Nate Cohn has a piece in the New York Times today in which he writes that the backlash to the new voting restriction law in Georgia is dramatically overstated because the partisan effects of those restrictions is minimal.

The latter clause of that sentence is true: studies of most voting laws in the past decade have not shown large, meaningful effects on turnout or on the partisan makeup of people who end up voting. The context of the election is generally much more important for driving government turnout. Here is Cohn’s recap of the evidence:

There’s essentially no evidence that the vast expansion of no-excuse absentee mail voting, in which anyone can apply for a mail absentee ballot, had any discernible effect on turnout in 2020. That shouldn’t be a huge surprise: Even universal vote by mail, in which every registered voter is automatically sent a mail ballot (as opposed to every voter having an opportunity to apply for one), increases turnout by only about 2 percent with no discernible partisan advantage.
Believe it or not, turnout increased just as much in the states that didn’t have no-excuse absentee voting as it did in the states that added it for the first time. Similarly, Joe Biden improved over Hillary Clinton’s performance by three percentage points in the states that added it, compared with 2.9 points in the states that did not.
A more rigorous study by political scientists at Stanford found that no-excuse mail voting might have increased turnout by a whopping 0.02 percent in the 2020 election. The study used a novel approach: The researchers compared the turnout among 65-year-olds in Texas, who were eligible to vote by mail without an excuse, with 64-year-olds in Texas, who weren’t. The turnout among 64-year-olds was indistinguishable from that of 65-year-olds, even though the latter group voted by mail in large numbers.
Like Georgia, Texas did not require an identification to vote by mail, but has a strict ID requirement for in-person voting.
The partisan makeup of the electorate didn’t appear to change, either. The Democratic share of voters appeared to tick up by two-tenths of a percentage point — enough to decide a very close election. But it’s also so small that it could just be statistical noise, with no effect at all. Social science methods just don’t offer the level of precision necessary to nail down whether this, or any, change might move the needle by a tenth of a point.’

This is a fair representation of the orthodoxy in political science, as far as I can tell. I have written similar pieces on the subject over the past couple of years, and covered the Stanford study Cohn refers to. He also posits some theories as to why a decrease in the convenience of voting might not decrease turnout, on average. They are worth reading.

In my view, Cohn’s factual discussion is all fine and unobjectionable on empirical grounds, except for one caveat. The study of absentee voters in Texas has actually caused a bit of grumbling among political scientists. The study worked by comparing absentee voting among treatment and a control groups — voters on either side of the 65-year age cutoff for eligibility in Texas. But, as Charlotte Hill, a PhD candidate in voting and elections at Berkeley points out, the control group is already made up of high-propensity voters. This creates a ceiling on the turnout effect you'd find among those voters, compared to the residual effects of increasing turnout that you might find with lower propensity voters. Charlotte goes into a few other examples of contrary evidence that doesn’t get mentioned in the piece here.

Aside from this, there are two broader issues that I want to raise. First, I think it is improper (or at least very messy) to project these conclusions onto the voting restrictions in Georgia. And second, I think the context, in this case, matters more than the factual debate over voting mechanisms. Republicans, after losing an election in Georgia due in large part to the mobilization of black voters, are now passing voting restrictions that could plausibly have a disproportionate impact on their ability to vote in the future. The fact that a discussion of this context is missing from Cohn’s piece is a rather large error of journalism.

Editor’s Note: This post was originally exclusive for paying subscribers. I ungated it for a public audience because of the magnitude of interest online. If you found it informative, consider sharing it and signing up for paid posts by hitting the buttons below:

Subscribe now

Share

What do you do when you don’t have all the data

First, one has to engage in a good bit of extrapolation of the political science consensus to apply it to Georgia’s new election laws> This is not nearly as black and white as it would seem.

While the most relevant political science evidence looks into the effects of no-excuse absentee mail voting and finds small effects, other types of restrictions/expansions do demonstrably affect turnout. The adoption of all-mail voting in west-coast states, for example, has been associated with a significant increase in the share of voting-eligible adults casting ballots.

There is also a strong relationship between mail-ballot usage and turnout, implying that the studies that put states into buckets based on their laws might miss nuances in the voting laws. Some states require absentee voters to get signatures from witnesses or a certified notary before they can send their ballots to the clerk of elections — a much higher burden of entry than in states that just ask you to sign the ballot and send it in. No study has grappled with this fact, which might explain why turnout changes were higher from 2016 to 2020 in states with less restrictive mail-in voting laws.

There are also (at least) three practical statistical issues with most of these studies that are conducted on aggregate data during only one or two election cycles. First, there is no ability to observe the counterfactual. Given the many confounders in predicting voter turnout and electoral competition, many studies — especially observational, correlational ones — are going to be underpowered. For example, Nate Cohn says that Virginia had the largest decrease in the cost of voting index, one measure of how convenient it is to vote, in the run-up to 2020 but didn’t see turnout increase more than other states. This is fine to point out, but I can think of many issues with the relationship — from the way the index is constructed, to the disparity in the enthusiasm of Northern Virginians, who lobbied for easier laws, and others — that might make the implications messier than they appear on paper.

Second, the relationship between voting laws and turnout in non-southern states might be less relevant to the dynamics of voting in the south. Non-white voters suffered most of the consequences from cutting polling places and resources in 2020 — would their turnout have been higher if those issues were fixed alongside increasing mail-in voting? There’s also more tension between these numerous minority groups and the Republican-dominated state legislature in the south than there is in other states. It is reasonable to expect there could be disparate impacts here, especially given the incompleteness of the data.

Finally, given the inability to control for a lot of these confounders, there’s a real risk that what has been true before won’t be true in the future. In Georgia’s case in particular, the new laws seem to place a higher burden both on voting by mail and submitting your ballot via dropboxes, which will now only be available during business hours. If a study hasn’t observed what happens when you making mail-in voting harder alongside the alternatives, how useful can the literature really be? There are tons of similar concerns here, from the laws preventing handing out water to people standing in hours-long lines to the inability to cast your ballot in-person at the wrong precinct, which roughly 10,000 Georgians did last year.

In a state that was only decided by 12,000 votes, all these “small” effects could easily add up. That’s true even if they’re as small as the political science says, which we’ve determined might not be the case in the future.

Voting restrictions are like, normatively bad, man

Lots of people are forgetting the context of Georgia’s new election law. This, I think, is their biggest mistake.

Perhaps the most important change from a security/suppression standpoint is that the state legislature will now play a larger role in election administration. The law kemp signed removes the secretary of state as chair of the state board of elections and allows the Republican-dominated legislature to appoint a majority of the board’s members, The state board can also now take over county boards of elections, if they decide to, meaning the state legislature has control over how elections are conducted at the local level.

Of these laws, Cohn writes that they “might prove to be very important. But for the purposes of this article, we are not considering them “voter suppression” provisions. They do not inherently make it harder for people to vote by restricting whether or how they can vote.”

I find this more than a bit obfuscating. It also illustrates the problem with the focus on the data when discussing voting rights. By conditioning on the legislature acting in good faith, analysts can discuss the impacts of laws as if changes are happening in a vacuum. But our prior should be that the legislature won’t act in good faith, given how Republicans behaved in 2020. The removal of the secretary of state from overseeing the elections can be understood as little more than a direct attack against the current SoS Brad Raffensperger, who stood in Trump’s way and worked tirelessly to uphold the integrity of the election.

Sticking to the political science evidence allows us to ignore the context of the data. In Cohn’s case, it also lets him ignore the backlash to his article. The problem with the Georgia law is not that it might decrease turnout or help Republicans, it’s that it was clearly intended to do so. The backlash to the clear attempts at voter suppression are being driven at least as much by the ethical and moral concerns of limiting political participation as it is by concerns that they might turn the state red by limiting the franchise. That, by the way, seems totally plausible to me: in close elections, small changes can make big differences.

The relevant point is not a contrarian take about political science. That can be situated within the broader conversation, but it should not be allowed to derail the whole debate. Rather, the relevant critique is that Republicans, in response to Donald Trump’s falsified claims of voter fraud in the 2020 election, are trying to wield their power over who gets to vote to shape the electorate into a more politically favorable group of voters. In a democracy — especially in the United States, which has had no cases of widespread voter fraud but a legacy of disenfranchisement of black voters— that is a clear ethical violation of our social contract and the principle of one person, one vote.

Should the response to attempted voter suppression really be all that different just because the laws might have a smaller-than-expected chance of succeeding?

Share

Politics by the Numbers

The focus on electoral math obscures the…