.049 1.180
MSA segment of less than 1 million -0.061 1.128
number of times moved (past 12 months) 0.737 0.467
Parents help with homework (always) -4.683 * 1.002
Parents help with homework (sometimes) -3.351 * 1.107
Parents help with homework (seldom) -2.452 ** 1.120

 (Marginal
 OLS coefficient Effect SE)

Number of days missed due to sickness 0.317 ** 0.140
Variable
Mother in household 1.272 1.122
Father in household 1.484 *** 0.863
Female -1.872 * 0.647
Last grade completed (9th grade) -2.295 2.141
Last grade completed (10th grade) -3.314 *** 1.720
Last grade completed (11th grade) -3.018 *** 1.695
Race (African American) 3.072 * 1.018
Race (Native American) -0.521 2.862
Race (Asian) 2.838 1.829
Race (non-white Hispanic) 2.563 * 0.977
Number in family -0.720 ** 0.409
Number in family (>5) -3.764 * 1.185
Family income ($20,000-$49,999) -0.304 1.010
Family income ($50,000-$74,999) 0.988 1.184
Family income ($75,000 or more) -1.737 1.158
MSA segment with 1+ million persons 0.037 1.180
MSA segment of less than 1 million -0.190 1.122
number of times moved (past 12 months) 0.732 0.467
Parents help with homework (always) -4.299 * 1.004
Parents help with homework (sometimes) -3.105 * 1.104
Parents help with homework (seldom) -2.282 ** 1.125

* Statistically significant at 1 %

** Statistically significant at 5%

*** Statistically significant at 10%

APPENDIX 3.
Stepwise Regression Results--Binge Drinking and Abuse/ Dependence on
Absenteeism in Past 30 Days (n=15,718)

 Days missed due to skipping

Alcohol (and other variables) OLS (Marginal Effect SE)

Binge drinking 0.611 * (0.040)
Race (Caucasian) -0.118 * (0.023)
Number of times moved (past 12 months) 0.089 * (0.015)
Parents help with homework (always) -0.311 * (0.034)
Parents help with homework (sometimes) -0.261 * (0.038)
Family income ($75,000 or more) -0.282 * (0.034)
Age of student (17 years old) 0.080 * (0.030)
Family income ($50,000-$74,999) -0.242 * (0.037)
Family income ($20,000-$49,999) -0.139 * (0.033)
Parents help with homework (seldom) -0.176 * (0.043)
Age of student (14 years old) -0.133 * (0.033)
Age of student (15 years old) -0.102 * (0.032)
Race (Native American) 0.218 ** (0.091)
Last grade completed (11th grade) -0.066 ** (0.029)

 Days missed due to skipping

Alcohol (and other variables) OLS (Marginal Effect SE)

Abuse/ dependence on alcohol 0.600 * (0.045)
Race (Caucasian) -0.111 * (0.023)
Parents help with homework (always) -0.318 * (0.034)
Number of times moved (past 12 months) 0.086 * (0.015)
Age of student (17 years old) 0.095 * (0.030)
Parents help with homework (sometimes) -0.267 * (0.038)
Family income ($75,000 or more) -0.270 * (0.034)
Family income ($50,000-$74,999) -0.236 * (0.037)
Parents help with homework (seldom) -0.190 * (0.043)
Family income ($20,000-$49,999) -0.135 * (0.033)
Age of student (14 years old) -0.151 * (0.033)
Age of student (15 years old) -0.117 * (0.032)
Last grade completed (11th grade) -0.069 ** (0.029)
Race (Native American) 0.214 ** (0.091)

 Days missed due to
 illness/injury

Alcohol (and other variables) OLS (Marginal Effect SE)

Number of times moved (past 12 months) 0.284 * (0.025)
Father in household -0.160 * (0.043)
Binge drinking 0.384 * (0.066)
Family income ($75,000 or more) -0.556 * (0.059)
Family income ($50,000-$74,999) -0.488 * (0.062)
Family income ($20,000-$49,999) -0.300 * (0.054)
Race (Asian) -0.426 * (0.094)
Female gender 0.127 * (0.034)
Number in family (>5) 0.142 * (0.049)
Age of student (14 years old) -0.140 * (0.043)
MSA segment of less than 1 million 0.088 * (0.034)
Parents help with homework (always) -0.091 ** (0.034)
Last grade completed (9th grade) 0.103 ** (0.042)
Race (Native American) 0.304 *** (0.014)

 Days missed due to
 illness/injury

Alcohol (and other variables) OLS (Marginal Effect SE)

Number of times moved (past 12 months) 0.282 * (0.025)
Father in household -0.160 * (0.043)
Family income ($75,000 or more) -0.548 * (0.059)
Abuse/ dependence on alcohol 0.358 * (0.074)
Family income ($50,000-$74,999) -0.483 * (0.062)
Family income ($20,000-$49,999) -0.295 * (0.054)
Race (Asian) -0.432 * (0.094)
Female gender 0.119 * (0.034)
Number in family (>5) 0.142 * (0.049)
Age of student (14 years old) -0.145 * (0.043)
MSA segment of less than 1 million 0.087 * (0.034)
Parents help with homework (always) -0.091 ** (0.034)
Last grade completed (9th grade) 0.097 ** (0.042)
Race (Native American) 0.297 ** (0.014)

* Statistically significant at 1%

** Statistically significant at 5%

*** Statistically significant at 10%

References

Bhatt, V. 2011. Adolescent alcohol use and Intergenerational transfers: evidence from micro data. Journal of Family and Economic Issues, 32(2), 296-307.

Cassady, P. 2001. Self-reported GPA and SAT scores: A methodological note. Practical Assessment, Research & Evaluation, 7(12).

Cook, P. J., & Moore, M. J. 1993. Drinking and schooling. Journal of Health Economics, 12(4), 411-29.

Dee, T. S., and Evans, W. N. 2003. Teen drinking and educational attainment: Evidence from two-sample instrumental variables estimates. Journal of Labor Economics, 21 (1), 178-209.

Gaviria, A., and Raphael, S. 2001. School-based peer effects and juvenile behavior. Review of Economics and Statistics, 83(2), 257-268.

Grant, B. F., Harford, T. C., and Grigson, M. B. 1988. Stability of alcohol consumption among youth: a national longitudinal survey. Journal of Studies on Alcohol, 49(3), 253-260.

Harrison, L., and Hughes, A. 1997. The validity of self-reported drug use: Improving the accuracy of survey estimates. NIDA Research Monograph, (167), 1-16.

Hausman, J. 1978. Specification tests in econometrics. Econometrica, 46(6), 1251-1271.

Heath, C. and Tversky, A. 1991. Preference and belief: Ambiguity and competence in choice under uncertainty. Journal of Risk and Uncertainty, 4, 5-28.

Henry, K. L. 2007. Who's skipping school: characteristics of truants in 8th and 10th grade. Journal of School Health, 77(1), 29-35.

Kearney, C. A. 2008. School absenteeism and school refusal behavior in youth: a contemporary review. Clinical Psychology Review, 28, 451-471.

Kremer, M., & Levy, D. 2003. Peer effects and alcohol use among college students. Working paper 9876, National Bureau of Economic Research.

Markowitz, S. 2001. The role of alcohol and drug consumption in determining physical fights and weapon carrying by teenagers. Eastern Economic Journal, 27(4), 409-432.

Midanik, L. 1988. Validity of self-reported alcohol use: a literature review and assessment. British Journal of Addiction, 83, 1019-1029.

Norton, E., Lindrooth, R., and Ennett, S. 1998. Controlling for the endogeneity of peer substance use on adolescent alcohol and tobacco use. Health Economics, 36(7), 439-453.

Reinisch, E. J., Bell, R. M., and Ellickson, P. 1991. How accurate are adolescent reports of drug use? Working paper, Rand Corporation: Santa Monica, CA.

Roebuck, C., French, M., & Dennis, M. 2004. Adolescent marijuana use and school attendance. Economics of Education Review, 23(2), 133-141.

Williams, J., Powell, L. M., & Wechsler, H. 2003. Does alcohol consumption reduce human capital accumulation? evidence from the college alcohol study. Applied Economics, 35(10), 1227-1239.

Wolaver, Amy M. 2002. Effects of heavy drinking in college on study effort, grade point average, and major choice. Contemporary Economic Policy, 20(4), 415-428.

by Wesley A. Austin, Assistant Professor of Economics, Department of Economics and Finance, University of Louisiana at Lafayette, PO Box 44570, Lafayette, LA 70504-4570. E-Mail: [email protected]. The author would like to thank an anonymous referee for valuable comments.

TABLE 1.
Descriptive Statistics (n=15,718)

Variable Mean Standard Deviation

Number of days drank-past year 12.567 38.202
Number of drinks in previous month 3.488 23.384
Binge drinking in the past 30 days 0.078 0.269
Abuse/ Dependence on alcohol 0.059 0.236
 classification
Female gender 0.481 0.499
Respondent perceives peer use of alcohol 0.607 0.488
Respondent perceives risk of harm 0.647 0.477
 from marijuana
Number of skipped school days 0.340 1.359
 (past 30 days)
Number of days missed due to 1.128 2.180
 illness (past 30 days)
Family income (less than $20,000) 0.134 0.240
Family income ($20,000-$49,999) 0.264 0.310
Family income ($50,000-$74,999) 0.314 0.300
Family income ($75,000 or more) 0.288 0.453
MSA segment with 1-I- million persons 0.421 0.493
MSA segment of less than 1 million 0.481 0.499
Age of student (14 years old) 0.254 0.435
Age of student (15 years old) 0.257 0.437
Age of student (16 years old) 0.251 0.434
Age of student (17 years old) 0.236 0.424
Mother in household 0.907 0.290
Father in household 0.714 0.451
Parents help with homework (always) 0.538 0.499
Parents help with homework (sometimes) 0.233 0.423
Parents help with homework (seldom) 0.117 0.327
number of times moved (past 12 months) 0.342 0.725
Last grade in (9th grade) 0.255 0.436
Last grade in (10th grade) 0.234 0.423
Last grade in (11th grade) 0.138 0.345
Race (Caucasian) 0.597 0.490
Race (African American) 0.144 0.351
Race (Native American) 0.014 0.121
Race (Asian) 0.033 0.179
Race (non-white Hispanic) 0.168 0.374
Number in family 3.234 1.576
Number in family (>5) 0.147 0.354

TABLE 2.
First Stage Estimates for Absenteeism in Past 30 Days (n=15,718)

 Number of days Number of drinks
Instrumental variables drank in past year in past month

Student is 16 years old 4.925 1.038
 (1.037) (0.621)
Peer use of alcohol 8.671 2.699
 (0.607) (0.364)
Risk of bodily harm from -10.416 -3.024
 marijuana use (0.6059) (0.364)
F-statistic of joint 46.11 24.14
 significance
P-value of significance level (0.000) (0.000)

 Binge Abuse/ Dependence
Instrumental variables drinking on alcohol

Student is 16 years old 0.259 0.074
 (0.007) (0.006)
Peer use of alcohol 0.059 0.043
 (0.004) (0.003)
Risk of bodily harm from -0.073 -0.057
 marijuana use (0.004) (0.004)
F-statistic of joint 69.66 34.35
 significance
P-value of significance level (0.000) (0.000)

TABLE 3.
IV/ OLS Estimates of Drinking on Absenteeism in Past 30 Days
(all three instruments) (n=15,718)

 Days missed due to skipping

Alcohol variables IV OLS

Number of days drank-past year 0.012 * 0.005 *
Marginal Effect Standard Error (0.001) (0.0003
P-value of overidentification test 0.407
Hausman statistic 4.905 *

Number of drinks in past month 0.041 * 0.006 *
Marginal Effect Standard Error (0.005) (0.0005)
P-value of overidentification test 0.374
Hausman statistic 5.968 *

Binge drinking 1.785 * 0.614 *
Marginal Effect Standard Error (0.229) (0.0416)
P-value of overidentification test 0.467
Hausman statistic 5.368 *

Abuse/ dependence on alcohol 2.342 * 0.602 *
Marginal Effect Standard Error (0.307) (0.0462)
P-value of overidentification test 0.617
Hausman statistic 5.869 *

 Days missed due to illness/
 injury

Alcohol variables IV OLS

Number of days drank-past year 0.013 * 0.004 *
Marginal Effect Standard Error (0.003) (0.0004)
P-value of overidentification test 0.946
Hausman statistic 3.705 *

Number of drinks in past month 0.045 * 0.002 *
Marginal Effect Standard Error (0.009 (0.0008)
P-value of overidentification test 0.898
Hausman statistic 4.488 *

Binge drinking 1.954 * 0.391 *
Marginal Effect Standard Error (0.392) (0.0671)
P-value of overidentification test 0.966
Hausman statistic 4.183 *

Abuse/ dependence on alcohol 2.568 * 0.384 *
Marginal Effect Standard Error (0.521 (0.0750)
P-value of overidentification test 0.995
Hausman statistic 4.337 *

* Statistically significant at 1%

TABLE 4.
IV Estimates of Drinking on Absenteeism in Past 30 Days Using IV
Pairs
 Days missed due
 to skipping

 Marijuana risk
 and peer drinking

Alcohol variables

Number of days drank-past year 0.012 *
Marginal Effect Standard Error (0.002)
P-value of overidentification test 0.407
Hausman statistic 4.905 *
Coefficient (Standard Error) of omitted IV 0.079 (0.042)

Number of drinks in past month 0.041 *
Marginal Effect Standard Error (0.005)
P-value of overidentification test 0.374
Hausman statistic 5.968 *
Coefficient (Standard Error) of omitted IV 0.086 (0.045)

Binge drinking 1.785 *
Marginal Effect Standard Error (0.229)
P-value of overidentification test 0.467
Hausman statistic 5.368 *
Coefficient (Standard Error) of omitted IV -0.041 (0.045)

Abuse/ dependence on alcohol 2.342 *
Marginal Effect Standard Error (0.307)
P-value of overidentification test 0.617
Hausman statistic 5.869 *
Coefficient (Standard Error) of omitted IV -0.061 (0.045)

 Days missed due to
 illness/ injury

Number of days drank-past year 0.014 *
Marginal Effect Standard Error (0.003)
P-value of overidentification test 0.946
Hausman statistic 3.705 *
Coefficient (Standard Error) of omitted IV -0.030 (0.060)

Number of drinks in past month 0.045 *
Marginal Effect Standard Error (0.010)
P-value of overidentification test 0.8984
Hausman statistic 4.488 *
Coefficient (Standard Error) of omitted IV 0.037 (0.063)

Binge drinking 1.954 *
Marginal Effect Standard Error (0.392)
P-value of overidentification test 0.966
Hausman statistic 4.183 *
Coefficient (Standard Error) of omitted IV 0.123 (0.069)

Abuse/ dependence on alcohol 2.598 *
Marginal Effect Standard Error (0.521)
P-value of overidentification test 0.995
Hausman statistic 4.337 *
Coefficient (Standard Error) of omitted IV 0.154 ** (0.068)

 Days missed due
 to skipping

 Age of student (16)
 and marijuana risk

Alcohol variables

Number of days drank-past year 0.014 *
Marginal Effect Standard Error (0.002)
P-value of overidentification test 0.997
Hausman statistic 3.897 *
Coefficient (Standard Error) of omitted IV -0.042 (0.034)

Number of drinks in past month 0.049 *
Marginal Effect Standard Error (0.010)
P-value of overidentification test 0.998
Hausman statistic 4.314 *
Coefficient (Standard Error) of omitted IV -0.051 (0.040)

Binge drinking 2.044 *
Marginal Effect Standard Error (0.356)
P-value of overidentification test 0.999
Hausman statistic 4.159 *
Coefficient (Standard Error) of omitted IV -0.038 (0.034)

Abuse/ dependence on alcohol 2.602 *
Marginal Effect Standard Error (0.462)
P-value of overidentification test 0.998
Hausman statistic 4.441 *
Coefficient (Standard Error) of omitted IV -0.030 (0.034)

 Days missed due to
 illness/ injury

Number of days drank-past year 0.014 *
Marginal Effect Standard Error (0.003)
P-value of overidentification test 0.998
Hausman statistic 2.896 *
Coefficient (Standard Error) of omitted IV -0.016 (0.050)

Number of drinks in past month 0.049 *
Marginal Effect Standard Error (0.014)
P-value of overidentification test 0.999
Hausman statistic 3.374 *
Coefficient (Standard Error) of omitted IV -0.025 (0.055)

Binge drinking 2.042 *
Marginal Effect Standard Error (0.530)
P-value of overidentification test 0.999
Hausman statistic 3.243 *
Coefficient (Standard Error) of omitted IV -0.013 (0.049)

Abuse/ dependence on alcohol 2.609 *
Marginal Effect Standard Error (0.685)
P-value of overidentification test 0.999
Hausman statistic 3.342 *
Coefficient (Standard Error) of omitted IV -0.004 (0.049)

 Days missed due
 to skipping

 Age of student (16)
 and peer drinking

Alcohol variables

Number of days drank-past year 0.009 *
Marginal Effect Standard Error (0.002)
P-value of overidentification test 0.999
Hausman statistic 1.875 **
Coefficient (Standard Error) of omitted IV -0.050 (0.041)

Number of drinks in past month 0.030 *
Marginal Effect Standard Error (0.008)
P-value of overidentification test 0.999
Hausman statistic 2.869 *
Coefficient (Standard Error) of omitted IV -0.057 (0.041)

Binge drinking 1.399 *
Marginal Effect Standard Error (0.382)
P-value of overidentification test 0.999
Hausman statistic 2.170 *
Coefficient (Standard Error) of omitted IV -0.047 (0.042)

Abuse/ dependence on alcohol 1.904 *
Marginal Effect Standard Error (0.522)
P-value of overidentification test 0.998
Hausman statistic 2.583 *
Coefficient (Standard Error) of omitted IV -0.040 (0.044)

 Days missed due to
 illness/ injury

Number of days drank-past year 0.012 *
Marginal Effect Standard Error (0.004)
P-value of overidentification test 0.999
Hausman statistic 2.076 *
Coefficient (Standard Error) of omitted IV -0.020 (0.060)

Number of drinks in past month 0.040 *
Marginal Effect Standard Error (0.014)
P-value of overidentification test 0.999
Hausman statistic 2.698 *
Coefficient (Standard Error) of omitted IV -0.028 (0.060)

Binge drinking 1.822 *
Marginal Effect Standard Error (0.622)
P-value of overidentification test 0.998
Hausman statistic 2.401 *
Coefficient (Standard Error) of omitted IV -0.016 (0.061)

Abuse/ dependence on alcohol 2.499 *
Marginal Effect Standard Error (0.861)
P-value of overidentification test 0.999
Hausman statistic 2.524 *
Coefficient (Standard Error) of omitted IV -0.006 (0.065)

* Statistically significant at 1 %: ** Statistically significant at 5%