Adolph & Handcock for the Democrats
Are the Democrats’ experts playing a shell game in their reports on the question whether illegal votes changed the outcome of the gubernatorial election?
Two reports written by expert witnesses for the Democrats have been posted at the Secretary of State’s web site.
While there are other points made in those reports which may be sound, one criticism of the Republicans’ expert witnesses’ reports is common to both and seems to miss the point.
Assuming the judge is correct that prevailing under RCW 29A.68.110 requires proof that “illegal votes” changed the outcome of the election, then the question is whether votes which can be shown to have been illegal votes caused Gregoire to be declared the winner of the election.
The question is not whether all illegal votes – known and unknown, proven and unproven – changed the outcome, but is instead whether the ones that are known and can be proven changed the outcome of the election.
Note how the two expert witnesses for the Democrats base their criticisms on the inability of the Republicans’ experts to know whether they have included all illegal votes in their analyses.
Note especially how Mr. Christopher Adolph shifts from talking about the need to include “all invalid votes” to “these alleged invalid ballots” which have been identified by the petitioners. The absence of a complete “census” of all invalid ballots is not only a flaw: It is possibly a bias designed into the data to favor the petitioners.
In a perfect world, we would have a set of data that included all illegal votes cast in Washington’s gubernatorial election.
This isn’t a perfect world, as evidenced by the careless way in which the election was administered in King County.
So, the question in this imperfect world is: What was the effect of the illegal votes which we do know about?
Here are relevant excerpts from the reports by the Democrats’ expert witnesses. Are they dealing with the actual question in this case or trying to shift to a different question?
Page 2 of the report written by Christopher Adolph:
Page 3 of Adolph’s report:
Page 4 of the report written by Mark S. Handcock:
Two reports written by expert witnesses for the Democrats have been posted at the Secretary of State’s web site.
While there are other points made in those reports which may be sound, one criticism of the Republicans’ expert witnesses’ reports is common to both and seems to miss the point.
Assuming the judge is correct that prevailing under RCW 29A.68.110 requires proof that “illegal votes” changed the outcome of the election, then the question is whether votes which can be shown to have been illegal votes caused Gregoire to be declared the winner of the election.
The question is not whether all illegal votes – known and unknown, proven and unproven – changed the outcome, but is instead whether the ones that are known and can be proven changed the outcome of the election.
Note how the two expert witnesses for the Democrats base their criticisms on the inability of the Republicans’ experts to know whether they have included all illegal votes in their analyses.
Note especially how Mr. Christopher Adolph shifts from talking about the need to include “all invalid votes” to “these alleged invalid ballots” which have been identified by the petitioners. The absence of a complete “census” of all invalid ballots is not only a flaw: It is possibly a bias designed into the data to favor the petitioners.
In a perfect world, we would have a set of data that included all illegal votes cast in Washington’s gubernatorial election.
This isn’t a perfect world, as evidenced by the careless way in which the election was administered in King County.
So, the question in this imperfect world is: What was the effect of the illegal votes which we do know about?
Here are relevant excerpts from the reports by the Democrats’ expert witnesses. Are they dealing with the actual question in this case or trying to shift to a different question?
Page 2 of the report written by Christopher Adolph:
The three independent problems in Gill and Katz are:
1. Use of a non-random, non-representative, and incomplete sample of invalid votes, which is useless for answering questions about the net effect of all invalid votes on the state-wide election outcome. (Emphasis added.)
Page 3 of Adolph’s report:
Are the petitioners’ data a census, a random sample, or a representative sample of invalid ballots? Neither Gill nor Katz claim they are. This omission is unusual and conspicuous. The first thing another scholar would ask about these reports is “Where did the data come from? Are the data a valid sample.” Unless the petitioners can make and support the claim that these data are a representative sample of invalid ballots across the whole state, it will be impossible to make even minimally valid scientific claims regarding the likely effect of these alleged invalid ballots on the election outcome. In particular, if petitioners over-sampled precincts that voted overall for Gregoire, then their method will tend to produce biased results suggesting, perhaps incorrectly, that Gregoire benefited from the inclusion of invalid votes.
Page 4 of the report written by Mark S. Handcock:
2.1 Assumption I: Their data set is a known fraction of the invalid voters
This assumption states that the list of invalid ballots in their data set is either: (1) a complete enumeration of the (sic) all invalid ballots cast in the election, or (2) a sample of the invalid ballots, where the probability of each ballot entering their sample is known.
The first approach would claim that the set of invalid ballots in their dataset is a true census of the invalid votes cast. This means that they have every invalid voter and have not included any valid voters as invalid voters. If this assumption was correct, then there would be no question of “design based” statistical inference from our sample to the population of invalid voters: this is the population of invalid voters. Assuming for a moment that we knew how these invalid voters had actually voted, then we would be able to say with certainty whether these votes had changed the election outcome. The findings obtained from our data would be the findings of interest, there would be no need to evaluate whether an inference can be made from these data to the set of all invalid votes (and by extension, to the election outcome). (Emphasis added.)
posted by Micajah | 6:19 PM
2 Comments:
You said:
"The question is not whether all illegal votes – known and unknown, proven and unproven – changed the outcome, but is instead whether the ones that are known and can be proven changed the outcome of the election."
I agree. I made the same point in a good discussion with a stat expert on Goldy's blog. He believed that the statistical analysis needed to show globally that all of the illegal votes changed the outcome, rather than just those identified by the litigants. I made the same point you are making.
However, rereading the language of section .110, I can see that a weak argument can be made for the Dem experts' position.
Section .110 reads: "unless it APPEARS that an amount of illegal votes has been given to the person .... after deducting therefrom the illegal votes that may be SHOWN to have been given to the other person."
One can argue that the word "appears" and "shown" are general enough, so that the plaintif/contestant, must show at a minimum that their sample of illegal votes is not cherry picked or totally biased. I doubt this argument will fly, given we live in an adverserial legal system, with thte burden on the parties to present evidence supporting their position.
I've skimmed the Dem reports, and while the Dems experts make some telling criticisms of the GOP expert's methods, the Dem reports are way to jargon laden, complex, and academic for a lay person to understand with ease. I say that as someone with some statistical training.
We may know early on Monday whether the judge will admit into evidence as expert testimony the "proportional analysis." I was surprised to see that the Frye hearing wasn't scheduled for an earlier date.
Jargon or not, it seems to me that the Dems' experts hammered often enough on the absence of evidence or research showing that it's valid to assume that illegal votes were cast in roughly the same proportions for each candidate as the legal votes.
The judge may let the evidence in, but I doubt it could meet the "clear and convincing" standard that he seems to be inclined to use. It's better than nothing, but not necessarily clear and convincing.
One of the Dems' experts plainly said that identifying the candidate who got the benefit of those votes is a hopeless case. Ironically, that seems to be the best argument for reading "appears" to mean something other than a showing by clear and convincing evidence.
Of course, there's always the option of calling one or two thousand people to the stand and asking them how they voted. Although it seems impractical, the judge may not believe impracticality forces him to accept some other way of showing how they voted.
Post a Comment
<< Home