3 Weeks Out – Understanding who votes

ClearPath's U.S. Midterm Series

Elections are about people. Understanding who votes is critical to winning. That doesn’t mean just knowing the numbers (which people will show up, what the demographic make-up will be, etc.). It means understanding them on a personal level, a human level. What do they care about? What are their fears, their hopes? What do they need and want for themselves, their kids, their parents? 

As we’ve said, elections are an emotional exercise. Strong campaigns address issues that matter to people. They forge deep, meaningful connections with communities over time. They reflect the communities they represent. While clear, data-driven strategy is a key ingredient to winning, connecting with people is paramount.

But which people? Should campaigns care about everyone equally? Well, no.  

1) “Likely voters”: Identifying who will vote is a good first step.

Over the last 10 years, the US set record LOW and record HIGH electoral participation rates. For our non-Americans readers, voting in the US is not compulsory, and registration is not automatic. (Elections are also on Tuesdays…it’s a long story.) Indeed, up until very recently, only 50-60% of adults eligible to vote would cast a ballot in Presidential years. In midterm years (that is, non-presidential elections like this one in a few weeks), it is even lower. In the 2014 midterm, only about a third (36%) of eligible adults voted—the lowest in 72 years. Just 4 years later, the 2018 midterms set the record for ballots cast in a non-presidential year. That record lasted only 2 years, as in 2020 a record number of Americans (158M) cast ballots. Say what you will about Donald Trump (and we will say plenty), he is undoubtedly a force for increased turnout (on both sides).  

To conduct electoral research in the US, or any country without compulsory voting, we must reduce our aperture to only those people who will vote. Developing strategy from “all adults” is useless, if only half of them actually show up. Instead, campaigns focus on “likely voters” (and further on their own target voters), lest they waste millions of dollars on ads to people who aren’t even going to vote. 

Identifying likely voters is tricky. There’s plenty of data science behind it, and it’s not always right. Some of the big missed projections in recent elections result, at least in part, from incorrect assumptions about who would vote. Still, an imperfect likely voter model is better than none at all. After all, the entire strategy and financial resources of the campaign rely on good targeting. Pollsters, strategists, paid and social media teams, TV analysts—we all need our likely voter models.

So what goes into a likely voter model? In the US we benefit from voter files—publicly available information about people’s voting history (i.e., whether they voted in a given election or not)—which helps. Regardless, what information is useful, what questions can you ask in a survey that go into likely voter models? Without getting into the finer details, these models typically incorporate information from three buckets:

1. Past vote behavior: Past behavior is a good predictor of future behavior. Did you vote in each of the last 4 elections? Then you’re probably voting in this one. Only vote in 1 of the last 4? Then we need more information about you. 

2. Stated intention to vote: People are terrible at predicting their own future behavior. This includes voting. But, a combination of questions (Do you have a plan to vote? How likely are you to vote? How motivated are you? etc.) in surveys help us refine the model if you don’t have information on their voting history, or to account for changing habits. 

3. Non-voting behaviors and characteristics: Educational attainment, income, race, geography, age, how often you attend religious services (“religiosity”), how close you feel to a political party, news consumption, whether your friends or family vote—each of these are, to one degree or another, correlated to or associated with a likelihood of voting and are, thus, potential inputs for likely voter models.  

Notably, likely voter models vary year to year. And state to state. And within each state year to year. They vary based on how competitive local elections are, who sits in the Oval Office, or if Mercury is in retrograde 4 weeks before the election.¹

Electoral participation is very high in states like Colorado, Oregon, and Washington where all registered voters are sent ballots via mail ahead of the election. It is also high in Minnesota where civic engagement is a state pastime. It is consistently low in states like New Mexico, West Virginia, and Indiana. And then there are states with laws and policies that actively attempt to prevent people from voting, like Texas, North Carolina, and Georgia.² 

Different demographics also vote at different rates. Whites, college educated, older, higher income, and Republicans are more likely to vote than other groups. Regardless of the year or the election context. This poses obvious and consistent problems for Democrats for whom minorities and young people make up disproportionately large shares of the electorate.

On the other hand, Republican strategists (mostly rich, white, arch-capitalists) in the late 70s and 80s made an unholy alliance, choosing to invest in white evangelical Protestants (born again Christians). They’re still collecting dividends. White born again Christians voted for Reagan by 50 points, by 40 points for Bush, and by 60 points for Trump. In addition to providing these extreme margins, white born again Christians punch above their weight at the ballot box—they make up just 14% of the total population but account for 28% of the electorate!

Republicans took the culture wars mainstream because they understand politics is about winning and about power. And they can do math. Republicans actively choose to politicize white evangelicals, and they consistently fight for evangelical Christian values when in office because the GOP knows that as long as they have this fervent bloc of supporters in their corner, their grip on power is hard to detach.

Creating your likely voter model and knowing the “math” of the electorate is critical to allocating resources and figuring out a campaign’s path to victory.

But models can fail. 

2) Models are not mysteries. They are assumptions. Quality, regular research tests and updates those assumptions.

In 2016, the Hillary Clinton campaign did not conduct state level polling in Wisconsin, Michigan or Pennsylvania for the last three weeks of the campaign. Instead, they relied on voter analytics and modeling—big data that could “predict” voting behavior. It didn’t. The three “Blue Wall” states fell for the first time in 24 years. 

Models are algorithms. And as any Millennial YouTuber or Gen Z TikToker can tell you, algorithms have inputs. They’re based on assumptions. Human assumptions. Human inputs. A model is only as good as its raw data. A model meant to predict human behavior requires human behavior as inputs. 

As we’ve said, dynamic campaigns build strong feedback loops with voters. Research is a major piece of that. Quality surveys are prerequisites to establishing the model. But, regular research is necessary to update the model to account for changes in behavior, or resulting from events or other campaign realities.  

The 2016 election is not the only recent example of modeling short-comings. In 2018, Democratic enthusiasm in primary elections dwarfed initial likely voter models. Campaigns updated LV models multiple times in 2018 to account for this Blue Wave, and later also the blowback effect of high Republican turnout to defend Trump. In 2020, massive spikes in turnout for both Republicans and Democrats were difficult to capture and varied state by state (e.g., in Montana the LV model, already predicting the highest turnout ever, undershot the actual result by an additional 10%!—sadly, all of it Republican).

Models also tend to break when unprecedented events happen (COVID, lockdowns, deep recessions, massive changes to election laws, etc.). These have difficult-to-predict impacts on elections and electorates, and therefore on likely voter models. 

At the end of the day, a model is a simplification. Smart campaigns know they are imperfect, but directionally useful. Other campaigns treat them like a palantír from the Lord of the Rings, able to show you the real truth. But to continue the metaphor, the palantírs were unreliable guides, often showing a “selective truth.”³ Campaigns that over-rotate to blind faith in their models are at great risk of disappointment.

Even more important than not treating your likely voter model as a sacred cow is understanding that it only helps us answer who votes. The model alone does not tell us who these people are.

3) Turnout is not strategy. We build smart strategies on understanding the people.

In Democratic politics, ‘Hispanic and Black voters are turnout audiences’ is a common refrain.

We hate this.

What that says is, we aren’t treating them as people. We aren’t listening to or respecting what they want or why they vote. It means ‘we don’t need to spend money persuading them to vote for us, we just need to turn them out’. It is built on math, not values. It is built on a sense of entitlement. It is built on…whatever the opposite of empathy is.⁴

Democratic strategists believe Black and Hispanic voters support Democrats at such high rates that simply getting them to the ballot box is all the campaign needs to do. Then they can use campaign resources on ads fighting over white voters who vacillate between D and R.

At best, this type of thinking reduces communities to a least common denominator. It oversimplifies and is dangerously short-sighted. Rather than party-building, community engagement, or even just being present all year round, Democrats show up for these audiences in September and October just in time to ask them to vote. Then election day hits, and Dems scurry back to DC. 

We are far from the first people to level this criticism. Black organizers and elected officials consistently ask the Democratic party to invest and campaign in Black communities (here it is after 2016; here it is again in 2020). These requests have not received an adequate response. And may be seeing the impact of this indifference. While Black women consistently vote for Democrats at 90+%, Black men have been shifting away from the party since Obama’s first term. 

As bad as Democratic flippancy has been with the Black community, it’s arguably worse with Hispanics.

Hispanics are 2022’s quintessential demographic example of the Democratic party refusing to focus on people. First, the Democratic Party wrongly lumps “Hispanics” together as a monolithic bloc. But Hispanics in Florida are different from Hispanics in Arizona, who are different from Hispanics in New York, or Colorado, or Texas. Mexicans in New Mexico are different from Venezuelans in Florida. Cubans are different from Salvadorans. Catholic Hispanics are different from Protestant Hispanics. Young Hispanics are different from older Hispanics. 

Second, up until very recently (last year in fact), the Democratic Party would default to talking to Hispanics only about immigration (here they are in the 2008 cycle, and Clinton in 2016, and Biden in 2020) as if by virtue of their heritage, this is the only thing that could matter to them. The reality is, Hispanics prioritize the economy, healthcare, and crime higher than immigration. Jobs, health, and safety. Fundamental needs. Maslow’s hierarchy. While Biden started to layer in economic and health in 2020, immigration was always, and infuriatingly, first. 

Third, Democrats ironically don’t even get immigration right with Hispanic voters, especially in border states. In Texas, Hispanic voters say Republican Governor Greg Abbott has the better immigration policy by a 9 point margin over Democratic candidate Beto O’Rourke. A sizable minority (30-40%) of Hispanics support increasing border security and making it harder for immigrants to overstay their visas. In focus groups, the nuance is starker—older Hispanic immigrants who migrated legally have an “if I did it the ‘right way’, then others should, too” mentality. 

Fourth, there is a tendency, particularly among the progressive wing of the Democratic party, to prioritize equity, even when it’s not the priority of the targeted “beneficiary.” For example, the recent rise of the term “LatinX” as a gender-neutral, pan-ethnic way to describe Hispanic people is a term that hardly any Hispanic person uses (only 3%, according to Pew). Its use arguably does more harm than good by signaling that Dems don’t understand (or even try to understand) the people they purport to represent.

Misunderstanding the Hispanic electorate—their needs, their values, their community, and even who constitutes this diverse bloc—is the result of Democrats defining Hispanics by their ethnicity, rather than understanding them as people.

Republicans get it. They take a tougher stance on immigration and the border, but lead with economic opportunity messaging for Hispanics. No surprise, Trump made gains with Hispanic voters nationally. A record number of Republican Latinas are running for Congress. Since a 2012 low-point, Republicans have strategically invested in Hispanic outreach, recruitment, and capacity building. And it’s working.

In the final weeks of the cycle, Democrats are outspending Republicans on Spanish-language ads 3:1. And they are prioritizing messaging on jobs, the economy, and abortion—all top issues among Hispanic voters. Though one final push before the election likely is not enough to make up for more than a decade of bad messaging reliant on bad assumptions, at least it’s a start. 

Turnout is not strategy. Campaigns are about people. Elections are about connecting with the right people in an authentic and relevant way. Building long-term capacity and bonds with people will win elections. It requires actually understanding them, listening to them talk about their needs, helping to fulfill them, and actively representing them. And not waiting until September of an election year to do so.

—

We’ve spent the last 6 weeks discussing the tenets of good campaigns using the US Midterms as our foil. That’s all well and good, but some of you might be asking about the elephant in the room. What does it mean to run a campaign against an opposition who actively undermines the election system and democracy itself? According to one analysis, 60% of Americans will be able to choose a candidate who believes Biden stole the 2020 election. Scary? Yeah.

Next week, we will tackle campaigning in the face of extremism. 

1. OK, that last one is bullshit, though a certain unnamed presidential candidate from a certain unnamed country once said in a private strategy meeting with one ClearPath partner in attendance that his "hat" was worth 10,000 votes, so there are a lot of strange beliefs out there.

2. The national Voting Rights Act of 1965 was passed to stop racial discrimination in elections, but has been systematically gutted by the Supreme Court. Yay, democracy!

3. Shout out to all our fellow nerds.

4. Indifference, apathy, disdain. We know; this just felt more dramatic.