Last Tuesday was perhaps the most exciting day of 2017 for election junkies. The long-anticipated special election in Georgia’s 6th Congressional district finally came to a close as Republican Karen Handel defeated Democrat Jon Ossoff by 3.74 points in the most expensive House race in American history. In another race that was surprisingly close, Republican Ralph Norman defeated Democrat Archie Parnell by 3.10 points in South Carolina’s 5th Congressional district.
Coalition Shifting in the Trump Era
One of the most interesting aspects of the 2016 presidential election was how Trump and Clinton reshaped their party coalitions. While Trump made massive gains among white working class voters, Clinton made significant inroads with college-educated whites. This resulted in a fascinating 2012 -> 2016 swing map, where rural counties shifted massively towards Trump while affluent metropolitan suburbs shifted heavily towards Clinton.
The question moving forward is how these coalitions will continue to develop during the Trump presidency. Will white voters continue to diverge based on education level? Or will voters view Trump as an abnormality and return to their pre-Trump partisan leanings? The answers to these questions can provide us valuable insight about the political environment heading towards the 2018 midterm elections.
Why County and Precinct Results Matter
One reason so many political pundits have been following special election results is that they serve as a leading indicator of the political environment heading into 2018. However, the obvious downside is that they represent only five data points. While some election analysts may extrapolate from the five topline results that Democrats are over-performing more in white working class districts like South Carolina 5 while struggling in affluent suburban districts like Georgia 6, they ignore potentially confounding variables like candidate quality, perceived competitiveness, and state political environments. Drawing conclusions about the battle for the House from state legislative races is also problematic because these elections are much less related to Trump and Congress.
By expanding our analysis to include counties and precincts, we can analyze hundreds of data points rather than a handful. This gives us a much more robust data set to test our original question: how are partisan coalitions developing during the Trump era?
Analyzing County and Precinct Results from 2017
Let’s analyze the four House special elections that have taken place this year that have featured a Democratic nominee facing off against a Republican nominee. More specifically, we want to look at the correlations between past election results and current election results.
Let’s examine how predictive the past two presidential results were for the results. We can set up five different predictors: 100% Obama, 75% Obama + 25% Clinton, 50% Obama + 50% Clinton, 25% Obama + 75% Clinton, and 100% Clinton. This allows us to test our original question of whether voters are treating Congressional elections as a referendum on President Trump, or if they are returning to their pre-Trump partisan leanings.
Montana has 56 counties and Kansas 4 has 17 counties, so we should be able to analyze those elections on a county-level. However, since Georgia 6 and South Carolina 5 have fewer counties, we’ll want to analyze them on a precinct basis.
Predictiveness of Various Vote Predictors on Democrat Special Election Performance
Note: All values are coefficients of determination (r^2)
|100% Obama||75% Obama + 25% Clinton||50% Obama + 50% Clinton||25% Obama + 75% Clinton||100% Clinton|
As we can see in this chart, the most accurate predictor is ‘25% Obama + 75% Clinton.’ On average over the four elections, 94.4% of the variation in Democrat vote percentage by precinct can be explained by a 75/25 weighted average of the past two presidential results. Very close behind in basically a statistical tie is the ‘100% Clinton’ predictor.
What these results tell us is that House elections are largely driven by voter preferences from the 2016 presidential election. For example, in Georgia 6, the majority of Romney/Clinton voters likely pulled the lever for Ossoff. However, a significant minority returned to their pre-Trump partisan leanings and voted for Handel. While Democrats may hope to win all 23 of the Clinton districts represented by Republicans, some of them (like TX-07) may be just out of reach because of the vestigial strength of the Republican Party in these places.
Applying the 75/25 Benchmark to Georgia 6
County and precinct results from this year tell us that we can best estimate a district’s political partisanship by calculating a 75/25 weighted average of its past two presidential results. In 2012, Romney beat Obama by 23.3 points in Georgia 6. Since Obama won nationwide by 3.9 points, Georgia 6 had a partisan lean of R+27.2 that election. In 2016, Trump beat Clinton by 1.5 points in Georgia 6. Since Clinton won nationwide by 2.1 points, Georgia 6 had a partisan lean of R+3.6 last November. Using our 75/25 weighted average, Georgia 6 currently has a partisan lean of R+9.5.
What Georgia 6 Tells Us About the Current Political Environment
Georgia 6’s partisan lean of R+9.5 suggests that in a neutral environment nationwide, a Republican would win Georgia 6 by 9.5 points. Since Handel won by 3.73 points, this suggests that the nationwide political environment is roughly D+6. This makes sense intuitively, since Democrats currently lead in generic ballot polls by about 7 points. Given that Democrats need about a 5-8 point lead on the generic ballot to win the House, that battle is probably a toss-up right now.
Previewing Competitive 2018 Seats
Using our 75/25 weighted average of the past two presidential results, we can determine the partisan lean of each of the 435 Congressional districts. This gives us a rough preview of which Republican seats are most winnable for Democrats, who need to gain 24 seats to win back control of the House. One caveat to keep in mind is that incumbency effects will affect race competitiveness, and we do not know for sure which representatives are retiring before the midterm elections.
The 24 Most Winnable Congressional Seats for Democrats
See spreadsheet calculations here
All margins / partisan leans are from the Democrats’ perspective
|District||Current Republican Representative||2016 Presidential Margin||2012 Presidential Margin||Weighted Partisan Lean|
Although a lot of attention has been given to the 23 GOP-held districts that Clinton won, several of them don’t make this list. For example, Republican Dana Rohrabacher represents a district that voted Clinton by 1.7 points in 2016, but is helped by Romney’s substantial 11.7 point victory in 2012. His future challenger may run into the same problem Ossoff faced – just enough #NeverTrump Republicans returning to their conservative roots to deny the Democrats a victory. On the contrary, several of the 24 most winnable seats for Democrats are districts that voted for Trump. This suggests that Democrats should give a second look at some of the white working class districts that swung heavily to Trump – a small amount of mean reversion here may be just what they need to win back the House.