Using Whole-Ballot Examination to Find Fraud

Parallel Election Whole Ballot Examination: Cross Party Line Voting

By Marj Creech
Dec. 10, 2006

One of the things a parallel election can do that an audit, a recount , or a conventional exit poll usually does not do, is to look at whole ballots from the individual voters. Recounts, audits, and conventional exit polls usually determine total numbers of votes for a specific race, but unless the ballots are photographed or photocopied, as is being done in Ohio for the 2004 ballots, whole ballot examination is unavailable.

Why would we want whole ballot information? Two reasons would be:
1) to study voting trends, e.g., did people who voted mostly a Republican ballot vote for one Democratic candidate, and
2) to find or disprove fraud, e.g., if a candidate “down-ticket,” say for a judgeship, gets significantly more party votes than the lead candidate, say for governor, in the official results, does the sample of ballots disprove or support this anomaly?

Another example: if a large number of voters voted BOTH against the one-man-one-woman marriage amendment (and so were pro-gay) AND for George W. Bush, in Delaware County Ohio in 2004--an unlikely event according to Richard Hayes Phillips and other researchers--does a sampling of ballots bear this out?

Ideally we would want a large sample of whole ballots from across the area being sampled. Photographing all the ballots from several precincts where the anomalies occurred shows something significant no doubt, but such ballots are not available to researchers for months, sometimes years after the election in question, due to lawsuits or stonewalling by election officials.

Timely study of ballots is available only through polling, either conventional exit polling, where data is obtained by the pollster for several candidates and issues for each individual voter, or through parallel election polling, where a sampling of a precinct’s voters is obtained on a secret, written “facsimile ballot.”

Granted that a parallel election may not attract a “random sample” of the voters—indications from four different elections in Ohio of from one to four precincts per election show that Democrats and/or left-leaning participants are sampled more heavily in parallel elections than Republicans or right-leaning participants. However, in whole-ballot investigation, that is not a problem, or at least, it can be taken into consideration.

For instance, in the highly contested 2006 Kilroy-Pryce Congressional race in the Fifteenth District of Ohio, Franklin County official recount results show that there were a number of voters who voted a straight Democrat ticket except they votedPryce(R) and not Kilroy (D) for US Congress. How likely is that?

Let’s look at the parallel ballots received by voters exiting from four precincts in Clintonville, a mostly professional-class neighborhood just north of the city of Columbus. In particular, let’s look at people who voted Republican for all candidates except for Kilroy, and people who voted for all Democrat candidates except for Pryce.

Note that only five candidate races and five issue contests were on the parallel ballot, in order to simplify and keep the ballot to one page.

First, a breakdown of voters by party preference as indicated on their ballot choices in each of the four precincts: (sorry the chart broke down in e-transmission, but I think you can figure it out--email me if not.)
Precinct in Clintonville
19C 19D 20A 20D UNK TOTAL
total number of parallel
voters, all parties 141 163 96 162 64 626

total number of Republican
voters, at least 4 out of 5 ,
all four precincts 23 21 7 30 5 86

Republican, 5 of 5 candidates 13 17 4 19 4 57
R, 4 of 5 candidates 10 4 3 11 1 29
R, 3 of 5 candidates 4 4 2 4 2 16

Total number of Democrat
Voters, at least 4 out of 5,
All four precincts 109 129 86 122 54 500

Democrat, 5 of 5 candidates 86 100 67 101 32 386
D, 4 of 5 candidates 23 29 19 21 22 114
D, 3 of 5 candidates 2 4 1 1 1 9

Neither party dominates
(voted third party or no vote
for candidate or candidates) 3 3 0 5 2 13

Neither party dominates
and 3/2 split (R & D)
combined, i.e.,
“Independent” Voters 9 13 3 10 5 40

Are Ohio Voters REALLY Independent?

1. Most Ohioans, some 70% or more, do not register as Democrats or Republicans (although they are only required to do so when voting in a primary; their last primary would reflect their most recent party affiliation). About half of the voters in this ward in Clintonville are registered as Democrat or Republican, but in actual voting, in this parallel polling sample, voters definitely lean either Republican or Democrat. By the definition of D’s being voters who voted either 5 of 5 or 4 of 5 Democrat candidates and R’s being voters who voted either 5 of 5 or 4 of 5 Republican, D’s were 500/626 or 80%, R’s were 86/626, or 14%.

The rest--40/626, or 6%--can be called Independent or No-Party voters. These percentages represent the way the parallel voters actually voted and may not correspond with their party registration. This midterm 2006 election is known as the election where Republicans reflected their dissatisfaction with their own party politicians. It would not be surprising to see R’s jumping ship for some races, while staying with their party for candidates they deemed still OK.

2. R’s polled were more likely to cross over to vote for the candidate of the other party (Third party candidate voting was so low in these four precincts that it can be ignored for this discussion.) by a comparison of 29/86 for the R’s and 114/500 for the D’s. If we add in the 3 of 5 candidate voters as showing a party preference, the numbers for the R’s are 45/102 and for the D’s123/509. The R’s sampled were likely to “jump ship” for 1 or 2 of the 5 candidates 44% of the time, while D’s jumped ship 23% of the time. Of course there could be a sampling adjustment needed, as pointed out by Pokey Anderson via phone conversation: are cross-over R’s more likely to vote in parallel elections, sensing that we parallel pollsters are on the “Lefty” side of the spectrum, and cross-over D’s less likely to vote parallel, sensing our disapproval? That is hard to gauge, but we do use secret written ballots in parallel elections, so that no one need feel embarrassed by their responses. Also it could be argued that the mavericks—the independent, non-party-committed voters are MORE likely to vote with us, of either party. Because no poll can factor in every possible bias, this conclusion that more R’s crossed over to vote for one or more D’s, than D’s crossed over to vote for R’s, at a ratio of almost double, is presented here as a probability, not a certainty.

If one uses the official results as a measuring stick versus the number of registered D’s and R’s who voted (data that, unfortunately, is not being made available to citizens until after January 15 from the Franklin County Board of Elections), one could see if this trend of R’s crossing over to vote for D’s bears itself out. Whole ballots may not be available from the BOE—in their case printing from the paper roll of the electronic voting machines—for months, or even years, if the 2004 election is a model.

We DO know the party registration, as determined by previous primary elections, of our parallel voters en masse, since they signed our parallel signature books. But I believe the actual ballots are a more accurate reading of which party’s candidates they REALLY favor.

Who Did Cross-over Voters Vote For?

Taking all the precincts together, Democrats who crossed over for one of five candidates, voted for these R’s: O’Donell (judgeship), 95 votes; Blackwell, 1 vote; Hartmann, 3 votes; and Pryce, 7 votes.

Republicans who crossed over for one of five candidates, voted for O’Neill 11 times, Strickland 9 votes, Brunner 6 votes, and Kilroy 1 vote.

Factoring in the larger number of D’s polled parallel than R’s, by 80% to 14%, or 5.7 to 1, defining R’s as those who voted 4 or 5 of 5 R candidates and D’s likewise, then crossovers comparing R’s and D’s can be weighted as follows:

O’Donell crossovers/ O’Neill 95/11, corrected for fewer R’s sampled = 95/(11)x (5.7) = 95/63, or about 1.5 times as likely for D’s to vote for the R judge, as R’s to vote for the D judge. Why D’s would vote for the R judge in Clintonville is unknown. Judges do not have a party affiliation next to their names on the official ballots, but it doesn’t make sense that D’s would confuse these judge names, though similar, any more than R’s would.

For governor the numbers are D’s one vote for Blackwell, R’s 9 for Strickland, so 9/1 R/D crossovers. Or corrected for fewer R’s polled, (9x5.7)/ 1 or R’s 51 times more likely to vote for Strickland as D’s for Blackwell. That would certainly help explain the landslide victory for Strickland!

For the 15th District, the D/R vote ratio is 7/1. Corrected for fewer R’s polled and it would be 7/ (1 x 5.7) = 7/5.7=1.2 times crossover D’s are likely to vote for Pryce, as R’s for Kilroy. This ratio is so small as to be insignificant, and in fact does NOT support the supposition that Dems voted for Pryce any more than Repubs voted for Kilroy.

For the secretary of state race, the numbers are 3 Dems for Hartmann and 6 Repubs for Brunner. Corrected for fewer R’s polled and we get 3 Hartmann’s/6(5.7)Brunner = 3/34= 1 to 10 times more likely for D’s to vote for Hartmann as R’s to vote for Brunner. This is also compatible with Brunner’s decisive victory.

For the Senate race, there were an insignificant number of crossovers.

Another odd combination of choices is: Strickland-Brunner-DeWine-Pryce (the judgeship race either not voted for or about evenly divided between the candidates), 21 ballots out of the 626. These were perhaps Republicans who were disenchanted both with Blackwell and with Republican control of the SOS office. Or this could indicate the strength of the candidates for people truly non-partisan.

More analysis of the whole ballots might provide further insight into voting trends. It would be more accurate to extrapolate the data gathered from these four precincts in Clintonville to a larger sampling area, if the precincts were more spread out geographically.

Also more parallel election polling sites would give us a better basis for looking for fraud and for debunking claims that R’s or D’s who otherwise voted a straight party ticket crossed over to vote for one particular candidate.

Anyone rigging elections will have to start vote-swapping for more than one candidate to be believable. Hopefully the complications of changing multiple races on each individual ballot will be their undoing. One can imagine a program that swaps votes for every tenth vote for candidate A and every tenth vote for candidate B, but these changes would not remain synchronized for the same ballot for very long, and would lead to anomalies of D’s or R’s crossing over. Whole ballot sampling could demonstrate the likelihood of such a crossover.