I am not sure if it is bad math or bad journalism, but contrary to popular accounts, it is highly unlikely that 450,000 voters in Minnesota split their votes between Dayton or Franken at the top of the ticket and a Republican legislator further down the ballot.
On November 6, 2014, in a Star Tribune article by Rachel E. Stassen-Berger and Glenn Howatt where they analyzed the results of the 2014 Minnesota elections, they sought to reconcile the difference between Democrats winning statewide and Republicans winning the House. They asserted that of the estimated 1,992,989 Minnesotans who voted, effectively 22.5% split their ballots. Conceptually and empirically, this is just incorrect.
First, keep in mind that no one can actually look at the individual ballots cast and therefore the Star Tribune’s article is purely conjecture. Conceptually, asserting that between a fifth and a quarter of voters split their ballots is unlikely. The political science literature is overwhelming in finding that partisan identification is a major driver and predictor of voting behavior. This has been true in the nearly 60 plus years of research into voting behavior. It is even more true today as the evidence mounts that voters now are more polarized and partisan in their voting than ever before. This is even true in Minnesota. Every since the Wellstone plane crash and memorial service there is powerful evidence of partisan voting, as evidenced by the close races or recounts in the 2008 senate and 2010 gubernatorial races.
Yet one might argue that Minnesota is different. With about 20 or so percent of the electorate not listing themselves as a Democrat or Republican perhaps one might say this high percentage of independents accounts for the split ticket voting. It might account for a small percentage of this, but there are similarly high percentages of independents across the US with little evidence of split ticket voting. Thus, Minnesota exceptionalism is not the answer.
Instead, the real answer has to do with now the geographic voting patterns in the state. A look at the state election results indicate that Mark Dayton for example, won 34 of the state’s 87 counties, with Jeff Johnson winning the majority at 53. Dayton (and Franken) racked up big wins in Hennepin and Ramsey counties, as well as several other traditional DFL ones, but lost elsewhere. Now look at where Republican legislators did well and won–in the counties where Jeff Johnson won. Such a geographic pattern can easily explain the apparent anomaly of Democrats winning statewide and Republicans winning at the legislative level.
Put simply, Democrat votes are concentrated in a few geographic areas of the state and there are more of them and they overwhelming voted for DFL statewide and legislative candidates whereas Republican voters are dispersed across the state and the voted straight party line for Jeff Johnson and Republican legislators.
The Star Tribune article thus conceptually and empirically got it wrong. Moreover, it also committed a variation of the classic ecological fallacy–falsely inferring characteristics about individuals based on aggregate or group behavior. Here they assumed individual behavior about voting based on overall statewide voting. Yet they did so without understanding the way the votes actually were distributed across the state and for the candidates.