Some people are convinced that George Bush stole the last couple elections. These claims are largely based on wild suppositions, the wildest of the suppositions being the claims regarding Ohio in 2004. In response to people's concerns regarding Diebold and possible vote tampering, a number of people are suggesting that after the election, election officials should do a sampling of ballots in every precinct to determine whether the machines are rigged. The proposals vary, but no one suggests more than a 10% sample of ballots be counted.
Under the sampling theory, we can know whether an election is fraudulent by looking at a mere 10% of the ballots. It's a ridiculous proposal, and will serve to increase voter cynicism and uncertainty about our vote counting process.
The proposal by the Election Defense Alliance has gathered a lot of attention and was brought to my attention after I spoke to the NIU law school Election law symposium in March. A person who is working with the EDA approached me about doing a pilot project in Champaign County using the EDA sampling proposal. I turned him down and referred him to our own vote count integrity program.
I offered a few points to him for which he had no response. Perhaps others do.
First, after a 10% sample, the results will all but certainly be different than the actual results. How much different will vary, but the likelihood of them being different by .5% is high. So for example, take the Myers Frerichs Senate race last year. Frerichs won by just under 1% (Frerichs - 27,149, Myers - 26,607, Parnarauskis - 1,894). Suppose we sample 5,565 ballots (10%) and we get the following results: Myers - 2,690, Frerichs - 2,685, and Parnarauskis - 190. This result is statistically not improbable. From a public relations standpoint it's a disaster. Certainly, those who understand or generally accept probability theory will not have an issue, except for those who have a vested interest in fomenting public dissatisfaction with the election process. A large number of other people will also have a problem because they simply don't understand the math. What we have done is not increased certainty, but rather increased uncertainty.
Second, the Election Defense Alliance suggests that they have a 99% chance of being accurate. That means, if they run this sample on every Congressional District, we will have four districts where the sample will show inaccurate results. In those areas, there will once again be a crisis of confidence.
Third, the random sampling will allow someone to program fraud into a system with a limited chance of being caught. So while the proponents of sampling speak of the high probability of 1% vote fraud being detected, they don't speak of the improbability of .2% vote fraud being detected. A computer programmer who wanted to drop one vote per precinct could have altered the Presidential election results in Florida, Iowa, Oregon, and New Mexico in 2000 and Wisconsin and Iowa in 2004. Sampling probably would not have caught this fraud.
Fourth, crimes are generally deterred by the probability of being caught. The Election Defense Alliance suggests that 40% of vote fraud would not be caught under a 2% sampling scheme as proposed in the Holt Bill. What that means is that anyone who rigs an election has a 60% chance of getting caught. I doubt that any computer hacker who is skilled enough to do all the programming work to steal an election would be willing to take that kind of risk. In fact, I'm not sure what criminal activity with a 60% chance of being caught is ever engaged in by reasonable people.
Fifth, the public has a right to know that their vote counting equipment is 100% accurate. Only a redundant recount in a sample of precincts can actually instill that confidence in people. When we conducted our redundant count, we were able to point to specific ballots that were likely the result of the vote discrepancies.
The Election Defense Alliance is sound in requesting some automatic triggers for redundant counts. I would fully support that. I was actually pleased when the election in the Village of Ivesdale went to a discovery recount. I wanted to know that indeed the votes were counted as they were cast. We found out that they were, and both the winners and losers in that contest can rest easy in the accuracy of the results.
Regaining the public's confidence in our election system is not beyond us. The most critical part of regaining that confidence is to provide facts instead of supposition and to engage in rational debate instead of lobbing wild accusations. Sampling is certainly not the answer.
While I agree with you that sampling is not the answer. However, as a Republican, I still have concerns regarding the Ohio election results in 2004. Those concerns center on the ballots cast and counted by Diebold, the absentee ballots acceptance/rejection system and the general Ohio administration of the election.
Oilman,
I am aware of some of the problems in Ohio. The problem in the debate about them is that they have been treated as the result of partisanship instead of basic incompetence. Ohio elections at the county level are controlled by a bipartisan election commission. The problems that have been detailed were the result of a joint effort of Republicans and Democrats.
Also, I don't like paperless ballots. But I haven't seen anything that leads me to believe that any results were tampered with.
Good post, terrible lead.
I've posted and linked to direct evidence of election tampering during the 2000 and 2004 election. Then it's ignored and a week later someone says "No has ever backed up these ridiculous charges".
After a while it transcends simple ignorance and moves into willful ignorance in which one chooses their own political agenda over basic democratic process.
It's disappointing to see coming from someone who has devoted so much of their time and energy toward supporting and improving the democratic process.
Perhaps it comes from the human failing of assuming that those above you are greater than or equal to you in terms of morality and ethics.
I agree with most of what is said here. The EDA plan is not a good idea for the reasons you state. EDA has a tendency to back some very poor ideas.
Far better election auditing schemes which use "redundant recounts in sample precincts" are discussed in depth along with pertinent mathematics at http://electionmathematics.org
Just to make a couple small corrections to what you said:
The Holt audit is 10%, 5%, 3% depending on margins. The Feinstein (S1487) audit is a 2% audit.
This statement is not precisely correct "The Election Defense Alliance suggests that 40% of vote fraud would not be caught under a 2% sampling scheme as proposed in the Holt Bill. What that means is that anyone who rigs an election has a 60% chance of getting caught."
The probability of "getting caught" (detecting corrupt counts when there is sufficient corruption to alter an election outcome) depends on the particular race, how close it is and the total number of vote counts in the race. i.e. the closer the race, the fewer counts need to be corrupt to alter the outcome and the less likely a2% audit would detect it.
This statement of yours is somewhat incorrect: "I doubt that any computer hacker who is skilled enough to do all the programming work to steal an election would be willing to take that kind of risk." In truth, there is no risk at all to the programmer. The only risk is that his hack would not be successful. i.e. there is virtually no way to trace a particular hack back to a particular person due to the lack security and accountability features of today's voting systems.
Please look at http://electionmathematics.org and click on the Election Audits page for a better analysis and a comparison of the effectiveness of the Holt vs the Feinstein vs. 99% success rate audits.
I agree completely with you that the EDA group is way off base, but their plan is not on the table anywhere to my knowledge, or is it?
I agree with most of what is said here. The EDA plan is not slightly a good idea for the reasons you state. EDA has a tendency to back some very poor ideas.
Far better election auditing schemes which use "redundant recounts in sample precincts" are discussed in depth along with pertinent mathematics at http://electionmathematics.org
Just to make a couple small corrections to what you said:
The Holt audit is 10%, 5%, 3% depending on margins. The Feinstein (S1487) audit is a 2% audit.
This statement is not precisely correct (i.e. not logical) "The Election Defense Alliance suggests that 40% of vote fraud would not be caught under a 2% sampling scheme as proposed in the Holt Bill. What that means is that anyone who rigs an election has a 60% chance of getting caught."
The probability of "getting caught" (detecting corrupt counts when there is sufficient corruption to alter an election outcome) depends on the particular race, how close it is and the total number of vote counts in the race. i.e. the closer the race, the fewer counts need to be corrupt to alter the outcome and the less likely a2% audit would detect it.
This statement of yours is somewhat incorrect: "I doubt that any computer hacker who is skilled enough to do all the programming work to steal an election would be willing to take that kind of risk." In truth, there is no risk at all to the programmer. The only risk is that his hack would not be successful. i.e. there is virtually no way to trace a particular hack back to a particular person due to the lack security and accountability features of today's voting systems.
Please look at http://electionmathematics.org and click on the Election Audits page for a better analysis and a comparison of the effectiveness of the Holt vs the Feinstein vs. 99% success rate audits.
I agree completely with you that the EDA group is way off base, but their plan is not on the table anywhere to my knowledge, or is it?
xian... do you still have a link to that information?
I've never been very comfortable with the all-electronic voting systems. I've worked with computers, with many years as a repair technician, to trust them with something as important as voting. While most companies do trust them to hold the bulk of their information they tend to maintain heavy back up processes to ensure against data loss and keep a great deal of hard copy information so that a system wide issue doesn't destroy them.
A single voting machine malfunctioning without any associated paper ballots can single handedly wipe out hundreds of votes or more. Something seen in the 2004 election. If one of the machines we have here dies, the paper ballots can be counted by hand instead and everybody's vote still counts. If there is a fear of tampering of the counting system all the paper ballots can be counted to ensure that the count was accurate without relying on the same possibly corrupted counting system.
I don't recall seeing anything that suggested a systemic stealing of the election via die bold machines in 2004. Flukes, malfunctions, and human error seemed to be the gist of the various problems. A great deal of it could have been avoided if they had an electronic system that still relied on paper ballots. I was even a bit hesitant about that until I saw it in action.
After years and years of debating the 2000 election I'm officially retired on the subject. :-)
--
Glock21 Op/Ed
Xian, you may have caught me at a bad time with those links in the past. Go ahead and post them again, or send them to me if you'd like. I've read lots of stuff on the "stealing" of the Ohio election in 2004 and find that all of it omits important information or makes wild assumptions. But I'll look at what you've got.
Sorry not to respond to this early. Chain of custody issues are critical as well and you're right to bring them into this discussion.
I'm convinced the 2000 election was stolen, but the people doing the "rigging" in that case were the SCOTUS.
The right way to do electronic voting is to use it for what it's good at, which is speed and keeping human fingers away from the ballots. The big problem with voting systems is what information security people call a violation of "least common mechanism". The votes (the idea that the voter is expressing) and the ballots (the medium by which the voted is expressed) are separable entities. Traditionally, that separation was done by counting. But with bad electronic machine design, there is no tangible record of the ballot itself, only the electronic copy, which is typically converted, aggregated, and stored in a database of some kind.
When the ballot and the vote are not separated, you make it easier to alter results.
I wrote a nice, long piece on this a while ago.
Loren,
I agree with you...electronic ballot counting is OK as long as there is a paper trail of some kind.