Recently, there has been a concentrated focus on the national mood and how it may impact the 2018 midterm elections. Many political analysts believe that due to a number of factors, including the current presidential administration, the influx of Democratic candidates on the ballot and a better understanding of climate change, gun control and minority rights, the Democratic Party will fair well in the upcoming election, taking seats previously held by Republicans. This phenomenon is referred to as the “Blue Wave.”
Curious about the merits of this claim, we decided to investigate whether this trend was evidenced in our model. To do so, we selected ten of the closest districts in our model, CA-39, NC-09, NJ-07, Fl-06, KS-02, WA-08, CA-21, CA-25, NJ-02, and ME-02, and recorded a number of different statistics for each. We collected data from the 2014 Congressional elections and the 2016 Congressional elections and recorded whether or not our district had an incumbent. Then we evaluated a few different predictions made by our model such as the Blair Partisan Index (BPI) a measure of how strongly a district leans toward the Democratic or Republican Party compared to the nation as a whole, SEER, which is similar to BPI but takes incumbency into account, and our models predicted percent chance of victory, and margin of victory for each district.
We found that all ten of these 2018 battleground districts had gone to Republican Candidates in both the 2014 and 2016 Congressional elections. The SEER predictions for each district also indicated that each district, without polling taken into account, would be predicted to go to the Republican candidate. Additionally, when we evaluated the BPI for each of the districts we had chosen we found that eight of the ten had a Republican leaning BPI of 2.5% or over. Our models numbers, coupled with a Republican voting history indicates that they are districts that by standard metrics should go Republican.
However, our model predicts that five of these ten districts will swing Democrat. ORACLE of Blair predicts that California’s 39th district will vote Democrat with a 0.10% margin of victory, New Jersey’s seventh district will go Democrat with a 0.3% margin of victory, California’s 21st district will vote Democrat with a 0.2% margin of victory, and New Jersey’s second district will vote Democrat with a 0.4% margin of victory. In California’s 25th district, Democrats also have a 50.1% of winning even though the predicted vote is split evenly 50/50. It is very telling that the ten closest elections, as predicted by our model, were all previously Republican strongholds.
In our bubble plot, we weigh the 2016 congressional vote against the 2014 congressional vote. Each bubble represents a district and the size of the bubbles represents the margin of victory for the current 2018 midterm elections. Because these are swing districts, all vote predictions are close to 50% and therefore almost the same size.
We were also curious to see what the predicted vote would be if we excluded polls and made predictions. Of the five swing districts predicted to vote Democrat in the 2018, ORACLE would have predicted every one of those districts going to the Republicans if we had not included recent polls indicating that voters were favoring Democrats in our model. This shows that districts with demographics and fundamental data that would typically favor Republican candidates are polling with higher numbers of Democratic support. Our table looks at SEER, which gives our prediction without looking at national mood or polling. Once we add polling to our prediction, we found many of the districts leaned more Democrat. While we may not be able to say for sure whether the “Blue Wave” is coming, there is evidence that previously Republican-held districts are now toss-ups.
Current Predicted Winning Party
Current Predicted Margin of Victory
2016 Congressional vote
2014 Congressional Vote
Difference in SEER and final prediction
# of polls