Close this search box.

AFL-CIO Analysis: Voter ID Confusion Cut Turnout By 35,000

Voter ID Mailer 1
The state sent mailers like this one to voters it said might lack ID

Pa. voters weren’t required to show photo ID at the polls in November, but the AFL-CIO says confusion over the law discouraged thousands of people from coming to the polls.

“We have said all along that this law had nothing to do with preventing fraud,” said Pennsylvania AFL-CIO President Rick Bloomingdale. “This law was always meant to confuse and intimidate legitimate voters, for the political advantage of the party who advocated for ID requirements. This analysis proves that even before being implemented, this law did exactly what its architects intended.”

The Pennsylvania labor union conducted an analysis of voters, comparing pre-election turnout predictions with actual voter turnout.

Analysts looked at turnout figures for roughly 758,000 registered voters that the state government determined in July 2012 may not have possessed the necessary photo ID to comply with the new law. The state came up with the names by cross-referencing registered voters with databases of those who have official ID (i.e. a drivers license).*

Then the Pa. AFL-CIO looked at the roughly 6.2 million voters who the state confirmed already had the necessary ID. They found a significant dropoff between the two groups.

The union concluded that between 35,239 and 36,613 voters who normally would have voted instead opted to stay home.

“Among voters disenfranchised in 2012, registered Democrats outnumbered registered Republicans by more than 2.5 to one,” the analysis continued.

In order to avoid skewed results, the analysis subdivided the electorate and made apples-to-apples comparisons based on each voter’s likelihood of turning out to vote. The analysis used propensity scores that were determined prior to election day. A score of 0 meant an individual was unlikely to vote at all. A score of 100 meant an individual was certain to make it to the polls.

The Pa. AFL-CIO relied on voter propensity scores from Catalist, a reputable data analysis firm that caters to Democratic and labor union clients – including President Obama’s campaign. Catalist uses public information like vote history and demographics to determine the likelihood that each individual voter will come to the polls. The calculations of likely individual voter turnout were conducted prior to the election and independently of the AFL-CIO study.

For example, the analysis looked at individuals with a vote propensity score of 10 or less and who were confirmed as having ID. They turned out at a rate of 10.3%. Next, the Pa. AFL-CIO looked at individuals with a vote propensity score of 10 or less who were on the state list of people who might lack necessary ID. They turned out at a rate of 5.5%.

Numerically, the biggest gap came between individuals who were given a vote propensity score of 40 to 49. Of that group, 36% of those confirmed to have ID came out to vote, compared to just 18.8% among those who the state said might lack ID.

Among voters with a propensity score of 90-100, the dropoff was minimal (95.8% turnout by those confirmed to have ID, 94.4% by those who might have been without ID).

The detailed analysis is here.

At PoliticsPA’s request, the Pa. AFL-CIO also provided a geographic breakdown. While Philadelphia accounted for more of the dropoff than any anywhere else, most came from Pennsylvania’s 66 other counties. Centre and Union counties were similar to Philadelphia in their rates of dropoff as a percentage of total votes. Allegheny, Columbia, Delaware and Lawrence counties also saw significant dropoff.

The new ID requirement was signed into law in March 2012 primarily with Republican support. Proponents said the measure would deter voter fraud, although no cases of in-person voter fraud have been reported in Pa.

The state began an ad campaign informing the public would need to show ID to vote, including television ads and mail pieces. Drivers licenses, military IDs, some college IDs and other forms of ID were to be accepted.

But a state judge delayed the ID requirement in October, saying it wouldn’t be in effect for the November election. The state did not advertise the fact that the requirement had been suspended – although some Democratic campaigns did. Those campaigns sought to rally supporters in opposition to the law, characterizing it as an effort to disenfranchise Democratic constituencies.

The law returned to court this month and the Pa. AFL-CIO, along with a number of other liberal groups, are actively campaigning for it to be thrown out.

Nils Hagen-Frederiksen says the Pa. AFL-CIO’s analysis is more about the court case than genuine study. He is the Press Secretary at the Governor’s Office of General Counsel and the administration’s primary spokesman on the voter ID issue.

“There is nothing ‘factual’ about this,” Hagen-Frederiksen said. “This is a play for public opinion.”

He pointed to several limitations of the study, including the fact that it is not known definitively which of the roughly 758,000 voters identified by the state actually lacked ID.*

Voters who didn’t have an individual voter propensity score from Catalist were not included in the analysis.

“It’s all based on a hypothetical, with no connection in any way to voter ID other than their speculation,” Hagen-Frederiksen continued.

“We are focused on the evidence and the law regarding Voter ID which is being debated inside the courtroom, under oath, not via press releases that are based on statistical speculation.”

Dr. Chris Borick is Director of the Muhlenberg College Institute of Public Opinion and Terry Madonna is the Director of the Center for Politics and Public Affairs at Franklin and Marshall College. Both said the Pa. AFL-CIO’s methodology appears to be good.

But Borick noted that it’s impossible to say for sure without knowing exactly how Catalist determines individual vote propensity.

“I looked over the AFL-CIO methodology and overall it looks fairly solid,” Borick said. “However the lack of specificity for their measure of propensity to vote makes it hard to evaluate the measures they employ.”

“The changes that they claim across groups look reasonable but ultimately the way that they calculate their propensity measure could affect those results.”

The results need to be divided by vote propensity because a simple comparison of the ‘had ID’ group to the ‘may not have had ID’ group may be misleading. For example, a 20-year-old college student from out-of-state is less likely to have Pa. ID. But his demographics indicate he is less likely to turn out to vote regardless of his ID status.

“I wouldn’t doubt that [Catalist] has a pretty refined measure that they use but ultimately it has to be transparent for a full evaluation,” Borick said.

Update: Another professor weighed in on the Pa. AFL-CIO’s analysis; and Catalist defended its vote propensity algorithm, which is declined to make public.

Dr. Rolfe Daus Peterson is the Associate Director of the Mercyhurst Center for Applied Politics. He said the Pa. AFL-CIO’s analysis was solid.

“They are using inferential statistics in an appropriate way in the discussion/summary. The 99% level of confidence is actually more rigorous than many social science studies use,” he wrote to PoliticsPA. “I think they are being cautious in their estimate (the real number could be higher).”

Here is Catalist’s statement:

“The Catalist 2012 National Vote Propensity Model predicts likelihood of voting in the 2012 election for all 180+ million registered voters in the Catalist national database. The model is informed primarily by vote history, and is augmented by factors like income, race, voter registration history, and consumer data. Catalist has been producing these types of vote propensity models over multiple election cycles, and they have been shown to be exceptionally accurate year after year.”

*The state’s list of people who might lack ID was not perfect. It was shown that it didn’t always account for people who used different names (eg. Mrs. Smith used ‘Margaret’ as her first name on her drivers license, but she registered to vote as ‘Peg’).

21 Responses

  1. Captdot-

    It’s not a presumption that people don’t have the “proper” ID. The state ran a comparison of voter records against the state’s database of drivers licenses and state IDs. There were hundreds of thousands of voters that did not have proper ID. The state acknowledged it in court.

    It’s a fairly straightforward matter to get the list of voters without state ID and compare their voting performance in past elections, against voters with IDs and having similar performance in past elections. Then you see if the two groups have different behavior in 2012. If you see a statistically significant drop off in the voters without ID, then you can infer and approximate how much voter turnout was depressed among this group.

    That’s how it’s done. They can’t call in hundreds of thousands of witnesses into court to ask them why they didn’t vote.

    However, they have called in representative witnessed who’ve testified about the different type of hurdles they encountered attempting to get a photo ID. These people demonstrate, in a different way, the disenfranchisement to be expected if this law was allowed to be in force.

    The 35,000 statistic is meant to demonstrate the effects of the confusion the State purposely generated with their ad campaign, in a attempt to suppress voter turnout.

  2. For all the people posting criticism, at some point you need to actually say if there is something about the analysis that you disagree with? Like in terms of the methodology or statistical analysis or the data sets?

    So far nobody has specified anything they find wrong with the study, and questioning the source or simply calling it ‘garbage’ or talking about leprechauns does not count.

  3. OK…here we have a “study” reported by the union, done for a union by a firm that caters to Democratic and labor union clients. There is a presumption that folks don’t have proper ID. How can anyone actually put any stock in this “study”. My liberal friends, please go back to your college research notes and understand that there is nothing scientific nor grounded about this “study”.

  4. Armchair-

    How many cases of in-person (as opposed to absentee ballot) voter impersonation fraud occur each year?

    The State itself has admitted that hundreds of thousands of voters do not possess the IDs that they claim are necessary under the new law.

    The ad was not only misleading (implying that people without ID could not vote), it was deliberately so.

    The update to this article makes mention of the higher statistical confidence level of the result. 99% is pretty significant.

    To put it in perspective, when they poll 500 or a thousand voters and issue a poll indicating a candidate favored with 57% +/-3% with 95% confidence, that means that there is 95% chance that the actual result will be in that range (54% to 60%), and only 5% chance it will be outside the range.

    So, for the 35,000 voters indicated by this study, the number is likely a good approximation, and the further you get away from it, you enter the range of 1 in a billion (or a trillion) chance the result is that far off.

    For the critics of the study, make your own statistical estimates. (However, I imagine that the vote suppressors have already done this and won’t dare show the results.)

  5. David Diano: 50/50 on the day. Right on Sestak; Wrong on statistics and science.

  6. Why is anyone disputing the findings of this study? Didn’t the PAGOP’s chief RACE HUSTLER Rob Gleason, basically verify this study?

  7. Tom-
    All your written words refer to made up numbers and unicorns and leprechauns. There is nothing “real” in your premise.

    Consider the size of the sample, and the methodology described, the numbers aren’t being “made up”.

    You can quibble with particular details of the formula to calculate the likelihood of people voting, but identifying likely voters isn’t rocket science given voter history and age. The model can be tested (and probably was) tested against previous elections and turnout relative to voter history.

    Dismissing a statistical study as “hypothetical”, misrepresents the nature and value of such studies.

    If a statistical study of Presidential turnout, based on propensity, predicted 75% turnout, it wouldn’t be dismissed as “hypothetical” with the other side arguing for 5% turnout. Someone might argue for 72% or 77% with similar formulas, but if the model doesn’t match past elections and common sense, it’s not going to hold up statistically.

    So, Tom, take the unicorn horn (or your head) out of the appropriate orifice, and stop showing the ignorance of a climate change denier.

  8. My premise is just as real as yours.

    I am also talking about your inability to process the meaning of a written word.

  9. Tom-
    I’m talking about voters being disenfranchised through deception and misinformation. You are talking about unicorns.

  10. So having less fake unicorns who are GOP supporters disenfranchised would be bad for the GOP? Terrible joke makes terrible sense.

    Do you know how to read?

  11. Tom-
    Well, that would hurt the GOP, as unicorns (with rainbows flowing from their behinds) are members of the GOP’s fantasy land.

  12. Voter ID isn’t about statistical studies. Voter ID is about Rob Gleason’s PAGOP doing some race hustling and Rob Gleason is PA’s # 1 RACE HUSTLER!

  13. True it is a bigger number, similar to past studies done by the AFL CIO in particular those calculating the ratio of unicorns to leprechauns.

  14. This number is a lot bigger than the number (zero) of voter impersonation fraud cases the PA GOP has found.

    But, my real question is: how many voters did the PA GOP expect to suppress with their confusing/misleading “show it” ad? Is 35,000 close to their statistical target?

    Of course the number would have been 10 times as high if the law had been in place.

  15. This is an absurd “study”. Without knowing the margin of error between expected turn out (using a propensity model) and actual turnout over an extended period, this number is meaningless. 35,000 people represent 0.6% of the total votes cast in the 2012 Presidential ballot. I’d bet that the margin of error using propensity to vote is over 2.5%, ie these results have zero statistical significance.

  16. This drivel wouldn’t make it past the first peer review of the shittiest statistics journal.

  17. The PAGOP under Rob Gleason is a travesty. It wants to reduce participation in voting so that it can game its dwindling base more power in this state than their numbers really deserve. The PAGOP is no longer the party of HOPE or OPPORTUNITY! It’s only appeal is to the worse instincts that we humans have: The Party is run by too many aged white males who should be retired TAKERS, but find it is too hard to leave the game, when it pays even when the Party and the State loses. When is the PRINCE OF DARKNESS himself, Brabender going to speak about the joys of Voter Suppression? Are the continued crys of disdain about Obamacare, really about having good healthcare policy in this state and nation or is it really a 2013 warning cry of Nigga, Nigga, Nigga: the PAGOP’s continuing reminder to White Folks about how much the PAGOP loves them and is protecting them from African-American parasites.

  • Do you agree that ByteDance should be forced to divest TikTok?

    • Yes. It's a national security risk. (60%)
    • No. It's an app used by millions and poses no threat. (40%)
    • What's ByteDance? (0%)

    Total Voters: 30

    Loading ... Loading ...
Continue to Browser


To install tap and choose
Add to Home Screen