THE SECRETS OF VOTER TURNOUT 2018

Transcription

1 Dish THE SECRETS OF VOTER TURNOUT 2018 AN ANALYSIS OF INDIVIDUALLEVEL VOTE HISTORY IN THE VIRGINIA GOVERNOR S RACE Comcast May 2018 Netflix!X!1

2 Overview VIRGINIA 17: WHAT HAPPENED Despite polls suggesting a tight race, Democrats swept every statewide office for the second election in a row. Ralph Northam wins the Governorship by by 9% In the House of Delegates, GOP loses 15 seats!2

3 Overview THE REAL STORY? TURNOUT In March, Virginia s voter file was updated with vote history from the November election shedding light on the extent of Democratic enthusiasm and offering clues about which way turnout will go this November. The news was grim for Republicans. Compared to the previous midterm baseline Democratic turnout +16.7% Republican turnout +0.5%!3

4 Overview THE REAL STORY? TURNOUT Amongst voters who lean towards one party but who don t participate in primaries, the turnout gap was even wider Democratic non-primary turnout +14.6% Republican non-primary turnout -9.0%!4

5 Virginia 2017 Elections THE REAL STORY? TURNOUT Compared to the previous midterm, that s an advantage big enough to shift the margin of victory to the Democrats by 7.5% And compared to the previous gubernatorial race, turnout in 2017 increased the Democratic margin of victory by 5%. Statewide, that s enough to explain nearly the entire shift away from Ken Cuccinelli in 2013 and Ed Gillespie s Senate bid in 2014.!5

6 Overview WHY IT MATTERS Turnout isn t the only factor driving a tough political environment for Republicans. Persuasion is an even bigger factor. That means Trump 2016 voters supporting Democrats this year. Virginia bucked this trend. The fact that Ed Gillespie didn t lose many Republican votes showed he ran the strongest possible campaign in a negative environment but it wasn t enough to overcome the Democratic turnout wave. Turnout shifts provide a floor for Democrats in Republican survival depends on strong, disciplined campaigns that keep defections to a minimum, and using analytics to identify and mobilize Republican voters at risk of sitting out 2018.!6

7 Two Virginias MAPPING THE DEMOCRATIC TURNOUT ADVANTAGE, BY COUNTY AND PRECINCT The Democratic turnout surge was not evenly distributed across the state. It was primarily concentrated in Northern Virginia, and in the state s urban areas. Republicans held their own on turnout in Trump Country counties in Southwest Virginia, but populationwise, these areas of over-performance were dwarfed by Democratic gains in cities and suburbs.!7

8 Two Virginias MAPPING THE DEMOCRATIC TURNOUT ADVANTAGE, BY COUNTY Compared to the midterm benchmark, Democrats outperformed Republicans on turnout by 19.6% in Fairfax County (NoVa suburbs) 21.3% in Loudoun County (NoVa suburbs) 20.5% in Henrico County (Richmond suburbs) 16.9% in Chesterfield County (Richmond suburbs) 20.0% in Virginia Beach City (Hampton Roads)!8

9 Two Virginias DEM SURGE: LOUDOUN COUNTY Pictured here: Brambleton, a fastgrowing suburban community west of Dulles Airport, in the competitive 10th Congressional District. Compared to the 2014 midterms, Democrats here outperformed Republicans on turnout by 38.9%!9

10 Two Virginias GOP MINI-SURGE: SOUTHWEST VIRGINIA Pictured here: Ridgeway, a rural community in Henry County, in the 9th Congressional district. Compared to the 2014 midterms, Republicans here outperformed Democrats on turnout by 6.5%!10

11 Implications for November MAPPING THE DEMOCRATIC TURNOUT ADVANTAGE, BY CONGRESSIONAL DISTRICT Democrats overperformed Republicans on turnout by at least 16% in 3 GOP-held seats enough to shift the margin of victory by at least 8% if repeated in November. VA-2 (Scott Taylor): +18.7%, VA-10 (Barbara Comstock): +18.4%, VA-7 (Dave Brat): +16.8% But the Democratic turnout advantage was over 12% in every other district except for VA-9 meaning that Democrats are energized across the board, not just in suburbs.!11

12 Implications for November TURNOUT METRICS BY CONGRESSIONAL DISTRICT District 2017 Dem Turnout Advantage Estimated Dem Margin Shift 17 Estimated Dem % Gain 17 Cook PVI VA % 6.7% 3.4% R+8 VA % 9.0% 4.5% R+3 VA % 5.6% 2.8% D+16 VA % 6.1% 3.1% D+10 VA % 4.8% 2.4% R+6 VA % 5.4% 2.7% R+13 VA % 7.6% 3.8% R+6 VA % 7.6% 3.8% D+21 VA-9 3.4% 1.9% 1.0% R+19 VA % 9.9% 5.0% D+1 VA % 10.3% 5.2% D+15!12

13 Demographics COLLEGE EDUCATION DROVE DEM TURNOUT The higher the share of college educated voters in a precinct, the more Democrats outperformed Republicans on turnout especially non-primary voting Democrats. Higher Turnout Than Expected More College Educated Voters!13

14 Demographics COLLEGE EDUCATION DROVE DEM TURNOUT In the most educated precincts, Democratic leaning non-primary voters outperformed Republicans by 34 points compared to the midterm benchmark. Higher Turnout Than Expected 34% GAP More College Educated Voters!14

15 Demographics AND EVEN MORESO AMONG WOMEN The Democratic turnout surge was magnified amongst women in college-educated precincts. Amongst women non-primary voters, the gap widens to 38 points. Democratic women were more likely to turn out than Democratic men, but amongst Republicans, gender made no difference to turnout. Higher Turnout Than Expected 38% GAP More College Educated Voters!15

16 Demographics MORE SURGE DEMOCRATS IN EDUCATED PRECINCTS Democratic Turnout Advantage, by Midterm Turnout Score and Precinct Education/Gender Male Female Democrats had their strongest turnout advantage amongst very low turnout voters in college educated precincts especially women. Midterm Likelihood to Vote More College Educated Voters!16

17 Demographics EMERGING DEMOGRAPHICS OUTPERFORMED Turnout exceeded midterm 2017 Turnout vs. Midterm Model, by Race/Gender benchmarks the most amongst nonwhites, particularly emerging voter blocs like Hispanics, Asians, and those of another race. Female turnout was stronger than male turnout in every group except for Asians. And overall, black turnout outperformed white turnout!17

18 Demographics EMERGING DEMOGRAPHICS OUTPERFORMED The data broken out by party is even more revealing Turnout vs. Midterm Model, by Race/Gender/Party Amongst Democrats, Hispanics and Asian voters were up by 30% or more relative to the midterm benchmark. Whites were up 15% and African Americans up 13%. White Republicans the lion s share of Republicans were nearly unchanged in their turnout patterns from the previous midterm.!18

19 Looking Forward KEY TAKEAWAYS Virginia shows us how bad it might be for the GOP with a nearly 8-point shift in vote margin due to higher Democratic turnout alone. The key demographics driving this higher turnout? Highly educated voters, women, Hispanics, and Asians. Republicans didn t do worse on turnout than they did in 2014, suggesting that they are no less motivated than in the past. Their problem is that Democrats are far, far more motivated. This pattern of higher Democratic turnout is seen in nearly every special election of the Trump era with similar demographic patterns. What about New Jersey? The turnout shift wasn t quite as extreme a 2 point shift in vote margin but the statewide contest wasn t as competitive. In highly charged, competitive elections, the Democrats turnout upside may be greater.!19

20 Looking Forward KEY TAKEAWAYS: FOR POLITICAL PRACTITIONERS Having a good handle on what turnout is likely to look like in November is critical to the ability to forecast outcomes and allocate resources across an unusually large and volatile set of battleground states and districts. The surge isn t affecting all voters, or all Democrats, equally. Good data about elections that have taken place in the current political environment is key. (The good news we have this data.) Simply letting pollsters telephone response rates tell us may not be enough. Trump s approval rating and the midterm ballot has improved in polls taken in 2018, but Republicans have continued to underperform in special elections to the same degree they did in Demographic and voter file modeling & analysis across a wider range of districts, rather than simple district-by-district polling, is an essential tool to keep your finger on the pulse and see if districts you didn t expect to be competitive are coming into play.!20

21 Looking Forward ANALYTICS & POLLING CAN WORK HAND IN HAND TO ANSWER THE QUESTION: WHO WILL VOTE? When it comes to turnout models, Comparing Turnout Scores Across Elections we no longer have to guess. We know from voter files who is likely to vote in a year like 2018 and how this varies across demographics. We can use this to improve the accuracy of traditional polling AND, at the same time, provide you with the most accurate Presidential Year possible list of targets to talk to. Midterm Year!21

22 INTERESTED IN LEARNING MORE? Using accurate turnout modeling informed by precise data, we re helping clients navigate a complex 2018 landscape with the industry s most advanced polling & analytics. Want to learn more about putting data and insights like this to work for your campaign or organization? We ve made it easy to book a 30-minute introductory call with our team: bit.ly/echelonintro!22

23 OUR JOB IS TO MAKE YOU SMARTER. We combine traditional opinion research, data analytics, and new technology to measure the competitive landscape. Our integrated offering gives you fresh perspective, maximizes your use of resources, and helps you see what s coming. MORE ACCURATE POLLING Reach hard-to-find audiences. Test messages in novel ways. ANALYTICS & TARGETING Understand your audience at the individual level. Match the right message to the right person. COMPETITIVE INTELLIGENCE Understand how specific audiences move in real time, complementing traditional research. ECHELON INSIGHTS!23

24 Methodology HOW WE CONDUCTED THIS STUDY Using Virginia voter file data from L2, we compared individual vote history in the 2017 election for Governor with a midterm turnout model trained on 2014 data. The model provides probabilistic scores, from 0 to 1, that a voter will turn out to vote. Using such a model (as opposed to actual vote history) allows us to carry forward the dynamics that existed in 2014 and apply it to the electorate of today, taking into account new voter registration, aging, and underlying demographic and population change. We also compared the results to gubernatorial election model trained on 2013 data, but opted to look at midterm model, as this is the election of interest in Performance relative to the midterm baseline is measured by taking the total number of 2017 votes within a population group, divided by the sum of their model scores. Relative partisan differences are calculated by dividing the Democratic performance score by the Republican performance score. Since Virginia does not have party registration, party identification is modeled. Voters are further classified as primary voters based on their history of voting in party primaries since Have questions about this analysis? Contact Patrick Ruffini at patrick@echeloninsights.com.!24

25