In a truly dominating performance, Democrat Ralph Northam won the Virginia Governor’s seat over Republican Ed Gillespie by 8.93 points last Tuesday. Since this was one of the only major competitive races in 2017, political strategists around the country are looking to its results to learn lessons heading into the 2018 midterm elections.
Correlation with Past Elections
One of the major questions heading into last Tuesday was how the 2017 race would mirror past Virginia elections, which had widely divergent geographic patterns. Although Democrats won Virginia narrowly in the 2012 Presidential, 2013 Gubernatorial, 2014 Senate, and 2016 Presidential elections, each Democrat took a different path to victory. Clinton made massive gains over Obama in the Northern Virginia suburbs, while Trump made massive gains over Romney in rural Southwest Virginia. In the 2014 Senate race, Gillespie did unusually well for a Republican in Northern Virginia thanks to his establishment credentials.
Using results from each of the 133 Virginia counties and independent cities, we can calculate correlations between previous election results and 2017 election results.
Predictiveness of Various Vote Predictors on Northam’s 2017 Margins
Note: Learn more about coefficient of determination (r^2) here.
|Predictors||Coefficient of Determination (r^2)|
|2012 Presidential Election||.906|
|75% 2012 + 25% 2016||.943|
|50% 2012 + 50% 2016||.966|
|25% 2012 + 75% 2016||.979|
|2016 Presidential Election||.983|
|2013 Gubernatorial Election||.939|
|2014 Senate Election||.855|
As we can see from this chart, the most accurate predictor is the 2016 Presidential Election. 98.3% of the variation in winning margin by county can be explained by the 2016 presidential results. This was a far more accurate predictor than the 2012 Presidential Election, where only 90.6% of the variation in winning margin by county could be explained by the 2012 presidential results. The 2013 Gubernatorial election was about halfway between the two presidential elections in predictiveness, while the 2014 Senate election was an unusually poor predictor of 2017 results. This is noteworthy because Gillespie ran as the Republican candidate in both 2014 and 2017.
What this tells us is that the 2017 Virginia Gubernatorial race was largely driven by voter preferences from the 2016 presidential election. This is consistent with results from the four competitive House special elections earlier this year, where Clinton vote percentage was a much stronger predictor (r^2 = .943) of Democratic House performance than Obama vote percentage (r^2 = .908). This suggests that the demographic and geographic trends from 2016 were not an anomaly – they are continuing to persist even when Trump is not on the ballot.
Gillespie Did Well in Southwest Virginia
Despite his establishment background, Gillespie emulated strategies from the Trump playbook, running ads focused on cultural issues like Confederate statues, sanctuary cities, and felon voting. Many analysts speculated that Gillespie was doing this to appeal to Trump’s white working class base that delivered him big gains in rural Southwest Virginia.
Gillespie ended up doing pretty well in the 36 counties and independent cities of Southwest Virginia. While previous statewide Republican candidates carried the region by 24 points in 2012, 26 points in 2013, and 20 points in 2014, Gillespie carried Southwest Virginia by 31 points. This was only several points off Trump’s gigantic 35 point margin in 2016.
Gillespie’s gains were fairly consistent throughout the region – he outpaced McAuliffe’s 2013 margins in 29 out of 36 Southwest Virginia counties and independent cities.
Gillespie Tanked in the Suburbs
Another major factor going into the election was how competitive Gillespie would be in the Northern Virginia suburbs right outside Washington D.C. These affluent counties have many college-educated whites and minorities and shifted strongly towards Clinton in the 2016 presidential election. This includes the three biggest counties in the state: Fairfax, Prince William, and Loudoun. Clinton won all three by large margins: 36 points in Fairfax, 21 points in Prince William, and 17 points in Loudoun. However, Democrats haven’t always done so well here. In 2014, when Gillespie ran for Senate, he held Warner to a 17 point win in Fairfax, a 3 point win in Prince William, and actually won Loudoun by less than a point.
This time around, Gillespie did really badly in the Northern Virginia suburbs. His margin of defeat was worse than that of Republican nominees in 2012, 2013, 2014, and 2016 for all three counties. Gillespie lost by 37 points in Fairfax, 23 points in Prince William, and 20 points in Loudoun. This suggests that Gillespie lost many of the voters who voted for him in 2014 and Clinton in 2016. This could be a bad omen for the 23 House Republicans who represent seats that Clinton won.
Gillespie also did really badly in the Richmond suburbs, which have historically leaned heavily Republican. Northam actually won Chesterfield County (Virginia’s 4th largest county), which no Democratic presidential candidate has carried since 1948. This represents a modest shift from the four previous statewide elections, where Republicans won by single digits. Northam also did well in Henrico County (Virginia’s 5th largest county), where he won by 23 points. Northam did a few points better than Clinton in these two counties, just like he did in the Northern Virginia suburbs.
Northam’s Strong Home Base Showing
Some of Northam’s strongest performances relative to 2016 came in his home base of the Norfolk / Virginia Beach / Eastern Shore area, where he served as State Senator for the 6th District from 2008 to 2014. Northam improved upon Clinton’s margin by 9 points in Virginia Beach City, 8 points in Mathews County, 6 points in Norfolk City, 5 points in Northampton County, and 4 points in Accomack County. This suggests that Democrats may be better off running candidates with already strong ties to a certain region.
Gillespie’s Improvements vs. Trump
Although the map of partisan shifts relative to the 2016 presidential election was a sea of blue last week, Gillespie improved on Trump’s margins in 36 of Virginia’s 133 counties and independent cities. Many of these were counties towards the Southern end of the state between Richmond, Lynchburg, and the North Carolina border.
Partisan Margin Shift (2016 -> 2017) by Precinct
Many of these areas have a high percentage of African Americans. The median county / city where Northam did worse than Clinton was 32.35% black. The median county / city where Northam did better than Clinton was 9% black. Looking at the state as a whole, there was an inverse relationship (r^2 = 0.259) between the percent of African Americans in a county and Northam’s overperformance relative to Clinton.
There are a few potential reasons why Northam may not have done as well as Clinton in heavily African-American counties. One possibility is that African American turnout was lackluster relative to that of other demographic groups. Patrick Ruffini of Echelon Insights found that the larger Clinton’s margin in a precinct, the greater the turnout improvement relative to 2013. However, this trend reversed dramatically at a margin of about D+50, suggesting that the turnout surge was not as impressive in heavily African-American precincts.
A second possibility is that Gillespie did better with African-American voters than Trump. Exit polls (although they can be problematic) corroborate this theory somewhat – Virginia black voters favored Clinton by 79 points and Northam by 75 points.
A third possibility is that the election’s focus on cultural wedge issues like Confederate statues helped Gillespie gain among white voters in the more Southern areas. While there was a significant correlation at the county level between Northam’s gains and black percent, there was very little correlation at the county level between turnout relative to 2016 and black percent. Perhaps white turnout in these counties increased enough relative to 2013 to balance out smaller increases in black turnout.
Prior to election night, there were some major concerns on the Democratic side about minority turnout in the post-Obama era, particularly during off-year elections. This may be problematic for Democrats down the line, but it’s also possible that increased turnout among college-educated whites can counterbalance that trend. This may also be less problematic for Democrats in their quest to take the House, where minorities are often packed into noncompetitive districts.
Trump undoubtedly played a big role in this election, based on the correlation between 2016 and 2017 results and geographic voting patterns throughout the state. Although gubernatorial races have been fairly independent of presidential voting in recent election cycles, perhaps we will see them become more nationalized just like Senate races in 2016. Nationalization could be a good thing for Democrats at the Gubernatorial level, where there are 9 Republican Governors in Clinton states but only 4 Democratic Governors in Trump states. However, this could present a major hurdle for them in the 2018 Senate elections, where Democrats must defend 10 Senate seats in states Trump won.