Where Did The Blue Wave Come From?

Turnout vs. Preference: Context and Caveats

Swirling around many analyses and think-pieces about elections is a central question: do Democrats/Republicans win elections by getting the people who agree with them to vote (turnout) or by convincing people who disagree with them to change their minds (preference)? Campaigns try to do both, of course, but is one more important in a given race?

We think preference shifts were a big deal in 2018. Before explaining our data, it’s important to address three key issues that arise with these sorts of analyses:

1. Who are the “people” we’re studying? Many excellent prior analyses, like those by Rachel Bitecofer and Yair Ghitza, use voter files to look at voters’ underlying party affiliation, to understand the behaviors of registered Dems and Republicans (and independents) from election to election. We use a different approach (described in more detail here) that subdivides people only by demographics, e.g., age, race, degree of education, etc. That means that if, within the same demographic group, some previously Republican voters stay home and some new Democratic voters show up, we will see that as a preference shift though from a party affiliation perspective, it could also be seen as a turnout shift.

2. What’s the geographic unit of measure? These analyses depend a lot on the geographical boundaries and the demographic categorizations. For instance, if college-educated voters move from one district to another in a state, that’s demographic change at the district level but not at the state or national level. In our work so far, we’ve looked at voters’ behaviors only at the national level.

3. What’s the comparator year? If there’s a change in demographics, turnout, or preference, the key question is: compared with what? Using 2014 as a baseline—comparing mid-term to mid-term—then there is a Democratic turnout surge, as Rachel Bitecofer argues. If you look instead at the changes from 2016 to 2018, focusing on the blue wave, voter preference was the more pronounced shift, as seen in this Yair Ghitza article. We take a few different perspectives, as noted below.

The Evolving Democratic Coalition

From 2010 to 2018, non-white voters were highly likely to vote for Democrats, with a >75% Democratic voter preference throughout all those elections. By contrast, white voters are more likely to vote for Republicans, remaining between 40% and 45% throughout, with a low in 2016 and an encouraging uptick in 2018.

There are more white than non-white eligible voters and white voters have higher overall turnout. We can see this clearly in a chart of the shares of the Democratic electorate, that is, fractions of Democratic votes. Though non-white voters are overwhelmingly democratic, they make up at most 40% of the votes Democrats get in any election.

Grouping by educational attainment, age and sex, we a different picture of the Democratic electorate. Young college graduates and older female college graduates are consistently democratic, but the younger voters are more Democratic leaning by about 10 percentage points. Older college-educated men were Republican leaning until 2018 when they shifted to slight support of the Democrats. We also see a very distinct shift: beginning sometime after the 2012 election, college-educated voters become more likely to vote for Democrats, and non-college-educated voters become more likely to vote for Republicans.

When we looked at the breakdown by race rather than educational attainment, it looked as if Democrats had gained with all groups from 2016 to 2018. But looking at educational attainment, it’s clear that Democrats lost ground with non-college-educated voters. Though the census data is not granular enough for us to do the same analysis broken down by race and educational attainment, it seems likely that the Democratic losses between 2016 and 2018 are largely with white non-college-educated voters, the so-called “White Working Class”, something we discuss in some detail in another post.

The movement of college-educated voters toward Democrats, while non-college-educated voters move toward Republicans, is also clear in the Democratic vote-share. In 2010, Democrats got 60% of their votes from non-college-educated voters but that is down to 49% in 2018.

Breaking Down The Changes

People tell various stories about how elections are won or lost. For example, one story about the 2018 blue wave is that disaffected Republican voters stayed home, while energized Democratic voters turned out in unexpected numbers. Another story is that significant numbers of voters shifted from voting Republican to Democratic.

We think there is a grain of truth in the first story but that the second is more convincing. To see why, we’re going to look at some tables that break down the changes in Democratic votes by our three factors: demographics, turnout and preference. Before we dive in, let’s consider just one row of such a table, looking at changes from 2016 to 2018:

We have:

1. The name of the demographic group (older female non-college-graduates).
2. The population (44.7 million) in the ending year.
3. The change in Democratic votes coming from changes in population (-120,000 votes). One important note here: if this is a group that votes for Democrats on average, increases in population lead to increases in Democratic votes and a positive number in this column. But if this is a group that votes against Democrats, an increase in population is a net loss of Democratic votes and the number in this column will be negative. Older female non-college educated voters tend to vote against Democrats so this net loss of votes comes from an increase in population.
4. The change in votes coming from changes in turnout (+693,000 votes). Again, increased turnout may lead to a positive or negative number here, depending on whether the group is more likely to vote for Democrats or Republicans. In this case, lower turnout is leading to a net gain of Democratic votes.
5. The next column indicates votes gained or lost from shifts in voter preference (-745,000).
6. The total votes gained or lost in this group over the time period (-171,000 votes, the sum of columns 3, 4, and 5).
7. The total votes gained or lost (column 6) as a percentage of the total electorate (-0.08%).

Looking at all the groups together tells a more complete story. Overall, changes in turnout does not appear to be a big net mover of votes. Some Democratic votes were gained by a drop in turnout among older non-college-educated voters, but a similar number of Democratic votes were lost by lower turnout among young college-educated voters. Similarly, demographic shifts play a small role—net gains among college-graduates outpace the aging of the non-college-educated groups. Nearly all the Democratic vote gains arise from changes in voter preference.

It’s instructive to look at the same groups but this time examining the changes from 2010 through 2018. Here we see that voter preference also plays the largest role. Over this time range, demographics begins to play a larger role as well. More people are getting college degrees and that increases the size of the college-educated electorate, which benefits Democrats on average. Again, the turnout changes are fairly small across all groups.

And it’s worthwhile to look at these longer-term changes by race as well as by educational attainment. Again we see that voter preference plays the largest role and that demographics is significant. One distinction here is a noticeable trend to higher turnout but one which generates more net votes among non-white voters than are lost among white voters, leading to a net gain of Democratic votes.

In our analysis, the blue wave—the electorate shift just between 2016 and 2018—was produced by voters who preferred Republicans in 2016, shifting toward Democrats in 2018. We also see that current demographic shifts, both in average educational attainment and non-white vs. white voters, lead to more Democratic votes. Turnout shifts are also largely good for Democrats over the long-term.

It’s important to note that this analysis is at the national level. Each district has its own version of this story and we plan to dig into those in more detail in the coming months.

Take Action

What’s the takeaway here, in terms of concrete actions you can take? Firstly, as always, it’s important to support organizations that fight for voting rights and work for higher turnout among minority communities and young voters. Though turnout was not the driver of the blue wave, it was the base beneath it. All those “extra” votes don’t matter, if the Democratic base doesn’t (or can’t) show up on election day. So consider supporting:

• Fair Fight, Stacey Abrams’ organization dedicated to fighting for voting rights in Georgia and beyond.
• BLOC, community organizers in Milwaukee, WI, working on improving black voter turnout.
• Campus Vote Project, an organization working on campuses across the country to register college students, an important and suppressed source of Democratic votes.

And stay focused on progressive policies where there is crossover support from independents and centrist Republicans. Data For Progress does extremely good work in this area and has identified policies that progressives support which also poll well with centrist Democrats and Republicans, e.g., capping credit-card interest rates or ending the war in Yemen.

Lastly, support local candidates who will rally turnout and engage undecided voters who may then support progressives up and down the ticket. This is part of how Blue Ripple Politics looks at all elections. Our current pieces on Louisiana, Kentucky and Mississippi highlight many great examples of this.

Want to read more from Blue Ripple? Visit our website, sign up for email updates, and follow us on Twitter and FaceBook. Folks interested in our data and modeling efforts should also check out our Github page.