Facile Turnout Stats on Voter ID: Wrong, the First Time

A simple before-and-after turnout test won't ever be up to the task. It's bad math that leads to bad legal analysis. Think of it as cotton-candy statistics: It looks tempting, at least at first, but there's absolutely no substance to it. It should not be used to gum up the conversation.
This post was published on the now-closed HuffPost Contributor platform. Contributors control their own work and posted freely to our site. If you need to flag this entry as abusive, send us an email.

Judge Richard Posner and Justice John Stevens wrote the 2007 Court of Appeals' majority opinion and 2008 Supreme Court plurality opinion, respectively, upholding Indiana's strict photo ID law against challenge. Their recent public musings about the merits of the dissenting opinions in those cases are sufficiently unusual to have provoked a flood of commentary.

One of these commentaries stands out. Hans von Spakovsky, who has served as a local election official, at the FEC, and at the Department of Justice, joined the mix again last week. In a piece titled "Right the First Time," Mr. von Spakovsky defends Judge Posner's original opinion upholding the ID law.

His primary argument ridicules the notion that ID has stopped some voters from casting their ballots, by pointing to Indiana's consistent turnout gains since the law was implemented. Indiana's law was implemented in 2006. But turnout increased 2 percent from 2002 to 2006 (including in counties with large minority populations), increased 8 percent for Democrats and 5 percent for black voters from 2004 to 2008, increased (including for black voters) from 2006 to 2010, and increased again for black voters from 2008 to 2012. Therefore, he claims, Indiana's ID law can't possibly have hurt voters, particularly minorities.

I don't know if Mr. von Spakovsky will talk about Kansas: After Kansas implemented a strict ID law in 2012, black turnout dropped by 2 percent, and Latino turnout dropped by 21 percent.

Conclusions about the role of ID from either set of numbers are, of course, nothing but garbage. They should fail Statistics 101 at any school in the country.

There are arguments to be made in favor of and against different kinds of ID laws (I've not been shy about where I think the balance lies for the strictest versions). There are disputes to be had about the empirical evidence of their impact, on populations including (but not limited to) racial and ethnic minorities. There are concerns to be addressed about confidence levels in turnout stats for small subpopulations (spoiler alert: the purported gains for Indiana's black voters are all within the margin of error). And there are questions to be raised about whether turnout is even the right measure of impact at all, or whether we should also be concerned about the 40 percent of eligible Americans who haven't generally voted.

But Mr. von Spakovsky's numbers are in a category of their own. He keeps trotting out (for example, here, and here, and here, and here) before-and-after turnout as evidence that ID works: Because turnout went up after ID rules were implemented, ID rules can't possibly be a problem.

This argument in "Right the First Time" was wrong, the first time. And it has been consistently wrong since.

Let me see if I can make the problem clear with an analogy. Picture the typical movie scene where a medieval army storms a castle. An attacking army runs at the walls. The defenders rain down flaming arrows and burning oil from the ramparts. And despite those efforts, the attackers eventually breach the stronghold.

Before flaming arrows and burning oil, some people got to the castle walls. After arrows and oil, even more people got to the walls. Which means nobody got hurt by the arrows and oil, right?

Doesn't make sense.

Or another: A fire is raging on a surburban block. The firefighters arrive. But there is even more fire damage after several more hours. So the firefighters had no impact?

The problem is that before-and-after figures alone don't tell us what we want to figure out. To know what work X is doing, we want to know what things would have looked like if everything other than X stayed the same, and X disappeared. To see how much better or worse things would have been with no arrows, or firefighters... or ID laws.

That's particularly hard to do when lots of things are changing at the same time.

Let's take 2004 to 2008. Mr. von Spakovsky says that turnout increased 5 percent for black voters in Indiana. What changed from 2004 to 2008? ID rules, yes. But also:

•The identity and quality of various candidates
•Competitiveness of various races
•The composition of the Indiana electorate
•Tactical and strategic shifts in two major wars
•The local economy
•The national economy
•The nature and placement of local ads
•Umpteen other election rules
•The location and staffing of local precincts
•The weather on Election Day
Indiana sports teams' records

And, of perhaps particular salience to black voters in Indiana, for the first time ever, an African-American candidate led a major-party ticket. That probably mattered.

And for the first time in a long time, the presidential campaigns made Indiana a battleground state, pouring unprecedented resources into mobilizing voters in the state. That probably mattered too.

Some of these factors likely caused turnout to change in big (indeed, very big) ways. Some of them likely caused turnout to change in small ways. Some may not have caused turnout to change at all.

But you can't know the impact of one changing part without filtering out the others. Maybe turnout would have gone up 3 percent without the new ID rule, and Hoosiers were so overjoyed by the new security measures that the ID law boosted turnout by another 2 percent. Maybe turnout would have gone up 7 percent without the new ID rule, and the ID rule blocked turnout by 2 percent. Or 5 percent. Or 10 percent. If you've only got before-and-after figures and don't account for any other context, it's completely impossible to tell.

After lots of elections have been held in lots of states, it may be possible to use advanced statistical tools to filter the turnout effects of strict ID rules from all of the other impacts going on. But we don't have enough information to draw these conclusions yet.

And the simple before-and-after turnout test won't ever be up to the task. It's bad math that leads to bad legal analysis. Think of it as cotton-candy statistics: It looks tempting, at least at first, but there's absolutely no substance to it. It should not be used to gum up the conversation.

Justin Levitt, at Loyola Law School, is an election law professor and national expert on the law of democracy.

Popular in the Community