What to Expect from Early Voting in the 2024 General Election

September 20, 2024

Early voting is nearly upon us for the November 2024 presidential election. North Carolina usually leads off with mail balloting starting on Sept. 6, or sixty days before the November 5 general election (legal disputes regarding RFK Jr. delayed North Carolina mail ballots such that Alabama led the pack this year). More states follow with mail and in-person early voting options, with a handy guide of dates and more available from the National Conference of State Legislatures.

As I’ve done since 2008, I will track early voting activity in the United States. This election I have moved my early voting tracker to the new Election Lab at the University of Florida site, which I hope will improve visitors’ experiences. Like poll averaging, there has been a proliferation of early voting trackers — some state specific and some national — since I first created a Google doc spreadsheet back in 2008. I believe my value-added is my experience and understanding of early voting dynamics. For those who want to learn more, my book From Pandemic to Insurrection: Voting in the 2020 Presidential Election covers the history of early voting, how it works, and the political battles that surrounded voting in the 2020 election. And with that I will impart what we may learn from early voting this election cycle.

Turnout

The most reliable and important information we can learn from early voting is turnout.

Eight states — California, Colorado, Hawaii, Nevada, Oregon, Utah, Vermont, and Washington — plus the District of Columbia run their elections primarily by mail. Election officials in these states send a mail ballot to every active registered voter. (Inactive registered voters — those who have not voted in a recent election — must request a mail ballot, at which time they become active voters.) Voters in these states still have options to vote in-person. Oregon and Washington — the two early adopters — offer emergency voting at an election office. Colorado innovated layering vote centers on top of mail balloting, which allows those who wish to do so an opportunity to cast an in-person ballot at special polling locations during the early voting period through Election Day. In Colorado and some other states offer special polling locations known as vote centers as a place where any voter may cast an in-person ballot and where eligible voters can do same-day registration — register and cast a ballot in-person.

Election laws and policies differ dramatically across the United States. As always, it is best to seek information from your local election officials for trusted information about elections.

These “all-mail” or “vote-by-mail” or “vote-at-home” states provide the strongest signal of turnout since upwards of ninety percent of voters will return their ballots before Election Day. In 2018, it was the vote-by-mail states that provided the first indicators of unusually high voter turnout when their early voting exceeded the number of votes cast in any prior midterm election, even when controlling for population growth.

There are other states that tend to have very high levels of early voting, such as Arizona (which has a permanent absentee ballot list) and North Carolina and Texas (where large numbers of voters tend to use in-person early voting). These states and others can also provide clues about overall turnout.

Turnout matters for a couple of reasons.

From the election forecasting perspective, pollsters broadly develop their “likely voter” models based on assumptions about overall turnout levels. Indeed, surveys with an “all adult” sample assume that all adults will register and vote while a “registered voter” sample assumes all registered voters will vote. Both of these implicit extremely high turnout assumptions are unrealistic. As Election Day nears, these all adult and registered voter pollsters will shift to likely voter models. (This shift used to happen exclusively in October, but many pollsters now release non-falsifiable likely voter models months in advance of an election.)

Pollsters build their likely voter models on turnout assumptions. As a poll consumer, we rarely know a survey’s target turnout assumption unless the pollster releases their survey estimates with low, medium, and high turnout scenarios — as some of the most transparent pollsters do. Early voting levels help us understand which of these turnout scenarios is most reasonable, especially when candidate support varies with turnout. There are suggestions this is currently the case, as Kamala Harris’ (and Joe Biden’s before her) support versus Donald Trump tends to increase as one moves from all-adult to registered voter to likely voter survey samples.

From an election night perspective, having good estimates of early voting activity assists the media and election nerds — of which I am one, naturally — in estimating the outstanding vote yet to be counted, which will matter when making election night winner calls in the close states. Knowing the size of the early vote can help us know if the election results that have been reported so far include early votes. This matters because early voters tend to be more Democratic than Election Day voters (except in the vote-by-mail states), which provides a nice segue to using early voting to make election forecasts.

Forecasting Election Outcomes

I know most of you don’t care about turnout and want to get to the good stuff: what can early voting tell us about who will win the election. If you skipped the turnout section to get here and if you want to be a good consumer of polls you should go back. I’ll wait.

Early voting can be used to forecast overall vote margins, but one must be careful about how one interprets early voting data.

The most important early voting fact to know is that Election Day voters are markedly different than early voters. Typically, early voting is a blue wave and Election Day is a red wave. If Democrats are leading in early voting — as they typically do — that is no guarantee Republican Election Day voters won’t swamp the early vote. If Democrats are losing the early vote — as the did in Florida’s 2022 midterm election — that is almost certainly a sign Republicans will win. Milage may vary across states, but these are generally good rules of thumb.

If early voters are different from Election Day voters, how might we forecast who will win an election from early voting?

A sound method to forecast election winners from early voting is to compare current early voting activity to activity in a past comparable election. This is an approach called “difference-in-difference.” It is similar to poll averaging where an adjustment is made to current polls for past misses by the same pollster.

To give an example. Suppose Democrats are leading by ten points among early voters in North Carolina. That doesn’t mean Harris will win North Carolina by ten points because we know Republicans likely will vote in large numbers on Election Day. The real question to answer is if the ten point Democratic lead is enough to withstand the expected Republican Election Day surge. If we know that in a past similar election Democrats had an eleven point lead among early voters, but lost North Carolina by four points, we can infer Republicans will likely overcome the Democrats’ ten point lead in early voting and win the election.

A key assumption of difference-in-difference models is that the past election is comparable. The idea is that if the current election (or poll) is held under the same circumstances as a past election, any difference in voting behavior among partisans reveals something about their current enthusiasm relative to the prior election. Passing a law to expand or contract early voting can affect how many people are voting early and the partisan composition of the early electorate.

Emergency mail ballot policies during the pandemic are a good example of when a comparison election may be absent. Mail voting doubled to record levels, primarily among Democrats who heeded public safety calls. In many states no comparable past election was available to draw a sound evaluation against. Fortunately, the vote-by-mail states were available to forecast turnout and election winners — because their voting method didn’t change much from the 2016 election.

The unusual pandemic circumstances will challenge 2024 early voting analyses, particularly in states that do not conduct vote-by-mail elections. Perhaps 2016 will be a better comparison election than 2020. I’d still caution making comparisons to 2016 because Donald Trump’s election fraud rhetoric changed Republicans’ voting behavior. Before 2020, Republicans tended to vote by mail (and on Election Day) and Democrats tended to vote in-person early. In 2020, Republicans embraced in-person early voting, eschewing mail balloting, while Democrats adopted mail balloting. There is some evidence from the 2022 midterm election that some of these behaviors represent permanent changes.

That partisans tend to use mail balloting and in-person early voting at different rates adds another layer of complexity to analyzing early voting. In most — but not all — states early voting begins with mail balloting. Later, in-person early voting options become available. The partisan composition of the early electorate is thus influenced by the availability of early voting options and how many people have used mail voting and in-person early voting.

Whew! We’ve covered a lot, so let’s recap before we move on:

1. Early voters are different than Election Day voters. More Democrats tend to vote early and more Republicans tend to vote on Election Day. Just because Democrats lead in the early vote does not mean they will win the election.
2. The vote-by-mail states are the best states to watch to forecast turnout and winners.
3. Timing matters. Mail balloting usually proceeds in-person early voting, and partisans tend to use these two voting methods at different rates. Don’t read too much into statistics until a generous amount of in-person early voting has been cast. We will get our best read on the election from early voting a week or so before Election Day.

The Meaning of Voter Registration

Up to this point I’ve been purposefully vague in talking about partisanship. There are three measures of partisanship available from voter files, in descending order of their accuracy:

1. Party registration in states with closed or semi-closed primaries.
2. Race in the few states that track it, if party registration is unavailable.
3. Early voting activities in Democratic and Republican strongholds.

Party registration is similar to self-identified partisanship often reported by pollsters, but it is not exactly the same. States have party registration to identify registrants eligible to vote in closed primaries (in which only party registrants can participate) or semi-closed primaries (where registrants who do not register with a party are allowed to choose which primary to participate in). States without party registration allow voters to choose which primary they wish to participate in (the top-two primary states are similar, but I do not want to digress).

Self-described independents on surveys are not entirely the same as people who do not affiliate with a political party (sometimes called “no party affiliates” or “decline to state”). As people get older they develop stronger partisanship and are more likely to register with a political party. Younger (and due to the growing diversity of the country, more diverse) voters are less likely to register with a party. The people who decline to register with a party are thus not pure independents like surveys; they tend to lean towards the Democrats because they are younger are more diverse.

There is an important early voting timing dynamic in states with party registration. The earliest of early voters are those who already made up their mind who they will vote for. They tend to be highly knowledgeable and attentive to politics, they vote at high rates, and they tend to be strong partisans, i.e., they tend to register with a party. No party affiliates tend not to share these traits. Thus, the share of younger and more diverse early voters with no party affiliation tends to start relatively small and then grows, particularly during the last week before Election Day.

Another early voting dynamic to watch for is so-called “Souls to Polls” in-person early voting mobilization drives run by African-American churches the weekend before Election Day. The share of African-Americans voting on this Sunday tends to spike. Indeed, it is possible to observe campaign-specific spikes that occur around major political events or campaign rallies held near in-person early voting locations. The 2008 Obama campaign pioneered holding rallies next to in-person early voting polling locations and providing preferential rally seating to people who sported an “I voted” sticker.

People register with a party to participate in a party primary. In some “semi-closed” primary states no party affiliate voters may cast a ballot in a party primary of their choosing. When they do, they often switch their party registration to that party. Indeed, anyone can change their party registration within allowed deadlines to do so. Given the greater interest in the Republican primary in 2024, we might expect that some registered Republicans are really anti-Trump no party affiliates or Democrats who registered to vote against Trump in the general election, particularly in semi-closed states. This is where polling could help, but few pollsters release crosstabs of respondents by their party registration.

What about states that do not have party registration? There are a couple of options. A couple of states like Georgia and South Carolina report the race of registrants who voted early. We might look at African-Americans for a sense of the partisanship of early voters. Lacking race, we might look at early voting activity among Democratic and Republican strongholds within a state, inferring relative greater interest as some meaning.

States to Watch

We will have early voting data from all of the battleground states and states with hot Senate races. We likely won’t have timely data from a couple of the more difficult states, like Wyoming and Missouri. Here are some highlights:

The Vote-by-Mail States

Watch the vote-by-mail states the most closely: California, Colorado, Hawaii, Nevada, Oregon, Utah, Vermont, and Washington — plus the District of Columbia. They will give us the best read on overall turnout. The battleground state of Nevada has party registration and I encourage you to follow Nevada political reporter Jon Ralston for his early voting takes. Colorado and Oregon also have party registration and will provide some indicators of partisan enthusiasm writ-large.

Florida

Florida Republicans changed state law to make it more difficult to request a mail ballot. There will be likely less mail ballot activity in 2024 compared to 2016. It remains to be seen what the partisan effect will be and if in-person early voting will compensate.

Ohio

Ohio does not have true party registration in that the state allows any registrant to vote in the primary of their choosing. Still, Ohio reports “party registration” on their voter file as the most recent party primary a registrant participated in. Such measures are correlated with party identification, but not as strongly as states that require formal party registration. Undoubtedly, there will be claims that Sharrod Brown is doomed because of high levels of voting among Ohio Republicans, but this will be a poor forecast of the actual state of the election. (Georgia does this, too, but for some reason there has been fewer bad interpretations of its “party registration.”)

North Carolina

North Carolina starts mail balloting very early and provides daily updates on its ftp site. For these reasons, North Carolina tends to attract a lot of attention. Beware that mail balloting usually comprises a single digit percentage of the state’s total and early vote. In-person early voting dominates mail balloting by far in North Carolina. I strongly recommend waiting for a few days of in-person early voting activity before drawing inferences.

Wisconsin

Regrettably, Wisconsin’s individual data is prohibitively expensive — costing $0.01 per record for the mail ballot data per time that it is purchased (really!). So, we’ll likely rely on daily state reports that have only aggregate numbers for each county.

Parting Words

Since I started tracking early voting activity lots of individuals and organizations have gotten in on the action. This is all good as far as I am concerned since we can check each others’ work and trade tips. Where possible, I link to the public sources so anyone can DIY. This won’t be possible everywhere since there are states that charge for their data.

I tend to work a couple of days ahead of the voter file vendors like TargetSmart and L2 who produce data for media organizations. These vendors analyze individual-level early vote data for campaigns that takes time to acquire and process. I collect statistics from any official or trusted source. As such, in the past I tended to report more early voting activity than some media sites like CNN and MSNBC that had contractual relationships with vendors. I assume they will do it again. Also note these voter file vendors may report modeled party affiliation, which at times provides some squirrely statistics in places like Michigan that does not have party registration.

I anticipate I will use data from Arizona and California political firms that have public early voting dashboards, as I have done in past elections. In these states early voting data is available only from counties, and it is more effort than I can manage to collect these data. Believe it or not, I do this by myself!

If you use my statistics, great! I just ask that you give me (and the organizations that I lean on) proper credit. These data aren’t always just lying around. I’ve written thousands of lines of code to process and visualize early voting data for you, taking into account quirks of these data I’ve learned over the years.

Lastly, thank you so much for subscribing to my Substack! I primarily use your generous subscriptions to give back to you by purchasing more data. So, please subscribe if you haven’t already done so.