I’m going to try to identify “polarizing elections”. These are elections where the voting public seems to be heading in different directions (sound familiar?). I identify the amount of “polarization” on a county-by-county basis, comparing the results to the previous election (and adjusting for the overall vote of the US). The further the county is from the expected vote, the larger its Polarizing Factor is. Sum up this number for all the counties, and you get a grand total for the US.
[First, a reminder about Adjusted Lean. This is determined by taking the party lean of a county and adjusting it by the national lean. Here’s an example: 2004 was won by the Republicans by 2.5 points. In the 3118 counties in the US, there was a wide range of voting results: some were strongly R, moderately R, even, moderately D, strongly D, etc. Let’s take a hypothetical county that was D+3.5 in that election. Its Adjusted Lean was D+6.0 (since the country was R+2.5, it was six points (2.5 plus 3.5) more Democratic than the country as a whole).]
The Polarizing Factor for a given election is simply the sum of the Polarizing Factors for each county. To calculate the Polarizing Factor of a given county, take difference between the Adjusted Lean from the previous election and the Adjusted Lean from this election. Multiply this by the number of voters in the county (thus adjusting for county size). Take the absolute value (i.e., make the number positive); otherwise the Polarizing Factors would just cancel each other out.
If the Adjusted Lean doesn’t change from one election to the next, then the Polarizing Factor will be zero. In this case, the county behaved just as we expected – it moved left or right in the same amount as the nation. On the other hand, if the Adjusted Lean changes a lot, then the Polarizing Factor will be high, reflecting that this county suddenly changed its behavior – perhaps going conservative when the rest of the nation was turning liberal, or vice versa.
Let’s look at two counties to see how this works. Here is La Plata County, Colorado:
| D | R | Lean | Adjusted Lean | Voters | |
| 2004 | 52.6% | 45.9% | D+6.7 | D+9.1 | |
| 2008 | 57.4% | 41.1% | D+16.3 | D+9.0 | 27979 |
At first glance, a slightly-Democratic county in 2004 became solidly D in 2008 (going from D+6.7 to D+16.3). But the Adjusted Lean takes into account national swing toward Democrats in 2008, and shows that the La Plata was pretty consistent: in both elections, about 9 points more D than the US as a whole. The difference in the Adjusted Lean from 2004 to 2008 is just a tenth of a percent (specifically, 0.142%). Multiply this by the number of voters in the county (to reflect the relative size of the county) and you get a Polarizing Factor of 39.7. What does this number mean? Technically, it’s the number of votes that would have to switch parties to make the Adjusted Lean unchanged. The lower the number, the more stable the voting behavior of this county is. 40 votes out of 28,000 is pretty stable.
Now let’s look at Perry County in Kentucky:
| D | R | Lean | Adjusted Lean | Voters | |
| 2004 | 46.3% | 53.1% | R+6.8 | R+4.3 | |
| 2008 | 33.2% | 65.2% | R+32.0 | R+39.3 | 10375 |
In this case, a slightly-Republican county swam against the D tide in 2008 and turned massively R. The difference in Adjusted Lean is 35.0% (39.3 minus 4.3), which results in a Polarizing Factor of 3460. At least a third of the voters in this county went against the grain. In 2008, this was not a stable county – voters changed parties quickly and decisively, and bucked the national trend toward D.
If an election consists of a bunch of counties like La Plata, it isn’t very polarizing. The country as a whole swung toward D or R, but we did so together: each county’s swing is similar.
If, on the other hand, if there are a lot of Perry Counties in an election, then things are not stable and polarization is occurring. Remember, Adjusted Lean takes into account the overall vote movement, so for every Perry County that polarized toward R, there is a corresponding county that overshot toward D in a big way.
Adding up the factors for each county, for the past five elections, and here’s what we get:
| 2000 | 2004 | 2008 | 2012 | 2016 | |
| Depolarizing Factor | 6,580,623 | 5,760,168 | 7,362,196 | 4,670,415 | 10,906,207 |
| DF as % of voters | 6.3% | 4.7% | 5.6% | 3.6% | 8.0% |
[Depolarizing Factor is the raw count. “DF as% of voters” adjusts for the number of voters in each election, which changes over time. In both cases, the larger the number, the higher the polarization.]
No surprise here – 2016 was massively polarizing. More than twice as polarizing as the 2012, and well above even 2008 (the election of the first black President).
Perhaps you’re all saying “well, duh”, but it’s nice to quantify it. Future posts will dig into these numbers.
