Audit Arizona (AUDIT AZ) Articles

Tuesday, February 21, 2006

Ballot-Box Fraud Committed by Poll Workers in Tucson, Arizona

Ballot-Box Fraud Committed by Poll Workers in Tucson, Arizona
At Precinct 324 on November 2, 2004

John R. Brakey was the Democratic Precinct Cluster Captain for four precincts in Arizona Legislative District 27, a part of predominately-Hispanic Congressional District 7. On Election Day 2004 he observed a multitude of irregular activities and was greeted with hostility by poll workers at two of the four voting stations he monitored. Brakey, a veteran civic activist, thought he "smelled a rat.” Accordingly, he felt impelled to initiate an audit of the voting process at one of these suspicious stations, Precinct 324.

All across the country, others were simultaneously beginning investigations of voting irregularities. So what was so different about what John Brakey did?

First and foremost, John Brakey went after a paper trail different than the ballots themselves. The ballots were locked up and could not be viewed without a court order, whereas most of the other documents could be obtained by Freedom of Information Act requests (or fished out of the trash the day after Election Day). What were these documents? First there were the complete “Signature Rosters” maintained by the county, which contain the names of all voters registered in each precinct, each voter being assigned a "Register Number" in alphabetical order. Second, there were photocopies of printed pages of these rosters that were ACTUALLY SIGNED BY VOTERS on Election Day. Third, there was the “Consecutive Number Register,” or "CNR." On the CNR, a poll worker (specifically, the clerk) HAND PRINTS the name of each voter who was issued a ballot at the polling station that day in the order of his/her arrival. Fourth, there were the "Notice to Voter Slips," on which poll-worker annotations (specifically by the judges) link the signatures on the Signature Rosters via the voters' Register Numbers to the names to be hand printed on the CNR. In principle, every voter who signed a roster would have had his or her name copied into the CNR at least once. Some voters’ names will appear twice (or more) because they spoiled a ballot and were issued a blank replacement ballot (or ballots). The spoiled ballots should then have been marked "spoiled" by the poll workers and set aside to be turned in at the end of the day, since every ballot issued must be accounted for.

Another way a voter name could appear twice on the CNR would be if a double vote were cast by this voter -- or by someone else in this voter’s name. A double vote is a way of “stuffing the ballot box” by voting more than once. Intentionally voting more than once in a federal election is a third-degree felony in most states and probably also violates federal election-fraud laws. The punishment varies from state to state but is usually up to five or 10 years in jail and fine of up to $5,000 or $10,000. Serious business for an individual, and even more serious if carried out as part of a conspiracy.

If there were such a conspiracy, how would one detect it? Frankly, neither John Brakey nor David Griscom, a retired Ph.D. physicist who began to assist him, had a clear idea what could be detected by means of the audit they had embarked upon. But before getting into what they found, let us consider in a little more detail exactly what it was that they did. They counted the number of UNIQUE NAMES on the CNR and also the numbers of UNIQUE SIGNATURES on each of the SIGNATURE ROSTERS – of which there were three: the “regular” Signature Roster, the Inactive Roster and the Provisional Ballot Signature Roster. For present purposes, the first two rosters can be lumped together under the rubric “rosters of voters whose ballots were accepted at the polling place.” By contrast, the PROVISIONAL BALLOT Signature Roster held the signatures of persons judged by the poll workers to have no assured right to vote at the polling place, but whose vote MIGHT be counted if subsequently approved by the Pima County Recorder. Accordingly, after filling out a PROVISIONAL ballot, the voter seals it in an envelope and prints and SIGNS his/her name and address on an attached affidavit. Upon receiving these provisional ballots, the Recorder checks the names and signatures against her records. If these voters (1) had NOT voted by mail-in ballot AND (2) WERE REGISTERED IN PCT 324, their ballots would have been counted. Otherwise, they would have been rejected.

As mentioned above, if everything had been done correctly, every name appearing on a signature roster would also appear at least once on the CNR. But one thing Brakey found was that there were three more unique names on the signature rosters than were unique names on the CNR. That was a small error, but when he asked the question of how many unique names there were on all signature rosters and CNR COMBINED, Brakey was shocked to find that there were 19 unique names on the CNR that were IN ADDITION TO the names signed on the various rosters. Combined with the three "extra" signatures, this made a total of 22 voters who made their signatures on a roster on Election Day but WHO WERE NOT CREDITED ON THE CNR AS HAVING CAST A VOTE! At the very least, this represents a gross error or misfeasance on the parts of the poll workers because the 19 “new” names on the CNR were not even close to being misspellings of the names of voters who actually signed in.

Another thing to come out of the Brakey-Griscom analyses was that the number of votes officially cast – combining all ballots accepted at the polling place (i.e., those actually put into the ballots box) with all those cast as provisional ballots – was 895, in exact agreement with the accounting on the "Official Ballot Report and Certificate of Performance" (the fifth and final part of the audited paper trail), which was signed by all seven poll workers. So far so good. HOWEVER, the number of UNIQUE NAMES on the CNR proved to be just 884. The inescapable conclusion is that 895 – 884 = 11 felony double votes were cast. But by whom?

There were 29 voters whose names appeared twice on the CNR and one whose name appeared thrice. Initially, all would be potential candidates for the double voters, but on the basis of documentary evidence, 20 of these voters could be reliably identified as individuals who likely spoiled their first ballot and were issued blank replacements. One of the remaining 11 instances of a voter being issued a second or third ballot was a person known from interviews and witness accounts to have voted only once, yet a second ballot was cast in her name as an alleged spoil replacement. In seven of the other 10 cases, the second or third same-name entry on the CNR was separated from its first occurrence by a number of places ending in a zero, specifically, 70, 80, 100, 100, 100, 120, 510. How likely is that?

People who run casinos (and some of the people who gamble) know the precise odds for such things as rolling “snake eyes” on a pair of dice (one chance in 36). Physicists like David Griscom do these kinds of calculations routinely (which is why their conventions are no longer welcome in Las Vegas). Suppose, for the sake of this discussion, that there were a slot machine with 885 numbers on each of seven cylinders. (Here, the number 885 is selected because it was the total number of ballots cast at John Brakey’s precinct 324 last November.) Suppose also that one of the jackpots could be won by having EACH of the 7 cylinders stop on a number ending in a zero. Assuming the machine was honest, you would have very close to one chance in 10 million of winning this jackpot. A different calculation pertains to the chances of seven double voters EACH INDEPENDENTLY casting his/her second vote at a precise moment causing the official record of this second vote to be separated on the CNR from his/her first one by a number of places ending in a zero. But the odds of this happening are about the same, close to one chance in 10 million. Dr. Watson, might have remarked, “Mr. Holmes, I suppose that this PROVES that these seven events CANNOT have been seven crimes committed independently by seven different criminals, but it COULD have been seven crimes committed in a deliberately systematic fashion by a single individual.”

“Elementary, my dear Watson.”

Suppose now that we had a slightly more normal slot machine with just three cylinders but that each cylinder had numbers running from 1 to 895. On a fair machine, your chances of winning the “three-number-one-hundreds” jackpot are one out of 895 multiplied by itself three times, or about one in 717 million! However, the mathematics is quite different in the case of three double votes out of total 895 votes cast each being SEPARATED by exactly 100 places from their first occurrence. Griscom's calculation of this case gives one chance in "just" 131 million. Nevertheless, these odds are proof "beyond the shadow of doubt" that these three double votes constitute a single crime committed by a single person -- the person (i.e., the clerk) that happened to be in control of the CNR. “But,” complained Dr. Watson, “why would this person give himself or herself away by deliberately making the numbers so regular?” Sherlock Holms replies, “We don’t need to answer that question to get a conviction, dear Watson, but we can infer that the criminal either wanted to be caught or, more likely, that he or she wasn’t familiar with THE FORENSIC VALUE OF STATISTICS.”

"In fact, we haven’t yet exhausted the forensic applications of statistics to the PRESENT case,” the legendary detective might then have announced to a nonplussed Dr. Watson. Here, Holms would be referring to the peculiar “rule of eleven” that infected the Precinct-324 poll records last Election Day: That is, there were exactly 11 voters who signed one of the two rosters for which “the voter's ballot was accepted at the polling place," but whose names are NOT LISTED ON THE CNR, 11 voters who signed the Provisional Ballot Signature Roster but are NOT LISTED ON THE CNR either, 11 voters who improperly signed BOTH the “regular” Signature Roster AND the Provisional Ballot Signature Roster, 11 registered voters listed on CNR who FAILED TO SIGN ANY ROSTER AT ALL, 11 phantom voter names appearing on the signed affidavits attached to 11 of the 59 provisional ballots that the poll workers submitted to the Recorder's office that DO NOT CORRESPOND TO A SIGNATURE ON ANY SIGNATURE ROSTER NOR TO AN ENTRY ON THE CNR, 11 felony double votes (already discussed), and 11 extra ballots issued as alleged spoil replacements (which might tend to camouflage, BUT DO NOT DISPROVE, the 11 double votes).

With regard to this “rule of eleven,” Griscom proposed a crude calculation of the probability of the number eleven popping up seven times in this context as independent random mistakes. It goes like this: If 11 out of 906 voter names (or 11 out of 895 ballots cast) could have been mishandled in a certain way, then surely it was possible that they could have been mishandled in the same way, say, 15 times. With this very conservative assumption, there would have been one chance in 15 that the number of errors of a given nature would come out precisely 11. But if, as is true in the present case, SEVEN DIFFERENT KINDS of errors were committed, there is just one chance in 15 raised to the 7th power that the NUMBERS of EACH of these seven kinds of errors would come out exactly equal to 11 AS SEVEN INDEPENDENT RESULTS OF RANDOMLY MADE MISTAKES. This calculates to one chance in about 170 million! It is the third independent “gambler’s odds” calculation implicating Pct 324 poll workers in a conspiracy to corrupt the election process. It should be remembered that INCOMPETENCE IS NOT A DEFENSE since incompetence can ONLY lead to RANDOM ERRORS, whereas the calculated odds STATISTICALLY RULE OUT RANDOM ERRORS!!!

Moreover, there were three MORE irregularities occurring in numbers close to, or related to, the number 11. First, as mentioned above, there were 22 voters who signed a roster but whose names do not appear on the CNR. Well, 22 = 2 ´ 11! Second, there were 10 blank entries in the CNR (just 1 less than 11), whereas no trained poll worker would have left so much as a single blank space by accident! Third, there were nine voters (2 less than 11) who were evidently told to vote provisionally EVEN THOUGH THEY WERE NOT REQUIRED TO. (That is, Recorder hadn't preprinted "Es" next to their names on the "regular" or inactive signature rosters, which she would have if she believed that those persons had been early voters.) If these three additional VERY-IRREGULAR irregularities had been folded in with the seven different ones mentioned in the previous paragraph, the odds of all 10 being random accidents would have come out to about one chance in about 570 BILLION (a number that Griscom recognizes to be about 44 times the age of the universe in years!!!).

But statistics is not the only evidence for possible criminal wrong doings. In fact, Griscom used the data that John Brakey had assembled to create a “balance sheet” similar to the Official Ballot Report and Certificate of Performance filled out and signed by the Pct-324 poll workers after the polls had closed. This does not involve “higher math.” Rather, it is very similar to the way you balance your checkbook.

The most common discrepancy between Griscom's balance sheet and the poll-worker-signed Official Ballot Report and Certificate of Performance involves poll-worker annotations on documents in the public record indicating that 98 voters had cast PROVISIONAL BALLOTS, whereas only 59 provisional ballots were submitted to the Recorder's office. The ONLY POSSIBLE RECONCILIATION of this discrepancy is to conclude that 98 – 59 = 39 provisional ballots were FED INTO THE (optical-scan) BALLOT BOX ON ELECTION DAY! Moreover, 36 of the 98 provisional voters listed on the CNR were NOT REGISTERED in Pct 324 and therefore their ballots would not have been counted had they been sent to the Recorder’s office in their sealed-and-signed envelopes. Since the Recorder did accept 48 of the 59 provisional ballots that WERE sent to her (11 were rejected), it is inferred that AT LEAST 36 – 11 = 25 of the provisional ballots scanned into the ballot box on Election Day were cast by individuals who could have been casting double votes after having already voted at their home precincts. Whether or not they were voting a second time, it was the Recorder’s prerogative to determine the VALIDITY of these ballots. The poll workers deliberately usurped this prerogative, thus destroying the integrity of the electoral process – 39 times!

The various irregularities committed by the Pct-324 poll workers clearly gave them both the means and the opportunity to steal votes in the presidential election. By Griscom's calculation, 113 votes out of the 895 ballots officially cast could have been shifted in one direction or the other, if that was the poll workers motive. (N.B. John Brakey has developed documentary evidence that there may have been still more stolen votes.)

If votes WERE STOLEN, on which candidate's behalf might this theft have been carried out? Well, Precinct 324 was registered 47% Democratic, while Republicans comprised only 22% of the registered voters. Yet, according to official Pima County records, Bush garnered 41.5% of the Pct-324 AT-THE-POLLING-PLACE vote, compared to Kerry's 57%. So should we conclude that fully two thirds of all Pct-324 voters who expressed No Party Preference decided to vote for Bush? Well, in fact, the 48 Pct-324 PROVISIONAL BALLOTS ACTUALLY ACCEPTED by the Pima County Recorder went just 19% for Bush and 81% for Kerry. This HUGE difference between the at-the-polling-place tally the accepted-provisional-ballot tally is simply too large to be an accident. Moreover, it can be argued that the provisional-ballot tally HAS TO BE a highly reliable representation of voter sentiment at Tucson Pct 324 last November 2nd, since it is difficult to imagine any means cheating on a provisional ballot that is sent to, AND ACCEPTED BY, the Recorder's Office in a sealed envelope with the voter's name and signature appearing on the attached affidavit.


John Brakey made a complaint to the Pima County Attorney's Office concerning the poll-worker shenanigans at Pct 324 that he and David Griscom had uncovered as a result of more than 1,000 hours of auditing. Accordingly, one very smart detective was dispatched from the county Attorney's office to make very careful notes of the specifics of the Brakey-Griscom complaint against these poll workers. Whether or not the county Attorney actually read these notes is unclear. What IS clear, however, is that she discussed the matter with the Pima County Director of Elections, who -- without taking the time to understand the details -- had summarily fired the two head poll workers at Pct 324 ...for "INCOMPETENCE." (Recall that the odds against the "incompetence defense" run from one chance in 131 million to one chance in 44 times the age of the universe!!!)

The Pima County Attorney ended up dismissing the complaint on the strength of assurances from the Director of Elections that there were no problems with voting anywhere in the county on Election Day 2004 -- other than a few incompetent poll workers who he had already fired.

Posted by Protect Democracy :: 2/21/2006 11:40:00 AM ::