#48#
[next in a series covering the Washington State 2004 gubernatorial election contest trial, being conducted in Wenatchee, Chelan County Superior Court , before Judge John Bridges. Previous postings in the series:
Day Three Wrapup
Day Three Morning Session
Day Two Wrapup
Day Two Morning Session
Day One Wrapup
Day One Morning Session]
As I noted, Mr. Goldstein already took the apparently obvious "want Frye with that" joke, and it's perfectly indicative of the pace at which I was able to follow the trial today--a half step behind. And by 3pm I was ready for a live cloture vote, so I missed much of one witness altogether. And it's quite late now. But I did catch a fair bit of the morning session and I do have some notes, and I can fill in the gaps from the above diary from Horse's Ass, dj's quite excellent Frye hearing breakdown, and Postman's seemingly way long trial dispatches today. I'm going to assume tonight that you're familiar with most of the players, which you would be after reading the others.
I picked up with counsel Burman doing cross on Polidata founder and well known data geek Clark Bensen. He was the middleman in the process, getting the data from the GOP/BIAW and passing it on to Katz. As a lay witness there's not much to poke Bensen on, but Burman does get some things accomplished. He asked Bensen about a lot of other data on illegal votes in counties beyond King. He wondered, did you provide any data on these? Bensen replied, "I wasn't asked to." He also wondered whose idea was it to go with proportional reduction to get a vote count, and that turned out to be the GOP. Implication: you were a data monkey who was handed some presorted shit before you flung it across the cage to Mr. Katz, is that correct? Perhaps I'm paraphrasing a bit there.
Next up was Katz, triggering the official start of the Frye hearing. Katz sure was a nervous fish, and I suppose he had every right to be; I get a knot in my stomach any time someone at work asks, "why is this number not the same as that number on this other report?"--even if I realize it's supposed to be different. That's not on TV in front of a judge. But it was clear he had rookie jitters. I did catch him quite a bit on video, and he was overly defensive, too quick a talker, too repetitive and vaguely evasive in his answers, and in both my personal and professional view unconvincing. As others have pointed out, the classic exchange came early, as Katz worried, "I'm not sure how this all works." Bridges helped him out by explaining, "I believe Mr. Burman's going to work you over." It's going to be hard for the right wing to hate this judge if he rules against them; he's been dynamite even in his exhaustion and frustration.
In response to being worked over, Katz does show a reasonable facility with the language and knowledge of the subject, but it's because of that, that I find his answers sounding more like grudging admissions and opaque denials. He proclaims proportional reduction the "best method, given the limitations of the data," and repeats the second clause several times to different questions. He claims as justification for his methodology, "the justification is that the actual information isn't available." He is forced to admit, "with great difficulty you might be able to get other data (referring to race, sex, etc.--other ways to divide the vote pattern). And he says (reasonably given his position) that not only felon votes but ALL illegal ballots could be expected to conform to the precinct level vote preference breakdown.
Then Burman zooms in and does a very nice job of both setting up his time with Democrat expert witnesses, and attacking Katz's methodology in scientific terms. In a nod to sample design, Burman asked whether the GOP had selected the precincts and given them to Bensen, who passed them on to him. Katz only copped to knowing where he got it, and not questioning the rest. Then he put Katz to the question: To determine if illegal votes changed the election, wouldn't you need either a full census of them, or a scientifically derived sample of them? After much hemming and hawing and back and forth, Katz finally admitted, yes, that's right. You don't know how they designed the sample, Mr. Katz? That is correct. You can't say if it was scientifically designed? No, can't say. Pretty good for a lay lawyer.
As much as Tacoma Tribune writer Ken Vogel wanted to paint a difference between the squeaky clean, officious sounding Auditor of Chelan County and birdlike, hesitating Bill Huennekens, that overstated difference pales in comparison to the gap between Katz and his former mentee at UDub, Christopher Adolph. Adolph had also never been an expert witness, never testified, and was grilled hard by Rossi counsel, but he was entirely unflappable and clear. Evidence of his ability to get through on layman's terms was when Bridges asked Katz later in the day if he bought Adolph's example of ecological inference bias: could we have predicted Ichiro's .372 batting average in 2004 based on the league average of .270? Katz was forced to admit he found no flaws in the analysis.
Adolph centered his objection on two prongs, as Burman had prefaced. Adolph criticized Katz's choice of a selective sample instead of a scientifically derived one, and also his choice to use ecological inference as the best method of getting an accurate answer. If they're going to make inferences, Adolph says, the data have to be validated. Asked by Burman if he would accept Katz's analysis--his former boss, now--for peer review, Adolph said no. Why? "Selection bias."
Adolph then goes into detail on the ecological fallacy. For fifty years, he tells us, social scientists have warned against relying on the method exclusive of any other information. Among his statements:
- aggregate conclusions do not necessarily describe individual behaviors
- science says confidence intervals for your conclusions are required
- Using the Katz method for estimating felon gender by precinct gender breakdown, the range should be between 44 and 50% male. In reality, the felons are 75% male
- Felons are much more likely to be male, and that's clearly not shown in the aggregate.
- No information in the presented aggregate data could solve the problem of ecological inference.
- and as almost an aside, "I'm not aware of any research into the voting behavior of ex-felons." Remember that statement.
Braden tries to take down Adolph using the ambitious former student technique, and also the unqualified tenure-tracked-but-not-yet-tenured-professor maneuver. Confronted with the Katz method as the best available test in the record (although why being better than nothing is a given here), Adolph says he thinks it's possible to design a survey that would do a good job of identifying all illegal voters. Quite a statement.
So in a little bit, Braden comes out on Adolph about there being no actual felon voting behavior data--most places don't let them vote until they repent, and some not until after parole. Except he uses the Uggen/Manza study that hit the news a while back, indicating in their opinion that had felons votes been cast in the 2000 Election, Al Gore would have won and the Democrats would have seized the Congressional Minority. I slept late, but I think I would have remembered that happening. Adolph is clearly waiting for that study to come out, and notes the difference--it's not data about how felons voted; it's data about how they'd vote if they weren't barred from voting. There are in fact no good data on how felons vote, since in most places it's frowned upon for cons to have the franchise.
It is at this point that my notes run out. As you have probably derived, Bridges denied the Democrats' Frye motion to exclude the evidence--but hinted that he thought the evidence wasn't all he'd expected-- and we moved on without full resolution. There was more testimony from a Dem witness and a GOP witness; Handcock and Gill. From what I can tell they both echoed back what their morning counterparts had said, although they claimed independent discovery.
I should be able to keep a better ear out tomorrow. Thank you, please come again! </apu>
Comments