METHODS, TECHNIQUES and RECOMMENDATIONS for CITIZEN EXIT POLLSTERS, BASED ON OUR EIGHT-COUNTY EXIT POLL STUDY of the Ohio Primary, March 4, 2008
1. GET MACHINE RESULTS
Someone MUST get the machine results for the precinct(s) surveyed as soon as they are available, preferably election night. Ohio no longer requires posting of the results at the polling site, which is a knife in the heart of transparency, since the precinct totals are the basis of the county totals and the state totals, and we don't want them to be changed further downstream. Chain of custody problems abound in election studies across our nation. In some cases the presiding judge might agree to provide the machine count totals when he/she runs the results tape at the closing of the poll, in optical scan counties with precinct tabulation. Be sure to determine which machines go with which precincts in multi-precinct locations.
Getting the machine counts later from either the County Boards of Election websites, or directly by email or by CD from the Boards, proved to be challenging, to say the least.
The machine results MUST be before the absentees are added in. The "unofficial results" will include the absentees (also called "early votes" or "mail-in votes," or "vote-by-mail," or "VBM"). You want the machine results WITHOUT the absentees added in, since we are checking the accuracy of the machines, and are comparing how people said they voted on the machines. And of course you do not want the "certified results," since they will include even more absentees (overseas absentees) and provisional votes that were deemed countable.
Optical scan counties also typically run their absentees through the scanner while the regular votes are being brought in from the precincts. The scanner will make a recording and a results tape by precinct of the absentees and then later of each precinct's regular votes. The machine separately tallies absentees and regular votes, so a machine count for each precinct without absentees should be easily available. However, Cuyahoga County wore me down by repeatedly providing everything but what I asked for. I finally extrapolated a total machine count for Democratic and Republican ballots, by getting the absentee list for each precinct and the provisional list, but I never did get a breakdown for each candidate for precinct machine votes only. Therefore the machine count for comparison to exit poll count is not as accurate for Cuyahoga as for other counties. Cuyahoga and Hamilton County election officials believe that referring requests for specific vote totals to their complex, and lengthy, website spreadsheets will get rid of the pestering citizen, but, in fact, a persistent citizen soon (or several hours later) discovers that precinct totals, sans absentees, by candidate, are simply not there. Trying to find the official who can actually provide the data needed is also part of the game.
Franklin County BOE's website actually did provide the candidate totals, by precinct, needed. There is a learning curve to navigating their spreadsheets, but once conquered, the information is there. Of course, Franklin County poll judges are often still posting their results at the precinct. Kudos to Franklin County. I reward them by accusing them of having rigged machines. But whether this is true or not (and rigged machines do not require the complicity of ANY election official, though it would help in a rigged recount or audit), citizens and auditors MUST be able to get all the data they need to verify an election. Transparency is necessary to prove an honest election, as well as a rigged one. An election unverified by the citizens is not a democratic election.
The smaller counties were much more forthcoming with the data I asked for, since it was often the Director or Deputy Director who answered the phone and who sent me the requested information by email. Sometimes they sent only what I asked for, no spreadsheet, simply the number of votes for each Presidential candidate for the precincts I requested. No public records requests were needed, though perhaps one for Cuyahoga might have eventually produced the needed info. Some patience is required in navigating even some of the smaller county websites. Some of the counties' election results websites seem to have been designed by the Director's Second Cousin Bubba. In one, the precinct names were not listed on subsequent pages of results. The election official I pointed this out to said that I only needed to count down on subsequent pages of contests, the same number of lines as on the original page. Despite this blending of low and high tech, this worked. Each county had it’s own way of listing the results I needed, or of not listing them. It was like solving eight different mysteries to get the machine results. I recommend sending one very patient person to the County BOE on election night and having them sit there until the unadulterated precinct results are obtained.
Note that the spreadsheets for each county differ slightly. I put in all the information I was sent , or that was on a website, that might have bearing on this exit poll comparison. Some counties provided the under/overvotes, some did not. Some had more candidates for the Republican Presidential candidates than others, not removing the names even though the candidate had withdrawn. John Edwards was on most counties' ballots, except for Franklin's. Cuyahoga still had Kucinich on the Presidential slate, but the optical scanners were set to count a vote for him and an overvote. Knowing this fact from a local activist, I recorded overvotes in Cuyahoga as votes for "other candidates." It made no difference in the comparison study to the exit polls, but does explain the high rate of undervotes in Cuyahoga. (Of course someone might ask what effect on the votes for Clinton or Obama does leaving in Edwards' or Kucinich's name have? My guess would be that Kucinich voters might favor Obama, but that is just a guess.) Some counties provided absentees by candidate, some did not. One website's results spreadsheet, Richland, has the odd heading, "times counted," next to "ballots cast." The County tech man explained to me by phone that "times counted" was how many times the machine "sees" the contest, or race. I thought it might represent over and uundervotes, where a voter skipped over a contest. But the tech guy said, "If a voter takes too long, the machine "time outs" and when the voter touches the screen again, the contest, or race is "seen" again, by the machine. This explanation sounds very odd to me. Does the Diebold (Primier) TsX screen save? Or go offline--which it should not, reportedly never being online! Is there a chance for miscounting in this feature? I believe that other TSx county websites also had this column on their voting spreadsheets. Further investigation into this feature might be fruitful.
Some Counties count the over and undervotes in their calculation of percentages, some don't. I did and also counted all other candidates listed. For this reason Obama's and Clinton's percentages don't add up to 100%. Also Franklin County had two issues on the ballot so we had issues only voters. These were not included in candidate percentages, of course. Very few voters would bother to be exit polled if they had not voted for any candidate for President, so that under and overvote voters are underpolled. This made an undiscernable difference in the comparison between the exit polls and the machine count.
Also it needs to be noted that the official Republican ballot in the Ohio Primary contains two slates of exactly the same candidates.Voters are asked to vote twice. The first was the "at large" delegates for all the candidates, the second was the Congressional District delegates for the candidates. Several people have agreed with me that having Republicans vote twice could have been confusing. For instance McCain received significantly more votes in the first slate than in the second slate, in the same precinct. One wonders why a voter would change their mind in two consecutive choices on the same ballot. Were they undecided and wanted to split their vote? Did they think the second slate was for their second choice? We decided to take the first slate for comparison purposes. For this reason, I feel that the comparison between the exit poll results from Republican candidates versus the machine count is less meaningful.
Finally please note that this is not a complete citizens' audit. A real audit would include examining poll books and signature books, absentees, provisionals, and counting numbers of voters and different types of ballots. Absentees and provisionals not counted vs counted, absentees returned, reasons for not counting, numbers of over and undervotes, unused ballots, ballot stubs, chain of custody (who is watching the ballots, who has access to the locked rooms where ballots are stored, etc?), the log of the voting machines, ALL of these should be examined in a complete audit. What we did in our exit poll analysis is a "spot check" of the machine count. But, in getting records from the Boards, you will be getting records that could be used in further auditing. For instance I received a complete listing of provisional ballots from Cuyahoga, who voted them, in which precinct, whether they counted and the reason for not counting them. Further analysis would be useful in determining who is disenfranchised and why.
BUT, getting simple machine count results for a few precincts for a few candidates should not be so difficult, in a purported Democracy.
2. WHERE CAN EXIT POLLSTERS STAND?
The rules for where exit pollsters can stand are not clearly understood by presiding judges. In Ohio we had a directive from SOS Brunner reiterating the findings of a judge in a court case right before the 06 election, saying that exit pollsters can stand within the 100 foot mark, but not "inside the polling place." Some of the presiding judges in the counties we surveyed did not read this directive or comprehend it when we gave it to them. They made our pollsters stand outside the 100 foot mark, in the sleet, making the job of exit polling , in practicality, impossible. Others, such as our presiding judge in Franklin County, called the Board of Elections, and got the opinion that we could be within the 100 foot mark, but had to be outside the building. We, upon consultation with our attorney, Bob Fitrakis, argued that our not being allowed "inside the polling place," hinges upon the meaning of "polling place." Is it the room where the voting is taking place, or is it the whole building, no matter how large? If the whole building is meant, since we were in a Senior Center, with diverse activities going on all day, by numerous people, coming and going, are we to conclude that seniors can do chair exercises and play cards inside "a polling place"? We only wanted to stand in the foyers, between the two glass entry doors, but some official in the Franklin County Board of Elections decided that exit polling must take place far enough away from the voters as to be ineffective, especially considering the very cold and inclement weather. We need a clarification of what "polling place" means, especially in a public building of many uses on election day. The ruling of the judge was that exit polling is an exercise of free speech, since the results are reported by the media, as you are reading now. If we are not allowed to exit poll in a manner and location that allows us to freely ask the opinions of the voters who just voted, free speech is denied. Determine what is the meaning of this directive and get it in writing prior to exit polling. Arguing on election day means lost opportunities to exit poll.
3. THREE TYPES of EXIT POLL
I have seen three ways to exit poll. The classical way, done by the professional pollsters in the 2004 election, involves asking every "nth," say sixth, voter a series questions about who they voted for, some of their characteristics, and their opinions on the candidates. Also included is the demographic data of each person polled, usually age range, race, and gender. Every person who refuses to be polled, for whatever reason, is also described by these same categories. Later the results are "corrected" to take into account any skewing for oversampling one gender, race or age group.
(Note: You can view every article as one long page if you sign up as an Advocate Member, or higher).