R-71 Wednesday update: Tally now tops 48,000
Election crews have now posted the latest Referendum 71 signature-check results, adding over 15,000 checked signatures, pushing the total to over 48,000. The tally now includes 43,147 signatures accepted and 5,142 rejected, with a slightly higher rejection rate of 10.65 percent.
A number of bound volumes, with 15 petition sheets per batch, were completed by the master checkers and added to the running total. The signature rejections were mostly due to the person not being found on the statewide voter registration database — 4,491 people. Another 388 rejections were because the signature on the petition did not match the one on file and 242 to disallowed duplicates.
The duplicate number was up substantially over the previous day’s 130. The duplicates are expected to push up the error rate as more and more signatures are checked.
Another 21 signatures are temporarily set aside while awaiting an electronic signature from the person’s home county. Typically most of the latter category is eventually moved over to the “accepted” pile.
Elections Director Nick Handy continues to say the signature-check is too close to call. Sponsors, Protect Marriage Washington, seek a statewide vote in November on the new “everything but marriage” law that gives expanded rights and responsibilities to state-registered domestic partners. The measure is on hold while the referendum is pending.
It takes 120,577 valid Washington voter signatures to qualify for the ballot. Sponsors brought in 137,689. That would allow a rejecton rate of no more than 12.4 percent.
The cumulative report is here and the more extensive report is here .
20 thoughts on “R-71 Wednesday update: Tally now tops 48,000”
If the current incremental increase in the cumulative error rate holds (i.e., it rises about .25% for each additional 15,000 signatures checked), then R71 will qualify by a razor thin margin, possibly a few hundred signatures. I am not sure, but maybe this would be the thinnest margin in WA history.
At that point, there will be some 120,000+ names that the Secretary will have refused to “master check” for no reason other than administrative cost and adherence to tradition. But if only a few tenths of one percent of that large accepted group are actually invalid, then R71 will have been wrongfully qualified and the Secretary’s willful refusal to master-check all signatures equally will have been the determining factor in R71’s erroneously qualifying. Given this, I cannot understand why the Secretary won’t at least consider doing a random master check on the accepted signatures to see if any are shifted to the reject column.
It is inconceivable that this won’t be the subject of a lawsuit if R71 qualifies, so why not just do it and save the time, expense and uncertainty? A random check could be done in a day and if no signatures were shifted, the exercise would demonstrate the integrity of the process. If signatures do shift, then that will alert the Secretary that the remainder need to be master-checked.
Won’t the Secretary or the Elections Division speak to this?
Susan Blakefield, I’m sure the Secretary of State is well aware that one or both sides will duke this out in court, but that’s not their position. As we’ve seen, one side has already approached the court and state agencies numerous times in order to turn out a more favorable and advantageous position than what’s mandated by state law, and we’ll see one or both sides approach the court again within the next 3 months. The Secretary of State, for all its stumbles in approaching information reporting, is neither a mediator nor a lawmaking agency, and ultimately shouldn’t be left holding the bag for adherence to rules or precedent without outside influence.
On the other hand, it’s quite odd what the duplication rate is doing. It’s a bit north of the margin of error for standard projections and seems to be sliding up a little faster than what modeling and stats have predicted. At the very least they should be commended for approaching a full check of signatures, since it would seem that this whole process is creating a statistics test-case (and I love stats!).
AJ
Your post is largely incomprehensible, so I will just say that this has nothing to do with the pro- and anti-R71 forces duking it out. This has to do very specifically with the way the Secretary is conducting the count. It is the Secretary who would be sued, not the proponents or opponents of R71. And since administrative costs are the stated reason for his refusal to master-check accepted signatures, I would ask whether a single day’s legal costs incurred in an inevitable lawsuit would not outweigh whatever marginal costs would be incurred by conducting a random master check of a sample of accepted names, which could be done in a day.
Double-checking all of the signatures, as opposed to just some, would not give an undue advantage to either side; it would simply make the count more accurate. The Secretary is not being asked to be a mediator or a lawmaking agency. He is being asked to conduct the count accurately, and that means treating the signatures equally. He shouldn’t do that as the result of “outside influence” but as the product of his own determination as to what law and equity require of his office.
As to your reference to precedent, David Ammons has acknowledged, there is no precedent sustaining the practice of double-checking only a select group of signatures. That is why there will be a lawsuit.
I agree AJ. This process, though clunky at times, has been intriguing to watch. I commend the Secretary of State’s office for attempting this level of transparency. The potential for embarrassments such as the revised reporting methods and “count adjustments” (etc.) is what makes most agencies shy away from doing this. As bad as these appear, we should take them for what they are: growing pains in the transparency process. I am very glad that they took the risk and we are able to gain a better insight.
Susan: Also keep in mind that if the Secretary of State were to change it’s *counting* policy (emphasis to differentiate it from the *reporting* policy) mid-stream for this issue and begin “master checking” accepted signatures as well, it would hand the Ref 71 sponsors a ton of ammo to use in the following (and inevitable) court battles. This isn’t to say that it is the “right thing to do” here, but the role of the counters is to follow policy, not establish it.
No matter which way this thing goes, it will get there by a razor thin margin (probably less than +/- 1000 signatures). Likely even less. You can rest assured that lawyers on both sides are prepping contingency lawsuits to be filed mere seconds after the signature count is finalized. Mark my words: When the count ends, the real fight will begin.
Susan B — in a post by David Ammons the other day, in a response to a suggestion that I posted about randomly double-checking 20,000 accepted signatures, he said he would talk to others in the SoS office about a random double-check. So let’s see what he says. Your logic, expressed by others as well, seems irrefutable — so what’s another 2 days of checking, compared to the certain cost of litigation if they don’t do any double-checking of any accepted signatures?
AJ — It does seem to be within the SoS’s discretion whether to double-check initially-accepted signatures. While nothing may require it, likewise nothing forbids it. And, once and for all, a significant random check (eg, 20,000 signatures = approx 15% of submitted signatures) would show if there have been statistically-significant errors in what signatures were accepted. As Susan B points out, a switch of only several hundred signatures (out of 137,000+) from initially-accepted to rejected (assuming there has been an error) will determine if the referendum qualifies or not.
Meanwhile, it is still hard to understand why someone — the State, or some of the “full disclosure” private requesters whom the State has proposed to add as litigant defendants in the federal lawsuit (Brian Murphy, Toby West, Arthur Nixon, or others?) — doesn’t ask the federal judge to dissolve the TRO NOW, and not wait for a hearing on Sept 3. From all of the postings on this blog, it’s clear that the federal judge should at least reconsider the balance of the equities/ harm to the public in determining whether to keep the TRO in place. Specifically, the federal judge apparently didn’t know, when he granted the TRO, that there would be a ‘perfect storm’ of two related- matters, one which was known to the judge but the second which seemingly was not known to the judge:
— As to the first matter known to the judge, apparently the judge was aware that by keeping the signatures secret, only the pro-R71 submitters and their allied observers at the signature-counting process would have information about each and every signature (name, address, etc), whereas the anti-R71 organizations and their allied observers would have ZERO informatoin about any of the submitted signatures. The judge was apparently untroubled by this disparity, instead finding that it was more important to eliminate the alleged potential harm in publicizing the signatures than it was to allow anti-R71 organizations access to the signatures.
— As to the second matter, this does not appear to be something the judge was aware of in terms of a ‘perfect storm’: namely, the judge didn’t know that there would be a disparity in the signature-checking process whereby initially-rejected signatures are double-checked — BUT no initially-accepted signatures are double-checked — AND that during this process there would be observers who could offer “comments and questions” to the checkers and double-checkers (per a posting by Dave Ammons). Even if such checkers are not bound to agree with the “comments and questions” of the observers, it’s clear that one set of observers (pro-R71) is highly advantaged compared to the other set of observers (anti-R71) because of the prior (pre-submittal) access to all signatures by the pro-R71 side. JUST IMAGINE sending 2 groups of ‘observers’ into a meeting and allowing them to comment on the documents being reviewed at the meeting — but one group has had access to/is aligned with an organization that submitted and saw those documents, yet the second group has been denied access to those documents before the review meeting. Does this seem fair? But that’s the result of the TRO…..and that may be something that would trouble the federal judge, and cause him to reconsider and dissolve the TRO NOW (not wait until Sept 3). Why isnt someone pursuing this matter?? After the signatures are counted, it may be too late……….
My last analysis shows that based on the rate of increase in duplication rate, the initiative is likely to qualify by about 200 votes. Susan makes a good point – running master checks only on rejected signatures unfairly biases the process towards qualification. I would love to see master checks of a couple of complete counts, to validate whether any of the previously accepted entries are reversed.
I agree with Susan. Statistically speaking, it is at best questionable and probably not appropriate for signatures which are initially rejected to receive a second check while signatures which are initially accepted to not receive any further scrutiny only because errors in the initially-rejected group will tend to be caught and corrected while errors in the accepted-at-the-first-check group will never be caught because your office decided not to look for them. What I’m concerned about is getting the most accurate possible count.
Like all processes, there is a certain error rate associated with each task. We have great people who care deeply about doing the best job possible, but they still make mistakes. That’s part of being human. The best we can do is to reduce that error rate as much as possible.
When we ignore potential errors in the group that accounts for 90% of the signatures, we’re not doing our best to reduce the error rate. I can’t imagine that a court would find otherwise–the only way that ignoring this source of errors could be upheld is if state law specifically mandates this particular checking procedure and if that law withstood a challenge based on accuracy and fairness–especially since errors of this type would invariably benefit one side, not the other.
As Susan suggests, we could reduce the uncertainty about this potential source of errors by taking a large-enough random sample from the larger accepted-at-the-first-check group and recheck them. By doing so, we can establish what the largest likely error rate is and have a statistical basis for deciding whether or not to recheck a larger group or even all of the signatures.
It looks as though this is going to be close enough that, if it were an election, automatic recounts would be triggered. Given that it’s likely to be that close, I would say that it is in the best interests of the people of Washington to do understand the likely sources of error in the counting process as soon as possible.
Also, is it possible to keep previous days’ data available online? It would be interesting and helpful to see the trends as the days go on. I’m especially interested in overall rejection rate and duplicate counts/rates, but it might be easier just to keep the entire page around and provide a link to it.
Why is the secretary of state continuous taking down the previous numbers everytime new fixed numbers come out? Where is the transparency? Why can’t I look at the numbers from the beginning of the counting to now?
Just want to throw in my 2 cents and agree with Stephen and others that the Secretary of State’s office should be commended for trying to make the process more transparent. I don’t necessarily agree with some of their policy decisions and signature checking traditions, but nobody can say that the office hasn’t been responsive to requests for clarification posted to this blog. And I salute you David Ammons, in particular, for fielding a barrage of questions and concerns with such patience and clarity.
I urge everyone to copy the secretary of state’s page since they continuously take down the old fixed numbers and put up new fixed numbers. How to copy web pages:
1. Press the Windows and Print Screen/SysRq buttons together.
2. Open windows paint, adobe photoshop or a similar program.
3. Press paste.
4. Save.
I want to agree with everyone here that the Secretary’s transparency is commendable. I have no criticism of the Secretary’s disclosure practices, and in fact, I think it is a model for other agencies.
On the disparity in treatment of the signatures, I have said my piece. My last 2 thoughts are:
1) in response to Stephen, changing the review process to give a second look at the accepted signatures would not be unfair. It would bring all signatures under the same procedure. The R71 proponents would have no argument that they are prejudiced by having the process be *more* accurate. We are not talking about singling out the accepted signatures for special scrutiny; we are talking about giving them the same scrutiny as the other signatures, in a double-checking process that the Secretary deems to be more accurate than single-checking. If master-checking this group of signatures yields genuine rejects, who is going to argue that those names should nevertheless count?
2) I totally understand why a reasonable administrator would not want to do a complete master check. It is more work. It delays things. And in all of WA history, it probably has never mattered to the outcome and thus would be a waste of time. But in this case, it matters. If R71 passes muster 200 signatures or anything like that, then a failure to apply equal scrutiny to some 120,000 names is going to be material. And millions of Washingtonians will doubt the integrity of the outcome. All of this is completely avoidable.
Thanks to all of the commenters today. Susan and others, I can now report that there will be some checks, perhaps in the neighborhood of 200, for both the accepted pile and the rejects, as requested by the official observers. This still doesn’t count as the full random or the second full every-signature check that some advocates request. As we have noted, the whole process is subject to judicial review, and if the court were to require a fuller re-check, or to require us to change our policy in this regard, then that’s obviously the judge’s discretion.
Paul Johns and Neil– the data is being released by volumes completed. so if you draw a line at the bottom of a day’s report, on the pdf Excel spread sheet, you continue right on as the next day’s volumes are added. it’s all building up, cumulatively, from Volume 1 to the very last one, Volume 623. If readers want the daily R-71 cover sheet report, as Neil points out, it’s easy enough to capture.
Jeannie–we appreciate your feedback. we’re very happy the material is helpful.
David–thanks for your replies and all your hard work. I’m glad to know that there will be at least some checks of the accepted signatures.
Am I correct in understanding that the signatures to be rechecked are only those requested by observers?
Statistically speaking, that is not as good as a random sample of a size sufficent to provide appropriate bounds for the false positive error rate.
I hope your office will decide to do such a check–doing so will give the voters more confidence in the process.
Second issue: thanks for the information on how the numbers are posted, but it doesn’t address the issue I had because it’s still impossible to go back to look at previous days’ data if you didn’t get it on that day. If you posted a small table that told which volume was the last done each day, that would solve the problem. It would be nice to post cumulative results by day, but those can be calculated from the spreadsheet if you know which volume was the last done that day.
By the way, it’s eaiser and quicker to save a web page by selecting “Save As…” or “Save Page As…” in the browser’s menu. This works in at least Internet Explorer and Firefox on Windows. I suspect it (or something very similar) will work in any browser–the feature has been in IE at least for many years.
Another question: Do the observer from each side have the opportunity to look at every signature? Are they in fact doing so? If so, that increases voter confidence (and should reduce the error rate).
Would it be possible to tell us how the observers work–what they get to see, and so on?
Thanks again for all your hard work.
J Scooter: good points.
Dave: please don’t bother explaining the observer process/policies. The explanation in the FAQ (which I’ve now read but hadn’t before) is quite clear.
The FAQ says each side gets only 3 observers to observe the work of 19 checkers and master checkers. (Using the FAQ’s numbers of 30 checkers and 8 master checkers hired, for two shifts a day–38 total divided by 2 is 19.)
That would mean that each observer is watching an average of 6 1/3 checkers.
Clearly the observers cannot watch each signature being checked. They cannot even watch all of the 4 master checkers on each shift (although they could come closer).
So observers can not take the place of a random second check of accepted signatures. That check still needs to occur for the sake of accuracy and understand the sources and likey sizes of errors.
A bit on the theory of errors in checking signatures.
For each signature, there are two possible actual validity statuses:
“actually valid” or “actually not valid” depeding on whether the signature is actually valid or not,
and two possible statuses as determined by checking:
“counted as valid” and “counted as not valid” depending on whether the checking processes counts the signature as valid or not.
In the vast majority of cases, the actual status and the counted status will match for each signature. In these cases, there is no checking error. Yea!
There are two types of checking errors. In both of these cases, the actual status of the signature is different from the counted status.
A “false valid” (or “false positive”) is when a signature which is actually invalid is counted as valid.
A “false invalid” (or “false negative”) is when a signature which is actually valid is counted as valid.
The procedure of having a master counter recheck signatures initially rejected reduces the “false invalid” error rate.
However, there is nothing in the process which reduces the “false valid” error rate. In fact, because we’ve done no additional checking, we don’t even know the potential range and probabilities of that error rate.
That is why J Scooter and I have asked that the Office Secretary of State do a random check on the signatures initially accepted as valid: we have no idea what the “false valid” rate is!
Once we’ve done this check, we will have an range and probabilities to tell us what this rate likely is. Only then will we be able to make a good decision about whether the rest of the signatures need a second check or not.
In my opinion, it is not responsible to ignore the issue of “false valids.” Doing so will almost assuredly to give an inflated count of “checked valid” signatures, which could result in the referendum being placed on the ballot even though it didn’t get enough signatures.
Paul Johns– Thank you and J Scooter for your string of comments. Re observers — it’s true that 6 people cannot observe each and every signature check and master-check, any more that the partisan observers in county elections offices can observe the actual tabulation of each ballot. for the relatively few ballots that are sent to the canvassing board, the scrutiny is much greater, of course. As I have noted before, the signature-verification process on I&R is assigned by the Legislature to the state Elections Division, and it is their duty to do their very best to “get it right,” knowing that the observers and potentially the courts will be watching. Regarding the second full check of the accepteds, again, that is a policy choice for another day if the longstanding approach is to be replaced by essentially are double-check of all signatures whenever there is any every-signature check. Either a judge or the Legislature (or voters by initiative) or perhaps even the OSOS by administrative rulemaking could change the process. As I have also pointed out, it is very rare to even have an every-signature check, since we usually are able to do random sample testing. At the pace we are going now, it will take a month to complete the process.
Comments are closed.