Usually, I believe what I read in the New York Times. But not today; at least, not Nate Cohn’s piece on the “collapse” of Democratic support among Southern whites in 2012.
Cohn notes that “white Southerners have been voting Republican for decades,” but asserts that “the hugeness of the gap [in 2012] was new.” Oddly, though, there is no indication anywhere in the 1,000-word piece of how many white Southerners voted Republican in 2012 or in any previous election.
Reputable surveys such as the American National Election Study and the Cooperative Campaign Analysis Project suggest that Barack Obama won 30 to 35 percent of Southern white votes in 2012. That would be roughly consistent with the performance of Democratic candidates in recent presidential elections — and roughly five times the share of the African American vote won by Mitt Romney.
The claim that “Southern whites’ loyalty to Republicans is nearing that of blacks to Democrats” is the subhead on Cohn’s piece in the print edition and the headline on The Upshot, the Times’ new “news, analysis, and graphics” site. However, even allowing for the natural tendency of writers and headline-writers to strain for newsworthiness, this is an egregious overstatement of what Cohn actually shows.
The closest Cohn comes to justifying the headline claim is to note that “In many counties 90 percent of white voters chose Mitt Romney, nearly the reversal of the margin by which black voters supported Mr. Obama.” A handsome national map accompanying the post (but not in print) does indeed show many Southern counties shaded orange. But the map actually shows counties in which “Obama won 20% or less of the white vote in 2012,” not 10 percent. So how “many” counties in the map actually support the claim in the text? And more importantly, what share of the overall population of the South do they include? If — as my rough perusal suggests — populous counties are conspicuously unshaded, the map may be quite misleading in more ways than one.
Moreover, it is quite unclear from Cohn’s piece how his “estimates” of white voting patterns (apparently covering multiple election years) were constructed. A note accompanying the map says it is based on “analysis of exit polls and U.S. Census data.” But of course, exit polls do not cover many counties, and Census data do not indicate who anyone voted for. Perhaps an ecological regression analysis relating vote shares and racial population breakdowns? But then why not use actual voting data for all counties rather than exit polls?
The Upshot is an immensely promising addition to the burgeoning analytical news sphere — as evidenced, for example, by David Leonhardt and Kevin Quealy’s terrific piece on the American middle class in cross-national perspective. But first-rate analysis requires more than pretty graphs based on opaque manipulations of data unsuited to address the central substantive points.