“Honey? Did you meet our new neighbors?”
“Not yet. Why?”
“I’m a little worried about them. They have an ‘I heart NPR’ sticker on their Prius.”
“Oh, God. Pass me the bottle.”
This is only one scenario, and a plausible one, of how the migration patterns of progressives can change the drinking habits of normal citizens.
This isn’t Yours Truly speaking. Science itself says so.
Or so claim Pavel Yakovlev and Walter Guessford in their peer-reviewed “Alcohol Consumption and Political Ideology: What’s
Party Got to Do with It?” (pdf) You’d be inclined to answer “Not much”, but then you haven’t taken the almost-desperate need of sociologist to research questions like this. And research they have, compiling statistic after statistic which prove, beyond all doubt, that a person’s mood and circumstance has something to do with whether they’ll take a tipple. Who knew?
News organizations are breathless over the pair’s “findings.” The Week leads with “Study: Liberals drink more alcohol than conservatives: Apparently, being liberal is thirsty work” and UPI chimed “Study: When a state becomes more liberal drinking increases” (via this site).
And just what are these “findings”? Foremost—and economists will want to take note—“Alcohol shipments are highly collinear with alcohol consumption”. The more alcohol bought, the more that is drunk. Counterintuitive? However, this is independent of politics, our main interest, so let’s push on.
A fellow named Berry allowed our duo to measure the political attitude of each State of the union. How States can have political attitudes is a vexing question better left unexamined. How Berry did it is, however, easily answered. Through equations like the one pictured above. Each State is awarded (via equations) its own unique number per year, which is labeled more or less liberal.
How does Berry handle States like New York, which in surface area is largely conservative except for the dangling carbuncle which is New York City? How dare you ask is the answer. Science, he says. So Science.
Anyway, our researchers calculated Berry’s Number for each State for each year between 1967 and 2003 and then averaged over the years. Yes, because not much has changed politically from 1967 to 2003. Result: one number per State. Then our pair took the amount of beer, wine, and booze flowing into each State for the same years; they took the population of each State for each of those years, then calculated the per capita average consumption of each type of alcohol. Then, in a step of statistical boldness, they averaged this average across the same time period. Result: one number per State of each of beer, wine, and booze per capita consumption. Yes, because drinking patterns (especially type) have not changed over this period.
You still following? Point is, all that manipulation allowed our guys to make plots like this (I left out wine, which looks like beer).
Now my dad would call this, and call it rightly, a “German Airplane”, i.e. a Messerschmitt. But then my old man isn’t a scientist. No, a scientist would say something like this, “The scatter plots shown in Figure 4, 5, and 6 suggest that average U.S. beer and wine consumption rises and spirits consumption falls as states become more liberal over time.”
It suggests nothing of the kind, because these plots are part silliness (the political leanings measure) and part unjustified averaging of heterogeneous data (both axes). And even if this weren’t a problem, the weird patterns in the scatter proves the futility and fallacy of using regression as causal explanation. Those lines are nowhere neat the dots. (Yes, of course, wee p-values were discovered.)
The authors do go on from these pictures to create a statistical model so freaky that it made my eyes tear, so I’ll skip discussing it. I can tell you it’s typical in its arbitrary, piece-meal construction and in the wildly speculative conclusions drawn from it.
How they take all this and conclude “that when a state becomes more liberal politically, its consumption of beer and spirits rises” says more about the authors’ preconceptions than it does about the drinking habits of citizens.
On the other hand, I’m perfectly willing to accept that the more progressive a person becomes they more they are driven to drink. These sad people have to cope somehow.
Announcing a new service. I will, for a modest fee, check whether a science headline or story accords with the evidence. Most don't.
— William M. Briggs (@mattstat) February 2, 2014