Drug Use by Workers,

Does It Reduce Productibity?


David F. Duncan, Dr.P.H.


Paper presented at the Annual Meeting of the

American Public Health Association

Philadelphia, Pennsylvania

November 11, 2002





Of the estimated 11.8 million adult illicit drug users in the U.S., 9.1 million or 77 percent were employed either full time or part time (S.A.M.H.S.A., 2002). Many find this a very distressing fact. We are warned that these employed drug users are a burden to their employers and a risk to the public. Mark de Bernardo, Special Counsel for Domestic Policy and Manager of Labor Law of the United States Chamber of Commerce, testified before a Congressional committee in 1987 that recreational drug users in the workplace are

Politicians, employers, marketers of drug tests, and the mass media echo similar views.

We are also told that drug users are costing society billions of dollars and that a large part of this is due to decreased productivity by drug users. One report (Harwood, Fountain, and Livermore, 1998) estimated that the cost to the nation of substance abuse in 1992 was $246 billion, but sixty percent of this was for alcoholism and only forty percent was due to use of illicit drugs – still, at $98 billion, a lot of money. A more recent and more focused report (O.N.D.C.P., 2001) estimated the costs due to illicit drug use in 2000 to be nearly $161 billion, with sixty percent or $110 billion of that amount being due to lost productivity. The implication often is that drug users in the workplace are less productive and are costing society and their employers those billions of dollars. As we shall see, this is not a valid implication.

These concerns have served as justification for both pre-employment and random post-employment urine testing for illicit drugs. Although media coverage might lead one to conclude that such testing is increasingly common, the fact is that a clear trend in the opposite direction appears to have emerged. More than ten years ago the Bureau of Labor Statistics (1991) reported that "overall about 1 of 3 establishments that reported having a drug testing program in 1988 said they did not have one in 1990." A more recent survey by the American Management Association (2000) shows that this is a continuing trend, with employee drug testing in 2000 being at its lowest level in a decade. The American Marketing Association’s annual surveys had revealed that most drug testing in American industry is due to government mandates requiring testing, not due to the business judgment of employers (American Management Association, 1996).

This doesn’t necessarily mean that employers are unconcerned about employee drug use or doubt the common view that drug users are poor workers, but many apparently have concluded that any benefits are not worth the costs of urine screening. One employer estimated that the cost of finding each positive result was $20,000. After testing 10,000 employees, he only found 49 positive results. A congressional committee estimated that the cost of each positive in government testing was $77,000 because the positive rate was only 0.5%. ("Workplace Substance Abuse Testing," 1992). These high costs for detecting a few drug users have begun to seem excessive to many employers.

Worse news for employers who use urine screening with their employees is the finding reported by Shepard and Clifton (1998) that companies adopting drug testing programs had lower levels of productivity than their counterparts that did not. "Both pre-employment and random testing of workers are found to be associated with lower levels of productivity" (p. 1).

What evidence actually exists to support the figures so widely cited in support of employee drug testing? The answer is three sources: the "Firestone Study," the RTI Study, and the ONDCP Reports. Let’s look at what these studies actually tell us and what they don’t.

The Firestone Study

Those figures given by Mr. de Bernardo above and (with some variation in the exact numbers stated) by many other advocates of drug testing usually cite, directly or indirectly, the Firestone Study. The problem is that there never was any Firestone Study, or at least none on this topic.

As Morgan (1988) has previously reported, the "Firestone Study" was actually an interview with the director of the Firestone Company’s Employer Assistance Program ("EAP"), a Mr. Ed Johnson, published in an internal company newsletter. Mr. Johnson made a number of statements about absenteeism, but these referred to alcoholics who had been served by the Firestone EAP. It is not clear whether these figures were actual statistics or informed guesses based on his clinical experience, but any generalization of them to all alcoholics employed by Firestone would be unjustified. More important still, they had no relevance at all to users of illicit drugs but that is who they are constantly being applied to.

The RTI Report

The "RTI Report" refers to an analysis conducted by researchers with the Research Triangle Institute. This analysis (Harwood, Napolitano, Christensen, and Collins, 1984) relied on the 1982 version of the National Household Survey on Drug Abuse, which included a question for the adult respondents about daily marijuana use. If the respondent stated that they had ever used marijuana daily for twenty days in one month, their household income was compared to another household without a daily user. The "user" households were found to have a household income 27.9% less than those households that did not have such an interviewee.

These results were then used to estimate the reduced productivity due to daily marijuana use. The prevalence rates of ever using marijuana daily were calculated by age and sex groups and were applied to the number of persons in the labor force from those groups. The number thus derived was multiplied by the expected income for each household. This led to their conclusion that "the loss due to marijuana abuse was estimated at $34.2 billion for 1980."

There are numerous validity questions regarding the reduced income figure itself. It relies upon interviewee estimates of household income – niterviewees who were not necessarily the household wage earner and who may, in any case, have little idea of what their household’s income really is. Furthermore, the income data were grouped in seven categories, with intervals as broad as $10,000, creating serious problems for deriving point estimates.

Calculation of reduced productivity required comparison to control household, to which data the same problems applied, but there also are matching issues. The control households were selected that had a respondent who resembled the marijuana use respondent in terms of age, education, marital status, and occupation. Respondents and households, however, were not matched in terms of race, social class, location, or number of residents in the household. This would mean that if an unmarried Black marijuana user living in an inner city slum had a lower household income than a non-user respondent of the same age, education, etc. who lived in the suburbs with her father, a physician, and her mother, a lawyer, the difference in their household incomes would be attributed entirely to the first respondent’s past daily marijuana use.

To their credit the RTI researchers clearly stated, that the association between marijuana use and reduced income was not necessarily causal. Did the marijuana users’ drug use cause them to earn less, or might lower income predispose one to use marijuana daily? Or are both due to some of the other factors that were left unexamined in the RTI study? There is no way to tell but the study is constantly cited as if it had proven that marijuana use caused the lower incomes.

Entirely ignored by those who cite the RTI Report as proof that drug use results in lower productivity is the fact that no effects on income were found for illicit drugs other than marijuana. Is it really plausible that marijuana has such effects on income, but that use of heroin or cocaine does not?

The use of income as a measure of productivity raises some serious questions. Are all persons with the same income equally productive? Is a physician at a free clinic less productive than one in private practice?

The use of employment as a proxy measure for productivity in other studies is equally problematic. Is a volunteer less productive than a salaried employee who performs the same tasks? Is a mother who stays at home to care for her child being unproductive?

The O.N.D.C.P. Reports

The "ONDCP Reports" were prepared by the Lewin Group. These reports prepared under a contract from the White House office of National Drug Control Policy and built on the prior work of Rice et al. (1990, 1991) to develop a more complete estimate of the economic costs of illicit drug abuse. The first report (Harwood, Fountain, and Livermore, 1998) estimated the costs for 1992, while the second (O.N.D.C.P., 2001) developed estimates for the years from 1992 thru 1998 with projections for 1999 and 2000. The economic costs of drug abuse for 2000 were estimated to be nearly $161 billion, of which almost $111 billion was "lost productivity. Productivity costs, however, included many costs other than reduced productivity by drug-using workers, such as deaths, health care costs, and law enforcement and incarceration costs. Lost productivity by the drug users was actually estimated at only about $25 billion.

These analyses used in the first report are more fully and completely described than those of the second report. The estimates of productivity loss due to drug abuse in the first report were derived from an analysis of data from the National Longitudinal Alcohol Epidemiologic Survey (NLAES). The NLAES (Grant et al. 1994) is a nationally representative household survey using face-to-face interviews of more than 42,000 persons aged 18 and older. The NLAES data can be analyzed to estimate current and lifetime prevalence of alcohol and illicit drug disorders among adults, based on DSM-IV clinical criteria (American Psychiatric Association, 1994). The NLAES also obtained information about the demographic characteristics of respondents and further information about their socioeconomic characteristics and labor force and employment experiences, including current employment status, hours of work, and personal earnings.

Respondents were categorized based on lifetime prevalence of drug dependence (other than alcohol). 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. Since such a positive result was, of course, taboo the DSM diagnosis of drug abuse was not included in the main analyses.

Analyses were performed separately for males and females. These analyses used logistical regressions and ordinary least squares regressions to examine whether persons with a history of drug dependence differed from the general population in terms of employment rates or earnings per hour (monthly earnings divided by hours worked). They ran these analyses on both a "reduced model" in which only basic demographic factors were included as control variables and a "full model" which included additional variables such as educational attainment, white collar status, and marital status. They also applied the microsimulation techniques that have been used in studies of the RAND Health Insurance Experiment (Newhouse and the Health Insurance Group 1993).

The results showed that for males drug dependence was not significantly associated with employment status in either the reduced or the full model. Drug dependence was found to have a negative and significant association with wage rates and earnings of males, although the estimated impact was lower in the full model than in the reduced model. Drug dependence was significant predictor of lower employment among females. Women’s wage rates on the other hand did not have a statistically significant association with drug dependence. Despite the inconsistencies in their results, the researchers developed multipliers, which they applied to the population prevalences of drug dependence in the population and arrived at an estimate of lost productivity due to drug dependence.

It is not clear whether these same multipliers were applied in the second report or if new multipliers were derived through some unspecified analyses. What is evidently different is that the second report developed the estimates of lost productivity using population estimates for the lifetime prevalence of "hardcore drug abuse" rather than for DSM-IV drug dependence. Hardcore drug abuser appears to have been defined as a person who reported having used either marijuana or cocaine on 100 or more days during their lifetime. It appears that the lifetime prevalences for marijuana and cocaine were added together without regard for the likelihood that some persons were being counted twice by virtue of having used both drugs on more than 100 days of their life.

Less Cited Studies

Kaestner (1991) analyzed data from the National Longitudinal Survey of Youth and found that increased illicit drug use (marijuana and cocaine) led to greater wages across age groups and genders. Gill and Michaels (1992) replicated Kaestner’s findings on the same dataset and extended the analysis to model likelihood of labor market participation. They estimated wage equations for users and non-users taking self-selection into account and found a strong positive association between drug use and higher wages. Register & Williams (1992) analyzed the same data, excluding female subjects and using a continuous measure of drug use and a different method for controlling for self-selection into drug use and the labor market. They found that " the net effect for all marijuana users...was positive".

Kaestner (1994a, 1994b) extended his cross-sectional analysis longitudinally through two waves of the NLYS and found results that are more complex. Cocaine use was positively associated with wages for females while marijuana use was negatively associated with wages for males. Timing however was important, with current marijuana use by males being positively associated with wages, while lifetime use was negatively associated.

Burgess and Propper (1998) examined data from a follow-up cohort of the NLYS followed to age 34-35 and found that socially deviant behavior (such as alcohol and drug use) in adolescence was associated with positive later life outcomes such as productivity and family formation. Heavy drug use in adolescence, however, was negatively associated with employment and earnings in adulthood.

MacDonald and Pudney (1998) examined data from the 1994 British crime Survey on two cohorts – 16-29 year olds and 30-59 year olds. In the younger cohort past marijuana use was negatively associated with employment but positively associated with career success for those who were employed. No effects were found for past hard drug use. No effects on career success among those employed were found for current use of either marijuana or "hard drugs," but current use of both was associated with greater unemployment.

In the older cohort, no evidence was found of any effect of current drug use on the wages of employed persons. Current marijuana use, but not current hard drug use, increased the likelihood of unemployment. Past "hard drug" use in this cohort was not associated with unemployment but past use of marijuana bore a weak negative association with employment.

Across both cohorts marijuana use by males had a positive association with occupational achievement that diminshed with age. For females no association was found between marijuana use and occupational achievement. For both sexes "hard drug" use was associated with greater levels of unemployment but not with achievement once employed. In general, the authors concluded that "any adverse effects of soft drugs appear not to be large or permanent."

Bray, Dennis and French (2000) used data from the National Household Surveys on Drug Abuse (NHSDA) to estimate simple descriptive statistics and analysis of variance (ANOVA) models of the relationship between symptoms of drug dependence and labor market outcomes for alcohol, tobacco, marijuana, and other (illicit) drugs. For males, they found that symptoms of dependence were associated with both lower employment rates and fewer hours of work. For females, they found that symptoms of dependence were associated with lower employment rates, but were not related to number of hours worked. They recommended that policymakers and researchers should consider the full spectrum of drug use and dependence rather than focusing on the simple use of an illicit drug.

In further analyses of this data, French, Roebuck, and Alexandre (2000) reported that male chronic drug users were less likely to be employed but that no such effect was found for female chronic drug users. "Perhaps the most important finding of this study, however, was the lack of any significant relationships between nonchronic drug use, employment, and labor force participation." They concluded that neither "light or causal drug use" was associated with any negative effects on employment. This led them to conclude "that employers and policy makers should focus on problematic drug users in the same way that they focus on problematic alcohol users."

In Conclusion

It seems possible at this point to draw five conclusions from this review:

    1. No study I found actually measures any impact of drug use or abuse on worker productivity directly.
    2. Employment rates and average wages have served as proxy measures for productivity in the studies examined.
    3. The studies examined provide some support for the hypothesis that drug dependence has a negative impact on these proxies for productivity but almost no support for the hypothesis that non-dependent use has any such effects.
    4. Accepting these proxies as measures of productivity, the impact of drug abuse on productivity, while quite large, is substantially smaller than is commonly suggested.
    5. That the impact of alcohol abuse on productivity as measured in these studies is much greater than that of illicit drug abuse.

These findings provide little reason to believe that urine-screening employees for illicit drug use will do anything to improve productivity. A urine screen cannot distinguish a drug abuser from a drug user nor can it distinguish the person who uses drugs on the job from the one who does not. Employers and society at large would be better off if they focused on identifying drug abusers and left the drug users alone. Drug abusers in the workplace can be far more adequately identified by monitoring absenteeisnm, late arrivals, and poor work performance than by testing their urine and these are matters any employer should be monitoring anyway.


American Management Association (1996). A.M.A. Survey on Workplace Drug Testing and Drug Abuse Policies. New York, NY: American Management Association.

American Management Association (2000) AMA Survey on Workplace Testing: Medical Testing: Summary of Key Findings. New York, NY: American Management Association.

American Psychiatric Association (1994). Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition. Washington, DC, American Psychiatric Association.

Bray JW, Zarkin GA, Dennis ML, French MT. (2000). Symptoms of dependence, multiple substance use, and labor market outcomes. American Journal of Drug and Alcohol Abuse, 26(1), 77-95.

Bureau of Labor Statistics (April, 1991). Anti-Drug Programs in the Workplace: Are They Here to Stay?" Monthly Labor Review, pp. 26-28.

Burgess, S. M., and Propper, C. (1998). Early health-related behaviour and their impact on later life chances: Evidence from the U.S.. CASE Paper #. London: London School of Economics, Centre for the Analysis of Social Exclusion.

de Bernardo, M. (1987). Statement. In: Proficiency Standards for Drug Testing Laboratories: Hearings Before a Subcommittee of the House Committee on Government Operations, 100th Congress, 1st Session, 91-92.

French, Michael T., M. Christopher Roebuck, and Pierre Kebreau Alexandre (2000). Illicit Drug Use, Employment, and Labor Force Participation. Southern Economic Journal, 68(2), 366.

French MT, Zarkin GA. (1998). Mental health, absenteeism and earnings at a large manufacturing worksite. Journal of Mental Health Policy and Economics, 1(4):161-172.

Gill, A., & Michaels, R. (1992). Does drug use lower wages? Industrial and Labor Relations Review, 45(3), 419-434.

Grant, B.; Harford, T.; Dawson, D.; Chou, P.; DuFour, M.; and Pickering, R. (1994). Prevalence of DSM-IV Alcohol Abuse and Dependence: United States, 1992. Alcohol Health & Research World 18(3):243-248.

Harwood, H., Fountain, D., and Livermore, G. (1998). The Economic Costs of Alcohol and Drug Abuse in the United States – 1992. Rockville, MD: National Institute on Drug Abuse. Accessed online 10/25/02 at http://www.drugabuse.gov/EconomicCosts/Index.html

Harwood, H.; Napolitano, D.M.; Christensen, P.L.; and Collins, J.J. (1984). Economic Costs to Society of Alcohol and Drug Abuse and Mental Illness: 1980. Report to Alcohol, Drug Abuse, and Mental Health Administration. Research Triangle Park, NC: Research Triangle Institute.

Kaestner, R. (1991). The effects of illicit drug use on the wages of young adults. Journal of Labor Economics, 9(4), 381-412.

Kaestner, R. (1994a).. The effect of illicit drug use on the labor supply of young adults. Journal of Human Resources, 29(1), 126-155.

Kaestner, R. (1994a). New evidence on the effects of marijuana and cocaine use on wages. Industrial and Labor Relations Review, 47(3), 454-470.

MacDonald, Z, & Pudney, S. (1998). Illicit drug use and labour market achievement: Evidence from the U.K.. Public Sector Economics Centre, University of Leicester. Accessed online 2/18/02 at http://www.le.ac.uk/economics/research/RePEc/lec/lpserc/pserc98-2.pdf

Morgan, J. (1988). The "scientific" justification for urine drug screening. Kansas Law Review, 36, 683-697.

Newhouse, J., and the Health Insurance Group. (1993). Free-For-All: Health Insurance, Medical Costs, and Health Outcomes; The Results of the Health Insurance Experiment. Cambridge, MA: Harvard University Press.

Office of National Drug Control Policy (2001). The Economic Costs of Drug Abuse in the United States: 1992 - 1998. Washington, DC: Office of National Drug Control Policy. Accessed online 10/25/02 at http://www.whitehousedrugpolicy.gov/publications/pdf/economic_costs98.pdf

Register, C., & Williams, D. (1992). Labor market effects of marijuana and cocaine use among young men. Industrial and Labor Relations Review, 45(3), 435-448.

Rice, D.P.; Kelman, S.; Miller, L.S.; and Dunmeyer, S. (1990). The Economic Costs of Alcohol and Drug Abuse and Mental Illness: 1985. San Francisco, CA: Institute for Health & Aging, University of California.

Rice DP, Kelman S, Miller LS. (1991).Economic costs of drug abuse. NIDA Research Monograph, 113, 10-32

Shepard, E. M., and Clifton, T. J. (1998). Drug Testing and Labor Productivity: Estimates Applying a Production Function Model, Institute of Industrial Relations, Research Paper No. 18, Le Moyne University, Syracuse, NY, p. 1

Substance Abuse and Mental Health Services Administration (2002). Summary of Findings from the 2000 National Household Survey on Drug Abuse. Accessed online 6/25/2002 at http://www.samhsa.gov/news/click3_frame.html

White, J., Nicholson, T., and Duncan, D. F. (2001). A demographic profile of employed users of illicit drugs. In: M. A. Rahim, R. T. Golembiewski, and K. D. Mackenzie (Ed.), Current Topics in Management, Vol. 6 (pp. 353-370). Amsterdam: Elsevier.

Workplace substance abuse testing, drug testing: Cost and effect (January 1992). Cornell/Smithers Report, Utica, NY, Cornell University.