Oregon’s civil-union law passed the legislature this year, and a group collecting signatures to put this law on the statewide ballot in 2008 missed the number of valid signatures by a very small margin–96 short of the required 55,179.
Now the law is blocked in federal court pending the outcome of a lawsuit challenging the state’s method of verifying signatures. And if the implications of the phrase “random statistical sampling method” are what they seem to be, the challenge isn’t coming any too soon.
Attorneys for the state said that procedures verifying signatures are applied equally to everyone. The state uses a random statistical sampling method to determine whether enough valid signatures are collected.
The Oregonian story is long on personal anecdote, including the old, tiresome tarradiddle about inheritance and medical decisions between gay couples, but short on detail as to how the state goes about verifying the signatures. “Random statistical sampling method” certainly suggests that they take some minority number of signatures and put them through a verification process–and then assume that if two out of ten of those signatures are invalid, then two out of ten of the whole number are invalid. And whether that nutshell I just gave is literally correct or not, this is a document full of math to ascertain that the process is definitely accomplished by estimates and not by verification of the whole number of signatures.
The signature verifications are made in stages. First, a random sample of 1,000 signatures is verified. If the petition does not qualify from the first sample, then the second larger random sample is verified. Qualification of the petition is then based on an estimate of M [where M=number of distinct valid signatures] made from the combined first and second samples. In the event that the petition is not qualified from the combined sample and a second submission is made as permitted by ORS 250.105(3),a random sample of signatures from the second submission is verified. Qualification of the petition after verification of the sample from the second submission is based on an estimate of M made from the samples in the first and second submissions. The methods for determining whether a petition has a sufficient number of valid signatures are described for the following three cases: after verification of the first sample, after verification of the second sample, and, when applicable, after verification of the sample from the second submission. For each of these three cases, examples are given to illustrate the numerical calculations and conclusions.
So in other words, the state “verifies” (unfortunate choice of word, considering how the process is accomplished) a “random statistical sampling” of signatures that are supposed to signify the position of a voter on an actual election question. But instead they’re treating it like a poll–if a random sample is in this proportion, then so must be the whole.
And that’s outrageous. It’s unfortunate in a way that this lawsuit should take place over this particular case–the protests will be wild and vehement, and the motives of anybody who thinks, as I do, that the state’s procedure is a fraud will be questioned to the hilt–but it’s high time this was looked at in detail. I certainly never knew they verified those signatures by poll rather than by count. I’m no statistician, but when I think of 96 short of 55,179, the phrase “not statistically significant” floats to mind.
I wish the judge a stalwart backbone and the plaintiff a good set of mathematicians.