How MinnPost calculates political leans of House districts

There are 134 state House seats up for election in Minnesota this year, but not all of those races are going to be hotly contested. So how do we decide which districts will be the most competitive and therefore most worthy of coverage? Enter PVI.

PVI, or Partisan Voting Index, was inspired by the Cook Partisan Voting Index, but it is a considerably simpler measure: rather than adjusting our political leans to some kind of statewide average, we simply average the margin of victory in each House district over the past three elections (so the 2014 PVI uses data from 2008, 2010 and 2012). Specifically, for each election, we subtract the percentage of the vote won by the DFL candidate from the percentage won by the Republican. (A negative PVI indicates DFL lean while a positive PVI district leans Republican.)

Redistricting made the calculation a little more complicated: voters in 2008 and 2010 were not voting in the same districts as voters in 2012. We solved the problem by taking precinct level vote totals by party and assigning precincts to the new districts based on where their geographic centers lay. Obviously this is an imperfect method: voters were not choosing among the same candidates in these artificially created districts. But it was our best idea for dealing with basically unknown districts.

How well does this work?

We calculated PVI for the 2012 legislative election. Here are the PVI numbers for each House district plotted against the actual 2012 election results:

If PVI perfectly predicted actual results, the dots would form a straight, diagonal line (and elections would be really boring!). But our goal is not to predict electoral outcomes, but rather to know which races will be hotly contested (and therefore presumably have a close result). By this standard, PVI measures up reasonably well: many of the races with low PVIs ended up with narrow margins of victory.

With the 2012 result in mind, we’re going to use PVI again as a big part of our decision on which House races to cover. You can see the 2014 numbers in “How Minnesota's House districts lean politically in 2014.”

You can also learn about all our free newsletter options.

Comments (1)

  1. Submitted by Kelly Guncheon on 09/10/2014 - 12:16 pm.

    Assumes constancy

    This seems to be of limited value. Were candidates exactly the same and there were no change in issue or the environment, this would be valuable, but neither of those are even close to being constant.

Leave a Reply