Apparently someone slept through their correlation and regression class. The point here is not just that large cities have both a high suicide rate and a larger number of country music stations, it’s also that this anecdote is the way that almost every course on introductory statistics introduces the problem of confusing correlation with causation. The only running joke as ubiquitous among number nerds is the spherical cow.
Granted that this kind of low-level innumeracy isn’t that unusual among bloggers. Just remember the next time he deigns to lecture us science-y folk about not taking him seriously about the genetic stupidity of black people, a point that depends entirely on advanced methods in correlation, that Sullivan has a slight problem with the basics.
Zifnab
This does still raise the question, what is the underlying cause of all this? What on earth would compel a large demographic of people to become so miserable, so apathetic, so lacking in self-esteem that they would actually listen to country music?
I think if we can find the root cause of all of this, we can nip this whole genre of music in the bud.
merciless
Thanks; that was my first thought too. “Correlation, not causation…again.”
My favorite example was a study that said more than 80% of the executives studied who had heart attacks, were coffee drinkers. Big headline: COFFEE CAUSES HEART ATTACKS! My friend read it and said, “I bet that well over 80% of those same executives wear dark-colored socks. Do black socks cause heart attacks?”
Vincent
Eh, I don’t really believe Sullivan believes that country music causes suicides. He probably just thought it was funny (which is indicated by the post title), so I don’t see a need to rag on him for this.
grumpy realist
I was just thinking of a similar argument, involving a) number of country music stations and b) number of pickles sold.
Therefore, listening to Dolly Parton causes one to like pickles.
grumpy realist
Oh, and has anyone run across a version of the spherical cow joke that doesn’t involve a physicist/mathematician?
(We were told the joke the first day of our QM class.)
Kynn
He’s probably just trolling his own readers for fun.
FAIR TAX! FAIR TAX!
Warren Terra
Cow? I learned it as ‘Ok, so we start by assuming a spherical chicken of uniform density …’
Haltelcere
I’ve never hear of a spherical cow before, but will try my darnedest to use it in a phrase at work tomorrow.
Robert Johnston
Sullivan is a recovering Republican. Confusing correlation for causation is a big first step up from confusing unverified personal anecdotes of welfare queens in Cadillacs for evidence of causation. Sullivan should be encouraged for the progress he’s making on his journey, not mocked for it’s humorously inadequate pace.
myiq2xu
Is a spherical cow anything like a holy cow?
jake
Fixed.
Dennis - SGMM
Post hoc, ergo propter hoc. Too bad that no one learns basic Logic anymore.
Peter ve
I have had a beard since 1983. There hasn’t been a nuclear war since 1983. Therefore, I can start a nuclear war by shaving.
FEAR ME!
Tom Hilton
On the other hand, if all I had to listen to was ‘country music’*, I would certainly kill myself.
*Or more precisely, what passes for ‘country music’ today, in the bland homogenized Clear Channel-dominated radio landscape. Real country music I wouldn’t mind so much.
RSA
You sometimes hear it in computer science, which I suppose is the math influence.
For what it’s worth, the paper was published in 1992, and we’ve gotten a much better understanding since then about why beta coefficients are useless in causal modeling.
Ted
Sullivan has a PhD in political science, but delves all too often in the more empirical variety. Take a look some time at his thoughts on the evolutionary reasons why women would want to have sex. Yes, you read that right.
As an aside, you’d think for that particular PhD, he’d have had to have taken tons of statistical analysis. I guess he didn’t retain much.
wvng
I’m with Vincent. It was a funny bit, not a cosmic truth. End of story.
But I know that I feel like committing suicide if I’m forced to listen to a country station – at least since they kicked the Chicks off.
:-)
Tony
It’s a correlation between airplay and suicide, not the number of stations. There is a difference.
My college communications prof, one of the best I ever had, said his analysis of music lyrics showed that country songs were the most depressed and hopeless of the group, by far.
So I say, “no wonder.”
Dood
I read the post and Sullivan didn’t say anything about causation. Methinks you jump the gun, John.
MNPundit
We got the “storks = babies” one in the game theory unit of my introductory Poly-Sci courses.
Ted
In other statistics, having a southern accent means you’re more likely to commit suicide, do meth, get a divorce, or be poor. There must be something about that drawl.
There, was that Sullivan-esque?
He does this shit all the time. Whenever some interesting but ultimately useless statistical correlation hits the news, Sullivan can be counted on to make himself look like an idiot in a link, and two lines or less.
Delia
You’d better watch out. The neocons will be looking to bomb you now that their case against Iran seems to be fizzling.
jcricket
Well said Ted. And very accurate, except the “two lines or less”. The initial post will be two lines. Then he posts a reader email (without comment), then he expounds for 1,000 words or more, with references to Kant, Hume, Burke, doubt, foreskin restoration and the horrors of Hillary Clinton.
He’s basically Hitchens without the sense of humor or nihilism.
jcricket
Also:
Fixed, per Sullivan’s anonymous barebacking adventures
douglasfactors
(1) Sullivan didn’t say anything about causation.
(2) I’m curious–what explains the correlation between the airtime devoted to country music and the suicide rate among whites?
demimondian
I don’t know what explains it, but I can think of several factors which could influence it. For instance, heavy alcohol consumption is highly correlated with suicide; it could be that there are more binge drinkers in areas with more country music air time. It could be that the country-music saturated areas exhibit less economic mobility, so that even if the effect is still present after controlling for economic class, it is still economically generated. It could be that people with affective disorders are more likely to listen to country music, so there are more country stations in places with a bunch of depressed people (because there’s more of a market for the product), and hence there are more suicides in those markets (because of the depressed people who drew the stations.)
Geoffbr
Someone seems to have slept through their humor class. It seems pretty clear from context that Sullivan linked to this for its comedic value, not because he considered a masterpiece of the scientific method.
Fwiffo
I was playing around correlating various CDC, census, etc. statistics with the Republican/Democratic lean of states. One thing I discovered was an extremely strong correlation between suicide and low population density, e.g. states like Alaska, Wyoming, etc. had the highest suicide rates, and densely populated coastal states had low rates. Only two states deviated much at all. New York had a lower rate than expected – I presume because it’s pretty big geographically, but most of the population is concentrated in very dense areas (more than other states, anyhow.) Florida was the other exception, which high rates of suicide. Florida has a much older population than other states, and old men have a suicide rate several times higher than other demographics.
jake
Um. What?
No. I don’t want to know.
I think he ought to rename his blog Andy’s Big Ol’ Ponderation Corner.
Delia
Hitchens has a sense of humor?
demimondian
Actually, jake, there’s question of why females gain from sex is an old and difficult one. If there’s no gain, then parthenogenetic species should dominate, and they don’t, so obviously egg-layers benefit. It proven to be extremely hard to figure out why a female would, on average, expend half of her limited reproductive potential creating males.
The Other Steve
Descartes walked into a bar one day. The bartender asked him if he wanted a beer.
Descartes replied, “I think not” and promptly vanished.
Cindrella Ferret
What? No country music fans here? Damn! I would be curious to know the musical tastes of the assembled crowd. Assume a circular crowd …
Equal Opportunity Cynic
The point here is not just that large cities have both a high suicide rate and a larger number of country music stations, it’s also that this anecdote is the way that almost every course on introductory statistics introduces the problem of confusing correlation with causation.
I don’t recall that particular anecdote in my college-level stats class, and I ate that stuff up.
But I’d rather discuss what you consider the subordinate point. Are there really a larger number of country music stations per capita in cities? I find that extremely hard to believe.
For example, major markets like NYC or LA might have a few, and have tons of other stations too — free form, dance music, foreign language, five different NPR stations, you name it. Small towns in West Texas have country and pop.
Tim F.
No, there are just a larger number of stations and ergo more total air time.
Equal Opportunity Cynic
From the study quoted in the post linked in Sullivan:
Our model explains 51% of the variance in urban white suicide rates.
Apparently they were comparing metro areas with other metro areas, so my West Texas example is irrelevant. One can easily imagine that NYC and LA have both higher suicide rates and more total airtime than Des Moines or Mobile, though. So it really comes down to how they measure “airtime” I suppose. I automatically assumed “airtime devoted to country music” was a relative rather than absolute measure, not to mention that a study controlling for Southernness would also control for urbanness.
But you could certainly be right.
Steve
Actually, Sullivan’s post is not that foolish — it doesn’t say what you imply it says. I agree with the posters who say that Sullivan is presenting it for amusement value. But besides that, the study which he quotes is NOT saying that the number of suicides by whites is higher in cities with more country music stations. That would be an obvious example of the fallacy from the introductory stats class.
The quote actually says that the RATE of suicides by whites is higher in cities where the PERCENTAGE OF AIRTIME devoted to country music is higher. This is a completely different observation, and because both variables are rates/percentages rather than absolute numbers, they do not automatically correlate with the size of the city.
I think there may be a more subtle statistical fallacy here. If the “multiple regression” analysis was a multivariate technique where a large number of variables are compared to each other in search of “significant” relationships, then it is very likely to produce false “significances” by chance. Such studies often consider a relationship “significant” if there is a 95% likelihood that the relationship is NOT accidental. When a study is run this way, 5% of the thousands of comparisons will look “significant” whether they are real or not.
grumpy realist
As to whether there are any Country Western fans here I think the question is what one considers “Country Western.” There’s the original real stuff, then there the plasticated stuff the Nashville churns out. (I’ve also got my suspicion that even with the “original real stuff” a lot of the stuff is the US equivalent of Scotland’s playing up of haggis–how else can I explain songs like “Blood and Whisky on the Highway”, “You’re the Reason God Made Kentucky.”)
My own musical tastes run to Early and Baroque music, opera, Japanese enka, foreign popular music (Francoise Hardy, Alla Pugachova, Jacques Brel, Nana Mouskouri, etc.) and some weird stuff like Dead Can Dance.
Steve
Sorry — a correction: The quote does not say whether “airtime” is a percentage or an absolute amount. As an ex-market researcher, I interpreted it as percentage because that would be the more useful, and more typical, way of measuring airtime.
Equal Opportunity Cynic
Steve,
You and I made the same assumption. Frankly I think the spurious example is so trivial that no one should be able to make a living as a researcher if they put that out as their research. But there was that right-wing writer someone linked the other day that denied the validity of statistical sampling, so I suppose you can make a living at anything if you find a big enough sucker to fund it.
ScottS
If you click a couple of links, you get to:
http://www.uta.edu/depken/ugrad/3318/music-suicide.pdf
the study was apparently written by Steven Sack of Wayne State University and Jim Gunlach of Auburn and was published by the a journal at UNC.
Your zeal to zing Sullivan is actually sort of pathetic. All of you.
Besides, correlation is not causation, but it is something. All historical chronological analysis is cause and effect correlations. You can’t prove that one thing led to another, but there can be some common-sense conclusions that assert that one thing did in fact lead to another. If correlations are irrelevant without perfect statistical analysis, the entire field of history is doomed. Its a damn shame that everyone feels the need to scientifically prove everything such that things for which good measurements don’t exist are no longer part of the public discourse.
Turning social studies into social science was generally good for psychology and mostly good for economics but come on. If a thousand donkeys take a shit on the road and it starts to smell bad, it probably isn’t the sewer line backing up. It was probably the donkeys.
None of the above should be taken as a validation of the country music study. I could care less. Maybe it was a junk article. Presuming that Sullivan vetted it for methodological soundness was either stupid, or, more likely, just lazy, but I expect better of Ballon Juice and its readers. And presuming that anything without scientifically objective levels of causation isn’t worth commenting on is, well… lame.
myiq2xu
Correlation? Co-relation? Are you implying that rednecks marry their cousins? Well they do, but only if they got really hot cousins.
But what has incest got to do with country music and suicide?
Uh, never mind.
Pb
Here are some more correlations to suicide statistics (by state) for you… high on that list are lower retail health sales, higher elevations, more women marrying young, higher divorce rates, having less people native to the state, having less people who voted for Gore in 2000, having more men than women, having a lower population density, having less muslims around, having more children in father-only households, and proportionately having less doctors. Voting Bush in 2000 is #19 on the list; % home units that are mobile homes is #20 on the list.
Ted
Also well said. Hitchens can be hilarious, in some circumstances. His formal debate with Dinesh D’Souza on ‘God’ was a treat.
Ted
Then maybe you don’t regularly read (or have long read) Sullivan. He tries to find fascinating ‘insight’ in everything, and statistical correlations are one of his favorite ways of doing that. He usually thinks these correlations actually mean something causal, or at least strongly associated. His early hyping of “The Bell Curve” is an excellent example, and one already alluded to by John.
The analogy up thread of the business executives’ black socks causing heart attacks is not far off the mark at all for him. Sullivan is basically a political astrologer
clone12
Since the paper did not include in-city migration rate in the paper (nor did it mention about looking at its effect), I wonder if country music is a proxy for in-migration rate of rural whites into cities, and whether the lower socio-economic status of this group relative to the general population leads to higher suicide rates.
I’m not sold on the “Country music nurtures suicide” hypothesis posited by the authors since a) all musical genres have downer songs and b) blacks, as they acknowledge, aren’t affected by country music at all- something you expect to see if country music truly does encourage suicide.
Anne Laurie
My favorite country-music quote comes from folksinger Ann Hill: “Everybody makes fun of country music until after their first divorce.”
jake
Um. I’m going to guess for the same reason guys do: A little critter with half your genetic code.
Because if she doesn’t she doesn’t get any critters with half her genetic code? Couldn’t you say the same thing about males? Suppose you’re a male cardinal. You’re bright red, predators can see you from miles off and you spend a lot of energy seeking a mate, chasing off rivals and hopefully not becoming a McCardinal burger for a hawk. Then there’s keeping an eye on the nest, feeding the young, etc and still dodging the hawks. All so he can get critters (some of them female) with half his genetic code.
I think the real question is why the hell anyone bothers.
Billy K
You all are talking about Country Music as if it were still 1977, with broken-down pickups, my wife left me, tear in my beer subject matter. If you’ve listened to any of the music country stations play nowadays, it’s mostly about family values and patriotism. Any drinking songs are about regret. When a man acts up, it’s not by cheating, it’s by playing hooky to go fishing. The shit is so watered-down and pandering I can’t stand it.
It’s not “Country Music” that’s causing people to kill themselves, it’s the Hannitized family values message for the exurbs.
Svensker
Is there a correlation between spherical cows and physicists/mathematicians?
(As to why women have sex — same reasons men do: it’s fun, and you get babies who are related to you.)
DR
BUT: Correlation often points to underlying causation; by that I mean that even if it does not establish causation, it usually points to an underlying common cause.
So the question is: what makes people both listen to Country Music AND Commit suicide?
I say religion, but then again, I’m biased :-)
Ed Drone
“Therefore, listening to Dolly Parton causes one to like pickles.”
You haven’t listened to country music lately, have you? Dolly Parton hasn’t been in the regular rotation on those stations for several years now, and only shows up in a “oldies” segment or when someone covers one of her old hits.
As for country music lovers here, I’m one (though not fond of modern country stuff). I’m a musician, though the country music I play is circa 1928-1938, and sounds a lot like “folk music” (don’t start that discussion, please — no one knows how to define it, but, like porn “I knows it when I hears it.”).
I also play rock and Civil War/Minstrel songs, and I’m not even schizoid
[yes, you are]
[no, I’m not]
[yes, you are]
[no, I’m not]
[yes, you are]
[no, I’m not]
[SHADDUP, YOU TWO!.
Ed
canuckistani
Can’t stand country music, in part because I’m a snob and don’t like all the red-state values and assumptions that get packaged up with it.
But I’ll make exceptions for Johnny Cash and Patsy Cline.
Jon H
“he point here is not just that large cities have both a high suicide rate and a larger number of country music stations,”
Er, what? Why would higher population result in a larger number of country music stations?
A radio station is not limited in the number of listeners it can serve so there is no necessity to add stations to deal with more population. Even if there were, there is limited frequency space for adding new radio stations, and the demographics of a large urban area may favor other genres anyway.
I don’t think there are many, and possibly not ANY, country stations in NYC.
I think it more likely that increasing population would encourage a shift from an amateurish independent/local country station to a big-corporate national advertising-package-sales centrally-programmed radio station, without necessarily increasing the number of stations in a given market.
Unless the demographics in a given area favor country over other music genres, regardless of market size.
Jon H
Billy K wrote: “You all are talking about Country Music as if it were still 1977, with broken-down pickups, my wife left me, tear in my beer subject matter. If you’ve listened to any of the music country stations play nowadays, it’s mostly about family values and patriotism.”
Nah, Country music nowadays is “Lifetime Radio”. As in, like the TV network.
Moral Panicker
1) You do not seem to have read the study linked or even Sullivan’s post as far as I can tell.
2) Many social variables (poverty, gun presence, Southern location, divorce) are included along with the presence of country music to ascertain an independent correlation between suicide and country music.
3) The presence of country music is measured by “the proportion of radio airtime dwevoted to country music,” not by the number of country stations or the level of hours of country music.
4) No, correlation is not causation. When I took statistics 1 the example was a town where there were more births when there more storks, because of course the town had expanded and there were more buildings where storks could build nests. No, this is not a proof of a causal connection but is shows a persistent and independent statistical relationship that is a basis for tongue-in-cheek speculation.
5) Maybe they should have used time-lagged variables to see if increases in divorce are follwed by increases in country music presence or vice versa to establish a time-senstiive correlation. My imagination is there is no connection either way and the authors of this study should have put in many more soical variables. Crime, POPULATION, unemployment, change in unemployment and other social variables.
6) Do I expect any acknowledgment from Tim F. that he is misrepresenting both the study and Sullivan’s reaction to it? No, we are on teh blogz!!!1 Just be judgmental to and sarcastic about people with whom you disagree by taking what they have said or written completely out of context! Then people will finally recognized that you are in fact teh genius and that every pointless musing in your head should be shared with completely random people through the stupidest medium imaginable.
7) Sullivan is clearly being light-hearted.