Sabato's Crystal Ball

For House Republicans, Past Performance Is No Guarantee of Future Results

What recent midterms tell us about how members of the presidential party fare

Kyle Kondik, Managing Editor, Sabato's Crystal Ball July 20th, 2017


As they dig their trenches to try to withstand what may (or may not be) a Democratic wave, Republicans may take heart in the performance of their current incumbents last year as a buffer against a potentially challenging environment next year.

Very few of the Republicans’ 241 House victories last year were small. Just 15 districts featured a GOP win in the single digits, meaning that 226 House Republicans had margins of victory of 10 points or more. So in order to capture a majority, Democrats must win several districts that most recently delivered decisive wins to the GOP.

However, the results from the last election may not actually be all that meaningful because House performance can vary greatly from one election to the next, particularly if there is a wave-style environment. The last three midterms are illustrative in this regard.

During the 2006, 2010, and 2014 midterms — all of which featured decisive House victories for the non-presidential party, which is relatively common in midterms — we looked at how the two-party vote shares for candidates from the president’s party changed from the preceding presidential year to the midterm. In other words, we looked at the Republican two-party vote share in 2006 compared to 2004 (under Republican President George W. Bush), and the Democratic vote share in 2010 compared to 2008 and in 2014 compared to 2012 (under Democratic President Barack Obama).

Over those three midterms, about 1,000 races — 1,001 to be precise — satisfied the requirements for inclusion in this study.[1] We used a compilation of two-party U.S. House results from Gary Jacobson, the distinguished University of California, San Diego political scientist.

On average, the presidential party’s share of the two-party vote declined by 5.3 percentage points in these three midterms from the previous cycle’s results. And, since we’re just taking into account the two-party vote, the actual average change in margin is double that, because when the president’s party share declines, the non-president’s party share grows by that same number.

So the average reduction in the presidential party margin in these thousand House races was actually 10.6 points. That kind of change turns a 55%-45% win for a House candidate from the president’s party into roughly a 50%-50% coin flip.

The median was a little bit lower than that, about a 9.2-point decline in margin. Of the 1,001 races we looked at, the president’s party share fell by at least some amount in close to nine of 10 of all the races, 874 total (or 87%). So the presidential party margin grew from the presidential election to the midterm in more than a tenth of these races despite the overall trend against that party. The reasons for these exceptions often included an underperformance by a weakened member of the non-presidential party; a new member from the presidential party elected in a presidential year building an incumbency advantage in a midterm reelection bid; or other factors. But generally speaking, the presidential party experienced an erosion in performance in the lion’s share of House districts from the presidential to the midterm year.

The declines varied by cycle, with the president’s party seeing their biggest average vote share losses in 2010 (7.5 points, or about 15 marginal points), with smaller but still sizable decreases coming in 2006 (4.6 points, or 9.2 marginal points) and 2014 (3.5 points, or about seven marginal points). Part of that probably had to do with the comparison to the previous presidential year House results. In both 2004 and 2012, the national House popular vote was fairly close — Republicans won it by 2.6 points in 2004 and Democrats by 1.4 points in 2012 — before the other party carried the national popular vote by a larger margin in the subsequent midterm — Democrats by 6.4 points in 2006, Republicans by 5.1 points in 2014. Meanwhile, Democrats won the national House vote by 10.6 points in 2008, which shifted to a 6.6-point Republican edge just two years later, contributing to the huge average Democratic vote share losses in 2010.

But these specific yearly averages reinforce a larger trend, which is that the president’s party often struggles in midterm House elections. The president’s party has lost ground in the House in 36 of the 39 midterms since the Civil War, and the average seat loss is 33. Democrats netted 31 seats in 2006 during George W. Bush’s second midterm, and Republicans netted 63 and 13 seats, respectively, in 2010 and 2014 during Barack Obama’s two midterms. The GOP gain in 2014 was relatively small in large part because the party already held 234 seats going into that midterm, and their 13-seat net gain gave them their biggest majority (247) since before the Great Depression.

The presidential midterm problem compounds whenever the president’s approval rating is underwater: Bush’s was -16 (39% approve/55% disapprove) in the RealClearPolitics average on Election Day 2006, and Obama’s was -3 (46%/49%) in 2010 and -11 (42%/53%) in 2014. Right now, President Trump is at -16 (40%/56%), and Democrats hold a seven-point lead in the RealClearPolitics national House generic ballot polling. So the threat to the Republican House majority is clear, even though of course those numbers have plenty of time to change.

But who does the threat loom largest against? And what kinds of seats might be in play? Here’s where going a little deeper into the vote share changes from the last three midterms may be helpful.

Remember, even in big wave years, most House races are not all that competitive. While there are all sorts of ways to measure competitiveness, House race ratings (done by the Crystal Ball or other handicappers) is one way to assess the districts that both sides saw as at least somewhat close and contested going into the election. These are the districts that are likely seeing the most competitive campaigns and the most outside spending.

Of the 1,001 races we studied over the three midterms, 238 were rated in competitive categories by the Crystal Ball in those years’ final House rankings. These were the races that were rated as Toss-up, Leaning, or Likely. The remaining contests were rated as Safe for the incumbent party. Perhaps unsurprisingly, the average drop in two-party vote share for the presidential party in the competitive races was greater than that in the ones that were not rated. For the competitive contests, the average vote share for the presidential party dropped eight points over the three midterms. And, let’s recall that we should double the eight points to fully account for the margin. So a competitively rated seat that provided the presidential party with a seemingly gaudy 58%-42% win in the previous presidential election would, on average, become basically a 50%-50% race. Safe seats, meanwhile, did not see as big of a drop, with a presidential party vote share drop of a little more than half of the rated seats, 4.5 points. This makes sense — rated races are ones likely to feature the strongest challengers, the most outside spending by both parties, and the most engaged electorates.

Rated races may include many open seats, too. As we’ll show below, whether a seat is open or features an incumbent is a very important factor.

Individual races often deviated far from this average. In the whole dataset, the biggest drop for the presidential party came in NE-3 from 2004 to 2006. In 2004, then-Rep. Tom Osborne (R), the former Nebraska football coaching legend, was romping to reelection, and he captured about 91% of the two-party vote. The following cycle, Osborne lost a very competitive Republican gubernatorial primary to Dave Heineman, who would go on to serve two terms as governor. Now-Rep. Adrian Smith (R, NE-3) captured the open seat, but he only won 55% of the two-party vote in the heavily Republican, rural district. The race seemed so competitive that then-President George W. Bush, who had won three-quarters of the vote in the district just two years prior, campaigned for Smith in the district just days before the election.

Seemingly secure ground held by the presidential party can become very insecure in a wave, particularly when there’s an open seat.

We divided up the House races from the three midterms we studied into four categories: Seats that featured an incumbent from the presidential party; seats featuring an incumbent from the non-presidential party; open seats defended by the presidential party; and open seats defended by the party that didn’t hold the White House.

The seats featuring incumbents from both the presidential and non-presidential party changed in similar ways. On average, the vote share for the party that did not hold the White House grew by 4.4 points in seats where it had an incumbent running for reelection, while the non-presidential party’s vote share rose six points in seats where an incumbent from the presidential party was running. If this pattern repeats in 2018 — again, a very big if at such an early point in the cycle — one would expect Democratic incumbents who won very close races in 2016, like Reps. Ami Bera (D, CA-7) and Rick Nolan (D, MN-8), to be able to improve their vote share in 2018 simply by virtue of the typical bonus that the non-presidential party receives.

Indeed, of 408 races that met the requirements for inclusion in this study featuring incumbents who were not members of the president’s party, 405 won reelection in the midterm. The three exceptions were: Reps. Lee Terry (R, NE-2) and Steve Southerland (R, FL-2), both of whom had significant, self-inflicted political damage and lost to Democratic challengers in 2014, the least “wavy” of the three elections studied, and Rep. Charles Djou (R, HI-1), who won a fluky special election in May 2010 thanks to a Democratic split in the all-party election for the seat. Djou then lost the seat in the regular general election later that year.[2]

Meanwhile, of the 499 races featuring incumbents from the president’s party, 81 lost over the three midterms studied. Given the average decline of six points in presidential party performance from the presidential year to the midterm year (or 12 points in terms of margin), the following members may be vulnerable next year:

Table 1: Potentially vulnerable House Republicans based on 2016 two-party margin of victory

Note: First-term members are in bold, retiring members are in italics.

Source: Vote by congressional district from Daily Kos Elections. 2016 victory margins from the Cook Political Report.

Perhaps more interesting, though, is what happened in open seats in the last three midterms. Granted, the sample sizes here are not that large — the presidential party defended 46 open seats that qualified for our list over the last three midterms, while the party that did not hold the White House defended 48 — but the difference in presidential party performance between the two is eye-popping, if not necessarily surprising.

In the open seats defended by the party that did not hold the White House, there was essentially no change in performance on average — the average change from presidential to midterm was a tiny .02 points in favor of the president’s party — meaning that the party that did not hold the White House typically paid little penalty for not having an incumbent in these seats. In fact, the non-presidential party only lost one of these 48 races over the three midterms: Delaware’s at-large seat, which now-Gov. John Carney (D) captured from the Republicans in 2010 thanks to popular then-Rep. Mike Castle’s (R) decision to run for Senate (he lost the primary in a surprising upset to Christine O’Donnell, which almost assuredly cost Republicans a Senate seat that year in addition to losing Delaware’s lone House seat).

Conversely, the results in open seats defended by the presidential party saw huge swings in favor of the opposite party. In such seats, the presidential party share declined about 11 points from the presidential to the midterm elections — or 22 points in terms of margin — and the president’s party only held 25 of the 46 seats included in the study over the three midterms. It was very rare for the president’s party to improve on the previous incumbent’s performance from the presidential to the midterm: The only midterm winner who did so in a significant way was now-Rep. Seth Moulton (D, MA-6), who defeated scandal-tinged Rep. John Tierney (D) in a 2014 primary and then markedly improved on Tierney’s narrow 2012 reelection margin. So typically, though not always, seats held by incumbents are easier to defend.

The results for open seats not held by the presidential party suggests that Democrats would have roughly even odds to hold their two most vulnerable open seats, NV-3 and MN-1, two districts President Trump won and that the outgoing Democratic incumbents only carried by about a point apiece. If a Democrat were in the White House, these seats probably would be easy pickups for Republicans, given the historical record.

There is not an enormous number of Republican open seats right now, and there’s only one that is both an open seat and that has an incumbent who won a relatively weak victory in 2016: FL-27, held by outgoing Rep. Ileana Ros-Lehtinen (R). Ros-Lehtinen holds the most Democratic seat held by any Republican — Hillary Clinton carried it by 20 points last year — and she only won by a little less than 10 points. Recreating the average open seat performance by the presidential “out” party would flip FL-27 from a 10-point GOP win to a 12-point Democratic win. This should be an easy Democratic pickup in the event of a good Democratic environment even though Republicans are typically strong down-ballot in South Florida. A Republican hold in this district would be an utter disaster for Democrats and probably would be suggestive of a midterm where Republicans performed quite well by historical standards.

But beyond FL-27, the other seven current GOP open seats[3] do not look like particularly attractive Democratic targets. Even including Ros-Lehtinen, the average GOP two-party vote margin of victory in these open seats was about 30 points. That’s stronger than the average 22-point loss in margin the presidential party has lost in open seats going from the presidential to the midterm, although there are several instances (all from the 2006 and 2010 waves) of the presidential party margin falling 30 points or more in open seats. Still, it may be that one or more of these seats will become a Democratic target next year. One longshot possibility may be TX-3, held by retiring Rep. Sam Johnson (R). The suburban Dallas seat is still very Republican, but Hillary Clinton cut Barack Obama’s 30-point loss there in 2012 in half, losing by 14 points — the district featured one of the biggest swings toward Clinton of any GOP-held seat in the country. And Johnson’s 28-point margin of victory suddenly doesn’t seem so impressive when one considers the huge drops in performance the president’s party sometimes suffers in open, midterm defenses.

Table 2: Open GOP House seats coming up in 2018 so far

Source: Vote by congressional district from Daily Kos Elections. 2016 victory margins from the Cook Political Report. Special election results from official sources.

We do already have a few examples of what’s happened in GOP-held open House seats in the Trump era. There have been four special elections in these seats — KS-4, MT-AL, SC-5, and GA-6. Republicans held all of these seats, but with significantly reduced margins from what the incumbents did last November. On average, the four former incumbents won by 23.7 points in the two-party vote margin in 2016. This year, the special election winners won by 4.7 points in the same measure, a drop of 19 points, very similar to the presidential party average loss of margin in the last three midterms in open seats (22 points).

So perhaps the overall message from this analysis is to watch, in particular, the GOP-held open seats, and how many more emerge over the next year. There will be more retirements, and they could make seats that were not very competitive in 2016 very competitive in 2018.

Depending on how many there are, these retirements could be a major factor in determining which party controls the House in 2019.

Beyond that, GOP incumbents who run again next year should heed the old cliché: run unopposed, or run scared. Their previous big wins may not protect them in the event of an anti-Republican environment next year. And for the Democrats, historical averages may be tantalizing but they provide no guarantees of easy victory, or victory at all. There will be a titanic struggle for each and every seat that is even marginally competitive.

No outcome is preordained.


1. Here’s how we decided which races to include: For each of the three presidential-to-midterm comparisons, any district that was uncontested by one of the two major parties in either election was excluded. We also excluded elections in states that use a “top-two” system (California and Washington) in which two members of the same party advanced to the November election. Districts that were redistricted between the presidential election and the midterm were also discarded, meaning that all of Georgia’s results and a handful of Texas districts are excluded from the 2004-2006 analysis because those states used different maps in those two elections. Finally, we did not include any results from Louisiana because of its unique, November “jungle” primary system that sometimes produces December runoffs. Understandably, this culled hundreds of races from our three-midterm study, but we still ended up with 1,001 total races over three time periods: 316 in 2004-2006, 357 in 2008-2010, and 328 in 2012-2014.

2. Our decision to exclude Louisiana from our study removes one of the rare losses for a non-presidential party incumbent in a recent midterm. In 2008, Joseph Cao (R) narrowly defeated indicted Rep. William Jefferson (D) in a heavily Democratic seat based in New Orleans. Two years later, Cao won only a third of the vote against now-Rep. Cedric Richmond (D, LA-2).

3. Even though the seat is currently open, we did not include UT-3 on this list of open seats. That’s because the seat, previously held by former Rep. Jason Chaffetz (R), will be filled later this year in a special election, and we’re assuming the winner of that special election will be on the ballot next November. Of course, it’s possible that the special election winner will not be renominated next year, but that seems very unlikely. Also not included is Rep. Mo Brooks (R, AL-5), who is running in a special election for Alabama’s vacant U.S. Senate seat. If Brooks doesn’t win the Senate seat, he could still run for reelection to the House in 2018, so we’re not counting his seat as open at the moment.

The New Dominion: Virginia’s Ever-Changing Electoral Map

Examining presidential elections in the commonwealth from 1968 to 2016

Geoffrey Skelley, Associate Editor, Sabato's Crystal Ball July 13th, 2017


In 2008 Barack Obama carried Virginia’s 13 electoral votes, becoming the first Democratic presidential nominee since Lyndon Johnson in 1964 to win the Old Dominion. Obama’s victory broke a run of 10 consecutive Republican victories in the commonwealth, and 13 of 14 going back to 1952. The 2008 presidential election started a new Democratic streak, which has now seen the party carry Virginia three consecutive times, with Hillary Clinton winning it by 5.3 percentage points in 2016.

Obviously, this party flip shows a shift in Virginia’s partisan leaning. But if we dig deeper, the nature of this realignment becomes more complicated. Some parts of the state have drifted inexorably toward the Democrats while others have moved unceasingly in the GOP’s direction. These changes aren’t best shown simply by looking at which party carried which region, city, or county in a given election. Rather, the shifts are better illustrated by comparing voting in localities to the national conditions, i.e. the national popular vote.

Table 1: Virginia’s relative partisan lean compared to national popular vote margin, 1968 to 2016

Source: Dave Leip’s Atlas of U.S. Presidential Elections

To start, just consider Table 1, which displays the national popular vote margin and statewide vote margin in Virginia from 1968 to 2016, covering all presidential elections since the passage of the Voting Rights Act of 1965. The last two columns in the table note the state’s relative partisan lean compared to the national circumstances. For instance, Richard Nixon won the 1968 national popular vote by a little less than one percentage point, but he won Virginia by nearly 11 points. Thus, the state’s relative lean was about 10 points more Republican than the country as a whole. Even when Obama carried the commonwealth in 2008, it was still slightly more Republican than the nation. In both 2008 and 2012, Virginia’s margin was closest to the national popular vote margin, with the latter almost perfectly aligning. But in 2016, Virginia took a step to the left while the United States stepped slightly to the right (at least in the popular vote margin). While Clinton won nationally by about two points — while losing the Electoral College — she won Virginia by a bit more than five, making the state three points more Democratic than the country as a whole. The 2012 and 2016 elections marked the first time since 1948 that Virginia was more Democratic-leaning than Republican-leaning relative to the national popular vote margin.

Changing metropolitan areas

The transformation in Virginia’s statewide partisan lean is the product of some major changes in the voting habits in different parts of the commonwealth. If we take a step down from the statewide level, we can start to really see these shifts by looking at Virginia’s three major metropolitan areas and the parts of the state that lie outside of them.

The three metro areas of note in Virginia are located outside of Washington, DC, around the state capital of Richmond, and in the southeastern corner of the state. Together, these are sometimes called the “Urban Crescent” because of their geographical curvature from north to south. The subsequent analysis is based on the Office of Management and Budget’s (OMB) definitions of the localities that make up these three metropolitan statistical areas (MSAs) as of 2015. “Northern Virginia” consists of the Virginia-based localities within the Washington-Arlington-Alexandria, DC-VA-MD-WV MSA, “Greater Richmond” consists of the localities within the Richmond, VA MSA, and “Hampton Roads” consists of the Virginia-based localities within the Virginia Beach-Norfolk-Newport News, VA-NC MSA. Any locality not included among these three regions is categorized as being a part of the “Rest of Virginia.” It should be noted that Virginia has an unusual local government structure that has resulted in separate status for 38 independent cities from the state’s 95 counties. Whereas in Illinois the city of Chicago is a part of Cook County, in Virginia the city of Fairfax is actually a separate entity from Fairfax County, which surrounds it. Importantly for our purposes, the votes from all 133 cities and counties are recorded separately. Thus, the term “locality” is used throughout this article as a general term covering both cities and counties. While the definitions of MSAs have changed over time — e.g. Culpeper County would not have been considered part of the Washington, DC metro area in 1968 — it’s easier to compare consistent geographical areas across time, so the OMB’s 2015 MSA definitions are used for every election from 1968 to 2016.

Using the same method to calculate relative partisan lean as employed above to compare Virginia’s statewide vote to the nation as a whole, we can see in Chart 1 below how the Urban Crescent and the rest of the state have shifted over time relative to the national popular vote. Reflecting the state’s overall Republican lean, the metro areas tended to be more Republican than not for much of this period. But this wasn’t always the case. Hampton Roads was the lone part of the Urban Crescent to vote more Democratic than the country in three of four elections from 1968 to 1980, though the 1968 election is complicated by George Wallace’s third-party run. Perhaps reflecting Jimmy Carter’s stronger appeal in Appalachia and in Southern rural areas, the relative lean of the Rest of Virginia was only narrowly Republican in 1976 and 1980. Meanwhile, Greater Richmond and Northern Virginia were reliably more Republican than the nation as a whole from 1968 to 2000. To some extent, we can see how the metros and the rest of the state did not move far from one another throughout this period, at least in comparison to recent election cycles. For instance, all four areas were more GOP-leaning than the United States from 1984 to 2000.

Chart 1: The relative partisan lean of the Urban Crescent and rest of the state, 1968-2016

Sources: Author’s calculations based on Virginia Department of Elections data. Metropolitan areas based on the Office of Management and Budget’s 2015 definitions for the Richmond, VA MSA, the Virginia Beach-Norfolk-Newport News, VA-NC MSA, and the Washington-Arlington-Alexandria, DC-VA-MD-WV MSA.

However, 2004 broke that pattern, when Northern Virginia voted five points more Democratic by margin than the country did. That election foreshadowed future shifts, with Northern Virginia moving roughly eight points toward the Democrats compared to 2000 and Greater Richmond moving four points toward the Democrats.

The 2008 cycle featured large shifts in each part of the Urban Crescent. Northern Virginia leapt from a D +5 area in 2004 to about D +11, Greater Richmond went from R +8 to R +1, and Hampton Roads moved from R +4 to about D +5. Only the Rest of Virginia (R +17 to R +18) moved away from the Democrats in what was a very favorable election for them. To some extent, this trend has continued. Whereas Northern Virginia and the Rest of Virginia were at most 11 points apart in relative partisan lean from 1968 to 1996, in 2016 they were 49 points apart: Northern Virginia was around 24 points more Democratic than the country as a whole while the areas outside the Urban Crescent were about 26 points more Republican. While Hampton Roads shifted toward the GOP by one point from 2012 to 2016, it still remains notably more Democratic than not. And Greater Richmond, once the most rock-ribbed Republican part of the state in presidential politics, had almost the exact same lean (D +7) as Hampton Roads in 2016.

Population shifts and the share of the statewide vote made up by each metropolitan area are, of course, another critical feature of Virginia’s changing voting habits. Chart 2 lays out the data regarding the percentage share of the vote for the Urban Crescent and the Rest of Virginia. It demonstrates just how much Northern Virginia has grown in the past 50 years or so, moving from just 21% of the statewide vote (again, based on the 2015 metropolitan area definitions) to 35% in 2016. Meanwhile, Hampton Roads and Greater Richmond have remained relatively flat in their shares of the vote, perhaps reflecting consistent but not explosive population growth like that seen in the DC suburbs and exurbs. As Northern Virginia has gone up, Rest of Virginia has gone down, sliding from 42% in 1968 to 29% in 2016. If one were to create an alternate universe where the vote percentages from 2016 were applied to 1968 vote shares for the three metros and the rest of the state, Trump actually would have carried Virginia by one point, 48%-47%, in 2016.

Chart 2: Share of Virginia’s presidential vote in the Urban Crescent and rest of the state, 1968-2016

Sources: Author’s calculations based on Virginia Department of Elections data. Metropolitan areas based on the Office of Management and Budget’s 2015 definitions for the Richmond, VA MSA, the Virginia Beach-Norfolk-Newport News, VA-NC MSA, and the Washington-Arlington-Alexandria, DC-VA-MD-WV MSA.

The sharp Democratic turn in the Urban Crescent, combined with Norther Virginia’s sizable population growth, have been basic factors in Virginia’s move from fairly reliable Republican territory in presidential elections to its current battleground or even slightly leaning Democratic status in such contests.

Bluing crescent, reddening rural

Digging even deeper, data for the 13 presidential elections from 1968 to 2016 demonstrate the partisan shift and changing influence of Virginia’s localities. Note that this analysis uses today’s existing cities and counties. Old administrative divisions that existed in the period of interest, such as extinct counties or independent cities — Virginia has had a lot of each — are calculated based on where they exist today.

Below is a GIF of cartograms of Virginia, going from 1968 to 2016. Each map contains the relative partisan lean of Virginia localities (based on today’s localities) compared to the national popular vote in each election. So adjust your eyes — the colors do not represent the straight-forward vote results but each locality’s lean relative to the country as a whole in each election. A cartogram is a useful way to examine a geographical area while taking an unseen factor into account, such as population. In this case, the localities are sized by the share of the vote they made up in each presidential election to better show their voting influence. (If the GIF is not playing automatically, please click on the map image to open it separately.)

Map 1: Cartogram of Virginia by relative partisan lean, 1968 to 2016

Sources: Author’s calculations based on Virginia Department of Elections data, map shapefile adjusted using QGIS’s Cartogram plug-in.

Notes: Election results for Manassas and Manassas Park in 1968 and 1972 were unavailable (they were part of Prince William County until 1975). Their sizes are weighted by their population as a part of Prince William County and they are colored by the overall partisan lean of Prince William County for those two elections. The City of Fairfax (in the middle of Fairfax County) is oversized due to the fact that Fairfax County’s judicial complex is located within its city limit, which affected the algorithm that altered the shapefile. To see images of the individual maps, click on the following: 1968, 1972, 1976, 1980, 1984, 1988, 1992, 1996, 2000, 2004, 2008, 2012, 2016.

These maps show the dramatic electoral changes that have taken place in Virginia over the past five decades. Note the growth of Northern Virginia, particularly in the share of the vote in Loudoun and Prince William counties, which are directly west and south, respectively, of Fairfax County. Meanwhile, note the bluing of that region, specifically from 2000 on. A reddening trend occurs in other parts of Virginia, particularly Southwest Virginia, which was once relatively Democratic. Buchanan County, located in Virginia’s coalfields, holds the honor of having the largest difference between its most Democratic- and Republican-leaning years over the 1968 to 2016 period: In 1984, while Ronald Reagan was winning a landslide reelection over Walter Mondale, Buchanan was 39.5 points more Democratic than the country as a whole. In 2016, Buchanan mirrored its neighbor to the north, West Virginia, and moved very sharply toward Trump — it wound up being 62.4 points more Republican than the U.S. That 102-point difference between its maximum leans for each party is far and away the most for any locality in Virginia from 1968 to 2016. Next-closest was neighboring Craig County, also a major mover in the GOP direction in recent years, at 91 points. As for the place that has seen the largest swing toward Democrats, Charlottesville was 64 points more Democratic than the country in 2016, but back in 1968 it was 15 points more Republican, a 79-point shift.

An additional way to look at these shifts is to consider the vote share and relative partisan lean of the 15 biggest localities in the state (by vote share) and the 15 smallest in each cycle in two comparative groups. Table 2 lays out these data for 1968 to 2016.

Table 2: Relative partisan lean and total vote share of the 15 largest and smallest localities in the state (by vote share)

Sources: Author’s calculations based on Virginia Department of Elections data

Interestingly, the 15 smallest localities in the state (again, based on today’s localities) were more Democratic-leaning than the 15 largest localities from 1976 to 1996, compared to the national popular vote. But starting in 2000, the 15 largest localities became more Democratic leaning, and the two groupings have only diverged further since then. Note that the smallest localities in the state of Virginia are not all deeply Republican today: the small city of Lexington is a fairly Democratic college town while Charles City County (confusing name, yes?) is plurality black and very Democratic — it was the only Virginia locality to back George McGovern in 1972. And not all of the 15 biggest localities at present are solidly Democratic. Trump carried Hanover County (14th-largest by vote share in 2016), a ruby-red GOP suburban/exurban county north of Richmond, by 32 points. But on the whole, the largest and smallest localities now reflect the urban-rural divide seen nationally in American politics: In 2016 Virginia’s 15 biggest places were, together, about 20 points more Democratic than the country as a whole, while the 15 smallest localities were about 15 points more Republican. And just as the country is becoming more urban, the share of the Old Dominion’s vote made up by the biggest places is increasing while that of the smallest areas is shrinking.


This article has presented voting data showing the significant shifts in Virginia’s voting behavior and its relative partisan lean over the 1968-2016 period. But it hasn’t exactly explained why this has happened. The answer to that question is partly reflective of the same forces that have changed politics throughout the country. For instance, as Northern Virginia grew rapidly, it attracted large numbers of highly-educated workers to serve in industries related to government, particularly federal contracting. College-educated voters have trended toward the Democratic Party overall, including white college grads. Based on the Census Bureau’s 2015 estimates, Virginia ranks sixth among the 50 states in its percentage of the population 25 years or older that has at least a bachelor’s degree.

Virginia has also become more diverse in many ways. It’s become more racially and ethnically varied since the 1970 census. Race and education are now the two strongest indicators of voting preference, so the fact that Virginia’s population has moved from being 19% nonwhite in 1970 to about 37% nonwhite today is surely a part of the story as well. The fastest-growing localities in the state, such as Loudoun and Prince William counties in Northern Virginia, have become dramatically more diverse since 1970. Loudoun was 13% nonwhite in 1970; today, it is 10 times bigger in overall population and is about 41% nonwhite. Prince William has seen even more dramatic changes: It was about 6% nonwhite in 1970; today, its population is roughly five times bigger (if you subtract Manassas and Manassas Park from its 1970 totals; they’re now independent cities) and the county is 54% nonwhite. The physical origins of Virginia’s population are now more diverse as well. In 1970, 63% of the state’s population had been born in the state; in 2010, that figure had fallen to just below 50%.

These are just some of the factors that have moved Virginia in the Democrats’ direction in national politics. It will be interesting to see where it goes next.

Generic Ballot Model Gives Democrats Early Advantage in Battle for Control of House

Alan I. Abramowitz, Senior Columnist, Sabato's Crystal Ball July 6th, 2017


Results of recent special elections have fueled speculation about whether Democrats have a realistic chance to regain control of the U.S. House of Representatives in the 2018 midterm elections. Although Republican candidates have won recent special elections for seats vacated by President Donald Trump’s Cabinet appointees in Georgia, Kansas, Montana, and South Carolina, the GOP victory margins in all four contests have been much smaller than those for the former Republican incumbents in 2016.

While Democrats have pointed to the substantial vote swing in these four special elections as a sign that a wave election may be coming in 2018, Republicans have emphasized the failure of Democrats to actually capture any of these Republican seats, especially the hotly contested seat vacated by Secretary of Health and Human Services Tom Price in Georgia’s Sixth Congressional District, as an indicator that they should have a good chance to hold onto their majority in the midterm elections.

In reality, the results of a handful of special elections probably provide little information about what to expect next fall. Fortunately, there is a better way of predicting the outcomes of midterm elections based on something called the “generic ballot.” This is a question included in numerous national polls asking voters about which party they prefer in the upcoming House elections. It turns out that the results of this generic ballot test can be used to accurately forecast the House seat swing in midterm elections.

The president’s party almost always loses House seats in midterm elections — this has been true in 16 of 18 midterm elections since World War II. However, these elections have produced a wide range of outcomes, from a gain of eight seats in 2002 to a loss of 63 seats in 2010. So what explains the variability in midterm results and what can we expect in 2018?

In addition to the results of the generic ballot test in early September of the election year, two other predictors go a long way toward explaining seat swing in midterm elections: which party controls the White House and the number of seats held by each party prior to the election. The president’s party matters because, regardless of whatever else is going on, voters tend to turn against the party in the White House in midterm elections. And the number of seats each party holds prior to the election matters because, all else being equal, the more seats the president’s party has to defend, the more seats it is likely to lose.

Table 1: Regression analysis of seat change in midterm House elections, 1946-2014

Source: Data compiled by author.

Table 1 displays the results of a regression analysis of Republican seat swing in midterm House elections between 1946 and 2014. All three predictors are highly statistically significant and the model explains about 90% of the variation in seat swing across these 18 elections. The coefficient of -.492 for previous Republican seats reflects the impact of exposure to risk — for every additional seat Republicans hold going into a midterm election, they can expect to lose close to half of a seat.

The coefficient of -13.7 for the midterm variable means that after controlling for the other predictors in the model, having a Republican in the White House costs the GOP an average of close to 14 seats. This coefficient also means that having a Democrat in the White House would be expected to result in a Republican gain of almost 14 seats in a midterm election.

Finally, the coefficient of 1.795 for the generic ballot variable means that for every additional one point advantage Democrats enjoy on the generic ballot in the Gallup Poll in early September of the midterm year, Republicans can expect to lose about 1.8 additional seats in the House.

Figure 1: Scatterplot of change in Republican House seats by predicted change in Republican House seats

Source: Data compiled by author.

Figure 1 displays a scatterplot of the relationship between predicted and actual seat swing in House midterm elections based on the generic ballot model. The results show that the model does an excellent job of predicting House seat swing. Most of the points are very close to the prediction line. However, a few elections, especially the highly unusual post-September 11 midterm of 2002, are farther off the mark. It’s worth noting that adding presidential approval or economic indicators to the model has no effect once you include the generic ballot. Whatever effect these have seems to be captured by the generic ballot.

It is especially encouraging to note that the model’s forecast for the 2014 midterm election, a Republican gain of 18 seats, was very close to the actual GOP pickup of 13 seats. This is significant because the 2014 midterm election occurred after the post-2010 round of redistricting in which Republicans were able to redraw House district lines in many states. Despite this, the model actually slightly overestimated Republican seat gains in 2014. There is little reason to believe, therefore, that the model’s predictions for 2018 will be thrown off by the effects of post-2010 redistricting.

Based on the results in Table 1, we can calculate conditional forecasts of seat swing in the 2018 midterm election, depending on the results of the generic ballot test in early September of next year. The other two predictors, presidential party and Republican House seats, are already set, though additional House special elections could subtly alter the House’s party makeup before November 2018 (for instance, UT-3 will have a special election to replace Republican ex-Rep. Jason Chaffetz in November 2017, though that seat is ruby red). The conditional forecasts are displayed in Table 2.

Table 2: Predicted change in Republican House seats by generic ballot polling

Source: Data compiled by author.

The predictions in Table 2 show that Democrats will need a lead of at least five points on the generic ballot in early September of 2018 in order to gain the 24 seats that they need to take control of the House. Based on the results of recent national polls, that number appears well within reach. On average, based on calculations from FiveThirtyEight, Democrats hold an adjusted lead of close to seven points on the generic ballot, mirroring that of the RealClearPolitics average. A lead of that magnitude would result in a predicted Democratic gain of close to 30 House seats, more than enough to regain control of the chamber. Given the model’s standard error of 11.6 seats, that forecast would give Democrats about a two-thirds chance of regaining control of the House.

So keep an eye on the generic ballot polling for 2018. If Democrats maintain a lead in the high single digits, that probably indicates they will have a decent chance to win the House or at least significantly cut into the Republicans’ majority. A bigger Democratic lead, into the double digits, would make a takeover more likely, while a smaller Democratic lead — or a GOP advantage — would put Republicans in a clearer position to preserve their majority.