Sabatos Crystal Ball

Join us for the American Democracy Conference in Charlottesville

Conference will be livestreamed next Thursday

UVA Center for Politics January 17th, 2019


Next Thursday, Jan. 24, the University of Virginia Center for Politics will host its 20th annual American Democracy Conference in Charlottesville.

The event will be held at Alumni Hall, 211 Emmet St. South, on the Grounds of the University of Virginia. It is free and open to the public with advanced registration, and members of the media are welcome to attend as well. We also will be livestreaming the conference at:

The schedule is as follows:

8:30 a.m. — Doors open

9 a.m. — Keynote address by former Gov. Terry McAuliffe (D-VA)

10 a.m. — Panel discussion on Donald Trump at the midterm of his presidency

Moderator: Christine Mahoney, Professor of Public Policy and Politics at UVA’s Frank Batten School of Leadership and Public Policy


Paul Begala, Democratic strategist and CNN political analyst

Jamelle Bouie, New York Times opinion columnist and CBS News analyst

Kate Obenshain, former Republican Party of Virginia chairwoman and Fox News and CNBC analyst

David Ramadan, former Republican member of the Virginia House of Delegates

11:30 a.m. — Lunch provided by Center for Politics

Noon — Keynote address by Karl Rove, former deputy chief of staff to President George W. Bush

This is the second part of the Center’s 20th anniversary edition of the annual American Democracy Conference. The first part was held last November in Washington D.C. and featured remarks from Sen. Ted Cruz (R-TX), House Intelligence Committee Chairman Adam Schiff (D, CA-28), and expert panels reacting to the midterm and looking ahead to 2020. A recording of that conference is available here.

The conference is being held in partnership with the University of Virginia’s Frank Batten School of Leadership and Public Policy and is part of “Democracy in Perilous Times: Unprecedented Challenges and Controversies,” an ongoing program series organized by the Center for Politics and the Batten School for the 2018-2019 academic year as part of the National Symposium Series. Information on future programming can be found here.

House 2020: Our initial ratings

Democratic overperformance in 2018 gives the majority party breathing room to start, but a GOP presidential win could put the House in play

Kyle Kondik, Managing Editor, Sabato's Crystal Ball January 17th, 2019



— Democrats start the cycle favored to hold the House majority, but a GOP presidential victory would open the door to Republicans restoring total control of Washington.

— Overall, roughly equal numbers of Democratic (47) and Republican (46) districts begin the cycle listed in our ratings in the Toss-up, Leans, and Likely categories. That means that more than three of every four House districts (79%) begin the cycle as rated Safe for the incumbent party.

— Presidential trends in these districts — both where they’ve been and where they may be going — were a major factor in the ratings. Consequently, some of the closest races in 2018 are not necessarily rated in very competitive categories to start this cycle.

Table 1: Initial 2020 Crystal Ball House ratings

Our first look at the House

The battle for the House in 2020 is at such an early point that, as of this writing, we don’t even really know how many seats the Republicans will actually need to flip the House next year. That’s because there’s a vacant seat, NC-9, that Republican Mark Harris appeared to narrowly win on Election Night, but credible allegations of fraud seem likely to eventually lead to a new election. So the House right now features a 235-199 Democratic majority, meaning that even if Republicans eventually do win NC-9, they will need to net at least 18 seats to re-take the majority next year.

That said, we are already starting to see activity in the race for the House. For instance, narrow 2018 loser Yvette Herrell (R) is already indicating she will seek a rematch with Rep. Xochitl Torres Small (D, NM-2) in a GOP-leaning swing district in southern New Mexico. Rep. Steve King (R, IA-4), whose rhetorical aid and comfort to white nationalists endangered him in his 2018 race despite his dark red district, already faces a couple of primary challengers and was just removed from his committee assignments by GOP House leadership. Two Republicans, Reps. Walter Jones (R, NC-3) and Rob Bishop (R, UT-1), have already suggested they won’t be running again in 2020; two others, Reps. Duncan Hunter (R, CA-50) and Chris Collins (R, NY-27), won under the cloud of indictments last November and may very well not make it through this Congress. All of these members represent Safe Republican seats on paper, although Hunter and especially Collins faced very difficult races last year.

The point is that the churn in the House never really ends, nor does the campaigning.

As we assess the House playing field for the first time, we think Democrats start as favorites to hold the House majority. There are a lot of little reasons that go into this assessment, but there’s one major one that really undergirds that analysis. Here’s goes:

Even if President Donald Trump is renominated and reelected to a second term in the White House, it is not at all clear from history that his reelection would provide sufficient coattails to net the minimum needed gain of 18 Republican House seats. Indeed, the last five reelected presidents saw their parties fall short of that net total in their election years, and the House hasn’t switched from one party to the other and then back again in consecutive elections since 1952-1954 (please refer back to our piece from November for more of this history). If Trump (or another Republican) does not win the presidency, it’s almost impossible to imagine the Republicans simultaneously losing the White House but gaining the House majority.

Right now, we don’t see either side as clearly favored to win the White House, meaning that we’re looking at the race for the presidency as something of a Toss-up to start. And if winning the White House is a necessary but not sufficient requirement for Republicans to win the House — which we basically think it is — then the GOP’s odds of winning the House are, by definition, less than 50%, because holding the presidency alone probably is not enough for them to win the House.

Table 1 above shows our initial House ratings.

Unsurprisingly, the most vulnerable seats on both sides are ones that generally were very close in 2018, will likely vote for the other party’s presidential nominee in 2020, or both. The 10 Democratic Toss-ups are all districts where first-term Democrats will presumably be seeking second terms in districts Trump won in 2016 by at least 3.5 points (and, in some cases, by double digits). The GOP Toss-up column includes the vacant NC-9, two of three remaining Hillary Clinton-won districts held by Republicans (PA-1 and TX-23, both of which were decided by close margins in 2018), and two Trump-won suburban districts (GA-7 and NE-2) that were also very close in 2018 and that could break against the president in 2020.

We are starting with more Democratic seats (27) than Republican ones (20) in the very competitive Toss-up and Leans categories, meaning that Democrats start the cycle with clearly but not decisively more vulnerable seats than the Republicans. That’s in large part because Democrats now control 31 Trump-won seats while Republicans only control three Clinton-won seats. From a Republican perspective, one can easily see how to win the House back: pick up 18 seats from those 27 most vulnerable Democratic-held seats, 21 of which voted for Trump in 2016 and very well could again. From a Democratic perspective, one can see avenues for supplementing their new majority, particularly if they are also winning the White House. Of the 15 GOP-held seats in the Leans Republican category, every single one was decided by a half-dozen points or less in 2018.

Still, just because a race was very close in 2018 does not mean it will be very close in 2020. For instance, we don’t list any of the Democrats’ newly-won California seats among the Toss-ups, even though several were decided only by a few points. The reason is that Clinton carried all of the new Democratic seats in 2016, and it’s reasonable to expect the Democratic presidential nominee to once again carry them in 2020, which probably would deprive GOP challengers of the oxygen they need to beat Democratic incumbents. Remember, Republicans had battle-tested incumbents in many of these seats in 2018, like former Reps. Jeff Denham (R, CA-10), Steve Knight (R, CA-25), and David Valadao (R, CA-21). Now Democrats will have the power of incumbency in those districts and, in all likelihood, the presidential wind at their backs. The opposite is true in two Minnesota seats Republicans won from Democrats in 2018, MN-1 and MN-8. Both voted for Trump by double digits, and if that repeats itself, it’s hard to see how Democrats can win those districts back, especially now that those districts have GOP incumbents (both were open seats in 2018).

We’re not going to go through our rationale for each of the ratings; there will be plenty of time to explore all of these districts in depth as the cycle goes along. But the 2016 presidential results, and our very early, subject-to-change expectations for 2020’s results, loomed large in formulating these initial ratings.

Generally speaking, the ratings reflect both where a district has been at the presidential level and where it might be going in 2020. Our best guess is that districts that are diverse and/or have higher-than-average levels of four-year college attainment may either not change much or continue to get worse for the president and Republicans over the next two years. Conversely, districts that are less diverse and/or have lower-than-average four-year college attainment may not change much or might get better for Trump and the Republicans. That’s been the trend in American politics even pre-dating Trump — though the president hyper-charged that trend — and we thought the 2018 results generally reinforced those trends (for more, check out recent Crystal Ball pieces by Alan Abramowitz and Noah Rudnick). So if a rating seems off to you, think about this basic framework, and the rating might make more sense (even if you still disagree with it).

The closing months of the 2018 midterm cycle were dominated by a consensus belief among major election forecasters (including us) that the Democrats were favored to win the House majority and Republicans were favored to hold the Senate majority. As we start the 2020 cycle, that same forecast — Democrats favored in the House, Republicans in the Senate — seems like the best initial bet, which is reflected in our ratings. How the presidential race goes, and whether one side opens up a clear advantage, is probably the most important factor in whether either chamber drifts into true Toss-up territory.

The Wandering Voters

How 2018’s gubernatorial results reflected 2016’s presidential trends

Noah Rudnick, Guest Columnist January 10th, 2019



— 2018 governors’ races continued along the same realignment patterns that emerged in the 2016 presidential election.

— An analysis of protest third-party votes for president in 2016 indicates those voters were likelier to pick the Democrats’ side in the 2018 election.

2016’s changes endured in 2018 gubernatorial races

In 2018, Democrats won over suburban voters to cruise to a roughly 8.5-point win in the national House vote, even as the rural areas remained red. In the following writeup, I look at where each party made gains this past cycle. I also look at how third-party performance in 2016 acted as a leading indicator of where there was a group of lost voters that would pick a side, in this case for the Democrats and against the party in the White House.

The lean of a county is defined as the statewide margin subtracted from the margin in the county. For example, if the Republican won a county by 5% in a state the same candidate won by 2%, then the lean would be 3%. If four years later in 2018, the Republican won the county by 8% while winning by 1% statewide, then the county’s lean is now 7%, and the difference in the leans is now 4%. Figure 1 below shows the change in the lean from 2014 to 2018 in gubernatorial races, with red becoming more Republican and blue becoming more Democratic relative to the overall statewide margin (the states that did not have gubernatorial races last year are not included).

Figure 1. Change in lean by county from 2014-2018

Note: Click on any of the images below to see larger, easier-to-read versions.

Those familiar with a county map will see that for the most part, the trends from the 2016 presidential election continued. The suburbs and cities became bluer as the rural counties continued lurching rightward. Figure 2 confirms this by showing a direct relationship that where the direction and magnitude of the margin change from the 2012 to 2016 presidential contests went, the lean change in the governor’s elections from 2014 to 2018 followed.

Figure 2. Gubernatorial lean change vs. presidential margin change by county

So, one way to figure out what happened in 2018 was to use a continuation of the trends in the margins, but there’s an even more revealing way to look at the new set of Democrats in the midterms. My personal theory is that there are now ex-Republicans, especially those from the suburbs, that on balance voted for Democrats in 2018 and make up the new swing voters. But these voters didn’t vote for Hillary Clinton, and that’s not why the margin swung from 2012 to 2016. These voters backed third-party candidates in 2016, opting not to vote for either of the two unpopular major-party candidates.

In order to test this theory, the first thing I did was devise a measure to compare minor-party performance levels across the years. I collected the third-party performance for every county in the United States in presidential elections from 1980-2016, with the invaluable help of Dave Leip’s election atlas.  I then used a relative measure in order to compare the performance in one year not only to all other counties that election, but also within that county for all years going back to 1980. For example, if third parties in a county received 4.5% of the vote and 3% nationwide, then the performance here would be ((.045-.03)/.03) =50%. To compare over the years, I took the median since 1980 and compared each year to the median in that county. Figure 3 below shows the map of how third parties as a whole performed in 2016 relative to their historical baseline, with blue being a higher performance than usual and orange being a lower performance.

Figure 3. 2016 Relative third party presidential performance

One can see some obvious patterns on this map. For instance, much of New Mexico is shaded dark blue, indicative of Libertarian nominee Gary Johnson’s relative strength as a third-party option in his home state, where he once served as governor.

The next map in Figure 4 is a combination of Figure 3 and the change in the presidential margin from 2012 to 2016. Dark red is where third-party performance was down relative to the historical baseline and the margin shifted towards Republicans, dark blue is the same but a shift towards Democrats. The light red, on the other hand, was where third-party performance was higher than usual but the margin still shifted rightward. Light blue is a high third-party share and shift towards Democrats.

Figure 4. 2016 third party and margin combined status

Figure 5 uses these categories and plots out the relationship with the measure from Figure 1, which charted out the change in the lean of a county from 2014 to 2018. Going up, the county got more Republican, but heading downward, a county became more Democratic at the gubernatorial level.

Figure 5. 2018 shift in county lean by third party/margin status

First, you can see that the marginal change by who gained had the largest impact, but third parties mattered here as well. In the category of those that moved towards Republicans in 2016 and where third parties underperformed, the lean moved 12% towards Republican as the median. But where third parties did better than their historical average, they moved only 6% more Republican, a sizable and significant decrease of several points. For the counties that moved Democratic from 2012 to 2016, counties where third parties were down only moved 1% more Democratic. Among counties where third parties were higher, the shift was only 3% more Democratic, a little higher but not significantly. The magnitude, or how much they shifted, matters as well.

Figure 6 looks at this measure by first creating an adjusted measure that centers all the states to compare them by setting the median county third-party performance to zero. I then plotted those points on the x-axis and the shift in the lean from 2014 to 2018 governor races on the y-axis.

Figure 6. 2018 governor lean shift vs. adjusted relative performance

The inverse relationship here shows that as relative third-party performance increased, or as more people than normal voted third party, there was more movement to the Democrats. It also means that counties with a lower-than-normal third-party vote, the vast majority of which were rural areas, continued moving more Republican.

What does all of this tell us? In summary, that the 2016 election set in motion a split that broke heavily from the previous midterms, especially on rural/suburban lines. But we can also see that beyond a one-to-one equivalency in the margins, and especially in areas where Republicans lost ground, voters who went to a third-party candidate in 2016 seemed to break harder for Democrats in 2018, giving that party an edge. This affects 2020, as this new swing bloc could decide the election if it again breaks for the Democrats after opting out of the major-party choice in 2016 — or it could opt-out again or even drift back to the Republicans. It also means that the campaigns will have to adjust to new rural or suburban voters in their coalitions and how best to appeal to those that have strayed into the gap between the parties.

Noah Rudnick is currently a student at the Ohio State University and a contributor to Decision Desk HQ. Follow him on Twitter @rudnicknoah.