With the new state House and Senate district lines in place, MinnPost has analyzed the political leaning of each district, based on the three most recent House elections: 2006, 2008 and 2010.
To determine the political leanings of the new districts, we drew inspiration from the Cook Partisan Voting Index (CPVI), which measures how strongly a geographic area leans toward one political party. Rather than comparing to the national averages, however, we averaged results from the three most recent legislative elections in Minnesota.
Each district is assigned a PVI value, which demonstrates its political leaning. R + 12 denotes that a district voted more for a Republican candidate than a Democratic candidate by 12 percentage points. Likewise, D + 2 would represent a two percentage point leaning toward the Democratic Party. Even means a political leaning of less than one percentage point.
The PVI represents a difference in percentage of vote won in a given election. If the Republican candidate won 45 percent of the vote, while the Democrat won 47 percent, the PVI would be D + 2.
Using geolocation software, we matched each voting precinct to its corresponding district in the newly released district lines. We then calculated the PVI of each district for each of the three most recent state House elections: 2006, 2008 and 2010. Finally we averaged these numbers to find an overall PVI for the districts, which is the number shown in our chart and map.
Each PVI is found using the following formula:
100 * ((Votes for a republican candidate / Total number of votes) – (Votes for a democratic candidate / Total number of votes))
The PVI uses sign of this number for the party and the absolute value for the percentage point lean. Our formula shows that a positive number denotes a Republican lean, and vice versa.
As an example, if a district had 120 votes for the republican and 80 for the democrat, its PVI would be R + 20.
100 * ((120/200) – (80/200)) = 20, which is positive, so we find R + 20.
To determine the “viability range” of the vulnerable districts displayed in the chart, we looked at the most recent election for which we had known results (2010) and compared the results to the average of the previous three house elections (2004, 2006 and 2008). The vast majority of cases where a seat changed party hands (more than 75 percent) occurred within 0 to 10 points of advantage. For that reason, we decided to focus on these districts as the most likely to be in play in this election.
For more technical details, check out our project on GitHub.
Gleaning more information from the method
It should be possible to do some trend analysis. Suppose the PVIs for each district for each year be plotted on a graph of PVI vs year. The data could be color coded by, say, region of the State.