As recently as 2015, automatic voter registration did not exist in the United States. Yet today, 16 states plus the District of Columbia have enacted (though in several cases, not yet implemented) some version of AVR. Almost overnight, it has become a core part of the agenda for those who want to make it easier for more people to vote. This year alone, AVR bills have been introduced in 39 states.1 Where they can’t convince the legislature, AVR advocates sometimes take their case to the people — Alaska, Michigan and Nevada have all enacted the policy via ballot measure. And someday, AVR could become a national mandate: It was a centerpiece of H.R. 1, the voting-rights bill passed earlier this year by the newly Democratically controlled U.S. House of Representatives.
AVR is meant to help people register to vote without needing to remember to do so — states just automatically register eligible citizens whenever they first interact with a government agency, usually the Department of Motor Vehicles, though some states include other agencies as well. Say you’re in California and you’re renewing your driver’s license. Unless you opt out, you’re going to be on a voter roll by the time you leave the DMV. AVR proponents say this can inject thousands of new voters into the electorate and help achieve near-universal voter registration.
That logic assumes, though, that being unregistered is the main thing that stops some people from voting. There’s another option: that unregistered people are mostly those who were never going to vote anyway. If that’s the case, turnout wouldn’t change much no matter how many more people got registered.
To find out whether states that automatically register voters saw an increase in electoral participation, FiveThirtyEight collected registration and turnout statistics from all eight jurisdictions that implemented AVR in time for the 2018 general election (Alaska, California, Colorado, the District of Columbia, Georgia, Oregon, Rhode Island and Vermont). To our knowledge, this is the first time this data has been made public for multiple states, and therefore the first opportunity to see how registration trends changed.
Our analysis shows that people who were registered through AVR do vote — but not necessarily at the same rate as those who register themselves. Let’s run through the numbers:
- From the day they first implemented AVR through their 2018 voter-registration deadline,2 those eight places automatically registered around 2.2 million new voters. (Although, as we’ll talk about in a minute, some of those voters surely would have registered without AVR.)
- As many as 6 million existing voters who interacted with a government agency had their voter registrations automatically updated — for example, by replacing an outdated address.3
|Jurisdiction||start of AVR data||New AVR Registrants||AVR updated registrations|
|Alaska||March 1, 2017||46,088||321,303|
|California||April 1, 2018*||624,871||1,607,328|
|Colorado||March 1, 2017||240,968||634,369|
|D.C.||June 27, 2018||7,685||40,505|
|Georgia||Sept. 1, 2016||702,635||2,335,739|
|Oregon||Jan. 1, 2016||516,695||Up to 907,494**|
|Rhode Island||June 11, 2018||6,835||37,810|
|Vermont||Jan. 1, 2017||26,852||9,696|
- In the four jurisdictions that reported turnout among the newly registered voters, between 42 percent and 54 percent of them cast a ballot in 2018.4 Voters who registered through a means other than AVR turned out at rates between 46 percent and 76 percent in these places, though some of these people did have their registrations updated through AVR.
- In Colorado, Oregon and Vermont, overall turnout was significantly higher than turnout among people who were registered automatically, which means people who registered themselves were much more likely to vote. However, in Rhode Island, the AVR and overall turnout rates were about the same. And in the District of Columbia, turnout was actually higher among those who had been automatically registered than among the general population.
|Jurisdiction||AVR Voters||AVR Turnout Rate||Overall Turnout Rate*|
|Vermont**||2,058 to 11,754||8 to 44||57|
Now for the caveats. A kind of statistical counterfactual haunts these numbers: We can’t know how many of these voters would have registered (or updated their registrations) without AVR. For one thing, new eligible voters are created every day as people turn 18 or become citizens, and many of them could be captured by AVR before they get around to registering themselves to vote. And even those who were previously eligible might have added themselves to the rolls without AVR. Take, for example, someone who just moved to Colorado or a past nonvoter who’s fired up about the 2020 race: They both may have planned to register themselves but the DMV got to them first. And it’s possible that AVR registrants who would have registered even without AVR wound up turning out at higher rates than those who wouldn’t have registered otherwise, which would drive up turnout numbers for the whole group.
But based on what we do know, AVR appears to deserve at least some of the credit for this boost in civic participation. For example, according to Colorado state election officials, the state saw a big increase in the number of people registering to vote at the DMV after spring 2017, which they attribute to the implementation of AVR. A report by the Brennan Center for Justice, an advocacy group that supports AVR (among a number of other causes), found that AVR was directly responsible for increased registration rates in several states. And then there’s the behavioral economics of it all. Reed College professor Paul Gronke told FiveThirtyEight that social science research has generally found that an opt-out system (like AVR) is more effective than an opt-in one (like having to actively register yourself).
The story, though, doesn’t end there. Each state implements AVR a little — and sometimes a lot — differently. What makes the system work better in some places than in others? It’s a big question for voting-rights advocates as they push to export the most effective versions of AVR to even more states.
How to register as many voters as possible
On March 16, 2015, Oregon Gov. Kate Brown made history by signing the nation’s first AVR law — and also, apparently, one of the most successful. According to our data, Oregon registered the most voters per capita per day5 of any jurisdiction in our sample. The state has gone from 73 percent of eligible voters being registered at the time of the 2014 election to 90 percent for the 2018 election.
Oregon’s success may be at least partly credited to the fact that AVR registrants are given the option to opt out not at the DMV, but in a notice mailed to them after the fact6; voters have 21 days to return the notice and cancel their registration. By contrast, most other states ask people if they want to opt out at the time of the transaction itself.
This kind of delayed system is controversial. Many AVR supporters prefer the back-end system — where voters are sent an opt-out notice after their transaction is over — because they believe it results in more registrations. Jeanne Atkins, who was secretary of state when AVR first passed in Oregon, said that a system like Oregon’s also minimizes the burden on DMV employees, who would otherwise have to walk voters through more of the registration process.
But others, such as election-law expert Christy McCormick, prefer a front-end system — where voters are asked if they want to opt out at the time of the transaction — because it puts the agency in voters’ hands. For example, what if an Oregonian wants to purposely stay unregistered for political or religious reasons? He might miss the mailing and get registered against his will. For some people, automatic registration could even be a safety issue. Voter registration records (including home addresses) are public information in some states (including Oregon), and although Oregonians can apply to keep their information confidential, someone who gets registered without realizing it and then misses the opt-out notice might not know that their address was publicly available — which could be dangerous for victims of stalking or those escaping domestic violence.
But registering voters is only half the battle. Oregon’s system hasn’t translated to a high number of new votes. Only 42 percent of automatically registered Oregonians turned out in the 2018 election — well below the 70 percent of all Oregon registered voters who returned a ballot. That means Oregon was either the worst or second-worst jurisdiction of the five we examined at converting new registrants into voters.7 Oregon struggled with this in 2016, too: 47 percent (118,494 out of 251,209) of automatically registered voters that year cast a valid ballot,8 whereas 80 percent of all registered voters did.
It’s easy to venture a guess about why the numbers are low: By definition, AVR-registered voters are those who were previously unregistered, suggesting many aren’t all that interested in politics, or at least aren’t already in the habit of voting. But not all places have such low AVR turnout rates.
How to get new voters to turn out
In the District of Columbia, only 46 percent of all registered voters turned out in 2018, but 54 percent of automatically registered voters did. And in Rhode Island, the difference between the overall turnout rate and the AVR turnout rate was less than a percentage point. What did these places do differently?
Councilmember Charles Allen, who introduced D.C.’s AVR law, credits the city’s high AVR turnout rate to the fact that the district sends postcards to every registered voter just before an election to remind them to vote. And in Rhode Island, a DMV employee verbally asks customers if they want to opt out, which Atkins theorized may make those who don’t opt out more likely to know they are registered — and therefore more likely to vote — than under a system like Oregon’s.
However, AVR advocate and progressive activist Sean McElwee disagreed that a one-time bureaucratic interaction is enough to get someone politically engaged; he believes campaigns are the best way to activate voters. Indeed, one way that AVR can actually lead to higher turnout is that automatically registered voters’ data is more up to date, making it easier to contact them and urge them to vote. Thus, AVR turnout may be a better measure of how successful campaigns were at targeting the pool of automatic registrants than how successful the government’s implementation of AVR was.
It could also be that D.C. and Rhode Island just got lucky. Other states with front-end opt-outs (notably, Colorado and Vermont) did not see high AVR turnout rates,9 and most AVR jurisdictions mail new registrants something — either an opt-out form or a simple reminder — saying that they are now registered (though they do it immediately after the transaction, not necessarily right before the election).
To that point, Whitney Quesenbery, the co-director of the Center for Civic Design, an organization that consults on the usability of forms and systems related to the voting process, offered one last theory for D.C. and Rhode Island’s high turnout rates: Both places rolled out AVR during the summer of 2018. That means that everyone who was automatically registered in those jurisdictions went through the process within a few months of the election. Accordingly, all new voters there received two or three reminders of their registration status in the thick of campaign season — first, when they registered at a government agency; second, when they received confirmation of their registration in the mail; and, in D.C., third, when they received the usual pre-election postcard.
What’s the data not telling us?
Most of these numbers tell the story of AVR’s successes, but not its failures. California stands as a cautionary tale of how AVR can create serious problems when government agencies are incompetent or unprepared. There, the legislature dropped AVR in the lap of an understaffed, overcrowded DMV with an outdated computer system the department described as “a 40-year-old dinosaur,” and the state rushed to get AVR in place before the June 2018 primary. Not surprisingly, the rollout was a mess, exposing the system to foreign hackers, glitching out on users and failing to register some people in time to vote. Users complained that the interface was confusing, and DMV employees were not properly trained to answer their questions.
An audit found that the system created nearly 84,000 duplicate voter registrations in its first five months, which county election officials had to laboriously fix one at a time. More than 171,000 people were not registered with the political party they should have been. Worse, the system introduced inaccuracies into 23,000 people’s voter records, including 4,600 people who opted out but were registered anyway. Most seriously, the state admitted in October 2018 that employees making sloppy data entries had wrongly registered an additional 1,500 people — including at least a few noncitizens.
“While California’s system design and rollout was particularly mismanaged, it demonstrates the problems that [AVR] risks generating,” Lisa Dixon, the publications director for Lawyers Democracy Fund, an election-integrity advocacy group that’s opposed to AVR, wrote in an email to FiveThirtyEight. Basically, any time you give more agencies access to the voter database, more agencies could screw it up. And in addition to the California fiasco, Lawyers Democracy Fund says there are plenty of other reasons to oppose AVR. For one, it can bloat the voter-registration rolls, which can slow down voter processing and make for longer lines at polling places. (Although it’s worth noting that having inaccurate or out-of-date voter records, which AVR can help remedy, also creates problems at polling places.)
Automatic voter registration began as a way to get more voters on the rolls — and the data we have so far suggests it has succeeded. However, we should be careful not to overstate the case for AVR. Doubtlessly, at least some of the people who registered and voted under AVR would have done so in its absence. And there are still valid reasons to oppose AVR, as it can cause serious problems in addition to its benefits.
And in their own ways, both sides of the debate agree that AVR is not a cure-all for the problem of low turnout. As our data shows, many people who were automatically registered still did not vote in the 2018 election. Organizations like campaigns still have to finish the job.
Including some that already have AVR, as part of a never-ending quest to perfect the policy.
Except in California, where data is only available by the month. Therefore, instead of the data running from April 23 through Nov. 6, 2018, the numbers we collected for California are from April 1 through Oct. 31. In addition, we may have missed some people who registered through AVR in Colorado, as the state began to roll out the program at select locations in February 2017 — before the time period covered by our data.
The exact number is probably slightly less; Oregon only reports these numbers by year, so the 6 million figure includes Oregonians who had their registrations updated between Election Day and New Year’s Eve 2018.
In addition, Vermont reported a turnout rate of 32 percent among the combined population of automatically registered new voters and existing voters whose records were automatically updated. Election officials there told FiveThirtyEight they could not be more specific, but simple math reveals that Vermont’s turnout rate among only new registrants must have been somewhere between 8 percent and 44 percent.
That is, the number of AVR registrants adjusted for population and the length of time AVR had been in place. Excludes some voters registered automatically in Colorado, where the state rolled out AVR to some locations in February 2017, before the period covered by our data.
Experts I talked to were unanimous in saying that Oregon’s opt-out method produced more registrations; Colorado even recently switched from a front-end opt-out system (where you’re asked if you want to opt out at the time of the transaction) to a back-end one (like Oregon’s, where you are only given the opportunity to opt out after the fact) for this very reason. However, the Brennan Center study cast doubt on this assertion: “Perhaps surprising to some,” it found, “Oregon’s use of a back-end opt-out system does not produce higher registration rates than states that chose a front-end opt-out model.”
Of course, it still had the highest raw number of new voters, in part because its system had been in place the longest.
These are updated numbers sent to FiveThirtyEight in April by the Oregon secretary of state’s office; in the immediate aftermath of the 2016 election, Oregon reported that 43 percent of automatically registered voters had returned a ballot.
Although we can’t rule out the possibility that Rhode Island and D.C. communicated more clearly with voters than Colorado and Vermont did.
Nathaniel Rakich is FiveThirtyEight’s elections analyst. @baseballot