Less Anti-Drug Propaganda, Less Drug Use

Students for Sensible Drug Policy juxtaposes data on drug use by teenagers with spending on the federal government's anti-drug ads and finds a clear association:

 

As SSDP concedes in the fine print, this comparison does not really prove anything, but it is similar to the logic drug warriors routinely use to demonstrate the dangers of drugs and the benefits of prohibition. In any case, as the evaluations of the National Youth Anti-Drug Media Campaign quoted by SSDP show, there are other reasons to believe the ads are ineffective at best.

Comments

[Eryk Boston]

The correlation is weak. They simply make it look similar by tweeking the range of the horizontal numbers until you get a similar angle of descent for both graphs. This is lying with statistics 101 but you’re right that it is the same type of nonsense used by drug warriors and every other con artist trying to convince a nation of people who learned math in government schools.

[thoreau]

Horrible data analysis.

Great way of turning the tables, however.

[ed]

Wouldn’t a true juxtaposition place one graph over the other and involve different-colored lines and thingies?

[Eryk Boston]

correction…..vertical numbers

[Eryk Boston]

ed….you can’t do that since there would be different and incomparable terms on the vertical bar. That’s the reason why such nonsense can be made to look however you wish it to look.

[Taktix]

This reminds me of the Flying Spaghetti Monsterism chart that shows a direct correlation between the shrinking number of Pirates vs. the Global Temperature increase.

Praise HIM!

[Brian24]

What’s wrong with kids these days? Have they become so lazy that they can’t even be bothered to do illegal drugs anymore?

[Chuck]

I get a correlation coefficient of r=0.9568, which is significant at the usual confidence levels (p

[dubya]

Cut and run! Cut and run!

[Lamar]

Eryk Boston:

You aren’t the first person to say that these charts show a weak correlation. The SSDP beat you to the punch…and they put the charts out!

[Robert]

The best comparison would not plot time at all, but spending on the horizontal axis and drug use on the vertical axis, and include data from before 2001. I bet the correlation would be good.

Still more interesting might be to do the above using drug use data delayed 1, 2, or more years after the spending data, or vs. rolling avg. of spending going back some # of yrs.

[Todd]

If they performed regression analysis and came up with a correlation of .8 or more, than they might have a legitimate correlation, then again it might still be considered spurious. Once again though, still better than some government data.

Anyone ever heard this quote from a pol before? “We enacted (such and such a) law and (pick your drug) use fell by 39.8%,” with no graph or information of any sort to back it up and giving no source?

[Chuck]

Gott damn HTML!! The last part of my post was supposed to indicate a p-value of less than 0.01. Forgot that the stupid intraweb thinks that a “less than” sign is something completely different.

[TrickyVic]

Who’s lying? SSDP didn’t create or modify any of these numbers. They just put them side by side. The drug use numbers came from http://monitoringthefuture.org/ and was funded by research grants from the National Institute on Drug Abuse, a part of the National Institutes of Health.

It still proves nothing but it’s fair game in the drug war.

[amr]

I get an r-sq of .9156 of spending vs percent reporting use.
Not super, but it’s only 6 data points.

According to my regression, 98Million ad budget gives us 0 reported drug use and 0 Million of ads gives us -175% reported drug use

[Lamar]

AMR:

I had my own regression, but then again, I am on vacation.

[Thrall]

I use drugs. HAHAHA stick it, current governmental administration.

[amr]

I’m a doofus.
Ad bugdet of zero gives us 9.8% reported use.
Ad buget of -175 Million gives us zero reported use.

[Ruthless]

These charts remind me of the one tracking workplace accidents. Seems they were declining at a faster rate BEFORE OSHA came along. No surprise.

[Madog]

As noted, the charts are definitely designed to make the decrease in drug use match the decrease in spending. As a percentage, spending dropped about 46% and drug use dropped about 24%.

[Larry A]

As noted, the charts are definitely designed to make the decrease in drug use match the decrease in spending. As a percentage, spending dropped about 46% and drug use dropped about 24%.

1. Both charts show a downward trend, showing that less spending is associated with less drug use.
2. The charts show that drug use dropped about 1% for every 2% drop in ad spending.

[Grand Chalupa]

A drop of $85 million over the course of the Bush administration?

Like I said before the election, if they’re both gonna spend money I’ll take corrupt assholes over sincere do gooders any day of the week.

[Warren]

The correlation is very strong. The weakness is the small sample size. I would think the data for both graphs goes back over twenty years. It’d be interesting to take a look at the full sample set. I seem to remember drug warriors talking about how drug use went up when the propaganda budget went down, suggesting the correlation use to run the other way. As we all agree, this small set proves nothing. But if, as I suspect, the long term xi-sq approaches zero, then that is good evidence that the two are uncorrelated. Which would indicate that the ad campaigns don’t carry any weight when people are deciding if they want to do drugs.

Yeah, I’m stunned too. I mean I always thought I started using recreational drugs because of all that peer pressure. And keep using them because I’m now powerless to stop. Guess I was thinking for myself all along… Who knew?

[MainstreamMan]

Whatever the correlation here… remember there is no causality implied.

The argument can be made that spending was reduced in response to reduction in usage. This would require other sources of evidence to demonstrate.

[Kwix]

MSM,
Exactly the point. Correlation != Causation no matter what the ONDCP may tell you.

[Shelby]

This is the most compelling argument I’ve ever seen for increasing funding of the ONDCP.

[Ashish George]

If you wanted to really scare kids away from drugs without being too dishonest, show them–say, in fourth grade or so–Requiem For A Dream and tell them, “Worst case scenario…”

[thoreau]

Sadly, the money taken out of the propaganda budget was probably spent on something else, and I’m sure that somebody in ONDCP will show that the rise in some other type of spending is correlated with the recent drop in spending.

Of course, that argument is no stronger than the argument implied by these charts, and that’s sort of the point of the charts. I find it to be a clever explanation of the pitfalls in statistical analysis. “Correlation is not causation” is a mere mantra. Provocative demonstrations of statistical abuse like this are much more powerful, in some ways. And they also show the problems with the arguments made by drug warriors.

Bravo, SSDP.

[thoreau]

CORRECTION:
“…correlated with the recent drop in drug use by teens.”

[MainstreamMan]

In case you don’t understand Thoreau’s “provocative demonstration of statistical abuse” comment, I refer you to the comment above the graphs…

“The National Institute on Drug Abuse says the Drug Czar’s ads may actually cause more teen drug use.

Monitoring the Future’s data seems to support this conclusion.”

They don’t.
In any way.
At all.
No matter how you look at the numbers presented.

I can’t second the “bravo.”
Far more people will believe the SSDP is telling the truth than will recognize the flaw in reasoning. And that is why they presented the graphs as they did.

Dishonest is dishonest.

[thoreau]

MSM-

I guess I thought it was so obviously a joke that I didn’t think of it as deceitful.

If nothing else, it raises questions about the efficacy of such propaganda efforts. This corrrelation doesn’t really prove anything, but if the ONDCP’s propaganda efforts were effective then you’d at least expect a weak correlation with the opposite sign. The fact that the exact opposite was found suggests that ONDCP efforts have no real connection to actual drug use by teens.

[MainstreamMan]

Thoreau,

“if the ONDCP’s propaganda efforts were effective then you’d at least expect a weak correlation with the opposite sign.”

This has too many assumptions embedded in it for me to agree with you. As you know, demonstration of program efficacy are far more complicated than that. It is not that these graphs “don’t really prove anything.” It is that they don’t have any meaning at all without a lot more contextual information that would allow for their interpretation.

As for the obviousness of the joke, I think the posts by Chuck and amr say something.

[MainstreamMan]

For a comparison of the way the ONDCP presents their argument to that of the SSDP you can go here…
http://www.whitehousedrugpolicy.gov/publications/predict_drug_use/predict_drug_use/heavy_drug_use.pdf

Disclaimer… I do not support current drug policies, I just think that the SSDP are doing more harm than good when they use such weak arguments.

[dhex]

i thought it was an obvious joke too.

[thoreau]

MSM-

Fair point, I guess the expectation of a weak negative correlation is indeed contingent on some assumptions.

I thought Chuck and amr were running with the joke, on the theory of “Hey, how few kids would use drugs if the ONDCP shut down? Let’s find out!”

The efficacy of a joke like this depends entirely on presentation, and whether they make it clear that they’re just poking fun and not trying to pass off this useless “analysis” as something serious. They’re just making a point about the weak metrics often used by bureaucrats, and pointing out that such sloppy methodologies can give any conclusion that you like.

Whether that point comes through depends on how it’s done.

[thoreau]

I think their point was made pretty well by this thing put below the graphs on the SSDP site:

WARNING: While attempting to infer causation from the correlation between the Drug Czar’s ad budget and teen drug use may seem ridiculous, it is certainly no more absurd than the statistical gymnastics performed regularly by the Drug Czar and his staff. The difference? They use taxpayer dollars to fund their stretching.

[Lamar]

“Far more people will believe the SSDP is telling the truth than will recognize the flaw in reasoning.”

I guess this is the end of the honest disclaimer. Saying, “this doesn’t prove anything” doesn’t mean “this doesn’t prove anything” anymore. Apparently, to MSM, it means, “this proves our case conclusively and dishonestly, even though we state in print that it doesn’t prove anything.” Heck, it is time consuming to have to READ.

“And that is why they presented the graphs as they did.”

Ooops, I take back everything. When I wrote my earlier post, I didn’t realize (1) that you were on their board of directors, and (2) that their intent means a goddam thing.

“Dishonest is dishonest.”

Yes, and when you tell the truth in plain English, they call it, “not dishonest.”

[P Brooks]

“This *sizzle, sizzle* is your brain on charts.”

Good grief. Lighten up, for pity’s sake.

[Chuck]

Mainstream Man–

What exactly is it my post was supposed to have said? The very first post on this thread stated that “the correlation is weak.” Whatever else it may be, it is *not* weak by the usual measures–it is in fact statistically significant even with the small sample size. The purpose of my post was to point that out, period. Anybody who read more than that from my post needs to check the interpreter in their head for malfunctions.

I did not attempt to find a regression line, nor to extrapolate using that line, since (a) it is a small dataset, and I am suspicious of small sets regardless of what the r value is, and (b) it is *always* bad practice to try to use a regression line to make predictions outside the range of the original data. Correlation certainly does not imply causation, but it is a worthwhile exercise to hoist the ONDCP, FDA, DEA, etc., on their own petard for a while and make THEM explain why the kinds of analyses they like to do are meaningless. It’s only dishonest if the SSDP makes any inferential claims from this data. They have not, and if others have done so it’s not their problem.

[MainstreamMan]

“It’s only dishonest if the SSDP makes any inferential claims from this data. They have not, and if others have done so it’s not their problem.”

Sorry, but I took “The National Institute on Drug Abuse says the Drug Czar’s ads may actually cause more teen drug use….Monitoring the Future’s data seems to support this conclusion” to be an inferential claim based on the data they presented.

Look around their site a bit more… they make many unsupported claims… one of them being “it is certainly no more absurd than the statistical gymnastics performed regularly by the Drug Czar and his staff.”

It is more absurd. When the ONDCP presents data they include methodology, justifications for that methodology, complete data sets, etc…

Things that are lacking in the SSDP presentation of this data.

[Ethan]

I think the causal nexus may be even more complex: the “American credibility and moral standing abroad” chart looks just like the two charts above.

[MainstreamMan]

Lamar,

Sorry, but I see lots of dishonesty in this disclaimer…

While attempting to infer causation from the correlation between the Drug Czar’s ad budget and teen drug use MAY SEEM ridiculous, it is certainly NO MORE ABSURD than the statistical gymnastics performed regularly by the Drug Czar and his staff. THE DIFFERENCE? They use taxpayer dollars to fund their stretching.

“May seem” would typically be used ostensively quite differently than “is.” In most cases “it may seem” would be followed with a justification for doing it anyway.

“No more absurd”… well I pointed that out already.

“The difference?” There are lots of differences more important than funding involved in using what the SSDP has presented to make an inference of causation and what the ONDCP does.

[MainstreamMan]

Lamar,

“Apparently, to MSM, it means, “this proves our case conclusively and dishonestly, even though we state in print that it doesn’t prove anything.”

Now who is being dishonest? When did I come close to making such a claim?

[MainstreamMan]

Chuck,

Mia Culpa.

[MainstreamMan]

Chuck,

That is, for implying you were making any inference.

[Lamar]

Well, I didn’t mean it was a quote. Perhaps we should create paraphrase or interpretation quotes. I guess the question comes down to, why did they “release” the graphs if they don’t mean anything.

[MainstreamMan]

Lamar,
The graphs have independent meaning.
It is the comparison that is meaningless.

“Perhaps we should create paraphrase or interpretation quotes.”

The problem is that I don’t feel you fairly paraphrased my statements. (Given your “it is time consuming to have to READ”, I found it worthy of a response).

Maybe we should create !hyperbole! marks or ;( Strawman 😉 marks.

[Chuck]

MSM–

No biggie. And your points about the disclaimer are good ones, IMO. Maybe I should RTFA more closely. I think I know what they were getting at (now who’s doing the interpreting?) but it could have been said much better.

[legalize sensi]

The U.S. Government makes too much money policing and locking up my brothers and sisters for the drugs Whitey brings in. Peace.

[TrickyVic]

Drugs law have helped us reach number 1 in the world for prison population. If your in the business of putting people in prison, job well done. Set your sights higher for 2007.

note: must be read in a partly sarcastic tone.