Last year, U.S. workers peed into one drug testing company’s cups about 9.1 million times. And last year, as in other recent years, analysis of about 350,000 of those cups indicated drug use. Most often, the drug of choice was marijuana, followed by amphetamines and painkillers.

The data are a little patchy, but the best estimate is that about 40 percent of U.S. workers are currently subjected to drug tests during the hiring process. Intuitively, that seems like a good idea: A sober, addiction-free workforce is probably a more productive workforce and, in the cases of operating forklifts or driving 18-wheelers, a safer workforce too.

But some of this cup-peeing might be for naught (and that seems to be something that other countries recognize: Drug testing is far more widespread in the U.S. than anywhere else). In many situations, drug tests aren’t capable of revealing impairment on the job, and the cost of finding a single offending employee is high. Besides, as the country takes a more and more permissive stance toward marijuana, and as the painkillers doctors prescribe are abused more and more often, there are gray areas that arise. What role should drug testing play in the workplaces of 2015?

Contemporary workplace drug testing owes its existence to the policies of Ronald Reagan, who in 1988 signed an executive order that led to legislation requiring federal employees and some contractors to be tested. The typical American employer wasn’t required to do anything differently (and still isn’t), but some large companies took this as a cue. A new market bloomed in response. “These … policies fueled the development of a huge industry,” writes SUNY Buffalo’s Michael Frone in his book Alcohol and Illicit Drug Use in the Workforce and Workplace, “comprising drug-test manufacturers, consulting and law firms specializing in the development of drug-testing policies and procedures, and laboratories that carry out the testing.”

This industry has relied on superficially intuitive arguments for drug testing: It’ll make employees use drugs less often and it’ll ensure a more efficient workplace. But those arguments have some significant holes.

First, as Frone writes in his book, there isn’t any proof that drug tests reduce drug use. In fact, a stronger deterrent effect might be that casual drug users choose not to work for companies that will test them. (Those employers might be missing out: More than half of Americans said they have tried marijuana, which is a big pool of talent to ignore.) “It’s become sort of a game,” Lewis Maltby, the president of the National Workrights Institute, told The Washington Post. “Employers know that it doesn’t mean anything. Anyone who smokes pot will just stop for a few days. It’s an empty ritual that nobody wants to be the first to give up."

There’s also the concern of spending not-insignificant amounts of money to pinpoint a very small portion of the working population. “For some employers the cost to find a single drug user can be high,” says Frone. And identifying those workers might be misguided to begin with, considering that “a positive test result cannot determine use or impairment at work and [that there’s a] general lack of evidence that drug testing has an impact on performance or safety,” Frone says.

Also, drug testing’s binary, drugs-or-no-drugs mechanism fits better with the delineations of legality that were common when Reagan signed that executive order. Today, marijuana’s legal status is a confusing patchwork of local laws, and legally-prescribed painkillers are more and more frequently abused. Both of these drugs can show up on tests, but it’s not clear what some employers should do after detecting their presence. Five years ago, The New York Times reported on the story of a woman who was fired from her job after testing positive for a painkiller that her doctor prescribed. Other workers have been terminated under similar circumstances, even when the medication in question was meant to treat job-related injuries.

According to Frone, there are three main reasons why drug testing remains so common. One is that companies still mistakenly believe in its effectiveness. Another is that some insurance companies might give discounts to employers who test. But the third is more political, more symbolic: Some companies use it to project a clean-cut, anti-drug image.

So what does a reasonable drug-testing regimen look like? Frone’s view is that it should be kept in place for jobs in which safety is the concern—forklift operators, truck drivers—but phased out elsewhere. The evidence that it deters drug users and enhances efficiency at work is simply too thin. “There are many potential ... causes of poor productivity, such as family problems or emotional problems or dysfunctional personalities, that collectively have a stronger impact on employee outcomes than drug use per se,” says Frone. He’s in favor of a constructive, communicative approach, in which job performance is monitored and therapy referrals are doled out, when necessary, after conversations between a worker and his or her employer.