Or at least that is what I am led to believe by articles like Mic’s A Shocking Thing Happened When Tennesee Decided To Drug Test Its Welfare Recipients, which describes said shocking thing as:

1 out of 812 applicants tested positive for drugs. One. Single. Person. Tennessee conservatives suspicious that welfare recipients are a bunch of drug-addicted slackers were proven dead wrong. Big surprise!

After instituting dehumanizing drug-testing requirements to welfare recipients on July 1, 10 people total were flagged for possible drug use and asked to submit to testing. Five others tested negative, and four were rejected after refusing. As Think Progress notes, that means that just 0.12% of all people applying for cash assistance in Tennessee have tested positive for drugs, compared to the 8% who have reported using drugs in the past month among the state’s general population. If you assume the four people who refused were on drugs, it’s still a paltry 0.61%.

In other words, the plan intended to verify right-wing beliefs that welfare recipients are a bunch of drug-addicted slackers looking for a handout has demonstrated exactly the opposite.

The article has 11,000 notes on Tumblr right now, I’ve seen it all over my Facebook feed as well, and the same story has been taken up, with the same editorial line, by a host of other news sources. Jezebel: State Drug Program Busts A Whopping 37 Welfare Applicants. Wall Street Journal: Few Welfare Applicants Caught In Drug Screening Net So Far. New Republic: Red States’ New Tax On The Poor. Daily Kos: Tennessee Just Wasted A Lot Of Money Drug Testing Welfare Recipients. ReverbPress: Another GOP Fail: 0.2% Of Tennessee Welfare Recipients Found To Use Illegal Drugs. Mommyish: Results Of State Drug Testing Prove Gross Assumptions About Welfare Applicants Are Wrong. Washington Post: Scott Walker’s Yellow Politics.

These stories all make the point that we have many stereotypes about the poor, and one such stereotype is that the use lots of drugs, but in fact these sorts of welfare programs find them to use fewer drugs than the general population, and therefore we should stop being so prejudiced.

And if they were found to use only two-thirds, or half as many drugs as the general population, this might indeed be the lesson.

But look at the numbers in the quoted Mic article. Welfare users use only about one percent as many drugs as the general population. Really?

No. Not really at all. According to legitimate research in this area, poor people use as many drugs as anyone else and probably more. The National Household Survey on Drug Abuse found that illegal drug use was slightly higher in families on government assistance (9.6%) than families not on government assistance (6.8%). The National Coalition For The Homeless notes that about 26% of them use drugs, which is about 2.5x as high as the general population. I crunched some data I have from the hospital I work at, and it shows that poor people (defined as people who get health insurance through an aid program) have moderately higher rates of drug use related problems than the general population. So these articles are reporting a drug use rate in the Tennessee population about one percent of that ever reported in any comparable poor population anywhere else.

Kate from Gruntled and Hinged brings up another curious inconsistency. The false positive rate for drug tests is – well, it depends on the test procedure, but it’s usually at least 1%. So if every single welfare user in Tennessee was 100% clean, we would still expect between 1% to 5% positive drug tests. Instead, they got 0.12% positive drug tests. This isn’t just suspiciously good, it’s impossibly good.

So what’s going on here?

Before I explain, here’s a collage of the stock photos displayed above some of those news stories I linked to.

I now have a picture on my website called urine_collage.png

If you’re familiar with the state of the American media, you won’t be surprised to learn that urine was not involved in the ovewhelming majority of this program’s drug tests.

So how did they test people for drugs?

They gave them a written test, where the test question was basically “do you use illegal drugs or not?” You can see the exact procedure on the sidebar here.

And lo and behold, the overwhelming majority of people answered that they didn’t.

A more accurate stock photo they could have used

Now the numbers make sense. It’s not that only 0.2% of welfare recipients use drugs. All this tells us, if anything, is that 0.2% of welfare recipients are on so many drugs they can’t figure out how to check “NO” on a form.

Why would the government do something like this? As best I can tell, the plan was originally to give everyone urine checks, but in Florida the courts decided that urine-checking people without prior suspicion was unconstitutional. The Republicans were pretty attached to their “drug test welfare recipients” plan and didn’t want to look like they were wimps who backed down just because of one little court case, so they decided to give people the written test in the hopes of having prior suspicion for the people who said yes. Sure, it made no sense, but they could still tell their constituents they were drug testing those welfare recipients, and in principle they’d won an important victory. Or something.

Which raises another interesting question – how did Florida’s urine-based program do before the courts struck it down?

According to the media, abysmally. MSNBC: Drug Testing Welfare Recipients Looks Even Worse, “[Florida Governor] Scott’s policy was an embarrassing flop. Only about 2 percent of applicants tested positive, and Florida actually lost money”. TBO: Welfare Drug Testing Yields 2% Positive Results, “Newton said that’s proof the drug-testing program is based on a stereotype, not hard facts.” ATTN: Why Drug Testing Poor People Is A Waste Of Time And Money, “Florida tested welfare recipients for four months before its drug test mandate was thrown out by the courts. Only 2.6 percent of welfare recipients tested positive. The rest of the Florida’s population use drugs at a rate of 8 percent. So, again, welfare recipients used drugs less than everyone else.”

Now we’re merely at one-quarter of the drug use rate people with good methodologies find. Improvement!

So I looked up exactly how this works. Apparently welfare recipients were asked to pay for their own drug tests, and would be reimbursed if the results came back negative. 7000 welfare users did this, but 1600 declined to do so – numbers that were not mentioned in most of the pieces above.

Opponents of the program say that maybe those 1600 people could not find drug testing centers near them, or couldn’t afford to pay for the tests even with the promise of reimbursement later, or something like that. I am sure that some of them did indeed decline for reasons like those.

But also, people on welfare don’t have very much money [citation needed]. If I were a welfare recipient, and they were going to drug test me and not reimburse me if I came out positive, and I was on drugs, I would decline the hell out of that test.

Suppose that the poor in Florida use drugs at the same rate as the poor in various studies and surveys – about 10%. We have 8600 welfare recipients, so we would expect 860 drug users. Of the 7000 who agreed to testing, we know that 2.5% are drug users – that’s 175 people. That in turn would suggest that of the 1600 who refused testing, about 685 were drug users – 40% or so. That would imply that about 80% of drug users versus about 12% of nonusers refused testing.

These numbers seem pretty reasonable to me. Most welfare users want to keep their benefits, so the majority will agree to testing, but a few will inevitably fall through the cracks because they can’t reach a testing center or because they have moral objections to the tests. On the other hand, clued-in drug users will realize that for them, testing means a major inconvenience and monetary charge without any likely corresponding gain. So we would expect drug users to decline testing at a higher rate than nonusers. In order to use the Florida data to say that welfare recipients in general use drugs at a rate of 2%, we would need to assume that drug users were no more likely to refuse drug testing than nonusers, even though the testing rewarded non-use with money but punished use with a loss of money.

(note that there are some different numbers in different places for Florida. I assume that these represent different years, stages of testing, parts of Florida, etc, but I’m not sure. The only one that is seriously different from what I’m saying above is the one that says “only 1% of people declined testing”. After some search, I’m pretty sure that’s referring to that only 1% of people made appointments for testing, then cancelled later. But I am less confident in the Florida numbers than in the analysis of Tennessee)

So the Florida numbers are consistent with welfare recipients using drugs less, more, or the same amount as the general population.

So I have a question for you guys.

How come Brian Williams is being dragged over the coals for lying in the media, but everyone who publishes these kinds of articles gets off scot-free?

If I understand correctly, Williams said that his helicopter got shot at when he was in Iraq, but in reality he was just in a helicopter in Iraq at the same time as some other helicopter nearby was getting shot at. This is obviously stretching the truth, but it seems to me it could have been worse. No important policy decisions are going to hinge upon exactly which helicopter Brian Williams was in. And he didn’t get it infinitely wrong – for example, there was, in some sense, a war in Iraq.

On the other hand, discussions of how many poor people use drugs is pretty important for all sorts of policy questions, and these people completely dropped the ball. So why does nobody get reprimanded for this kind of thing?

You might argue that Brian Williams’ actions were obviously malicious and deceitful, but that screwing up drug numbers is an excusable mistake. I say it’s exactly the opposite. Brian Williams did exactly what I unfortunately do all the time – unthinkingly tell a story the much cooler way it should have happened, the way it happened in my head – rather than the way it actually did happen (my colleagues elsewhere in the psychiatry blogosphere go further and call this “normal brain function”).

On the other hand, I have more trouble imagining a situation in which I would accept the claim “only 0.1% of poor people use drugs, which is barely one percent of the rate in the general population” without wanting to do a little more research to see if it is true. If your reporters are capable of making this mistake honestly, get better reporters.

But I’m not sure it’s honest. A lot of these sources admit they took their story from a Think Progress piece on the issue. Think Progress does mention that the tests are a sham, although only in one sentence that is easy to miss. Either the secondary reporters didn’t read Think Progress thoroughly, or they consciously decided not to mention it.

But even if it was an honest mistake, I still have trouble excusing their arrogance. I mean look at that Jezebel article. The writer says this proves that people who think welfare recipients use drugs “consider ‘facts’ troublesome” and that their “entire social philosophy boils down to ‘Ew, poor people.'”

You’re saying that’s not as bad as a helicopter-related embellishment?

Yes, okay, drug testing welfare applicants is in fact probably a bad idea. It’s a bad idea because the courts have banned doing it in a way more effective than asking them politely if they use drugs or not, but it was a bad idea even before that. It’s a bad idea because drug tests have frequent false positives, but it’s a bad idea even without that. It’s a bad idea because quitting drugs is really hard and denying people benefits isn’t going to help.

But if, in the service of proving this to be a bad idea, you decide it’s acceptable to fudge the numbers to make your point, horrible things happen. First, you contribute to a culture of telling lies and lose the opportunity to protest when the other side does it. Second, you make it harder to trust you on anything else.

But most important, tell one lie and the truth is forever after your enemy. I recently argued that we need to reform suboxone prescribing laws, because it’s the best anti-addiction medicine we’ve got and right now poor people can’t access it. . Why should anyone listen to me now? They can just answer “Actually, that would be a waste of money. As per an article I read in Jezebel, pretty much no poor person has ever been addicted to drugs.” Then the laws don’t get reformed and people die.