Skip to main content
Menu
What We Know About Voter ID Laws

At a rally last month, President Trump endorsed voter ID laws, saying this to the audience: “Only American citizens should vote in American elections. Which is why the time has come for voter ID, like everything else. You know, if you go out and you want to buy groceries, you need a picture on a card.”

Many jumped on the head-scratching line about groceries, but the more significant piece of the president’s speech may have been his unfounded implication that many noncitizens are voting. Trump’s comment is likely to add fuel to the long-running debate among citizens, elected officials, courts and researchers about the motivations behind and impact of voter ID laws. With some credible new evidence on these questions from political science — and the midterms fast approaching — it’s a good time to take stock of what we do and don’t know.

Estimating the effects of voter ID laws is a tricky business, but the most credible estimates suggest the laws’ turnout effects haven’t been large enough to swing many elections.

First, while voter ID laws are often discussed as a single class of laws, they differ in subtle but crucial ways. Different laws require different kinds of ID, for example. In Wisconsin and Virginia, only a photo ID will pass muster. In Ohio and Arizona, identification without a photo — like a utility bill — will suffice. States including Texas don’t allow student IDs, a policy that clearly hinders younger voters.

Second, if voter ID laws reduce turnout, they are likely to do so both by preventing some people who show up at the polls from voting and by deterring others from showing up in the first place. While we can sometimes count the number of people who are turned away at the polls, we are left to infer the laws’ deterrent effects. Also, it’s difficult to untangle whether a given change in turnout was the result of the law itself or some other state-specific factor. Turnout was down in Wisconsin in 2016 relative to 2012, for instance, but was that because of the state’s voter ID law, or was it because of the absence of President Obama on the ballot, changes in mobilization by the political campaigns or other factors?

So, when we evaluate research on voter ID laws, it’s critical to assess the strength of the underlying research design. Summarizing a wide range of studies in a 2017 review, Benjamin Highton concluded that “a small number of studies have employed suitable research designs and generally find modest, if any, turnout effects of voter identification laws.”1 Some of the newest evidence reinforces that conclusion and uses high-quality administrative data that can address at least some of the problems that have bedeviled prior studies.

For example, Bernard Fraga and Michael Miller have a new working paper that examines Texas in 2014 and 2016. In 2014, Texas had a strict law that required voters to use specific forms of photo ID. In 2016, however, a federal court forced Texas to provide an alternative way to vote for those who showed up at the polling place lacking appropriate ID. As a result, voters in 2016 could file a “reasonable impediment declaration” and then vote after providing a different form of ID such as a utility bill or a birth certificate. Fraga and Miller collected those declarations and matched them to the Texas voter file, allowing them to characterize the group of more than 16,000 people who would have been prevented from voting under a strict voter ID law.

That’s a small slice of the electorate in a state in which nearly 9 million ballots were cast — though admittedly, this estimate is a lower bound: There were many voters who were unlikely to know about this alternative channel. But this research design is valuable because it enables individual-level characterizations of precisely who would be barred from voting under the strict policy in place in 2014. Which brings us to something else we can say confidently about voter ID laws:

Voter ID laws disproportionately disenfranchise minority communities.

Fraga and Miller found that black voters constituted 11.4 percent of those voting in Texas in 2016 with ID but 16.1 percent of those voting without ID, which shows clear evidence of a disparate racial impact. Likewise, Latino voters made up 19.8 percent of those voting with an ID but 20.7 percent of those voting without an ID. So even if voter ID laws haven’t swung election outcomes, they can deny thousands of people their right to vote — denials that fall disproportionately on black and Latino citizens. Whether voter ID laws swing elections is far from their only important consequence.

Those disparate impacts are clear from a second newly released study, too, which also used individual-level records to provide a more granular view of precisely who is affected by voter ID policies. In Michigan’s 2016 general election, voters who arrived at the polls without ID were able to vote after they signed an affidavit. Researchers Phoebe Henninger, Marc Meredith and Michael Morse2 collected these affidavits to identify a set of voters who would have been turned away under a stricter policy, like the laws in Georgia, Virginia and Wisconsin. By their calculation, about 28,000 voters — or 0.6 percent of 2016 Michigan voters — lacked photo identification.

Those 28,000 voters were more nonwhite and more Democratic than the Michigan electorate overall. Henninger and her co-authors estimated that nonwhite voters were between 2.5 and 6 times as likely as white voters to lack voter ID. And while Michigan doesn’t record partisan registration, the researchers’ model-based estimates suggest that more than 70 percent of those filing affidavits would be Democratic primary voters.

Older voters can be affected by voter ID laws, too.

In another study, this one published in 2017 by the Journal of Empirical Legal Studies, I teamed up with Meredith, Morse, Sarah Smith and Jesse Yonder to estimate the effects of a policy change in Virginia between the 2013 gubernatorial election and the 2014 midterm election. Virgina went from a law that required identification but accepted IDs without photos to a law that demanded specific forms of photo ID.3 In 2014, there were just 474 provisional ballots cast statewide for lack of photo ID, which is a small fraction of the 2.19 million voters who voted in the election.4 So implementation matters: That muted impact may have been partly the product of a statewide mailer telling registered voters without driver’s licenses about the new law.

Still, the provisional ballots were more common in precincts where there were more voters without a driver’s license and more voters over 85. That makes sense: Those are places where fewer voters are likely to have the requisite photo ID. As Matt Barreto and his co-authors have found in multiple surveys, elderly voters — like black and Latino voters — are more likely to lack photo ID than the voting population overall. In fact, in the Virginia study, the share of voters in a precinct who were 85 or older was much more strongly associated with the percentage of provisional ballots cast by voters lacking ID than were other demographic factors, such as the share of voters who were black or Hispanic. That, in turn, leads to another observation about voter ID laws:

The long-term effects of voter ID laws are likely to differ from the short-term effects.

As Nicholas Valentino and Fabian Neuner found via survey data, voter ID laws are likely to produce a strong emotional response among Democrats and thus lead to a counter-mobilization that galvanizes them to vote, at least in the short term. But that may fade, along with the one-time effect of states informing voters about the laws as they are being phased in.

Over time, the partisan effects of voter ID laws have the potential to shift as well. Some age cohorts are reliably more Democratic or Republican than others. The silent generation, which has been pro-Republican relative to the generations before and after, is now filling the ranks of America’s oldest voters. As members of that GOP-leaning generation stop driving, they will be less likely to have photo IDs, and so more likely to be affected by voter ID laws. It’s plausible, then, that the partisan impacts of voter ID laws may shift to affect more GOP voters. The same would be true if the leanings of black or Hispanic voters shifted.

While the impact of these laws is debated, the intent is clear.

In 2012, a Republican legislative leader in Pennsylvania made headlines by saying that the state’s voter ID law — which was later overturned by the courts — was “gonna allow Governor Romney to win the state of Pennsylvania.” But even when the law’s authors are more circumspect about their motivations, the evidence is clear: It’s Republican legislatures and legislators that tend to pass them. There’s also a racial dynamic: Seth McKee found that Republican legislators are more likely to back voter ID laws — and Democratic legislators less so — as their districts have more black voters. It’s also no accident that the states whose voter ID laws make headlines — Wisconsin, North Carolina, Virginia — are often swing states with diverse electorates.

It’s important, too, to underscore that Trump promoted voter ID laws for a reason: They tend to poll reasonably well, especially among Republicans. A 2016 AP-NORC poll found that 79 percent of those surveyed favored requiring all voters to provide photo ID, with Republicans especially supportive.


It’s certainly not a consensus, but the weight of recent research suggests that even if voter ID laws have limited effects on which party wins specific elections, they still affect tens of thousands of voters in larger states, particularly black, Latino, Democratic and elderly voters. And importantly, these laws’ long-term impacts may well differ from their immediate effects upon implementation. So as politicians, lawyers and social scientists continue to debate these laws, the very effects themselves are likely to change beneath our feet.

Footnotes

  1. Highton noted, though, that strict photo ID laws are “a relatively recent phenomenon,” so it will take further election data to estimate their long-run effects.

  2. All of whom have been affiliated with the University of Pennsylvania, where I work.

  3. Those without ID could cast provisional ballots, but they were counted only if the citizen then brought an acceptable photo ID in person to the city or county election board within three days.

  4. By contrast, in 2013, nearly 500 provisional ballots were cast in Fairfax County alone, primarily by voters who requested an absentee ballot but appeared in person anyway.

Dan Hopkins is a professor of political science at the University of Pennsylvania, and his research focuses on American elections and public opinion.

Comments