16 June 2014

Driven to Data Distraction

OK, twist my arm. It's Cantor time.

David Carr titles today's piece, "Eric Cantor's Defeat Exposed a Beltway Journalism Blind Spot" (my dead trees version has a slightly different headline). Much of his writing is about the evisceration of newspaper staffs, but he also, obliquely, takes aim at the data bits of elections. Specifically, campaign internal polls. Only good news can be had.
Data-driven news sites are all the rage, but what happens when newspapers no longer have the money to commission comprehensive, legitimate polls? The quants took a beating on this one, partly because journalists are left to read the same partisan surveys and spotty local reporting as Mr. Cantor's campaign staff, whose own polling had him up by more than 30 points.
[emphasis added]

There may well have been nothing wrong with Cantor's polling per se, in the sense that Cantor's operatives didn't purposely skew the data to please Cantor and campaign management. What a candidate tells the public is a different story, of course; but serious ones want serious, accurate, polling data. Otherwise, he's sailing blind. As with all stratified random sampling, getting the strata right is the most important task. Only then can you draw proper inferences. I'll go out on a limb, and postulate that Cantor's campaign didn't spend much time polling in the truly redneck counties (too many miles for too few responses, and it's been reported in the post mortem that he campaigned not much), and that Brat (and the Right Wingnut surrogates like Ingraham) did. They all got in their pickup trucks with gun rack and plastic Jesus and headed off to vote that Communist Cantor off the ballot. I mean, isn't that a Jewish word?

As asserted here much of the time, data is subservient to policy whenever they disagree. Believing one's own propaganda is fraught with terror. In Cantor's case, one might chuckle that he was hung by the lynch mob he created. One could also conclude that there's a scary bit of Nazi out there in the South. As if it ever left. The real question will be whether the 65,000 or so who voted in the primary represent the thoughts (I use the word with trepidation) of most of 7th/VA? Or could it be Something Else? Read on.

In a previous missive I asked, plaintively, what Nate might have thought. Well he did have some thoughts. He takes the earthquake analogy to the ends of the earth. An interesting read, but I didn't see that he did any better than any other pundit in explaining the defeat. It clearly wasn't in any of the available data before the election, as Carr points out. But what was Ingraham doing out there?

If there were ever a case where policy beat the crap out of data, this is it. No rational Republican would cut off his nose to spite his face by giving up a seat of such power. But the yahoos in 7th/VA were happy to do so. Or, as Will put it, Cantor was: "hoisted with his own petard". Or not?

Some data.

According to the Cook PVI, 7th/VA isn't even the most right wing district in VA, at R+10. But, anyone who's lived in the DC environs knows, there's two Virginias, and the Old South version is increasingly paranoid, even as it gathers up moolah from the socialist bureaucrats living in the DC commuting counties.

Let your fingers do the walking through the Google pages. In an unsuccessful attempt to find the precinct level vote cast (did Brat really win by getting out the redneck vote?), I did find this bit of irony:
Virginia has an open primary process, in which registered voters do not have to be members of a party to vote in that party's primary. With the Democrats in the 7th District having already nominated their candidate at a convention on June 7, they were free to vote in the Republican primary on Tuesday. The 17,900 additional voters casting a ballot in this year's Republican primary relative to in 2012 could be a result of Democrats voting in an attempt to unseat Cantor.

Even so, only about 10% of the district's voters made the decision. About time we had mandatory voting?

Could it be? Democrats can play dirty? Or did Ingraham, et al, persuade the gun rack brigade? We won't know with any certainty until the general election, if then. A Democrat false flag operation would certainly explain the discrepancy between polling and outcome. But so would the gun rack brigade. An educated surmise could be drawn if one knew which precincts accounted for those "extra" 17,900 votes. Did they happen in Blue Leaning precincts or Red Leaning ones? Were these votes spread across all precincts? Didn't find the numbers, alas. I've been (attempted) to be polled, and the first question is "what's your party?". Cantor's pollsters would likely have skipped those that said "Democrat". Or the long-hair with tattoos who said, "fuck off". And so forth.

Finally, The Times has this graph (thanks R?), which appears to show Cantor getting the DC ex-urbanite vote, while Brat took the downstate ruralists and some suburban Richmond. Could be a mix of Democratic counterinsurgency and Tea Party get out the vote. This Fairfax County FAQ says affiliation is not tracked.

Cantor's defeat is open to much speculation. Love it.

No comments: