Reply #226: RE- exit polling evidence highly admissible

[davidgmills]

With expert testimony.

And certainly relevant on the issue of whether Kerry won instead of Bush. Goes straight o the heart of an election contest suit. In fact part of the proof necessary for an election contest is to show that the outcome was questionable and likely wrong.

The proof is subject of course to expert testimony that the exit poll was flawed or inaccurate. But that takes rebuttal expert proof.

Hypothetically, suppose you had a highly accurate exit poll and you could get experts like Zogby and Harris and Gallup and about 20 statisticians, mathematicians, computer scientists, and social scientists from the premier universities in the country to testify that the polls almost surely were indicative of widespread computer fraud. Suppose it is widely covered by the media as it would probably be if this were blockbuster evidence.

Now suppose that the defendants can only come up with rinky dink local yokels to counter this testimony.

The judge is going to be in a real box and if it is a jury case (I have no idea if election contests are jury cases or not), maybe the jury goes your way. All hell breaks loose.

By the way, I am not talking about any rinky dink exit poll either. Here is a copy of an article I submitted to Informationclearinghouse for publication yesterday. They have published a number of my articles before but who knows whether they will publish this one.

...............

Why We Must Re-exit Poll Ohio Now

The stalemate. Late on election day, and well through election night, Kerry was winning the exit polls, causing many to predict his victory. Then something happened. Despite being substantially ahead in many of these election polls, his lead dramatically disappeared.

The day before election day, some pollsters had Kerry ahead, some had him slightly behind, some had it a dead heat. Virtually no pollster had him behind by 3%, Bush’s ultimate margin of victory, and that certainly was not the consensus of the polls. Most had Kerry ahead in critical swing states.

Immediately, because people were told that exit polls are extremely accurate, something appeared to be amiss. Soon there were arguments on both sides about whether the exit polls were accurate or not. In short, a stalemate began to exist between those who believed the exit polls were accurate, and those who believed they were not, and that is where we still are eight weeks later.

The reason for the stalemate. The reason for this stalemate is due to a polling phenomenon that statisticians call “clustering.” Before I describe it further, let’s look at some basic background information on polling.

In pre-election polling, potential voters are called at “random.” In pre-election polling, the uncertain variable is who will actually vote. The “likely voter” variable causes most of the margin of error prior to election day. Pre-election pollsters rack their brains trying to figure out who will be “likely voters.” While exit polling solves the problem of trying to figure out who the “likely voters” are, it creates one of its own, one called “clustering.” The experts say that exit polls are not “random” polls, but instead are polls of “clustered” communities.

An example of putting marbles in a jar will illustrate the difference between the two. If one puts 500 red marbles and 500 blue marbles in a jar and one mixes the jar thoroughly until the marbles are totally random, then when 50 marbles are randomly taken from the jar, the expectation is that one will get very close to 25 reds and 25 blues. Then it is very easy to predict that the jar is full of half reds and half blues with a high degree of accuracy.

If however one puts “clusters” of 50 reds and “clusters” of 50 blues in a jar and the jar is not thoroughly mixed, but left in “clusters,” then it is easy to see how if 50 marbles are taken randomly from the jar, one might get 32 reds and 18 blues, and consequently, be totally wrong about what percentage of reds and blues there are in the jar.

The experts say that pre-election polling is like the first example; exit-polling like the latter. They believe that many precincts where the polls take place are clusters of Democrats and clusters of Republicans and that these clusters must be accurately taken into account or their predictions will be off.

The best known analysis of the exit polls has been done by Dr. Freeman of the University of Pennsylvania. When Dr. Freeman did his first analysis, he was accused of having “failed” to take into account the “clustering” effect when he came up with his now famous 250 million to one odds of something being terribly amiss. When he was informed of his “mistake” he factored in a “clustering” correction that reduced his odds to 650,000 to one. Later his critics claimed his “clustering” correction was still not adequate and that he had still underestimated the “clustering” effect. (Experts refer to it as a “design effect”, but I will use the term “cluster effect”).

But just how much should one correct for the “clustering” phenomenon? No one knows. Others, as well as I, have many questions about it.

To begin with, the effect of clustering should only occur with small sampling. Using the example of the marbles in the jars again, one can see that if you sample 50 marbles in 1000 clustered jars, the 35 reds you get in one jar will be offset by 35 blues in another and the effect at some point will cancel itself out. The higher the population, the more samples will be required; thus as our population has increased, the clustering effect should be headed towards extinction. Surprisingly though, the "experts" say the clustering effect is heading in the other direction. According to expert pollsters, pollsters used to account for a 30% clustering effect in 1996 (the number Dr. Freeman used in his corrected analysis), but now claim that for 2004 it has increased to between 50-80%.

My second concern is whether we really are in fact polling in "clustered" places. More and more I see graphs of a purple America, not a blue and red one. Yes, there are a few truly red areas and a few truly blue areas, but I have a hard time believing that the average polling place is much greater than a 70/30 split. Probably most are less than that. This would certainly reduce any cluster effect.

Third. If there is a cluster effect in exit polling, it should be evident when compared to pre-election polling which is supposed to be random. Random sampling and clustered sampling are not supposed to produce the same results. But in a study done on Democratic Underground by a mathematician (who uses the pen-name, Truth is All), he compared the very last pre-election polls with the latest "non-corrupted" (meaning mixed with actual results) post election polls, and instead of the numbers being different as one would expect if there were a clustering effect, they were almost identical. He is one of the mathematicians who believes that the cluster effect concept is bogus, at least as it applies to polling. He was not attempting to show the cluster effect was a bogus concept by the study. I pointed it out to him. He thinks the cluster effect is so bogus he doesn't even want to engage in the argument. But his results should be evidence that the claimed cluster effect doesn't exist or has cancelled itself out due to the volume of samples.

My fourth problem is a business one. If you include the cluster effect in your calculations, it means that margins of error are easily 3.5% or more in some of the best of situations. I cannot imagine networks paying the millions of dollars they pay for election night to only be within a 3.5% margin of error when the polls close. You would not be able to call any of the swing states on anything like the timetable the networks want. I don't buy the argument that they pay all this money for other data. Sorry.

Fifth. Another mathematician on Democratic Underground, (who uses the pen name macdonald and who believes in the concept of clustering) has reworked the numbers using the highest numbers for the clustering effect. His conclusion is still the same as Freeman’s (i.e seriously amiss) but with not nearly the 650,000 to one odds. But even he has no idea what the real numbers for cluster effect should be; he's just using the latest ones the experts have thrown out.

Sixth: Why were exit polls apparently so accurate in the past but not now? To be fair, some dispute this notion.

Seventh: Why are they still accurate in Europe? Presumably they have political clusters there and they can get it right. Again to be fair, some dispute just how accurate they are.

Eighth: Could it be that whatever factor you use for clustering is only as good as the honesty of the last election?

Ninth: Could it be that gradually increasing usage of computers accompanied by computer fraud has caused these numbers to inflate?

Tenth: Another mathematician on Democratic Underground (who uses the pen name jwmealy) has shown that the exit polls of those states with paper ballots were far closer than those that used paperless systems. He suspects computer fraud.

Regardless of its reliability, or lack of it, the exit poll data is still driving the uncertainty about the election, with uncertainty likely to persist.

The best solution to the stalemate -- Re-exit poll Ohio now in a do-or die poll. Ohio was the key swing state in the election; Kerry was leading in the exit polls by over three percent until exit poll numbers were “contaminated” with actual numbers; it is also the state where several lawsuits have been filed questioning the election results.

Exit polling Ohio now would be random and not clustered. It would also eliminate the variable of who is a likely voter and who is not. A proper exit poll now of 16,000 Ohio voters would produce an accuracy, I am told, of 1%. I am also told it would not be unbearably expensive.

Given all the questions raised in Ohio about computer fraud and questionable ballots, it would give us a really good idea of what the voters of Ohio intended. To me, a highly accurate exit-poll is nothing but a very valid check of the system. If the exit poll results do not square with the actual results, a red flag is thrown up and people are almost obligated to check the system out.

Suppose for example the news tomorrow was that three highly respected pollsters, Zogby, Harris and Gallup had, in a joint venture for the good of the country, exit-polled 16,000 voters in Ohio and found that of Ohioans polled, 52% had voted for Kerry and 48% for Bush? Suppose they also said their poll was accurate within 1%.

What would happen when the news got out? Not only would the media be covering it, but all judges in the Ohio election contest suits would instantly take a different viewpoint on their cases. Would Kerry then look like a sore loser if he got active in the Ohio lawsuits? Would what Conyers is doing be taken seriously? Might some senators balk on January 6?

I think the answer to all these questions is yes.

I am a lawyer, so I tend to look at what is going on from a legal perspective. From the standpoint of a lawyer, every lawyer knows of the “harmless error” rule. If you want to have a case overturned, you must convince a court that the error was not harmless and that it likely made a difference in the outcome of the case. This harmless error rule, whether consciously or not, is in every judge’s mind. If you want to prove that there was fraud in an election, you also have to prove that the fraud likely caused a loser to win and a winner to lose. Unless the margins of victory are incredibly small, how can you prove that there is a good chance the outcome of the election was in doubt? I say you need an exit poll, and one which the scientific, mathematical and statistical community almost universally accepts as valid. Right now the status of the 2004 election is the electoral equivalent of the “harmless error” rule.

Certainly, the litigation in Ohio is staring the “harmless error” rule in the face. If you want really good discovery, you need the judge to be on your side and hammering the hell out of the opposition. If he doesn't, they will stonewall the hell out of you. To get the judge's attention, you have to have him believing that your case probably will make a difference in the outcome of things. If the judge believes that the possibility for making a difference is slight, he is likely not to give you much help.

But a really accurate exit poll, which showed Kerry won or should have won, changes the dynamics, especially given what has happened in the Ukraine. A judge will see the double standard instantly if the Republicans start to argue that a really legitimate poll was flawed.


Did the Ukraine elections get set aside for any reason other than the exit polls? No.

In hindsight, I think one of the major mistakes of the 2000 election, is that extremely accurate exit polls weren't taken. If the Supreme Court of the US had been faced with extremely accurate exit polls of Florida that showed Gore won Florida by 2%, maybe they would not have decided the way they did. Judges are political animals. It would have been a much tougher decision for the moderate voices on the court.

Moreover, if a new exit poll of Ohio shows the opposite of a Kerry victory and Bush wins the poll 52% to Kerry’s 48%, would most of the major election controversy go away? In a heartbeat.

So I say, for the good of democracy, let us re-exit poll Ohio with reputable pollsters right now and get it done before January 6.



David G. Mills is an attorney who practices in Memphis, Tn.