Synopsis
Alcohol and drug abuse can interfere with an individual's employment and productivity. Both drug and alcohol abuse may interfere with a person's ability to work (physical and/or mental impairment); ability to find a job (lack of skills or lack of experience); and, potentially, willingness to find a job. Wages among workers with alcohol or drug abuse problems may be lower than among similar workers without such problems. Shortfalls in productivity and employment among individuals with alcohol or drug abuse disorders accounted for estimated losses of $80.9 billion in lost productivity. Of this, it is estimated that $66.7 billion resulted from alcohol problems and $14.2 billion resulted from drug problems. These estimates were arrived at through analysis of data from the National Longitudinal Alcohol Epidemiologic Survey (NLAES; Grant et al. 1994) using the microsimulation techniques that have been used in studies of the RAND Health Insurance Experiment (e.g., Newhouse and the Health Insurance Group 1993; Manning et al. 1987; Duan 1983). The analyses reported below examine the employment and productivity effects of alcohol and drug abuse disorders.
Introduction
Chronic and severe alcohol and drug abusers are often thought to have great difficulties obtaining and keeping stable employment. This is seen in populations that enroll with publicly funded treatment providers. In the most severe cases, alcohol and drug abusers may be institutionalized, homeless, and/or spending much of their time committing crimes to generate the income necessary to maintain a drug habit. Drug or alcohol abusers may be frequently intoxicated or high and unable to work if they had a job. Many severe alcohol and drug abusers have never held a full-time job for any significant period of time. It is common for alcohol and drug abusers to move from one part-time job to another. Moreover, it is believed that many more alcohol and drug abusers are either temporarily out of work because of their alcohol or drug use (they have been fired or quit) or that they are "underemployed" in part because of their alcohol- or drug-abusing lifestyle.
Although the negative relationship between the presence and severity of an alcohol and/or drug abuse problem and an individual's labor supply and wage rate may seem logical, empirical evidence has been mixed, particularly in the case of drug abuse and dependence. Analysis of the relationship between alcohol or drug abuse and productivity is complicated by a variety of measurement and conceptual issues. These issues include the following:
- How to measure alcohol and drug abuse (consumption versus clinical measures);
- How to model the several potential interrelationships of income and alcohol or drug use or problems (e.g., as income increases, consumption may increase if drugs and alcohol are normal goods; and disability from alcohol or drug abuse or dependence may impair ability to work);
- How to measure lagged or cumulative effects over time;
- The extent to which individuals with alcohol or drug abuse disorders are underrepresented in general populations that are studied through surveys; and
- The role of co-occurring mental disorders, which can often be found among individuals with alcohol or drug disorders.
Viewing the reductions in productivity associated with alcohol and drug abuse as a cost to society is somewhat controversial. Some economists see the decision to consume or even to abuse alcohol or drugs as a rational, utility-maximizing choice. If an individual chooses to consume drugs or alcohol, even though such consumption reduces productivity, the loss is strictly to the individual and is at least offset by the perceived benefits the individual obtains. Following this line of logic, no costs are incurred by society because the drinker or drug user experiences all of the effects. The effects are said to be "internal" to the user, in contrast to effects that might "spill over" to those around the user through drinking and driving, alcohol- or drug-related crime, or poor performance (below expectations for the given wage) on the job. These are usually termed "external costs" and are widely regarded by economists as the only costs germane to decisions about taxation designed to correct for "spillover" effects that are associated with problem alcohol consumption.
However, the position taken in this report (and in essentially all prior major alcohol and drug abuse cost-of-illness studies) is to estimate all such costs, both internal and external. There are several reasons to justify this approach. First, such costs are often transferred to family members and the rest of society and thus are not completely internal (see chapter 3 and chapter 7 of this report). Second, in a more theoretical line, it can be argued that the choices made by drug and alcohol abusers are made under uncertainty with incomplete information. Abuse or addiction, and the associated negative consequences, are not desired outcomes, but rather are usually unintended outcomes resulting from inaccurate perceptions regarding the risks associated with alcohol and drug consumption. As such, the reduced productivity resulting from abuse or addiction may be viewed as costs to society and not solely as elements of a private decision calculus. Finally, documenting the costs that are borne by individuals with alcohol and drug abuse disorders may help support more informed individual decisionmaking about behaviors that could lead to the development of such disorders.
Estimates reported in this study are based on analysis of a recent national survey that has many strong attributes recommending it for this purpose: the NLAES. The NLAES data set allows analysts to categorize individuals based on the presence of alcohol and drug use disorders as well as major depression (one of the most prevalent mental disorders and the one most closely associated with alcohol use disorders). Those with multiple conditions can be distinguished from those with an alcohol, drug, or mental disorder diagnosis only. This categorization allows analysts to explore the impact of alcohol and drug abuse and comorbidity on productivity using two important labor market outcomes: employment and personal earnings.
The next section reviews previous studies on the impact of alcohol and drug abuse on labor market outcomes and briefly discusses the issue of comorbidity of alcohol or drug abuse and mental disorders. Following this literature review is a description of the NLAES data and the construction of study samples, estimates of the impact of alcohol and drug abuse on employment and personal income from the data set, and application of these estimates to compute the productivity losses associated with alcohol and drug abuse.
Previous Studies
A number of studies have analyzed the impact of drug and alcohol abuse on various aspects of work and productivity, including education, labor force participation, earnings or income, occupation, and job mobility. Findings of the most recent studies are summarized in table 5.7.
In general, studies of the impact of alcohol abuse or dependence on wages or income have found significant negative effects ranging from about 1 to 30 percent. Most of these studies use data from the Epidemiologic Catchment Area (ECA) survey. The ECA data allow categorization of individuals on the basis of a clinical definition of alcohol abuse or dependence, as opposed to a definition based simply on alcohol consumption patterns. The studies in table 5.7 that categorized individuals based on alcohol consumption found either positive or no effect of alcohol consumption on wages or income: People who consume alcohol did not have lower average income, but rather higher income. Studies using the ECA data have also found that a lifetime diagnosis of alcohol abuse or dependence (i.e., a diagnosis of alcohol abuse or dependence at any time in one's life) is associated with a reduced likelihood of working full-time and that symptoms of alcohol abuse occurring before age 18 are associated with a 1.5-year reduction in educational attainment.
Recent studies of the impact of drug use and abuse on wages or income are more numerous than those that analyze the impacts of alcohol abuse. The majority of these studies use data from various years of the National Longitudinal Survey of Youth (NLSY). These data allow for the categorization of individuals based on drug use (type of drugs used and the level of use), rather than on the basis of diagnostic criteria, and focus on individuals of ages 18 to 30. The findings from studies using the NLSY generally indicate that drug use is associated with a positive deviation in wages. Explanations given for this finding include the existence of unobserved characteristics that affect the observed relationship between drug use and wages (e.g., that individuals who use drugs, for unobserved reasons, are more productive than average), and the idea that drugs are a normal good, and because wages are a primary component of income, the relationship observed is simply an income effect (i.e., reverse causality). Several studies have attempted to address these issues with sophisticated econometric modeling (Kaestner 1991, 1994b); Register and Williams 1992; Gill and Michaels 1992) or by using longitudinal data to estimate models of drug use and wages that can control for individual characteristics over time. These studies still find positive or insignificant effects of drug use on wages.
Studies of the relationship between drug abuse and income using the ECA data have found the effects to be negative, but insignificant for males and positive for females. One study, however, estimates the effects separately for males ages 18 to 29 and males ages 30 to 45, finding a significant and negative effect of drug abuse or dependence on personal income for males in the 30-to-45 age group. These studies differ from those using the NLSY because they use a measure of drug abuse or dependence, as opposed to "any" drug use, and they include persons over age 30 in their samples.
The effects of drug use on labor force participation have been addressed by two recent studies, one using the diagnostic data available from the ECA and the other using the drug use data from the NLSY. These studies find significant negative effects of drug use or abuse on employment or labor supply for males (Kaestner 1994a; Buchmueller and Zuvekas 1994) but insignificant or positive effects for females (Kaestner 1994a).
The co-occurrence of alcohol or drug abuse and mental disorders has been shown to be quite common. Regier et al. (1990), using data from the ECA on persons both in institutions and in the community, estimate the lifetime prevalence of mental disorders to be 37 percent among persons with an alcohol disorder and 53 percent among persons with a drug disorder. None of the studies of the impact of drugs and alcohol on wages listed in table 5.7, however, have attempted to disentangle the effects of mental illness and alcohol or drug abuse among individuals with diagnoses of both conditions. Because mental disorders have large negative effects on wages and income (Bartel and Taubman 1979; Rice et al. 1990) and because the comorbidity of mental illness and alcohol or drug abuse is common, the true impact of alcohol and drug abuse on income may be either larger or smaller than estimated, depending on whether or not an indicator for mental illness is included in the specification. Of the studies described in table 5.7, those using data from the ECA typically have included indicators for the presence of mental disorders, whereas those using the NLSY have not.
It is important to incorporate psychiatric comorbidity in estimates of the impact of alcohol and drug abuse on reduced productivity. The discussion of psychiatric comorbidity in chapter 4 of this report suggests that assigning causality to alcohol and drug abuse or mental illness among persons with both types of problems is very complicated. This analysis has employed regression analysis in order to estimate the independent contributions of alcohol, drug, and depression disorders on earnings and employment.
Data Source
This section provides a description of the primary data source used in this analysis, the construction of study samples, and the key variables used in the analysis. This analysis of the relationship between alcohol and drug disorders and productivity (employment and earnings) has been performed using the 1992 NLAES (Grant et al. 1994). The NLAES is a nationally representative household survey using face-to-face interviews conducted by the U.S. Bureau of the Census for the NIAAA. More than 42,000 people ages 18 and over were interviewed in depth about their attitudes, beliefs, and behaviors with respect to alcoholic beverages and illicit psychoactive drugs. Additional questions examined symptomatology of depression, which is one of the most frequently co-occurring disorders with alcohol and drug abuse.
NLAES data can be analyzed to estimate current and lifetime prevalence of alcohol and illicit drug disorders among adults, based on clinical criteria from the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV; American Psychiatric Association 1994). Identification of individuals meeting clinical criteria for alcohol and drug disorders was based on responses to questions about alcohol- and drug-related behaviors, experiences, and related symptomatology that correspond to clinical criteria for diagnosing abuse of and/or dependence on alcohol and/or psychoactive drugs.
NLAES also obtained information about the demographic characteristics of respondents and further information about their socioeconomic characteristics and labor force and employment experiences. Information about current employment status, hours of work, and personal earnings are particularly useful for this analysis. The earnings variables exclude receipt of unearned income, such as dividends and social welfare benefits. Variables usually employed in analyses of labor force participation, employment, and earnings were collected in the survey. These include age, gender and ethnic background (basic demographic characteristics), marital status, presence of children in household, and "human capital" factors (such as educational attainment and primary occupation).
The survey achieved a more than 90-percent interview response rate. The survey used a complex sampling design, with oversampling of young adults and the African-American population. The sampling design requires analyses to employ specialized statistical routines that adjust tests of statistical significance for these factors. As in the primary publications of NLAES findings (e.g., Grant et al. 1994), tests of statistical significance reported in this study have been performed using the SUDAAN software (Research Triangle Institute 1992).
The Prevalence of Alcohol, Drug, and Mental Disorders
In examining the relationship of alcohol and drug disorders to employment success, this analysis has examined both lifetime and past-year prevalence of the disorders in question. However, for theoretical reasons, the focus has been directed to lifetime diagnosis over current or past-year diagnosis. Theoretically, a history of a major health problem (be it a general health problem, alcohol or drug problems, or a mental health problem) could affect current labor market success by having previously interrupted either one's educational attainment or the expected progression of one's career. Of course, many health problems do not have major or lasting effects, or such effects are successfully remediated through clinically assisted or self-rehabilitation efforts. Thus, current labor market success could be affected even with no apparent current impairment.
The prevalence of alcohol and drug dependence and of major depression for the NLAES sample appears in table 5.8. In the total population analyzed in this study (adults ages 18 to 64, not full-time students), 21.3 percent of males and 10.2 percent of females had ever met criteria for alcohol dependence. The male and female rates for lifetime drug dependence among the 18- to 64-year-old population were 4.5 percent and 2.9 percent, respectively. Prevalence rates also have been calculated and reported for major depression, because mental illness is commonly co-occurring with substance abuse disorders. (Depression was the only mental disorder assessed in NLAES; it is one of the most prevalent mental disorders.)
The ever-alcohol-dependent population is also subdivided into those with early and later initiation of drinking. About 3.6 percent of males and 1.5 percent of females ages 18 to 64 had ever been alcohol dependent and had started drinking (more than just "sips") before their 15th birthday (termed "early drinkers" in this report). Analyses reported below indicate that this subset of the ever-alcohol-dependent population appears to experience the most substantial and/or lasting effects of alcohol problems on labor market success.
The prevalence rates of alcohol and drug dependence pertain only to individuals that ever met the diagnostic criteria for "dependence," whether or not they had also met criteria for the diagnosis of "abuse." An extensive series of exploratory analyses revealed that a diagnosis of "abuse" for alcohol or drugs was rarely significantly related to differences in employment success or in earnings or wage rates, and in a few regressions it actually had a positive association with these outcomes.
Co-occurrence of alcohol, drug, and depression disorders raises the possibility of interactions among the disorders - that is, persons with multiple disorders might experience effects that differ from the sum of the individual effects from the several problems that they experience. This hypothesis was tested. Although there was some evidence from the logistic and ordinary least squares regressions to support this for males (but not at all for females), there were relatively small numbers of observations (on the order of 100) in certain of the disorder-combination cells. Because of this sparseness of observations, the analysis did not examine interactions between alcohol dependence with early and later drinking initiation and other disorders.
Analyses
Analytic Approach
The primary hypothesis of this analysis is that persons meeting criteria for a diagnosis (either prior or current) of alcohol or drug disorders will have impaired productivity in the workplace and therefore are likely to have lower than expected
- Employment (i.e., having a job and/or hours of employment);
- Wage rates; and/or
- Earnings.
This analysis has taken a straightforward approach to modeling the relationship of the primary employment-related outcome measures (specifically, employment, earnings, and wage rates) with respect to the following:
- Standard demographic and socioeconomic predictors of employment outcomes; and
- Indicators of alcohol, drug, and mental illness disorders.
The analytic approach is generally consonant with the analyses identified and summarized briefly in section 5.3.3. The analysis is built upon single-equation analyses that essentially assume that causality flows from the independent, or predictor, variables on the right-hand side of the equation to the dependent, or outcome, variables on the left-hand side of the equation. This is the primary approach that has been taken in the study of this problem. The variables that are used in the regression analysis later in this section are identified in table 5.9.
A central issue in this analytic approach is whether the dependent variables and the independent variables (notably employment outcomes and alcohol and drug disorders, respectively) are simultaneously determined by the individual. In fact, it is theoretically possible and empirically likely that certain employment-related values and alcohol and drug consumption behaviors may be simultaneous, which would dictate alternative modeling approaches for estimation of those relationships.
Specifically, standard neoclassical theory of consumer and labor market behavior predicts that individuals with higher wage rates might well have higher consumption of most goods and services (including alcoholic beverages and even psychoactive drugs). This tendency is known as the "income effect" of higher wages. However, theory and empirical studies recognize that the "income effect" can well be offset by other factors, including the "substitution" effect of employees working less in order to "consume" more leisure. This analysis recognizes the theoretical possibility that individuals with higher wage and employment income may consume more alcoholic beverages and psychoactive drugs - which indeed a number of prior studies have found.
This analysis is directed at alcohol and drug disorders rather than consumption levels. "Disorders" refers to the specific sets of symptoms related to intensive and/or extensive use of alcohol and/or psychoactive drugs that meet established clinical standards for existence of alcohol or drug disorders. As noted above, NLAES was designed around the DSM-IV clinical standards to allow potential diagnoses to be assigned to NLAES respondents based on their answers to specific sequences of questions concerning alcohol and drug experiences and symptoms.
Disorders are theorized to be less subject to the endogeneity and simultaneous equations problems that exist with consumption variables. Specifically, users do not generally decide to have alcohol and drug disorders, although they do make choices about levels of consumption before the disease process is initiated. This analysis maintains the hypothesis that alcohol or drug dependence or abuse are undesired, undesirable, and generally unexpected consequences of consumption. Although dependence and abuse are certainly related to levels and patterns of consumption, there is increasing evidence that dependence and abuse are disorders that have components of genetic vulnerability (NIAAA 1997). Thus, two individuals who have identical drinking (or psychoactive drug use) patterns may have quite different probabilities of experiencing alcohol and drug disorders.
In a formal sense, measures of alcohol and drug disorders are more appropriate for empirical analyses as "exogenous" (i.e., predetermined) variables than are measures of alcohol or drug consumption because disorders are undesired and undesirable outcomes that happen with uncertain probability.
NLAES has further detail about whether the respondent met criteria for alcohol/drug dependence/abuse either within the year preceding the interview or at any time previously. The analyses explored the various alternative measures. This analysis has primarily used "lifetime" measures, meaning whether the individual had ever met the diagnostic criteria in his or her life. However, exploratory analyses also examined the impact of "current" disorders (i.e., meeting diagnostic criteria for the disorder during the year prior to the interview). This report presents estimates for the measures that were most robust and consistent in the empirical results. Limited discussion is given to the findings for the alternative measures.
- Table 5.7: Studies of the Impact of Alcohol and Drug Abuse on Productivity
-
Table 5.7: Studies of the Impact of Alcohol and Drug Abuse on Productivity Study Data and Sample Principal Findings Alcohol Studies Berger and Leigh 1988 Quality of Employment Survey (1972-73), males and females age 18+ Positive effect of current alcohol consumption on wages. Estimated wages of drinkers exceed those of nondrinkers by 8-57% for men and 26-40% for women. Heien and Pittman 1989 National alcohol survey (1979) Problem-drinking status endogenous. No significant effect of alcohol consumption on family income after accounting for this endogeneity. Mullahy and Sindelar 1989 ECA (New Haven site only), males ages 25 to 59 Symptoms of alcohol abuse at young ages are associated with a 1.5-year reduction in educational attainment. No significant effect of current symptoms on current earnings. Rice et al. 1990 ECA (multiple sites), males and females ages 18 to 64 Lifetime diagnosis of alcohol abuse is associated with a 1-9% reduction in personal income for males and a 1-19% reduction for females. Magnitude depends on age. Mullahy and Sindelar 1991a,b ECA (multiple sites), males and females ages 18 to 64 Lifetime diagnosis of alcohol abuse/dependence is associated with reductions in the likelihood of working full time (7-19%) and personal income (3-23%). Mullahy and Sindelar 1991a ECA (New Haven only), males ages 22 to 64 (mainly ages 30 to 59) Lifetime diagnosis of alcoholism is associated with reductions in the likelihood of working full time and in household income (17-31%). The magnitude of the estimates are sensitive to the model specification. Drug Studies Rice et al. 1990 ECA (multiple sites), males and females ages 18 to 64 Lifetime diagnosis of drug abuse/dependence associated with a 1-9% reduction in personal income for males (not significant) and positively associated with personal income for females (not significant). Kandel and Davies 1990 NLSY (1984 and 1985), males ages 18 to 27 Use of marijuana is associated with an increase in employment gaps and the number of weeks unemployed. Use of cocaine increases job mobility, gaps, and unemployment. No effect of recent drug use on current wages controlling for previous years' wages. Kaestner 1991 NLSY (1984) males and females ages 18 to 27 Use of cocaine and/or marijuana has positive effect on the wages of males and females after controlling for selectivity and endogeneity of drug use. Register and Williams 1992 NLSY (1984), males ages 18 to 27 Predicted use of marijuana or cocaine negatively associated with employment, but long-term and on-the-job use are positively associated. Overall positive effect of marijuana use on wages. No significant effect of cocaine use on wages. Gill and Michaels 1992 NLSY (1980 and 1984), males and females ages 18 to 27 Use of drugs significantly reduces the probability of employment. No significant effect of "hard" drugs on employment. Drug users have higher wages but lower returns to human capital characteristics than nonusers. Kaestner 1994b NLSY (1984 and 1988), males and females ages 23 to 30 Positive effect of drug use on wages in cross-sectional model for males and females. Negative effect of drug use on wages for males (not significant) and positive effect (significant) for females in longitudinal fixed-effects model. Kaestner 1994a NLSY (1984 and 1988), males and females ages 21 to 30 Marijuana and cocaine use have negative impact on labor supply in a cross-sectional analysis, particularly among males. Longitudinal analysis indicates no effect of illicit drug use on labor supply for males or females. Buchmueller and Zuvekas 1994 ECA (multiple sites), males ages 18 to 45 Drug abuse has a significant negative impact on the likelihood of employment and on the income of males ages 30 to 45 but no effect on the employment or income of males ages 18 to 29. Buchmueller and Zuvekas 1996 ECA (multiple sites), males ages 18 to 45 Drug abuse or dependence has a significant negative impact on income among young and prime-age males. Although moderate drug use is associated with higher income among young males, this disappears among older males.
- Table 5.8: Lifetime Prevalence of Alcohol, Drug, and Depression Disorders
-
Table 5.8: Lifetime Prevalence of Alcohol, Drug, and Depression Disorders, Persons Ages 18 to 64, by Gender Diagnosis NLAES Percent With Lifetime Diagnosis Males Females Alcohol dependence 21.3 10.2 and early drinking 3.6 1.5 and later drinking 17.8 8.7 Drug dependence 3.7 2.2 Major depression 9.9 13.3 Source: Analysis of the National Longitudinal Alcohol Epidemiologic Survey (NLAES) (adults ages 18 to 64, excluding full-time students).
- Table 5.9: Variables Used in the Regression Analyses
-
Table 5.9: Variables Used in the Regression Analyses Dependent Variables Primary Alcohol and Drug Variables Control Variables - Employment in past month
- Earnings in past month
- Wage rate (earnings/hours worked)
Variables used in reported models
- Alcohol dependence and early drinking and later drinking
- Drug dependence ever
Variables tested in models not reported
- Alcohol dependence: current; prior to past year
- Alcohol abuse: current; prior to past year
- Alcohol consumption (current)
- Drug dependence: current; prior to past year
- Drug abuse: current; prior to past year
Reduced model
- Gender
- Age
- Ethnic background
- Rural/urban residence
- Children in household
- Depression disorder ever
Full model adds these variables to the reduced model
- Educational attainment
- Skilled profession
- Marital status