Health Effects Of Illegal Drug Use
The literature that seeks to determine the impact of drug use on health is reviewed. As cannabis is the most widely used illegal drug, the focus is on the relationship between cannabis use and health. Much of the research in this area is contributed from epidemiology and is focused on the mental health effects of drug use. There is a smaller and more recent literature contributed by economics. A distinguishing feature of this literature is the utilization of methodologies designed to identify causal effects of drug use. In addition to studying the direct effects of drug use on health, the economics literature also considers the impact of drug use on labor market outcomes. Because a significant cost of poor health resulting from drug use is considered to be reduced labor market success, understanding the evidence regarding the indirect health effects of drug use is of significant interest. The indirect health effects of illegal drug use which originate, from effects on crime, violence, traffic accidents, etc., are not discussed. The relative contribution of illegal drug use to the economics costs of risky behavior more generally is discussed by Cawley and Ruhm (2011). They find that illegal drug use makes a modest contribution to these costs.
Medical And Epidemiological Literature
The earliest attempt to identify the causal impact of cannabis use on mental illness is by Andreasson et al. (1988) who study a cohort of more than 50 000 18 to 20-year-old Swedish conscripts. The authors find that the postconscription risk of developing schizophrenia is increasing in the number of times cannabis is used before conscription. This was a controversial finding and prompted a raft of epidemiological studies on the relationship between cannabis use and mental health more generally. This literature is so large that there is now a large number of studies dedicated to reviewing it.
In their 2003 review, Degenhardt et al. (2003) conclude that there is a modest but significant association between earlyonset regular or problematic cannabis use and depression later in life, although there is little evidence of an association between depression and infrequent cannabis use. The authors go on to conclude that even if the association between cannabis use and depression is assumed to be causal, regular cannabis use can only explain a small proportion of depression in the population. Macleod et al. (2004) review more than 200 studies based on longitudinal data that seek to determine the psychosocial impact of cannabis use. They conclude that although there is evidence of associations between cannabis use and various measures of psychosocial harm, the extent of the associations and the strength of the evidence is not always large. Furthermore, the authors conclude that the causal nature of the associations is far from clear.
Many of the overview studies have focused on the relationship between cannabis use and psychosis. Arseneault et al. (2004) conclude on the basis of their review of previous research that cannabis use is likely to have a causal role in the development of psychosis but the magnitude of its impact is unclear. Kalant (2004) concludes from his review of previous studies that there is more evidence for a causal relationship running from cannabis use to psychiatric problems than there is for reverse causality, i.e., psychiatric problems leading to cannabis use. Henquet et al. (2005) review seven studies and conclude that cannabis use has a causal effect on later psychosis. They note, however, that the effect is not very large and the mechanism underlying the causality is unclear. Semple et al. (2005) provide an overview of 17 case–control studies that examined the association between cannabis use and schizophrenia or schizophrenia-like psychosis. They also conclude that cannabis is a risk factor for psychosis but indicate that it is not clear whether cannabis is a precipitating or a causative factor in the development of schizophrenia. Hall (2006) argues on the basis of his review of studies that there is a strong association between cannabis use and psychosis, but it remains controversial whether the association is causal. Moore et al. (2007) present an overview of 11 studies on psychosis based on data from seven cohort studies. Although they find that there is an association between cannabis use and psychosis, they are unable to rule out spurious correlation resulting from unobserved confounding factors as the underlying explanation for this association.
In their recent review, Hall and Degenhardt (2009) argue that previous research on the relationship between mental health and illegal substance use has produced mixed findings, with some papers reporting a positive association between cannabis use and mental health problems and others reporting no association. McLaren et al. (2010) review the methodological strengths and limitations of major cohort studies that have sought to determine the causal nature of the relationship between cannabis use and psychosis. The authors conclude that, on the basis of the current studies, no inference can be made about a potential causal relationship from cannabis use to psychosis. Discussing a variety of papers Werb et al. (2010) conclude that the research to date is insufficient to conclusively claim that the association between cannabis use and psychosis is causal in nature. The fact that populationlevel rates of psychotic disorders do not appear to correlate with population-level rates of cannabis use suggests that these two phenomena may not be causally related.
Econometric Studies – Direct Health Measures
In examining the relationship between mental health and cannabis use, the literature from epidemiology cited above has attempted to identify the causal effect of cannabis use by controlling for observed factors that may be a source of confounding. However, as noted by Pudney (2010), the potential for unobserved common confounding factors makes inference regarding the causal impact of cannabis use difficult. In contrast, economic research routinely makes use of statistical techniques designed to account for unobserved confounding factors in studying the impact of one outcome on another. Despite the potential to provide strategies for addressing the issue of unobserved confounders, and thus better assess the health risks faced by drug users, there are very few contributions from the economics literature on this issue. As detailed below, the economic studies that do attempt to tease out causal effects suggest that there may be risks to both mental and physical health from using cannabis.
Williams and Skeels (2006) and van Ours and Williams (2011) use Australian data to study the impact of cannabis use on physical and mental health, respectively. Williams and Skeels (2006) find the probability of reporting very good or excellent self-assessed health to be 8% lower among those who consumed cannabis in the past year compared with those who had not, and 18% lower for those who reported weekly use. Along similar lines, van Ours and Williams (2011) find that cannabis use increases the likelihood of mental health problems, with the probability of experiencing mental distress increasing with the frequency of past year use. Although each of these studies considers a single dimension of health, there is significant evidence that poor mental health is correlated with poor physical health. van Ours and Williams (2012) investigate the impact of cannabis use on health in a framework that accounts for the potential for shared frailties in the domains of physical and psychological well-being, as well as selection into cannabis use. Their analysis of Amsterdam data suggests that cannabis use reduces the mental well-being of men and women and the physical well-being of men. Although statistically significant, the magnitude of the effect of using cannabis on mental and physical health is found to be small.
van Ours et al. (2013) is the only study to address both the potential for common unobserved confounders and reverse causality in studying the health impact of cannabis use. Their analysis of the relationship between suicidal ideation and cannabis use is based on a 30-year longitudinal study of a birth cohort. They find that intensive cannabis use – at least several times per week – leads to a higher transition rate into suicidal ideation for susceptible males. There is no evidence that suicidal ideation leads to regular cannabis use for either males or females.
Econometric Studies – Indirect Health Measures
In addition to their stock of human capital, a person’s labor market productivity is determined by their health capital stock (Grossman and Benham, 1974). Drug use is conjectured to reduce labor market productivity through its deleterious effects on an individual’s stock of health. Although intuitively appealing, empirically assessing the validity of this conjecture is complicated by the fact that individuals choose, or self-select into, drug use. Specifically, there may be important unobserved determinants of wages or employment that also influence the decision to use drugs. An example of an omitted variable particularly relevant in this context is an individual’s discount rate. Individuals who discount the future heavily are more likely to use drugs because they place little weight on the future negative health consequences of their drug use (Becker and Murphy, 1988). They are also more likely to choose jobs with little investment in on-the-job training, and that consequently pay relatively high current wages but relatively low future wages. This may give rise to a positive correlation between drug use and wages even if drug use is negatively causally related to wages. Similarly, individuals with strong preference for leisure may also be more likely to use drugs if drug use and leisure are complements in the production of euphoria. Such a relationship would produce a negative correlation between drug use and labor supply even in the absence of a causal effect of drug use on labor supply.
The empirical strategy pursued by the first-wave studies for estimating the causal impact of drug use on wages and employment is instrumental variables. Three of these studies draw on data on 18 to 27-years old from the 1984 cross section of the National Longitudinal Survey of Youth (NLSY) and all three studies found evidence that, rather than reduce wages, drug use increases wages. Kaestner (1991) finds that for males, drug use measured as past 30-day use of cannabis, lifetime use of cannabis, past 30-day use of cocaine, or lifetime use of cocaine, raises hourly wages. Similarly, male wages are found to be increasing in the frequency of cannabis use in the past 30 days by Register and Williams (1992). Gill and Michaels (1992) report that the use of any drugs in the past year or any hard drugs (cocaine, heroin, inhalants, psychedelics, other drugs, other narcotics) in the past year increases the hourly wage rate received in a combined sample of males and females. The estimated magnitudes of the wage effects are quite large. For example, Kaestner (1991) estimates that males who have tried cannabis earn 18% more than otherwise similar males who have not tried cannabis, Register and Williams (1992) estimate that using cannabis on one more occasion per month increases hourly wages by 5%, and Gill and Michaels (1992) find that drug users earn approximately 4% more per hour than nonusers, and that hard drug users earn approximately 10% more per hour than nonhard drug users. Moreover, both Kaestner (1991) and Gill and Michaels (1992) report that the premiums for drug use are attributable to unobserved differences between the users and nonusers and not differences in returns to human capital and other characteristics.
Kaestner (1994a,b) uses the 1984 and 1988 waves of the NLSY to compare cross-sectional and longitudinal estimates of the impact of cocaine and cannabis use on labor supply and wages, respectively. He finds that the results based on the 1984 data, which show that cannabis and cocaine use increases wages and cannabis use decreases hours spent working in the sample of males, cannot be replicated using the 1988 data. Moreover, when unobserved differences that affect drug use and labor market outcomes are controlled for through a fixedeffect estimator, drug use is found to have a negative but insignificant impact on wages for males (Kaestner, 1994b), and mixed, although generally insignificant, effects on hours worked (Kaestner, 1994a). The overall conclusion reached by Kaestner is that drug use does not have a systematic impact on labor supply or wages.
The counterintuitive and inconsistent findings of the above studies motivated a second wave of economic research into the impact of drug use on wages and labor supply. Taken at face value, most of the second-wave studies tend to find evidence that nonproblematic use of drugs (light to moderate use, or the use of soft drugs) has no impact on labor supply, measured by employment or hours worked, but that problematic use (heavy use, or the use of hard drugs) does, although Burgess and Propper (1998); DeSimone (2002); Zarkin et al. (1998) and van Ours (2006) provide counterexamples. Similarly, most of the second-wave studies find that infrequent or nonproblematic drug use has no impact on wages, whereas problematic use does have negative wage effects. Once again, there are also exceptions to this generalization, such as MacDonald and Pudney (2000). It is noteworthy that many of these studies (especially those based on US data) tend to treat drug use as exogenous to labor market outcomes.
Focusing on the studies that are more rigorous in their efforts to address the potential endogeneity of drug use, the results are mixed. For example, although van Ours (2007) finds that using cannabis at least 25 times in one’s lifetime reduces the wage of prime-age males, the use of cocaine is found to have no effect, and MacDonald and Pudney (2000) are unable to detect any impact of either hard or soft drug use on their proxy for wages, that is, occupational attainment. Similarly, with respect to the employment of males, DeSimone (2002) finds that both past year cannabis and cocaine use reduces the probability of employment, whereas, MacDonald and Pudney (2000) find no employment impact of soft drug use (which includes cannabis) and van Ours (2006) finds no impact of cannabis or cocaine use on employment. Finally Conti (2010) introduces cognitive ability as additional variable in a wage equation with cannabis use as explanatory variable, showing that this causes the effect of cannabis use to become insignificantly different from zero.
Given the conflicting nature of the empirical findings, it is simply uncertain as to whether there are negative labor market consequences of drug use in general, and cannabis use in particular. Furthermore, it is unclear as to whether this literature should be interpreted as reflecting a lack of robust evidence of a negative health effect of drug use, or as reflecting the presence of a productivity improving effect of drug use that is confounding the negative health effects.