clock menu more-arrow no yes mobile

Filed under:

Is it time to do away with job applicant drug testing?

The research linking passing a drug test to performing better at work is weak.

If you buy something from a Vox link, Vox Media may earn a commission. See our ethics statement.

A drug counselor tests urine for traces of drugs.
Testing urine for traces of drugs.
Photofusion/UIG/Getty Images

The job market is now so tight that some firms are complaining they can’t hire as many workers as they want to — and indeed, official figures now show more empty jobs than jobless people seeking work. Despite this, pre-employment drug testing continues to be a hiring requirement for many workplaces.

Rules that seemed to make sense when workers were easy to find become less attractive when labor is in short supply. Firms have already started to reconsider their policies against hiring the formerly incarcerated, a reform that will help ex-criminals and their families, and probably also reduce recidivism: Former criminals might not remain former criminals for long if they can’t find a way to make an honest dollar.

The next policy on the chopping block might be drug testing. Much of the current drug-testing regime dates to the 1980s, when many employers adopted “drug-free workplace” policies under strong urging from the federal government and drug-war advocacy groups; pre-employment testing was a key part of those policies.

An estimated 70 percent of companies with more than 2,500 employees use pre-employment screens, and they cover some 40 percent of all jobs.

But as it happens, the research suggesting that screening workers for drug use leads to better job performance is surprisingly thin.

Two studies of postal workers, done in the late 1980s, near the peak of the crack epidemic, tested job applicants — but didn’t show the results to people making hiring decisions. Both found that employees who had tested positive before employment were absent more frequently and were more likely to be fired than those who tested negative. In one of the studies (but not the other), there were also elevated rates of accident and injury among the “positive” group.

In each case, the smaller group of employees who had tested positive for cocaine performed worse on the job than the larger group that had tested positive for cannabis. Neither study included any significant number of opioid or methamphetamine positives, two of the drugs that most concern us today.

A third, smaller study at a large hospital found somewhat worse outcomes for the drug-positive group. But perhaps due to the small sample size (N = 180, of whom only 22 tested positive), the differences were not statistically significant.

And that, as far my search of the literature found, is it. Since none of those studies involved a workplace that actually rejected drug-positive applicants, job-seekers had no incentive to abstain before being tested.

However, in a workplace that uses pre-employment testing as a screen — the usual practice today — employees have every reason to avoid drug use in the days immediately preceding a test, even if they plan to use drugs later. (Three days will suffice to clear the body of the metabolites of most drugs; the exception is cannabis, which may remain detectable for weeks among those who have been using heavily.)

Whether the predictive value of a positive test result is the same when applicants have incentives to actively avoid testing positive isn’t obvious.

No one has properly analyzed the costs and benefits of drug testing

Do two generation-old studies, both of postal workers, constitute anything like an adequate basis for policies applied today to such a wide range of workplaces with diverse workforces? To my knowledge, no one has done anything resembling a benefit-cost analysis, either from a company perspective or from a social perspective.

No one has done an analysis that takes into account the costs of having to recruit more candidates — or relax other hiring standards — to fill the available jobs from the smaller pool of drug-negative applicants. Nor has anyone attempted to measure the external costs of denying employment opportunity to people who use drugs. (By “drugs,” I mean “drugs other than alcohol,” because urine tests do not detect alcohol.)

Here’s one of those unmeasured external costs: As I mentioned, cannabis takes an unusually long time to clear the body. If job seekers who are aware of that fact use other drugs (including the synthetic cannabinoids sold as “Spice” and “K2”) in place of cannabis, both the workplace costs and the social costs of pre-employment testing might be very high indeed. The synthetics are much more potent and much riskier than cannabis.

All of this suggests that it’s time for companies to reexamine their drug-testing policies. One way to start that reexamination would be to use data currently collected in the field to do some fresh research on whether pre-employment testing actually works. One specific kind of study would be especially useful: Studying the job performance of applicants whose tests actually showed some drug use, but below the cutoff that defines a “positive” test.

Everything is already in place to conduct such a study. When a job applicant deposits a urine specimen, the specimen goes to the testing company’s laboratory, along with thousands of others. For each drug covered by the test, the heavily automated process spits out a numerical value, generally calibrated in nanograms of the target substance per milliliter of urine.

We should move beyond crude measures of “passing” or “failing” tests to look at the details

But the company paying for the test doesn’t usually see that numerical value. It learns only whether an applicant has passed or failed the test. Passing — testing “negative — is defined as having less than some specified nanogram level of each drug. The federal Department of Health and Human Services sets the cutoff levels for federally required testing (of federal employees, some contractor employees, and workers in safety-sensitive jobs such as truck driving). Many drug-testing providers use those federal cutoffs for all of their tests, although some private companies specify their own.

Clearly, some of the “negative” tests will involve values just below the cutoff, and some of those applicants will get hired. And that sets up your experiment: Compare the workplace performance (attendance, productivity, injuries, sick days, health care utilization, disciplinary actions, turnover) of those just under the cutoff with otherwise similar employees hired at the same time who tested at zero.

If the truly “drug-free” employees turn out to be better workers than those who barely escaped being screened out, it’s reasonable to conclude that those who are actually screened out would have been worse-than-average employees. We might still debate the details of drug-testing policies, but the basic premise that they work on some level would be sustained.

But if that isn’t the case — if the people who were just below the cutoff perform as well at work as those who were at zero — then there’s little reason to think that those just above the cutoff (potential employees who were screened out) would perform worse. Such a finding would suggest that the current cutoff value is too low, at least for that specific workplace.

(Of course, coming to work under the influence of drugs raises different questions from off-the-job drug use. But the drug most likely to be involved in on-the-job intoxication is alcohol, which, again, urine testing doesn’t detect.)

It might also make sense to compare those who tested zero with all those whose test results were non-zero, or try to identify a “dose-effect curve” between tested drug level and job outcomes: How does performance change as the measured drug level increases? A more adventurous firm might want to go further still, hiring a sample of applicants with test results well above the cutoff, to study whether they actually perform worse on the job than those who passed the drug test.

It’s possible, though far from certain, that hiring applicants who test positive for opioids is riskier than hiring those who test positive for other illicit drugs. Most people who take prescription opioids actually have prescriptions for them, and (possible legal problems under the Americans With Disabilities Act aside) it’s ethically dubious to screen out people on the basis of the medicines they have to take.

Moreover, the standard panel covers heroin but not the fentanyls, and the last thing you want to do is incentivize workers to switch from prescription opioids or even street heroin to fentanyl to be able to find jobs.

If the opioids really stand out from other drugs, that fact will show up in the experiments suggested above, and employers can try to figure out what to do about it after the experimental results are in.

Even employers who want to keep pre-employment drug testing for other drugs might consider no longer screening for cannabis, which currently accounts for about half of all positive test results. Some firms, especially where cannabis is now legal under state law, have already decided to leave cannabis out of their screening programs.

Perhaps it will turn out that pre-employment drug testing is cost-justified, at least for some jobs — especially safety-sensitive ones — even when labor market conditions are tight.

But drug-war passions are fading and job applicants are hard to find. In that shifting social context, companies and public employers, including the military, might want to take a second look at whether pre-employment drug testing actually does the job it was designed to do.

Mark Kleiman is a professor of public policy at NYU’s Marron Institute. His book on drug policy is Against Excess. He is the co-author, with Jonathan Caulkins and Beau Kilmer, of Marijuana Legalization. He blogs at the Reality-Based Community and can be found on Twitter @markarkleiman.

The Big Idea is Vox’s home for smart discussion of the most important issues and ideas in politics, science, and culture — typically by outside contributors. If you have an idea for a piece, pitch us at