At the Kulongoski & Dean rally last night, Earl Blumenauer summed up this voting rights issue perfectly:

[Paraphrasing here] "If we can provide a printed receipt for every $20 ATM withdrawal, we can provide a printed receipt for Democracy."

blumenauer's exactly right. democrats have cared about counting votes accurately for years. sadly, the republicans in power keep giving money hand over fist to companies making machines that just don't make sense.

What exactly do you expect county election offices here in Oregon to do that they are not doing already? Here in Washington County the election office runs tests with thousands of sample ballots before and after the actual count to verify that the machines are working correctly. They run samples for each type of ballot that they will be counting (in the May primary here in Washington County we had hundreds of different ballots and each had to be checked to make sure that the scanning machines were set up correctly). They even do some random additional recounts to make sure that things match. What else do you need them to do before you are confident that the results of our elections are valid and accurate?

I am much more concerned with the lack of tracking for our voter registration drives, and the potential privacy issues with our current registration forms. We had documented abuses from Republicans in 2004 here in Oregon. With the increase in identity theft, we need to look at how we can improve the security of our voter registration system.

When voters have doubts as to whether the voting process has been corrupted, that is a problem in and of itself. A hesitant individual who in the past wondered if his or her one vote counted, now needs to worry if the count has been rigged. Loss of faith might become loss of involvement.

Knowing how simple the electronic tampering could be, and having lived long enough to see my share of political scandals & coverups, yeah, I am very suspicious and have been since 2000. Or was it the Ollie North era? Nixon? Viet Nam? The Kennedy assassination? Can you be born a cynic, or is it a conditioned response?

Gee, could this voting machine coup be why the Administration haa always seemed utterly unconcerned about public opionion? hmmmmmm

Lol. The voting machine conspiracy always brings out the nuts from the left.

So the Deomcrats say "If we can provide a printed receipt for every $20 ATM withdrawal, we can provide a printed receipt for Democracy." Okay - does Ted have a proposal for this? I mean, your party has been in the governor's chair for a long time.

I know that it is easy to always blame the Republicans at Washington and feel that you do not have to provide an alternate solution (not you guys - Dem. politicians). However, don't start that here in Oregon. If there are problems then you had better have some solutions and an answer as to why you, having been in power for so long, have not pushed the issue.

Republicans are likely going to get creamed in a few weeks, but what if they don't? If you lose you immediately start claiming the process is rigged. Give me a break.

Compare the 2000 presidential election and the 2004 Washington governor's election and you will see the difference. Remember the "Sore Loserman" stickers? Well, you are about to win and you still are complaining.

As the Secretary of State's communications person, I have to correct some inaccuracies in this post.

First, Secretary Bradbury did not bring the charges about voter registration fraud Mr. Hinds discusses from the 2004 election. Complaints were filed with our Elections Division which we addressed by passing on to the Attorney General's office to investigate. After a long investigation, the AG determined there was not enough evidence to charge anyone with a crime.

Secondly, every county in Oregon switched to using optical scan machines to record every ballot in 2004. The machines are required by law to be certified - and tested by county clerks five days before an election as well as on Election Day before the tallying of votes begins. If a recount is requested, it is done by hand, by human beings, using the paper ballots submitted by voters. This last requirement (a hand recount) is the result of legislation first proposed by Bill when he was in the Oregon legislature. He is proud that this piece of legislation made it into law as it is a safeguard Oregonians can count on for the security of their ballot.

Third, there is not one Diebold machine in use for optical scanning of ballots in the state of Oregon.

Finally, every vote by mail ballot that comes into a county clerk's office is inspected by an actual human being, with its signature matched against the signature on the voter's registration card - yet another example of the care and attention to security that goes into our election system.

If anyone has doubts about election returns, they should file a complaint with our Election Division and request a recount. The complaint can not be anonymous - you have to bring the complaint openly, using your name. We investigate every complaint brought to us and do everything possible to assure Oregon's voters that their vote is counted accurately and securely.

At a time when the use of high-tech voting equipment in other states may be making voters fear that their vote is secure, the relatively low-tech solution of vote by mail should give Oregon voters a real level of faith in the security of each and every vote cast in this state.

Mr. Hinds,

I'd be interested in your response to mconely above.

What is lacking in the Oregon election system is an independent verification of the machine tallies. I am not proposing a significant change from what oregon election officials already do. I am simply suggesting that voter confidence would be enhanced by instituting a process whereby a scientifically sampled hand-counted verification be done of the paper ballots to verify the machine count. This is a pretty low-tech non-labor-intensive way to make sure that the machine count is accurate at a statistically significant level. There is currently an administrative recount provision in Oregon in which a sample of precincts is used to validate the count. Because of the clustering effect of sampling precincts rather than ballots this is not a statistically sound methodology. Sampling precincts requires more hand-counting of ballots (and more work) than simply sampling ballots. The key advantage is that a scientific sample of ballots yields results which are statistically relevant.

“If anyone has doubts about election returns, they should file a complaint with our Election Division and request a recount. The complaint can not be anonymous - you have to bring the complaint openly, using your name.”

So let’s say someone in a household is being intimidated to mark their ballot a particular way by another member of that household. Do you seriously believe they would risk exacerbating the situation by sending a signed complaint to the elections division?

Give me the privacy of the old-fashioned voting booth any day. Time to ditch vote by mail.

I like vote by mail. It's the only way I can manage to read through all the initiatives, bond measures, tax proposals and other choices that should rightly be made by our legislature in a representative democracy instead of our current mob rule. I also like that the paper ballot IS our receipt.

But I could live with it & would still vote if I had to use a machine. Buckman Res's call is interesting in light of attempts in Congress to get rid of ballot-by-mail. I fear if Oregon is forced to go back to ballot booths we wouldn't be up to the challenge, especially if we were ordered to do it just prior to the '08 election. The lack of machines & trained staff would make the lines in '04 Ohio look like the HOV lane. Just a thought.

mconley -

As the Secretary of State's communications person your misleading comment demonstrates either a lack of intelligence or intellectual honesty. Put in simplest terms, the dangerous flaw in our VBM system is that the votes are actually "counted" in the technical sense that the counts are accumulated on, and the final counts reported from, electronic tabulating machines running proprietary software from the same corporations making the so-called "black boxes", combined with the fact that in the VBM system there is no possibility of exit-polling.

And despite what misleading commentators like paul say, it is a scientific fact that exit polling is the gold standard for independent, public election auditing, if it is properly done, because people are asked what they have actually done moments after they have done it rather than what they may do some days in the future. (By tradition all pre-election polling stops the weekend before an election.) And exit polling can be properly done, it's just that pollsters like Blumenthal and commentators like paul (if he is indeed the Prof. Paul Gronke his email suggests) who come from a political science background like the formal training in quantitative methods to understand why and how. More on that in a following comment.

In short, you have given objective reason here to suspect that neither you nor Secretary Bradbury belongs in your governmental role if you believe your post was a responsible comment to the discussion here. As a pragmatic matter, to request a recount requires some suspicion based on other evidence, exit polling being among the most predictive and persuasive, that the reported count is suspect. You and Bill Bradbury commit a fraud against the electorate if you proudly argue that providing a mechanism for a recount in a system which renders that mechanism all but meaningless is an acomplishment.

paul -

One or more folks using this same name and email address have commented here before on the 2000 and 2004 election results viz a viz the exit polls. The nature of this site is such that we have no way of knowing who actually posted those comments and/or if any of those people is Prof. Paul Gronke, so I preface this with two caveats: First, my comments mainly will address the overall theme of the last post by "paul". Second, because that post takes shots at the qualifications of an academic, it is fair and important to assess whether the putatitive commentator "Paul Gronke" would himself have the qualifications to make such a comment.

I take up the latter first. And make no mistake, this is not a defense of Freeman, but rather a critique from first principles on the position that "paul" has advanced.

The publicly available information for "Prof. Paul Gronke" resume indicates that he has no academic training in mathematics, much less thesubspecialties of probability and statistics that are used as tools by pollsters and political scientists. His publication record includes no substantive work in these scientific disciplines.

Blumenthal's academic credentials, on his website, are just a BA in Poli Sci and course work towards a Masters degree at the Joint Program in Survey Methodology (JPSM) at the University of Maryland. His years doing polling do not make him an expert in the mathematics and science of polling, only some who is experienced in offering political color commentary about polling. He is only a (possibly) skilled user of tools whose complexities and weaknesses he has no publicly acknowledged formal training, and no publicly demonstrated ability, to understand. He is at best marginally equipped, a priori, to actually understand what discrepancies between pre-election polling and exit polling are meaninful, when and if they arise. In particular, by his own voluminous, meandering words about "weighting", he seems to have only a superficial understanding of the relationship between a priori and a posteriori statistics. (It is hard to tell, because most of his word count is spent on petty criticisms of others commentaries, rather than the science, in an apparent attempt to defend his own professional status).

The only scientific question about weighting is whether the assumptions about the a priori statistics implied in a weighting of a sample are commensurate with the a posteriori statistics of interest. Period. As anyone who is actually trained in probability and staticstics can explain, it most certainly is possible to make those two things line up so that a poll is a good estimator (in a precise mathematical) sense of an underlying variable - or of another statistical outcome as the certified vote count in fact is. paul, and Bluementhal more so, despite their bluster, demonstrate no knowledge or competence in this underlying science. For the most part, they are merely users of tools developed by other disciplines. And as with most tool users, it is arguable that their public comments reflect an incomplete, anecdotal knowledge of the fundamental theory and limitations of those tools that is derived from limited empirical experience.

(Their experience, incidentally, is nothing more than a type of statistical sampling. Based on that, if we are persuaded to be dimissive of polling results by their generally negative arguments, as informed by their apparently apprenticeman's knowledge of the science, we should be equally suspect of their views in the first place.)

What the readers need to understand here is that the results of the 2000 elections, and then the 2004 elections, became a direct challenge to the priesthood around public opinion reporting that the academic and professional political science community had created post WWII. Struck by the anomalies in election results some, including people in mathematics and engineering disciplines who were more skilled than the political scientists in the science of samping and estimation, started to look more closely at the rather unsophisticated and ad hoc techniques on which "experts" like Blumenthal have built their careers, and pointed out obvious theoretical problems. No priesthood takes well to having their franchise threatened, and we are seeing that battle being played out now.

And lest folks think this is a reach, recent stories about the GOP GOTV efforts provide much of the rest of the picture. Starting in the 80s, the GOP embarked on a determined effort, routed in the real science of these things, to make sure their minority status in the electorate would still result in a 50%+1 vote count for their side. Put another way, their actual goal became winning elections in contrast to actually becoming a majority party representing core American values. Part of that was targetting, which is nothing more than understanding the a priori statistics of the electorate and deciding how to effect the entire process from campaign to vote counting to produce their desired a posteriori statistics. They set out to make that happen by working to get elections just close enough through message framing and voter suppression, and then working to thwart independent means of audit elections which could raise flags, so that folks like paul and Blumenthal pursuing their own professional agenda can be relied on to distract when skilled practitioners in the science say that the results are just unusual enough to require further scrutiny before being accepted.

Remember, the purpose of elections and elected government is to insure the peaceful transfer of power. Given those high stakes, in a time when even habeas corpus, the bedrock of democratic civilization, has been legislatively abolished, why should any thinking person right now have confidence in election results? And given that, why are elections with paper trails, public counts, and well designed exit polls and other audit means (including scrapping VBM as being inimical to restoring trust) not the answer? What other agendas are critics of these apparently simple solutions actually putting ahead of restoring confidence in our elections?

paul -

How could you possibly claim to know my intent, such as misdirection beyond whatever may have been stated, and where in your comment do you prove such an assertion? The epitome of poor reasoning and argumentation, don't you think? (Of course, it works for the psycho right wing in this country precisely because of the people they are hoping to persuade.)

As far as why I framed my comments as a I did, it was precisely because I don't make assumptions in the absence of valid evidence. And I did not want critics to be able of accusing me of criticizing the "Paul Gronke" when in fact none of us have any way to know if one or more people are posting under the name "paul". I instead addressed my comments to the fictional "paul" who is nothing more or less than the totality of the comments by the possibly one or more persons posting under that pseudonym.

With regard to the publicly avaiable of credentials of the Paul Gronke:

Ph.D.: U. Mich Political Science M.A. U. Essex Western European Politics B.A. U. Chicago Political Science Other Methods training IPCSR, U. Essex Summer Schools for Quantitative Methods

I seen nothing in this that indicates any training in the mathematical or statistical sciences. Nor does the C.V. indicate any published research in the mathematics of statistics, sampling, and estimation. To be blunt: Just because a master carpenter swings a hammer and drives nails for 30 years doesn't mean the master carpenter has any knowledge of the metallurgy and physical chemistry of the composite materials and alloys used to make the handle and head of the hammer.

The same goes for Blumenthal, only more so. The guy has built a business in the role of tool user and communicator. He is not an expert in the science by any means.

As far as your other four cited sources, to which specific claim in each case are you referring? I'll just comment on what seems to be the most probable in each case:

• Walter Mebane - With regard to the Florida optical scan issue, that was quickly recognized to be a non-issue within a month of the 2004 election even my non-experts, including anyone (such as myself) who lived in Florida at some point in their life and understood the basic demographics. There was no mathematical sciences question involved, and there was no exit-poll discrepency involved.

• Jaskeet Skehon - I'll just quote his comment from his paper Data Troubles: Explaining Descrepancies between Official Votes and Exit Polls in the 2004 Election which in fact supports exactly what I have argued here: My goal here is not to put forward a definitive explanation for the biases found in the 2004 exit polls, but to point out that we don’t know what the best explanation is. The simplest model, Republicans simply have a higher nonresponse rate orthogonal to other covariates, does not appear to account for the observed patterns.. The question is whether the "biases" are in the exit polls or the reported tabulations and the theme of this paper, which includes no mathematical methods, is justifying the assumption the exit polls were wrong rather than assuming the null hypothesis that the reported vote counts were wrong and proving that could not be the case. Of course, this approach insures that that the quoted "conclusion" is correct by virtue of being a tautology given the assumptions.

• Henry Brady - His work on the Palm Beach County focused on why the reported vote count differed so markedly from the pre-election polling. That is this is a case where statistical methods suggested a problem requiring explanation, analogous to how a variance between reported results and exit polls would, and he, like Skehon, conclude it was the butterfly ballot. This supports my general contention that polling has a roll in as an audit check on reported results, and in no way contradicts my thesis that properly done exit polling (or even exit polling in which sampling problems are understood and can be correct) serves a vital role in insuring public confidence in the vote count.

• R. Michael Alvarez - Surely you are joking in offering him in this context. He is the co-director of the initially controversial MIT-CalTech Voting Technology Program formed in the wake of the 2000 election to design new voting technology. It is more than a little fair to say that the mission of these two universities includes a healthy dose of self-interest: Namely, insuring they have the social status of leading efforts to produce new voting technology, rather than the matters at issue here. They have, in fact, adapted their mission over the years as they came to realize that gearheads sometimes don't always understand some of the human dimensions of voting in our democracy I alluded to in the last paragraph of my post. In fact, one of their affiliated faculty (Ron Rivest) has become a rather strong critic of electronic voting machines a la Diebold, and now is an advocate of paper trails. Without the 2000, 2002, and 2004 discrepancies between reported votes and pre-election polling and exit polls, it is quite plausible to argue these guys wouldn't even be doing this work.

So have we disposed of the whatever it is "paul" is trying to argue about my contentions about the type of expertise, and lack thereof in some cases, with regard to the specific points I have been making? But thanks "paul" for providing references to people who support my general thesis about the value of polling and who in no way contradict my specific thesis about the value of exit polling.

(More in the next post on the topic of obfuscation, and how it is important to understand when folks don't know what they don't know.)

I felt it important to take up "paul"'s attempt to interpret, and misrepresent, the mathematical points I argued because it so nicely illustrates why tool users don't always know what they don't know. (Or at least communicate that what they do know.)

He notes there are two sources of error in a survey estimate in the way he has been taught as an non-specialist in mathematics and statistics to understand the mathematical tools his discipline uses. This is much as how an engineer is taught to use formulas and techniques based on the physics, chemistry, and mathematics involved, without really needing or having a deep understanding of those sciences. In particular, he states One is a result of pure random error. This is reflected in the "margin of error" that is reported in most published reports on public opinion data. This is what Ask keeps referring to as "scientific facts." That is not at all what I am referring to as "scientific facts". And in fact, this sentence is so superficial with regard to the real science on it's face as to be all but irrelevant to this discussion.

"Margin of error" as reported is close to what he says it is, but not quite. It is the range that a reported statistic has with some probability (typically 95%). This bound is given based on some almost always unrealistic assumptions about the sources of such errors. I am not making any such assumptions, but in fact commenting on this limited understanding of the mathematics by users of the tools and how those with more sophisticated mathematical skills focus on characterizing the sources of error to derive more useful information from whatever sample statistics they have at hand.

More specifically, the "scientific fact" that I'm referring to is that: 1) We can understand the nature of the error in a result as a combination of the sampling process and whatever assumptions we make about the population we are sampling. 2) We can also understand when the descrepancies between the results of any two such experiments may be meaningful and when they aren't. And finally, 3) we can understand the nature of sampling errors in exit polling well because the variable we are estimating ("how did you vote"), is more closely linked in time and psychology to the variable of interest ("the vote you cast"), than the variable we are estimating ("how will you vote") in any pre-election poll.

And make no mistake, the tabulation of votes cast is also a sampling just like pre-election and exit-polling. It is just a sampling with a larger sample, different types of errors, and for which a particular statistic (total votes) is used as a decision variable for other purposes.

So his second comment But he displays his own lack of familiarity with how polls are actually conducted, on the ground, when he ignores the second, and in many cases, largest source of survey error--bias due to poorly worded questions, poorly trained interviewers and poorly selected samples. is simply silly. Not only am I not ignoring these sources of error, as a tool user he doesn't understand my comment that practitioners of the science look at the problem differently than the formulaic way tool users such as he does. And they know how to consider these sources of error when deciding whether the discrepancies between any pre- or exit-polls differ from reported results. That is, they value any alternative sampling to the vote counts because that sampling provides an independent estimate which can be used, with sufficient skill, to assess the validity of the primary estimate to some degree and in a way that is qualitatively different than if no such independent estimate were available.

This particular paragraph is factually wrong, misstates my argument, and in fact doesn't actually support the conclusion it states

Ask uses a lot of big words and technical jargon to obscure the central point: these latter effects have been demonstrated time and again with respect to the 2004 exit polls. The interviewers were poorly chosen, and the people that they interviewed were not a random sample of voters leaving the booth.

First, what has been demonstrated is the empirically tautologous results that these effects exist: Duh, they always exist. There has been no scientific demonsration these effects actually explain discrepencies between any polling and reported vote counts. That is a critical difference and if anyone is trying to obscure something, it is "paul". (But my point is that we can't know intent on this, and that folks in "paul"'s discipline have given us little reason to believe they even understand this is the case.). Second, the sampling is always "random", the samples just don't have the particular "random" characteristics required for the tool users to apply the tools in the limited ways they have been taught and get viable answers. That in no way means that skilled scientists can't understand the particular characteristics of the "randomness" involved in this case, and use that understanding to study the plausibility of the vote counts when they differ from pre- or exit-polling. Which is why this:

There is no amount of statistical mathemagic that can change this fundamental point.

certainly plays well as a throw-away comment in an era when intellectualism itself is under attack, but is not a concrete, supported conclusion of a valid argument.

I'll conclude by just noting that this comment:

However, I also believe that the overwhelming majority of empirical evidence indicates that there is no credible evidence that the 2004 election was stolen.

what it says it is: A belief and not a conclusion supported by scientific argument. The "evidence" is simply whatever facts exist. There has been no analysis of the evidence which disproves the 2004 election was stolen in the face of sampling evidence which is still unexplained (see the Sekhon quote above). Again, the approach of one skilled in the science would be to assume the 2004 election was stolen, and show the evidence proves there is no possible way to this could be the case.

So we arrive back at my main contention that what we are seeing play out in part is that political scientists and pollsters are having their priesthood status challenged since the 2000 election. We need to fix the system, part of that fix are paper-based balloting that can be counted and recounted before the eyes of any concerned party.

But another key part of that fix is improving exit polling rather than being cowed by unsupportable claims that exit polling is not useful in auditing reported vote counts by folks who find their cultural franchise threatened when the tools they have been given to use are not quite up to the circumstances.

I think "ask" has tried this tack before, ask, with similar results.

http://www.mysterypollster.com/main/2006/05/words_worth_rem.html

paul -

Until then, I see no reason to continue. I've dealt with enough conspiracy theorists in my time.

Whatever makes you happy. Frankly, I remain unconvinced who the person or persons are that post under the name "paul", since this has hardly been an exchange that I would have expected from a serious scholar. Right down to the final classic, but intellectually dishonest tactic of throwing the "conspiracy theorist" charge.

For the record, let's be clear on three points:

1) My main point is and always has been that exit polls are an important source of data that serves as an independent check on reported vote counts. And that exit polls can be well done, although even the data from less than perfectly done exit poll can still have value in the skilled mathematical scientists.

2) You started the line of attack on not "overstating" credentials with your comments about Freeman. Although I had initially and previously stuck to substantive matters about exit polling, I merely made sure folks here understood the same type of information about the credentials of some of those who are making claims that overstate or misrepresent the facts once you pursued this tactic. It is this type of behavior I have been observing, particularly in the wake of the 2004 election, that led to my second point that there is something else going with the political science/polling community that is not about the mathematics and utility of exit polling.

3) I responded in an intellectually responsible way with facts to your argument. In particular, I responded to the substance of the work by every one you cited. And in each case the work you cited did not refute my primary thesis 1) and therefore implicitly lending support to my observation 2).

Bye bye.

a -

I'm not sure whether your comment is critical or supportive because it is so brief. I do appreciate you bringing it to light here, though, because I had pretty much lost interest in the topic right after I made the post and before the answers we see there now were posted. I'll respond on the substance to what was at the cited link:

1) My point there was essentially that Lindeman provides an example of something happening with the political science/polling community that has nothing to do with the mathematical facts of the exit poll controversy.

2) Lindeman misrepresent the position of more sober commentators as defenders of "exit poll inerrancy", when in fact no one I refer to, nor myself take any such position. And he takes the disreputable position of asserting that I didn't engage his rather superficial paper on the substance. I most certainly did, just not in a way he liked. My argument was essentially that his thesis is simplistic and mused why that may be the case. One can't engage substance that is mispresentative and superficial any more than that.

3) Liddle's comments need to be contextualized to understand that there isn't as much there as she postures. First, she doesn't cite which of her work to which she is actually referring, but I think we should guess it may be her work responding to the USCV work calling the Mitofsky explanation that the discrepancy in Ohio between exit polls and reported votes were due to sampling problems. The mathematical substance of her paper is a fairly simple undergraduate exercise in the algebra of one estimation equation and in statistics. At the bottom line, one only needs to read the conclusion of her paper to see that she presents her work ultimately as supporting the conclusion we still don't know why the discrepancy exists. In addition, the USCV folks added "Appendix B" to their paper acknowledging the substance of her work, pointing out that she does not refute their thesis that the exit polls may have been correct, and demonstrating that the issue remains most ambiguous (but leaning towards their thesis) in precincts that went red.

All -

I can give you one concrete example of why exit polling is useful:

We know of one scenario where exit polling in Ohio can help us develop an awareness and fuller understanding of uncounted votes.

Anyone who voted on a provisional ballot because they were purged from the roles would be expected to have accurately answered an exit pollster as to how they voted. And they may not have understood that they voted provisionally, or what that meant. We know for certain those provisional votes were not counted.

I'm not claiming that provisional ballots would have made the difference in the election, nor even that exit polling would be our only or best source for information about the incidence of people being forced to vote provisionally and not have their votes counted. Only that exit polling is an independent source of data that is valuable in helping us reach a fuller understanding of what goes on in an election and highlights when reported vote counts deserve much closer scrutiny because they are improbable (as opposed to impossible). And finally, that to this day the discrepancy between the exit polls and the reported vote counts remains unexplained, and as of mid-June 2006 the thesis that it is due to sampling problems unproven in the literature I've seen (I could have missed something published after that because I lost interest in the topic until "paul" popped off about it in this thread).

Does anyone else find it more than a little suprising that the political science/pollster community would object so angrily to that position? Particularly at a time when even the mainstream media has sensitized the public to problems with electronic voting machines in the polling place, and exit polls can help bring confidence-building information to the table on election night?

Electronic Voting Machines Could Skew Elections Researchers, Candidates Have Little Confidence in Machines Designed to Make Elections Easier to Call http://www.abcnews.go.com/WNT/Technology/story?id=2596705&page=1

<hr/>