How Frequently Do People Use Different Drugs?

How Frequently Do People Use Different Drugs?Charles Fain LehmanBlockedUnblockFollowFollowingJan 25One of the most frustrating things about Tell Your Children, the anti-marijuana tract from former New York Times reporter (and spy novelist!) Alex Berenson, is that its most interesting points are almost entirely unrelated to its thesis.

That central idea — roughly, that marijuana causes psychosis and schizophrenia, and psychosis causes violence, therefore marijuana causes violence—is compelling neither prima facie nor post hoc.

That Berenson spends so much of the book on this strained argument is frustrating precisely because of its otherwise useful criticisms of marijuana and its proponents.

TYC does a great job, for example, pointing out that medical marijuana is mostly a stalking horse for legalization.

While the drug has medical uses, they appear to be few and far between — it’s certainly not as medicinal as some make it out to be.

Similarly, Berenson highlights some alarming figures about the frequency of heavy marijuana use.

For the sake of brevity, I’ll quote from his Wall Street Journal op-ed from TYC’s roll-out:[T]he number of Americans who use cannabis heavily is soaring.

In 2006, about 3 million Americans reported using the drug at least 300 times a year, the standard for daily use.

By 2017, that number had increased to 8 million — approaching the 12 million Americans who drank every day.

Put another way, only one in 15 drinkers consumed alcohol daily; about one in five marijuana users used cannabis that often.

These figures come from the National Survey on Drug Use and Health (NSDUH), administered annually by the Substance Abuse and Mental Health Services Administration to 70,000+ Americans.

NSDUH is an odd creature: roughly half of its sample is 12 to 25 year-olds, presumably because SAMHSA is more interested in the drug-use behavior of young adults than the general population.

Still, it’s the best data source we’ve got for (the often-unreliable) self-reported trends in drug use, so it gets used/cited frequently.

Berenson’s claim about the high number of high-frequency marijuana consumers surprised me, not because I doubt it, but because I’d never thought to check.

So I wanted to dig into what, exactly, the frequency distribution of different substances look like.

Everything I do here is based on NSDUH 2017, with standard weights (ANALWT_C).

You can get a .

tsv and the codebook here.

I’ll try to flag variable names as I go.

I’ll also say right out of the gate that these are data for 2017, and as such represent near-current-year trends in drug use.

NSDUH has data going back to the early 2000s, but I’ve worked with this dataset for the very simple reason that loading the full time-series data takes *forever*, and this is just a blog post.

Still, I would caution against drawing too strong of a conclusion from one year’s worth of data about use.

That said, here’s a distribution of the number of days in the past year (MJYRTOT) that marijuana users used in 2017.

(It technically includes marijuana and hashish, but uh, I don’t know a lot of hashish users.

)A couple of noteworthy features here.

You can start to see the u-shaped curve that Berenson mentions — lots of low-frequency users, and lots of high-frequency users.

Also, you can see that NSDUH respondents like to approximate their own use, which is why you get those big spikes.

With such a big domain, it’s hard to see trends.

For ease of interpretation, I’m going to bin the x-axis into sets of sixty days or so.

This has the added benefit of delimiting 300-plus-day-a-year users, Berenson’s definition of “daily use” from earlier.

(This and all subsequent charts do not include non-users.

)So yes, about 20 percent of marijuana users in 2017 were users in the 300–365 day range.

We also know, from the last chart, that most of that is because of people who report 365-day use.

But how does that compare to other substances?.Over on Twitter, SB Nation’s Seth Pollack asked me how alcohol (ALCYRTOT) compares.



my tweets delete every month — part of why this post is going on Medium!) Here’s what that chart looks like, with the same binning:Alcohol lacks the same U-shape: most drinkers drink a little (0–60 days = at most a day or so a week of drinking), and those who drink more are about evenly distributed.

There are way more daily marijuana smokers than alcohol drinkers.

But what about other drugs?.Here, I’ve done the same thing, but with six major substances — marijuana, alcohol, cigarettes, cocaine (COCYRTOT), heroin (HERYRTOT) and methamphetamine (METHDYSYR).

(Funnily, NSDUH only asks about past-month cigarette use; I’ve just multiplied that number (CIG30USE) by 12 for an annual estimate.

)What to make of this?.We already know what alcohol looks like.

Unsurprisingly, the overwhelming majority of cigarette smokers smoke daily; by contrast, past-year cocaine users use it very infrequently.

I’m not sure why the latter fact is true — notably, I did not look at crack, which is delimited in NSDUH, and may show a different pattern.

Actually, the similarity between cocaine and methamphetamine suggests that both drugs’ low frequency of use may be a function of their low market availability — although, as I and others have noted, use of both drugs is increasing.

While noting the limitedness of this data, the similarity between heroin and marijuana’s use patterns is interesting.

Both drugs have a large proportion of infrequent users, and then a sizable minority — 15–20%— who use daily.

Marijuana is not a “downer” in the same way that heroin is — it’s classifiable as both a stimulant and a depressant.

But both have depressant and analgesic effects, which is part of why marijuana advocates routinely argue that the drug may be a substitute for prescription opioids for chronic pain; they point to evidence that states with medical laws saw lower increases in their rates of opioid overdose (Berenson accurately notes that subsequent analysis shows this trend vanishing; other research has limited the medical/opioid abuse connection).

There’s an interesting question, then, of the two’s substitutability in a social context.

Do people heavily use heroin, or perhaps more usefully prescription opioids, for similar reasons to those that cause them to heavily use marijuana?.(Principally, people could heavily use one of two different drugs—especially both of which have notable sedative effects—for the same reason while the non-heavily using population could use them for different reason.

) Marijuana and heroin aren’t pharmacologically similar, but their growing use may be a product of common social pathologies manifesting in distinct chemical expressions.

Assessing this claim on anything more than a speculative level would require more data than I’m going to look at here (at the very least, to tease out the year effects, but also to look at whether opioid and marijuana use are more comorbid than each’s comorbidity with other drug use).

So I want to turn to another question: how do different groups use?NSDUH demographic variables kind of suck — a function of a survey which focuses primarily on the young adult demographic and tacks on older adults as kind of an after thought.

Also, if you chop the survey up too much, you stop having anything remotely approaching representative.

Still, here’s sex, race/ethnicity, urbanization, and age.

Sex (IRSEX): men are more common among high-frequency users of all these drugs, except for cigarettes (apparently women smoke more heavily!).

Not a big surprise, given that men engage in more risky behaviors generally.

Race/ethnicity (NEWRACE2) for the three most common groups: Non-Hisp.

White, Non.


Black, and Hispanic.

I’m pretty sure N is way too small to draw reasonable conclusions about heroin/meth.

Cocaine use is more frequent among Black respondents (which, as I’ve written elsewhere, fits with higher rates of cocaine OD death among African Americans).

White respondents are more likely to be heavy alcohol and cigarette users, interestingly; marijuana appears to be about equally popular across races, with slightly more frequent use among African Americans in the middle use groups.

Here’s urbanization (COUTYP4).

Rural folks smoke cigarettes more frequently; everyone drinks about the same; everything else is hard to make generalizations about, given the comparatively difference in sizes of N.

Lastly, one of NSDUH’s wonky age variables (CATAGE).

Again, serious concerns about the sub-group sample sizes for the bottom row, and top row basically shows that use goes up with age (except with marijuana, where the heavy users concentrate among 18- to 34-year-olds).

Looking at all of this, I have three takeaways.

First: I think there is probably something to concerns about marijuana dependency, especially if men are more likely to be heavy users than women (because heavy marijuana use may be comorbid with the kind of problems low-status men currently face).

Second: the heroin/marijuana analogy is worth further exploration, although I’m absolutely not committed to it.

I dig in on the analgesis and depressant aspects of both drugs because those effects can obviously be connected to social/cultural desires for escape, disconnection, and suppression of experience.

Jill Jonnes (I think) talks about the upper/downer cycle in drug macro-history (society vacillates between uppers and downers, i.



If we’re in a downer period, it would make sense for there to be core groups of marijuana and opioid users.

Third: I want to look at longer-run trends in cocaine use, and crack cocaine use, to try to understand why most cocaine users are infrequent.

Cocaine overdose deaths are about as common as prescription opioid overdose deaths, but people appear to use cocaine infrequently.

Does that mean that the cocaine crisis really is just about fentanyl adulteration?.Are most deaths from speedballing, while cocaine alone is (relatively) safe?.With production continuing to increase in Colombia, the cocaine problem will only get worse — what explains people’s cocaine consumption choices?(Update: thanks to Michael Josem (@MichaelJosem) for catch a mislabelled axis in the first graph.

It’s now been fixed.


. More details

Leave a Reply