Study completed on May 3, 2008
Calculations and report written by Marj Creech, Election Defense Alliance.
Assisted by Victoria Parks, who organized the citizen poll takers.
See attached pdf chart of summary data.
Volunteer citizens conducted written exit polls in eight counties in Ohio, many at multi-precinct polling sites, making a total of 20 precincts, on the day of the Primary election in Ohio for both Democrat and Republican Presidential candidates, on March 4, 2008. 1787 surveys were collected, an average of 31 % of the voters in those precincts. A spreadsheet contains the raw numbers of both the exit poll surveys and the machine count for corresponding candidates. The machine counts were obtained ( except for Franklin County, which posted the results at the precinct that was surveyed) from the county Boards of Election, since Ohio no longer makes precinct posting mandatory. Obtaining these results from the County Boards proved to be the most time-consuming and difficult part of this study. In two of the eight counties, the board would not provide machine counts without the absentee ballot totals already added in. Despite the difficulty in getting machine counts, comparisons of the parallel results from the exit poll proved that citizen exit polling is a highly accurate check of machine counts, when taken in a summary of all the precincts together. In the summary, the discrepancy between exit polls and machine counts was fewer than five percentage points, in all candidate contests. But when broken down by Optical Scan counties vs. Touchscreen (DRE) counties, the discrepancies between exit poll and machine counts of the Democrat candidates, showed a 2 point spread discrepancy for OS counties, but a 9 point spread for DRE counties. There could be many explanations for this discrepancy, but a possible one would be that touch screen machines in some or all of the three DRE counties that were surveyed, were manipulated or rigged to give Clinton a larger margin of victory than the exit polls indicated.
Franklin County, in the precinct of RENOLDSBURG, 1-F, provided the largest discrepancy between machine count and exit poll, a whopping 27 point loss for Obama. 111 exit surveys were collected out of 254 voters, which is 44% polled.
163 Democrat ballots (64%) and 82 Republican ballots (32%) were cast, with 9 Issues Only ballots (4%).
Of the 163 Democrat ballots, 81 voters filled out surveys (50%). Of these 81 ballots, 45 (63%) affirmed, by written affidavit, that they had just voted for Obama for President, on the ES & S IVotronic touch screen machine and 26 (36%) affirmed that they had voted for Clinton. 5 voters for each candidate checked that they had voted provisional, and these votes were not counted in the totals, since these voters voted on paper ( to be optically scanned later if it was determined that their vote was legitimate), and not on the voting machine on election day. In stark contrast, the posted results at the precinct about an hour after the polls closed, showed the machine count to be 79 for Obama and 79 for Clinton, or a 50% -50% split. (Edwards was not on the ballot in Franklin County, in contrast to all of the other counties surveyed.)
Before incorrect machine counting can be alleged, other explanations should be ruled out. Were we oversampling Obama supporters, who had a high correspondence to the African American population at this precinct, roughly estimated at 30 to 50%? Victoria Parks and I were the main pollsters, who were there all day, and for much of the day we missed voters exiting at one of two possible doors because the presiding judge had told us we could not exit poll in the foyer. We had disputed this ruling, but to avoid controversy we had backed off from one door which was in the judge's direct line of vision. Was this the "White voter" door? Not likely.
Another explanation was that the cross-over Republican voters who voted for Clinton, as recommended by Rush Limbaugh and another right wing radio personality, were leery of signing an affidavit that they had voted for Hillary Clinton. There was much controversy over the legality of Republican cross-over voters, who did so only to keep the contentious Democratic Party Primary going as long as possible since Clinton needed a strong showing in Ohio to stay viable as a candidate.
So should we propose a "reluctant Clinton voter" hypothesis? But there would have had to have been at least 18 of these reluctant Clinton voters to make the exit poll count 50-50, who refused to fill out a survey when we asked. As it was, I can recall only a handful of voters who refused to fill out the survey. And since we were not insisting on signatures, these refusers must have had other reasons, such as being in a hurry, or not liking exit polls.
Did we have exuberant Obama voters, who were eager to tell the whole world how they voted? This is possible, but there would also have been exuberant Clinton voters. One might expect that the perceived preferences of the pollsters may have come into play, but all of us were white women, mostly of middle age, the classic profile of Clinton supporters, not Obama supporters. (In reality, we had voted for several different candidates, but we did not discuss our preferences with the voters.) If we were scaring off Clinton supporters, why would we not scare off McCain voters or Tiberi (R Congressional candidate) voters? We got a higher percentage of Tiberi voters than the machine count indicated, but were exactly the same (59%) for McCain. The school bond passed by a percentage of 59% to 37% in our exit poll , compared to the machine count of 48% to 50%, but this could be explained by our overpolling Democrats ( 50 % of the D's polled to 37% of the R's polled), who are generally more liberal on increasing spending. In conclusion, the Obama/ Clinton percentage discrepancy between exit poll and machine count is not easily explicable by our polling methods; the best explanation seems to be an incorrect machine count of the votes the voters touched on the machines for Democratic Presidential candidates.
In an election system with citizen oversight, a recount of the paper trail in this precinct would be allowed by citizens. But in Ohio a Presidential candidate has to order a recount of that contest, and no Democrat or Republican candidates ever seem interested in bringing up a possible fraud charge. Of course even if citizens were allowed a recount, the paper rolls might be torn or not available, as has been the case in other Ohio recounts, and, even more troubling, there is no reason to believe that the paper roll ballots are what the voter actually voted, since very few check the roll when they vote.
COMPARISONS WITH THE OTHER TOUCHSCREEN COUNTIES' EXIT POLL DISCREPANCIES in OHIO's PRIMARY on MARCH 4, 2008
See attached chart of summary data and spreadsheets of data of all 20 precincts
Reynoldsburg, a suburb of Columbus, in Franklin County, showed the greatest discrepancy between exit poll and machine count, only in the Democratic Presidential race, but Richland, with 144 Democrat surveys showed Obama losing 4 percentage points between exit poll and machine count. Stark County shows NO discrepancy, but the survey totals are too low to be significant, (due to the presiding judge chasing the single pollster away from just inside the door of the building out into the sleet, where voters were very reluctant to stop and fill out a survey.)
Ashland County, in one precinct (Mifflin 2) shows a point discrepancy going the other direction, with Clinton doing better in the exit polls than in the machine count, but the number of surveys collected is smaller (35) than those collected in Franklin (81). In the sister precinct (Mifflin 1), Clinton drops 6 points in the spread. Trumbull also shows Clinton doing better in the exit poll, but only 6 Obama supporters stopped to be surveyed. Also Trumbell's machine count is tainted by the inclusion of absentee ballots, which the Board of Elections said they could not separate out for our audit. The demographic in these rural counties is fewer minorities than in the cities, such as Columbus, which is located in Franklin County. An interesting study would be to compare the participation of African Americans versus whites in surveys in rural areas where minorities are scarce, versus participation where the racial mix is more equal. Also the influence of having all white pollsters in a heavily white area versus in a heavily black area, or in a racially balanced area needs to be studied. (See recommendations report (published in another OpEd paper) re: asking for and recording demographic data both of those polled and of those who are not polled.)
(Note: You can view every article as one long page if you sign up as an Advocate Member, or higher).