Split Difference: How many ticket splitters will it take to flip the Senate?

Flip The Senate

Fourteen races will determine whether Democrats can flip the Senate or whether Republicans keep their majority in 2018. Thirteen of them are in states won by Donald Trump in 2016. Ten of these Trump-state races have Democratic incumbents. And every one of them will be won or lost based on the success each side has in turning out its base and in persuading those with less of an allegiance to either party.

This report uses new voter file data to identify a crucial constituency in these Senate races, a universe of voters called “Splitters” – voters who have exhibited a tendency to vote for candidates of both parties. A broad but essential lesson from this analysis is that there is wide variation by state in how many Splitters are likely to vote in 2018, and this will shape Democrats’ winning coalitions. For these key Senate races, we have ranked each state’s voters on their share of Splitters – from Super Splitters to Seldom Splitters.

Among our findings:

  • In every Senate battleground race, more reliably Republican voters are likely to turn out than reliable Democrats in 2018. On average across these races, reliable Democrats are expected to make up just 27% of voters.
  • The expected share of Splitters varies widely by state – from Super Splitter West Virginia where it’s more than half of the 2018 electorate to Seldom Splitter Michigan where this share is just nine percent.
  • Splitters in the Senate battlegrounds tend to be older, less educated, and less diverse when compared to the overall likely 2018 electorate, reliably Democratic voters, and those with less predictable voting behavior.

There are some people who on the Tuesday after the first Monday in November vote regardless of the contest. They don’t need to be called by a campaign or get a mailer. Their behavior is set. If you are reading this paper, you may be one of them. There are others who are just as certain to vote, but the uncertainty lies in for whom they’ll vote. They are known as ticket-splitters – voters who have exhibited a tendency to vote for a candidate of one party for one office and a candidate of another party for another office. Finally, there is a small number of likely voters that we don’t know enough about. They may be young or have a limited record of turning out, but how they are likely to vote is up in the air.

Based on data from the progressive data firm Catalist, we have an idea of the voters likely to turn out in this year’s Senate battleground races. The most glaring takeaway is that Base Democrats—reliably Democratic voters—are expected to make up on average just 27% of voters across these marquee Senate races. By contrast, Base Republicans are expected to make up 40% of the electorate across these races. In each of the states, more Base Republicans are likely to vote than Base Democrats. Second, Splitters—those voters who have exhibited a tendency to vote for candidates of both parties—are expected to make up on average 20% of voters. An estimated 13% of voters fall into the unknown category.

This multidimensional electorate makes amassing a winning coalition a complex task. But the stakes are high. Last year, few observers gave Democrats a chance at winning back the Senate in 2018. The Party was just three seats shy of the majority, but the map looked too tough. Ten Democrats are up for re-election in Trump states—states he won by an average of 15 points—and just a few Republicans are vulnerable.1 But Trump foundered out of the gate, Republicans tried and failed to repeal the Affordable Care Act, and then Democrats started to win. First in state legislative races, then in Virginia and New Jersey, and later in Alabama where Doug Jones cut the Senate margin to two seats.2 Today, Democrats have a narrow, but viable, path to winning back the chamber this year—one that runs through 14 battleground states (note: this report excludes North Dakota from its analysis because it does not have voter registration).

Base Democrats—reliably Democratic voters—are expected to make up on average just 27% of voters across the marquee Senate races in 2018.

What is a certainty is that for Democrats to chart this path, they must win across a diverse range of state electorates. This report focuses on one key difference between these electorates, the share of Splitters, and classifies them by this measure. The likely impact of Splitters varies widely by state, from West Virginia where these voters are expected to make up more than half of the electorate to Michigan where they may make up less than 10%. And while this report analyzes the decisive role of Splitters, it is also important to note that these voters are only determinative because of Democrats’ diverse base that makes up the bulk of winning coalitions. The bottom line is to attain a majority Democrats must win 12 of the 14 battleground Senate races. To do this, they must take an electoral path broad enough to carry an army of Splitters while also appealing to those base and less known voters.  

Methodology

This report uses voters who are likely to vote in 2018 for its analysis. By studying actual voters likely to participate in the election, rather than Census data that measures the broader population, we can better understand the likely impact of specific voter groups on this November’s election. This data is accessed through the voter file maintained by Catalist, a leading progressive data and analytics firm.

Looking at the likely 2018 electorate, this report focuses on Splitter voters who have a history of voting for candidates from more than one party. To isolate this group of voters, we leveraged two of Catalist’s models—a statistical tool for estimating voters’ characteristics or behavior. The first is a 2018 vote propensity model, which estimates the likelihood that an individual will vote this year. The second is a ticket-splitter model, which estimates the likelihood that a voter has a tendency to vote for a candidate of one party for a certain office and a candidate of another party for another office.

By combining the vote propensity and ticket-splitter models, we isolated likely 2018 voters who are more prone to split their ticket for each of the 2018 Senate battleground states. Sorting by share of Splitters out of the likely 2018 electorate in each state, these battlegrounds can be divided into one of four categories:

  • Super Splitter States: The Splitter vote is determinative here. Splitters are expected to be a majority of the vote in 2018, and it is a category of one, the Mountaineer state of West Virginia.
  • Strong Splitter States: Splitters determine elections in these states. They are expected to make up, on average, one-in-four voters in 2018.
  • Sometimes Splitter States: Splitters can turn elections here. On average, an estimated 14% of voters will be Splitters here in 2018.
  • Seldom Splitter States: Races in these states are less likely to depend on Splitters. They are expected to make up 9% of voters in these states in 2018.

It is important to note, data on Splitters comes from a national predictive model that is then applied to individual state voter files. In short, this data should be considered estimates; while models are rigorously tested, not all Splitters are swing voters and not all swing voters have been identified here as Splitters.

Finally, demographic profiles in subsequent sections include data on race, age, and educational attainment. This data is based on a combination of voter file records and predictive modeling from Catalist.

The Battleground States

The geographic scope of this analysis covers 13 states that will decide control of the Senate in 2018. As mentioned, this report excludes North Dakota because it does not have voter registration, which limits opportunities for data analysis. While all 13 of the Senate battleground states can be expected to be competitive, Democrats’ approach to winning them will (and should) be different. This analysis zeroes in on a key difference between states: just how large Splitters’ likely vote share will be this year. Across all 13 states, 20% of likely 2018 voters are Splitters. But this share varies by state, from more than 50% to just nine percent. States are categorized here, as seen in the map below, by the likely share of Splitters in this year’s election.

2018 Battleground Senate Races by Splitters

Image Alt Text

In addition to the share of Splitters, it is essential to understand the differing partisan contexts in these states. In the chart below, the red vertical bars indicate the margin between Base Democrats’ likely 2018 vote share and that of Base Republicans. And the purple bars illustrate the expected share of Splitters by state. As this chart indicates, Base Republicans are expected to outnumber Base Democrats in every state. This should not be interpreted as each of these states having an overall Republican lean; instead, it means they have larger bases of reliably Republican voters than Democrats.

2018 Battleground Senate Races

Looking at Splitters from a demographic lens reveals that they do not share the same characteristics as the overall likely 2018 electorate. Looking across the Senate battleground states, Splitters are 93% white, 16% are under the age of 40, and 36% are projected to have graduated college. By comparison, looking at all likely 2018 voters nationwide, 82% are white, 19% are under 40, and 54% are likely college graduates.

Both how many Splitters are likely to vote and Base Republicans’ likely vote share advantage will inform how Democrats work toward winning coalitions. In a state like Ohio, where the data indicates that Base Republicans will have a solid advantage and Splitters will be 26% of the electorate in 2018, winning Splitters will be essential to Democrats’ success. By contrast, in a state like Texas where the expectation is that Republicans will have a large base turnout advantage and there will be few Splitters, Democrats may be best served if they first focus on improving their base mobilization.

Super Splitter States

In Super Splitter States, Splitters are expected to make up more than half of state electorates. In 2018, only West Virginia, where Democrat Joe Manchin is running for re-election against Republican Patrick Morrissey, falls into the category.

2018 Super Splitter States and Expected Share of Splitters

Image Alt Text

This analysis estimates that Splitters will make up 56% of the 2018 vote share in West Virginia. This share of Splitters is more than double any other Senate battleground state this year, a margin that is attributable to West Virginia’s unique partisan culture. Democrats have a 10-point advantage over Republicans in voter registration in the state, but at the same time, voters here backed Trump by 42 points in 2016.3

The chart below compares Splitters likely to vote in West Virginia to all likely voters nationwide this year. Despite there being more registered Democrats than Republicans in West Virginia, Base Republicans—those likely to identify as Republicans and not split their ticket—are expected to have a 13-point vote share advantage over Base Democrats in 2018. Splitters here are 98% white, 16% are under 40, and 21% are projected to have graduated college. This makes Splitters in West Virginia far whiter and less educated, and marginally older than likely voters nationwide in 2018.

Image Alt Text

Given Base Republicans’ significant edge in expected vote share in West Virginia, the huge population of Splitters here is the reason why Democrats can still win in the state. This helps explain how Manchin won his last race, in 2012, when he took 61% of the vote on the same day that Barack Obama won just 36%.4 According to Catalist data, only 15% of 2018 likely voters are Base Democrats—voters who likely identify as Democrats and are unlikely to split their ticket.

Across all 13 battleground states, 20% of likely 2018 voters are Splitters. But this share varies by state, from more than 50% to just nine percent. 

Strong Splitter States

In Strong Splitter States, Splitters are expected to make up, on average, one-in-four voters in 2018. Five states fall into this category: Arizona, Florida, Nevada, Ohio, and Pennsylvania. Democrats are up for re-election in Florida (Bill Nelson), Ohio (Sherrod Brown), and Pennsylvania (Bob Casey), while the Arizona seat is open, and Republican Dean Heller is running for re-election in Nevada.5

2018 Strong Splitter States and Expected Share of Splitters

Image Alt Text

This analysis estimates that Splitters will make up an average of 24% of the vote in these states in 2018. Such a large bloc of Splitter voters can easily determine the outcome of races. For instance, in the 2012 Nevada Senate race between Heller and Democrat Shelley Berkley, exit polls show that Berkley won just 81% of voters who supported Obama that year.6 These ticket-splitting voters were a key reason why Berkley lost the election by one point, 45% to 46%.7

On average, Base Republicans’ vote share in Strong Splitter States is expected to be nine points greater than Base Democrats in 2018. This is the narrowest margin among the four categories covered in this analysis, but far larger than Base Republicans’ two-point expected margin when looking at the 2018 national electorate. Splitters here are 89% white, 16% are under 40, and 46% are projected to have graduated college. This makes Splitters in these states just marginally whiter, older, and less educated than likely voters nationwide in 2018.

Image Alt Text

When one-in-four voters are Splitters, which is the case in this category, these voters can be the decisive factor in a race. In 2016, Trump won four of the five states in this category, the exception being Nevada, and so the fact that each of these states is competitive again in 2018 is a testament to the impact of Splitters. By the numbers, 32% of likely voters across these states are expected to be Base Democrats, but the Party still needs to win over large numbers of Splitters to retain or flip these Senate seats in 2018.

Sometimes Splitter States

In Sometimes Splitter States, Splitters are expected to make up about one in seven voters in 2018. Five states fall into this category: Indiana, Missouri, Montana, Tennessee, and Wisconsin. Democrats are up for re-election in Indiana (Joe Donnelly), Missouri (Claire McCaskill), Montana (Jon Tester), and Wisconsin (Tammy Baldwin), while Tennessee is a Republican-held open seat.8

2018 Sometimes Splitter States and Expected Share of Splitters

Image Alt Text

This analysis estimates that Splitters will make up an average of 14% of the vote in Sometimes Splitter States in 2018. While this expected vote share is not as substantial as the previous two categories, it is still more than enough to tip a close election. In the 2012 Indiana Senate race between Donnelly and Republican Richard Mourdock, exit polls show that Donnelly won 15% of voters who supported Mitt Romney that year.9 These ticket-splitting voters helped Donnelly win in conservative Indiana by a 50% to 44% margin.10

On average, Base Republicans’ vote share in these states is expected to be 16 points greater than Base Democrats in 2018. But more so than any other category, this margin varies by state; Base Republicans’ edge is expected to be four points in Wisconsin but 35 points in Tennessee. And again, this does not mean Republicans will have a 16-point advantage in the overall makeup of the electorate, but rather that there is likely to be more solid Republicans at the polls than Democrats. Splitters here are 98% white, 19% are under 40, and 29% are likely college graduates. This makes Splitters in these states whiter and less educated than likely voters nationwide in 2018. But by share of voters under 40, Splitters here mirror the likely national electorate.

Image Alt Text

Fourteen percent of voters in these states are expected to be Splitters in 2018. By comparison, Base Democrats are expected to make up an average of 24% of the vote across these states. And so while the share of Splitters is smaller here than in the previous two categories, Democrats must bring them into their coalition to build toward wins in these places. In 2016, Trump won all five of these states, including Tennessee by a margin of 26 points.11 The fact that Democrats are in contention just two years later, across all the states in this category, is an indication that Splitters here could be turning back toward Democrats.

Seldom Splitter States

In Seldom Splitter States, data shows that Splitters are expected to make up, on average, just nine percent of voters in 2018. Just as in the other three categories, this nine percent figure should not be interpreted as representing all persuadable voters; instead, it is voters who are particularly likely to split their ticket. Just two of the 13 Senate battleground states in this report fall into this category, and one is a Democratic-held seat and the other a Republican seat: Debbie Stabenow is running for re-election in Michigan and Ted Cruz in Texas.12

2018 Seldom Splitter States and Expected Share of Splitters

Image Alt Text

While the other three categories are defined by how impactful Splitters can be, Seldom Splitter States are places where the base may have more real and potential impact on elections. For example, in Stabenow’s 2012 re-election bid in Michigan, exit polls show that she won 95% of Obama voters on her way to a 20-point victory.13 And in Texas, just over one in three voting-age Hispanic citizens voted the last time Cruz was on the ballot in 2012.14 In 2018, engaging and turning out this base Democratic constituency will be key to defeating Cruz on Election Day.

While the other three categories are defined by how impactful Splitters can be, Seldom Splitter States are places where the base may have more real and potential impact on elections.

On average, Base Republicans’ vote share in these states is expected to be 12 points greater than Base Democrats in 2018. Splitters here are 89% white, 13% are under 40, and 37% are projected to have graduated college. This makes Splitters in these states whiter, older, and less educated when compared to likely voters nationwide in 2018.

Image Alt Text

This category is the exception to the rule when it comes to the impact of Splitters on Senate battleground races in 2018. Among likely 2018 voters with available data, just nine percent of voters in these states are projected to be Splitters. By comparison, Base Democrats are expected to make up on average 31% of voters. Splitters here are likely to make up just a sliver of the 2018 electorate, but sometimes that is enough to decide an election; in 2016, Trump won Michigan and its 16 electoral college votes by just 11,612 votes.15

Conclusion

Remarkably, Democrats have a narrow, but viable, path to winning back the Senate this year. While this breakthrough is still an outside chance, the Party has a real shot at not losing any net seats in the chamber this year. Both of these potential outcomes were close to inconceivable 18 months ago.

Demographically, the crucial Splitter voters—those who have shown a willingness to split their ticket—are whiter, older, and have lower educational attainment than all likely 2018 voters. Most of these voters fall outside of the Rising American Electorate framework, which typically includes voters of color, young people, and single women, but they are nonetheless vital for Democrats in 2018 and moving forward. But it is also important to recognize that if Democrats benefit from a turnout surge this November it will be in large part because of the Party’s diverse base and less frequent voters.

If Democrats benefit from a turnout surge this November, it will be in large part because of the Party’s diverse base and less frequent voters.

The essential data point underlying this analysis is that on average just 27% of likely voters in the 2018 Senate battleground states are projected to be Base Democrats—voters likely to identify as Democrats and not split their ticket. Down the stretch to Election Day, Democrats can benefit from a strategic approach that expands on this share of Base Democrats to include Splitters and other persuadable voters.

This analysis also looked within likely state electorates to categorize states by expected share of Splitters, and the findings showed that Democrats face an electorally diverse map in 2018. This is especially true in the Senate races profiled here where Democrats are running in purple and red states. In Splitter-heavy states, such as West Virginia or Nevada, running to win over persuadable voters may be the single most important thing a campaign can do. Conversely, base-first states like Michigan might lean more toward campaigns that focus on coalescing support with solid Democrats. To enjoy broad-based gains, Democrats should chart a broad path forward, such that candidates have the freedom to run their races in a way that fits their states’ voters.

Appendix

The below chart provides state-level data for the expected share of Base Democrats, Base Republicans, and Splitters in the 2018 electorate. Data accessed through Catalist.

Image Alt Text

Topics
  • Elections49

Endnotes

  1. “2016 National Popular Vote Tracker.” Cook Political Report, https://docs.google.com/spreadsheets/d/133Eb4qQmOxNvtesw2hdVns073R68EZx4SfCnP4IGQf8/htmlview?sle=true#gid=19. Accessed June 19, 2018

  2. “Daily Kos Elections post-Nov. 8 2016 D-vs.-R special elections vs. presidential results.” DailyKos, https://docs.google.com/spreadsheets/d/1C2MVeM2K7WgqmJw5RCQbWyTo2u73CX1pI8zw_G-7BJo/edit#gid=546409300. Accessed July 30, 2018.

  3. “Voter Registration Totals as of June 30, 2018.” West Virginia Secretary of State, https://sos.wv.gov/elections/VoterRegistration/Pages/Voter_Registration.aspx. Accessed July 24, 2018; “2016 National Popular Vote Tracker.” See also, Cook Political Report, https://docs.google.com/spreadsheets/d/133Eb4qQmOxNvtesw2hdVns073R68EZx4SfCnP4IGQf8/htmlview?sle=true#gid=19. Accessed June 19, 2018.

  4. “Election Results Center.” West Virginia Secretary of State, http://services.sos.wv.gov/apps/elections/results/Default.aspx?year=2012&eid=13. Accessed July 25, 2018.

  5. “Battle for the Senate 2018.” Real Clear Politics, https://www.realclearpolitics.com/epolls/2018/senate/2018_elections_senate_map.html. Accessed July 25, 2018.

  6. “2012 Nevada Exit Polls.” CNN, http://www.cnn.com/election/2012/results/state/NV/senate/. Accessed July 25, 2018.

  7. “2012 General Election - Federal / State.” Nevada Secretary of State, https://www.nvsos.gov/silverstate2012gen/. Accessed July 26, 2018.

  8. “Battle for the Senate 2018.” Real Clear Politics, https://www.realclearpolitics.com/epolls/2018/senate/2018_elections_senate_map.html. Accessed July 25, 2018.

  9. “2012 Nevada Exit Polls.” CNN, http://www.cnn.com/election/2012/results/state/NV/senate/. Accessed July 25, 2018

  10. “2012 Indiana Exit Polls.” CNN, http://www.cnn.com/election/2012/results/state/IN/senate/. Accessed July 25, 2018; See also, “November 6, 2012 General Election.” Indiana Secretary of State, https://www.in.gov/sos/elections/2400.htm. Accessed July 25, 2018.

  11. “2016 National Popular Vote Tracker.” Cook Political Report, https://docs.google.com/spreadsheets/d/133Eb4qQmOxNvtesw2hdVns073R68EZx4SfCnP4IGQf8/htmlview?sle=true#gid=19. Accessed June 19, 2018.

  12. “Battle for the Senate 2018.” Real Clear Politics, https://www.realclearpolitics.com/epolls/2018/senate/2018_elections_senate_map.html. Accessed July 25, 2018.

  13. “2012 Michigan Exit Polls.” CNN, http://www.cnn.com/election/2012/results/state/MI/senate/. Accessed July 25, 2018.

  14. “Voting and Registration in the Election of November 2012.” U.S. Census Bureau, https://www.census.gov/data/tables/2012/demo/voting-and-registration/p20-568.html. Accessed September 7, 2018.

  15. “2016 National Popular Vote Tracker.” Cook Political Report, https://docs.google.com/spreadsheets/d/133Eb4qQmOxNvtesw2hdVns073R68EZx4SfCnP4IGQf8/htmlview?sle=true#gid=19. Accessed September 10, 2018.