Late last month Townhall Magazine published an article by Ed Morrissey about the Coleman-Franken recount battle. In it I am quoted rather substantially about my role in the recount as the statistical expert for Coleman. I thought I should write some thoughts about it.

I was contacted initially by the Coleman legal team only on January 20th, a mere days before the contest phase of the recount was to start. I had just a week between that first call and the deposition -- in which I was supposed to look at the data, come up with an analysis, prep for a deposition process (for the first time) and go through it. I had all of two meetings with lawyers beforehand, both on the weekend.

What I was asked to provide was a demonstration that the rate at which ballots were rejected by county officials differed in a statistically significant way from each other. Rejection rates are normally the domain of binomial or Poisson distributions, and we economists don't use those distributions as often as the normal or Student's t. So I took some time re-familiarizing myself with those tests, looking up some books, digging up my old college text, and arrived at what was pretty obvious looking at the data casually -- those rejection rates didn't vary due to chance.

So it's a very compressed time period. Some people seemed to enjoy having fun picking apart the deposition, but to that I would just say the following:

1. The decision to not call myself a statistician was mine. Any social scientist uses statistics, pretty much as an everyday activity. I don't use sampling theory daily, but something or other in statistics, sure. But what you might call a statistician would hold a PhD in statistics. I don't. The fellow we thought would be used by the Franken people didn't either -- he was a sociologist. But once some court accepts you, you're in, as best I can tell. He was, I had not before, and so the attack to exclude me was pretty standard procedure. (I taught college statistics for social sciences at the Claremont Colleges, and I teach economic forecasting at SCSU.)
2. What I was initially asked to look at was to show the distribution of rates of rejected absentee ballots among counties was not random. I assumed what they were trying to prove was the Bush v Gore point that voters did not enjoy equal protection of their absentee ballots. The contest panel, and eventually the Minnesota Supreme Court, did not accept the precedence of Bush v Gore. Without that, there was no reason for me to testify based on what I had given the Coleman lawyers at that time. Where many, even Ed, imply that I was denied the ability to testify because I was not an expert, that is not what the court said. They were not interested in the Bush v Gore argument, thus they had no need to hear of me regardless of whether or not I was an expert. They had decided, in Ed's words, to " on individual ballots rather than categories and generalities." From their decision:

   The only question that can be decided in an election contest is which party received the highest number of legally cast votes, and therefore is entitled to receive the certificate of election. The Court will be reviewing all ballots presented according to the uniform standard contained in Minnesota Statues Chapter 203B. It is irrelevant whether there were irregularities between the counties in applying Minnesota Statutes � 203B.12, subd. 2. prior to this election contest. The Court does not believe Banaian's testimony would assist in determining the issues properly before it.

   And with that fact's irrelevance, I became irrelevant to the court. As I am not a lawyer, I have no opinion on their decision.
3. At no point during the weekend before the deposition was I asked by the Coleman team whether I thought opening which ballots would lead to a Coleman victory. So when I told Ed my thoughts -- that had they opened the ballots I thought they could make an argument for, those from counties with excess rejection rates compared to the state average, they would not have enough net ballots to win -- that was my own speculation done actually the day AFTER I had been deposed. Had I gotten to the stand, based on what I knew when I was deposed, I had no answer. Based on what I did afterward, I did not think I could argue that even the statistical argument had any real chance of succeeding, particularly after the court had awarded Franken a net 87 additional votes in opening 351 ballots previously rejected. At that point, I thought, the game was up, though I admit to some cognitive dissonance over it -- I didn't really want to believe what I had found, and since I couldn't talk about it while the trial was going on, I pretty much buried that from my consciousness.
4. My analysis of those counties where rejection rates were statistically significantly above the state average, was that an 'extra' 1,924 ballots had been rejected. If those were distributed as the recorded vote was, Coleman would have gotten 841 additional votes and Franken 737 additional votes (the remainder to Dean Barkley and the other candidates.) But that called for an "add factor" -- you can't pick the ballots that were rejected wrongly based on a spreadsheet, which is all they gave me. And as I say, the court rejected that idea.

So to summarize: I did an analysis very quickly for a question the court didn't interest itself in, and when the Franken campaigned asked to keep me out, they ruled my testimony irrelevant. But had they asked me if I thought Coleman won, I would have said "probably not, based on the data."

Let me close with an agreement with Ed's premise for his article, which was butchered by the Townhall editorial staff when they chose the title. Franken did not steal an election. They played hard, harder than the Coleman team. As I said to someone after the deposition, it's one thing to take a knife to a gun fight, it's another thing to be the knife. Perhaps I was a fallback plan they came to late in the process; far be it from me to criticize the Coleman strategy when I know so little about it. But it appeared that after the decision to reject statistical argumentation -- which Ed argues should have been known to the Coleman lawyers based on the Rossi-Gregoire recount -- that the energy of the Coleman argument was lost, at least at that stage.

Labels: Coleman, elections, Franken, Minnesota