True the Vote? Or Skew the Vote?

-- By Brendan Fischer and William Dooling.

MADISON -- Claims by an out-of-state Tea Party group that the campaign to recall Governor Scott Walker is fraught with error do not stand up to even limited scrutiny.

Findings released this week from the Tea Party-led "Verify the Recall" effort allege that recall proponents fell short of the 540,000 signatures necessary to recall Governor Scott Walker. However, a cursory review of the pages they allege are erroneous actually include the information they claim is missing. Signatures the groups deem "ineligible" are very clearly legitimate. Some of the problems appear to arise from data entry errors on the part of True the Vote volunteers.

"Verifying" the Recall

Since early February, groups involved in the "Verify the Recall" effort have been recruiting volunteers from around the country to enter the Wisconsin recall petitions into a massive online database. "Verify the Recall" is a joint project between the Houston-based nonprofit "True the Vote" (a project of the Texas Tea Party group King Street Patriots) and the Wisconsin Tea Party groups Grandsons of Liberty and We The People of the Republic.

On Tuesday, February 28, Governor Walker declined to officially challenge a single signature on the petitions calling for his recall, instead requesting that the Government Accountability Board incorporate the "Verify the Recall" findings. Walker only referred to the involvement of the Wisconsin groups in his filing with the Board, perhaps to downplay the involvement of the King Street Patriots and their True the Vote project, which have been accused of a variety of voter suppression tactics.

The "data" put forward by the groups involved in Verify the Recall has served as fodder for right-wing media outlets to claim the recall petition collection efforts were riddled with problems, if not outright fraud. Evidence suggests the claims don't stand up to scrutiny.

A Look At the Data

The Texas-based "True the Vote" analyzed and released the results. Their full set of data is available here.

True the Vote states that volunteers from around the U.S. entered approximately 820,000 of the recall signatures into a database developed by True the Vote, and their analysis showed that only 534,685 recall signatures are eligible, almost 6,000 short of the required 540,000 mark. Their executive summary states they only processed 138,203 of the total 152,508 petitions submitted -- meaning over 14,305 pages were never entered. True the Vote boasts in their summary that each signature line was entered an average of 2.67 times. It is not known why they entered names and signatures nearly three times each into the database but failed to input the complete data set.

The group claims that less than the necessary 540,000 signatures are eligible "based on data available." According to their summary, out of the nearly 820,000 signatures reviewed, 55,608 were deemed "ineligible" and an additional 228,940 signatures required "further investigation," leaving only 534,685 eligible.

Their assumption appears to be that every single one of the 228,940 signatures they flagged "for further investigation" will turn out to in fact be ineligible. They are also assuming that the over 55,608 signatures they deemed "ineligible" would not be counted under Wisconsin law. They are further assuming that every signature on the 14,305 pages they never counted is invalid.

Do these claims stand up under scrutiny?

"Ineligible" Signatures Appear Valid

True the Vote says around 4,700 volunteers entered the data from scanned PDFs of the handwritten recall petitions into a database. They claimed that volunteers entered signatures from each page, after which True the Vote used its "proprietary software program to examine each signature for eligibility."

A cursory review of signatures that True the Vote considers "ineligible" strongly suggests they are not counting legitimate petitions. The following examples are not isolated -- many, many similar mistakes were quickly found in the True the Vote data set:

  • True the Vote discounts the signature of Mary Babiash (page 980) because she added the state abbreviation "WI" to her zip code. Her address is otherwise correct.
  • The group calls Clifford Winkleman's signature (see page 3) ineligible because it has a "bad sign date." The date section on his petition is completed, but it looks like he pushed hard when he wrote the "10" in "01/10/2012." Linda Winkleman, who lives at the same address, signed below him, and also on 01/10/2012.
  • The group would not count Joshua Epps' signature (see page 250) because he forgot to add the year when he dated his otherwise-valid signature.
  • The signature from Tyrell Luebkes (see page 983) would not be counted because he entered his city in the "street address" section, and vice versa.
  • They would not count Cheryl L Koch's signature (see page 497), saying she had a "bad sign date" of 1/91/12 because of a stray pen stroke behind the "9" on the correct sign date: 1/9/12

The Government Accountability Board told CMD that administrative rules require that all signatures be approached with a "presumption of validity," and state law requires a review of each petition "on its face." This means they would approach errors like the ones noted above with a presumption that the signature is valid, and would look at it in context -- for example, a valid street address written in the "city" section, and a valid city written in the "street address" section, would likely not disqualify the petition.

It does not appear that True the Vote and its Wisconsin Tea Party partners conducted their "independent" review by treating each petition with a presumption of validity and giving each signature an overall facial review, as required by Wisconsin state law and administrative rules.

True the Vote appeared to mark other signatures "ineligible" because of data entry errors on the part of True the Vote volunteers.

  • True the Vote considers Heath Beacher's signature (see page 497) ineligible because "1/6/ 2012" (note space) was entered into the database, causing the software to mark it with a "bad sign date." The date on the original, handwritten petition is legible and valid. This data entry error is enough to invalidate her signature.
  • Keith Iverson's signature (see page 259) is given a "bad sign date" of "Dec 2, 2011," apparently because the volunteer entering the information into the True the Vote database wrote "Dec." rather than "12." This flags his signature as "Invalid."
  • Kelly Ullrick signed a recall petition with her full name, but the True the Vote volunteer entering her into the database (see page 984) failed to include her last name. The software tagged this as an "incomplete record," and thus ineligible, due to a missing last name, even though her last name is clearly visible on the original petition.
  • Similarly, Alexandra Aulisi's signature (see page 9) is ineligible because of a "bad sign date," after the volunteer entering data wrote the letters "ll" rather than the numbers "11" for "11/15/ll" .
  • Meghan Walsh's signature (see page 168) is deemed invalid for a "bad sign date," because the True the Vote volunteer entered "!!/!^/!!" instead of 11/11/2011.
  • True the Vote considers Jocelyn Tilsen's signature (see page 2) ineligible as a result of a "bad sign date," because one of their volunteers entered "1/1O/12" (note O) instead of 1/10/12.

This is not a case of a few isolated examples being used to refute a much larger set of data. These examples were all identified after a very quick review of the results on the True the Vote website, and very similar errors appear to mark the rest of the group's "findings." Signatures marked "ineligible" for a "bad sign date" appear perfectly admissible. Despite True the Vote's claims that "each signature record was inspected 2.67 times," even a very cursory review should have caught these errors.

Review the data for yourself here.

Many Signatures Tagged for "Further Investigation" Appear Valid

Even if the 55,608 signatures True the Vote deemed "ineligible" with questionable methodology were not counted, it would have little impact, as hundreds of thousands would need to be struck to prevent a recall.

The main reason True the Vote asserts the recall campaign was a failure is because of 228,940 signatures they marked "for further investigation." Only after discounting all of these signatures were they able to claim that the recall campaign did not reach the requisite 540,000 signatures. The group made this claim without ever asserting that the signatures requiring "further investigation" should actually be discarded. A review of the signatures apparently flagged "for further investigation" in their report reveals many valid signatures.

True the Vote says the signatures marked "for further investigation" are those "that were partially marked through, illegible, possibly false, mismatched, or otherwise compromised." In their data set, it appears that signatures marked as "questionable" are those True the Vote believes require further review. Browsing through the entries deemed "questionable," it is difficult to believe any of these signatures would be struck.

Some recall signatures appear to be deemed "questionable" for imperfect handwriting (See: signature from Kristie Wherritt page 9010). Some appear to be flagged for date corrections. (See: signature from Chandler Charles page 9151). Some appear perfect in all respects (See: signature of James Schneider page 9002, signature of Brandon Kohl, page 9151).

It seems very hard to believe these signatures would actually be deemed ineligible under Wisconsin law, or under a fair, genuinely nonpartisan review process. Again, these are not isolated examples, but appear very similar to errors in the rest of the group's "findings."

Skewing the Vote

One of True the Vote's major claims is that "Donald Duck" signed a recall petition. But it is clear that anyone signing Donald Duck is not a friend of the recall effort.

True the Vote says their "involvement in the petition review is simply to help uphold the integrity of the process." But their misleading assertions, inaccurate data, and dishonest framing do little more than undermine and cast doubt on the process they claim to uphold.


I found errors in their first release of information about the Sen. Fitzgerald recall. The GAB's PDFs online were numbered from 1 to 3963. TTV's spreadsheet of names included the page number for each signatures. I wrote a simple program to analyze which pages did not result in any signatures. VTR/TTV's spreadsheet didn't have any signatures from 1,303 petition pages. (This included at least one 50-page numbering gap. TTV released its results knowing they couldn't fetch 340 of the GAB's PDFs). TTV eventually released an update. Hundreds of pages still did not result in any signatures, even though they seemed legible. TTV but never issued a correction press release, and the partisan Wisconsin Reporter news service had already sent out a release based on the first flawed analysis.

Brendan, You are really having a problem with this and have gotten this backwards. True the Vote issued their results on 2/28 stating their belief that there were more than the 540,000 eligible petition signatures needed to initiate a recall election. True the Vote became involved with verifying the recall signatures to preserve the integrity of the election process in Wisconsin when the GAB stated that they could not review the 1.3-1.5 Million recall signatures the gatherers claimed to have gathered. True the vote was not involved because of an interest in the outcome of the recall election itself. The True the Vote review determined that the petition gatherers had grossly overstated the number of eligible signatures and did a substandard job of verifying / validating the signatures before submitting them to the GAB: 557,000 signature lines were blank, 55,000 were ineligible and another 229,000 required further review. Since True the Vote only received 138,000 of the 152,000 petitions supposedly submitted, they projected that there were likely enough eligible signatures in the remaining 14,000 petitions and/or among the 229,000 signatures requiring further review to meet the minimum level needed for a recall. I also see that you tagged an old article to accuse True the Vote of wrong doing. You should have tagged the multiple articles stating that they were cleared of these false allegations over two years ago, but then again telling the truth to your readers does not appear to be your strength.

Bwilhm, I am not sure where I am not "telling the truth" to readers. True the Vote has, in fact, been the subject of multiple allegations of voter intimidation. Those allegations were serious enough to compel the U.S. Department of Justice to send personnel to Harris County, Texas to make sure that True the Vote's poll watchers did not continue their voter intimidation tactics. Just because a Republican County Clerk in Harris County Texas refused to pursue charges does not mean they were "cleared" -- especially when his investigation consisted of nothing more than calling up polling places and asking if there was anything wrong. Besides, it certainly seems like "voter intimidation" is an appropriate term to describe sending white Tea Partiers to minority voting districts to confront African-American voters and hover behind them as they cast their ballots. As for "telling the truth," the group you are defending is the same one that doctored a photo of an African-American woman holding a sign, changing it to read “I only got to vote once.” If you really want to get into "telling the truth," just look through a few pages of the signatures that True the Vote says are "ineligible" or that need "further review." As discussed in the article, the Texas group you are trolling for is making very, very dubious claims about what their data really shows. The only group doing a "substandard job of verifying / validating the signatures," to use your words, is True the Vote. I have not seen anything from True the Vote "stating their belief that there were more than the 540,000 eligible petition signatures needed to initiate a recall election." The group's Executive Summary, which I link to above, says there are only 534,685 signatures "based on data available," which certainly seems to imply that recall proponents did not meet the requisite 540,000 signatures. Mark Antill of True the Vote acknowledged to the Milwaukee Journal-Sentinel that recall proponents probably had enough signatures, but said his group did not reach a conclusion about whether there were enough recall petitions signed to trigger an election. And as stated in the article, both the signatures deemed "ineligible" and requiring "further review" should not have been marked as such, so even the 534,685 number was very misleading. You also say True the Vote "only received 138,000 of the 152,000 petitions supposedly submitted," which seems to imply that the GAB refused to turn over all the petitions. I have heard nothing about True the Vote not having access to all the petitions -- just that True the Vote failed to enter all of the documents into their database. Finally, you are wrong when you say that the "GAB stated they could not review" the signatures. They stated their review would have a limited scope, and that it is up to the candidates facing recall to bring challenges. Governor Walker has raised $12 million to protect himself against recall, and was granted 3x as much time as is allowed under state law to review the signatures. He voluntarily chose not to direct his sizeable resources towards doing a thorough review. By the Texas-based True the Vote getting involved in Wisconsin's recall process and releasing misleading results based on a shabby analysis of the data, they have done nothing to "preserve the integrity of the election process." The real impact of their effort is to unnecessarily cast doubt on the work of the over 20,000 Wisconsinites who gathered signatures throughout the cold, snowy Wisconsin winter, and to tarnish the work of these people (who actually live in this state!) with baseless accusations of fraud. So what am I getting backwards?

In addition, your article states: "'Verify the Recall' is a joint project between the Houston-based nonprofit "True the Vote" (a project of the Texas Tea Party group King Street Patriots) and the Wisconsin Tea Party groups Grandsons of Liberty and We The People of the Republic." Verify the ReCall's claim that they are non-partisan, is at odds with their origin. They claim that they are not "affiliated with, or in communication with, any candidate or public official, or with any campaign" (Note that they don't say 'party'). But they also write that: "This project is a joint effort between We the People of the Republic [] and The Wisconsin Grandsons of Liberty" []. The former also makes the claim of being nonpartisan and grass roots, while simultaneously claiming to "...serve as one of thousands of so-called "Tea Party" groups in America. The latter don't use the Tea Party moniker, but claims at once that they are "supportive of candidates who possess honesty and integrity and that are motivated by public service and not public power and prestige. " Whether you think Scott Walker fits that bill is another matter altogether. The point being that these are not organizations devoid of political motivation. These are groups that channel time, energy and money to support outcomes of elections. Tea Partiers may think they can have it both ways; they aren't a party, they are, they are grass roots, but heavily funded. Fortunately, their explicitly contradictory and confounding rhetoric is only bought by the same people that think exclusion of WI voters is "within the limits defined in the Constitution", the mission of the WGL.

I was involved in this project. Do you have any idea of how much work had to be done so quickly? We said that the Scott Walker recall had enough signatures so we did not go back to take a second look at the lines called into question. We entered each piece of data 3 times and the algorithm took the best of the 3 based on the accuracy of the person who entered it and which one had the clearest data. Entering the data is a nightmare because people have really bad writing in many cases. You have brought up some errors, but we entered data from 6 recall elections in a one month time period because your GAB would not do it. We did find some interesting data like 29 of your judges signed the recall petitions, despite the fact that they will have the recall come into their courts. It is amazing that rather than applaud some regular citizens stepping in to do what the government should have done, instead you all want to criticize.

Prompted at another web site defending what True The Vote had done, I took a look at their 'Duplicate Names Report'. ---- - At random, I looked at True The Vote page number 177. The first eight entries on the page point to link ending sw042814.png, and also to sw042764.png (where I figure I was going to find the duplicate signature to the first link) Sw042814.png points to Walker petition #42967, and sw042764.png points to …Walker petition #42967! That is, page 177 of the TTV duplicates report is telling me that there are eight ‘duplicates’, when in fact all they’ve done is somehow manage to duplicate a Walker petition page. (Amazing as that is, they also somehow manage to then NOT report signature #4 on that page as being a duplicate.) I didn't happen to pick the 'needle in the haystack; There are maybe *hundreds* of such entries where the 'duplicates' actually point to the same page. ---- Another page selected at random, TTV page 228, shows at the first entry that someone named Dawn Cotter signed twice – first on Walker petition #51563, and then again on page #51566. The second entry on that VTR page then reports THE SAME duplicate signature, but in the reverse order! That is, VTR counted her duplicate signature as two duplicates into their total. The third entry on TTV page 228 shows Robert W Cobb; original petition at sw051374.png, duplicate at sw065846.png. If you then scroll down to TTV page 281, you find a report for Robert W Cobb, original at sw65846.png, duplicate at sw051374.png. Again, not that I chose the page 'luckily'; TTV reported many duplicates twice (but not all; that would be too consistent, and make estimating actual duplicates too easy). Under state law, one is legit, and one gets thrown out. ---- Hard to guess how many of the 5000 or so are actually duplicates; maybe half? maybe fewer than half? (But it is also worth noting that even 5000 would NOT be a particularly large number of duplicates. Many of he 2011 senate recalls and the current senate recalls show duplicates in the half-percent to one percent range.)

One more note on 'duplicate' Robert Cobb. sw065846.png has Robert L Cobb, while sw051374.png has Robert W Cobb. Now that the TTV/VTR searchable data base is starting to work better, I looked him/them up in it. Even the TTV/VTR searchable data shows them as different signatures. Let me say that again - the 'duplicates' report from TTV/VTR that counts two entries for the 'duplicate' signature from Robert Cobb disagrees with the TTV/VTR searchable tool on Robert Cobb, which shows them as different signatures.

During the scandalous 2000 election, my name was, I assume, deemed a duplicate in favor of another person with my very unusual name, namely, me - or rather me from 15 years earlier living at an 15 year old address in CA. I had been registered to vote for more than ten years at my then current address- never missing a vote; however, when whatever firm was contracted to "streamline" voter information worked their magic, they effectively wrote me out in favor of the old me at my old location. Fortunately, the address I lived at 15 years earlier was my mother's house and she still lived there. Unfortunately, it was too late to reregister and two hours away in LA traffic. Imagine if my mom had moved? I would have never known that my polling place had "changed". I could argue that it was only because my mom was by then a homeowner that I managed to vote. That in combination with the fact that I live in the same city. When we lived in apartments, we never lived anywhere for 15 years, no matter how stable we were, buildings get sold, demolished, etc. Had I gone away to school, or moved off somewhere else, I still wouldn't have been able to vote. Or imagine all those poor people who have really common names, names shared with felons, like all the contentious not included African American votes in FL. Since then I have had the same thing happen again, in reverse. I moved back to the city I went to high school in, and once was told by my regular polling place that my name was not listed, nor were any member of my family's (though all our neighbors had no issues and the map placed us at that polling place.) They would not allow me to complete the hand-written alternative unless I drove home and got ID. It's infuriating, but I'm just glad that we aren't using e-voting.

One way or another, I've managed to locate in the Walker petition pdf files about 20 petition pages where I was the circulator. So I decided to check 'my petitions' against the True The Vote/Verify The Recall searchable data base. Out of the 20 petition pages, I found two problems: - Walker petition page #15865 looks to be missing entirely from the data base (though the first name on the page does show up in the data base as part of another page; she seems to have signed twice) - On Walker page #13564, the first five names look AOK, but the second five don't show up. You can download the database from the TTV web site as a csv file; this group of second five signatures are 'blanks' in the csv file. I don't know for sure, but I'd guess they're not counted in the TTV/VTR total number of signatures. So, 2 pages out of 20 with problems, 15 (or 14) signatures missing from a total of 173 on the 20 pages. (One other thing to note: when they first announce what they had found in the Fitzgerald petitions, they announced only 14600 "non-blank" lines. Then they issued an 'update' changing the number to 19400. Fitzgerald in his challenge said there were 20650 'non-blank lines'. So far as I know, TTV/VTR have never explained the discrepancy. In the searchable data base posted today for the Fitzgerald Recall, they're back to 14212 entries. Seriously, WTF?)

I have been unable to find the CSV file. Could you provide a link?