Mitofsky, "father of exit polling," dies at 72

[papau]

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=102x2491402

Mitofsky, "father of exit polling," dies at 72 discussion in LBN

http://www.cnn.com/2006/US/09/02/obit.mitofsky/index.html

Mitofsky, 'father of exit polling,' dies at 72
September 2, 2006
From Keating Holland
CNN

WASHINGT0N (CNN) -- Warren Mitofsky, considered by many to be the "father of exit polling," died of heart failure in New York on Friday. He was 72.

Mitofsky changed the way the media covers elections by pioneering the use of exit polls to project winners in U.S. elections beginning in the 1960s. He also developed many of the telephone polling techniques still in use today.

Mitofsky worked for CBS News for nearly three decades before leaving in 1990 to head Voter Research and Surveys, the first network exit poll consortium.

When Mitofsky joined CBS in the 1960s, political pollsters relied on house-to-house interviews to project winners of elections in the coming days.

Mitofsky developed the technique of canvassing people soon after they voted into a staple of modern news coverage, changing the way elections were covered and "called" by network news<SNIP>

[YDogg]

1. He was preceded in death by exit polling.

RIP.

[mikeargo]

4. Dead at 72...

Plus or minus five years.

RIP

[whistle]

2. The debate on the issues of exit polling was hosted by none other

...than Keith Olbermann back in November 2004

<snip>
Zogby Vs. Mitofsky
by Keith Olbermann

• November 24, 2004 | 9:18 a.m. ET

NEW YORK - It was a spectacular irony - a Republican senator using the word “fraud” about the presidential election. More spectacular still, he was visiting his condemnation of apparent election manipulation on the incumbent party. And beyond all that, he and others based their conclusions largely on the incredible disparity between the last exit polls and the vote count itself. Of course, Indiana’s Richard Lugar was talking about the presidential election in the Ukraine. But in so doing, he underscored that once again, the exit polls appear to have fulfilled the time-honored international tradition of the canary in the mine shaft. If only we could have used them in that way here.

“I don't think that exit polls can be used as a barometer for the accuracy of an election itself,” noted pollster John Zogby explained to me on last night’s Countdown, in what we think was his first full-scale television interview since the election. “At least until we find out if there's something broken with this round of election polls… I think that the gentlemen who are responsible for the exit polls should be fully transparent, release their data, discuss their methodology. Let us see what exactly it is that happened, and why it happened.”

It turns out one of those gentlemen doesn’t think anything happened.

In an unsolicited e-mail to Countdown, Warren Mitofsky wrote that he was “struck by the misinformation” in our program. He heads Mitofsky International, which along with Edison Media Research, conducted the election night exit polling for the television networks and the Associated Press. I referred to the variance among the early and late exit polls, and the voting. Insisting “there were no early exit polls” released by his company or Edison, Mr. Mitofsky wrote “the early release came from unauthorized leaks to bloggers who posted misinformation.”

Mitofsky compared those leaks to “the score at half time at a football game” and said the “leakers were reading complex displays intended for trained statisticians. The leakers did not understand what they were reading and the bloggers did not know they were getting misinformation.”

His defense of his work grew more strident. “The presidential exit polls released at poll closing time when they were completed had an average error of 1.9 percentage points. There were no mistaken projections by Edison/Mitofsky or any of the NEP members.” One more thrust: “All the professionals correctly interpreted the numbers.”


<more>
http://www.yuricareport.com/ElectionAftermath04/OlbermannZogbyVsMitofsky.html

[HysteryDiagnosis]

3. I'll believe it when I read about it in the polls. n/t

[Fly by night]

5. "Reluctant Bush supporters" expected to turn out en masse.

My own trip down the rabbit hole began two days after the 2004 election, when I wrote Mitofsky asking him four direct questions about the discrepancies in his exit polling. I never heard back from him (surprise, surprise).

I have been taught (here in the polite South) that if you can't say anything nice about your enemies, you can alway say they live in a beautiful state. Well, Mitofsky lived in a beautiful country, one with democratic traditions that I hope we can preserve, despite his best efforts to the contrary.

[OnTheOtherHand]

6. oh, no, sir

I suspect that Mitofsky's in-box was rather full in those days. For what it's worth, it seems from the other thread that at least some DUers did get answers to their questions.

I know that many DUers think they understand only too well what motivates a few of us to defend Warren Mitofsky's integrity. There isn't anything I can do about it. My parents raised us in the shadow of McCarthyism, and they taught us to defend people whom we had reason to believe were being attacked unfairly.

Mitofsky was never Saint Warren, but I have no basis for believing that he made "best efforts" to undermine democratic traditions, and considerable reason not to believe it. As far as I can tell, Mitofsky never believed that massive fraud accounted for the exit poll discrepancies -- and that unbelief is very widely held among survey professionals and political scientists.

I think a lot of people are confused on this issue because they believe that Mitofsky was somehow vouching for the integrity of the election. But I was startled to be reminded that John Zogby appeared on Olbermann's show, and was asked, "Do you believe it is appropriate to use those exit polls as a barometer or as a ruler against which you can assert that something may have been wrong with the elections?" And answered, "No. Remember, you know, exit polls, like any poll, are samples. I don't know what happened with the exit polls. I'm calling for more transparency, both in terms of the sampling and in terms of the data collected. But, no, I don't think that exit polls can be used as a barometer for the accuracy of an election itself, at least until we find out if there's something broken with this round of election polls."

A lot of people have taken a lot of crap for saying what John Zogby said, and it was all totally unnecessary. Some folks precisely fixated on the exit polls "as a barometer or as a ruler," instead of keeping their focus on the election itself. Some folks seem honestly to believe -- or have acted as if they believed -- that a few thousand exit poll interviews could tell us more about the Ohio election than the actual ballots. It has been bizarre. It remains bizarre.

[Fly by night]

7. When Mitofsky presented data for all of us to read ...

... that rebutted his own "reluctant Bush supporter" hypothesis in his own report as his only explanation for the final exit poll results (as I read it), he indicated (to me, at least) that he also had no plausible explanation for his own findings (as he released them). I do not wish to speak ill of the dead (any more than I already have), but I am clearly in the camp that believes that something was rotten in the 2004 "U.S. of Denmark" exit polls. Mitofsky's unwillingness to release his raw data early on and his strained explanations (which were not even supported by his own data) were further evidence (again, at least to me) that something was afoul. I might agree that if exit polls were not working in other democracies -- and if the exit poll discrepencies in the 2004 US elections were the only evidence for fraud in that election -- it might be possible to give some slack on this issue. But they aren't and they weren't.

In any case, it is sad that Mitofsky was burdened in his last two years with the awful knowledge that either a) his well-respected research methods for predicting election results (which heretofore had worked so precisely and accurately) had suddenly collapsed into irrelevance or b) that his still robust research methods had provided clear-cut evidence for the corrupt collapse of our own democracy. In either case, that must have been a heavy burden for a respected scientist to experience or for a proud American to bear.

But you and I may continue to respect each other, as we continue to agree to disagree on this issue. Thank you for that. RIP.

[Febble]

10. His own data

did not rebut "his own 'reluctant Bush supporter' hypothesis" - it is not even a term he used. What his own data do is support the hypothesis that the sample was biased - that a greater proportion of Kerry supporters were interviewed than Bush supporters.

What Mitofsky has not offered is any explanation for the underlying phenomenon by which Democratic voters appear more willing than Republican voters to take part in exit polls - what he has offered is evidence that this is the case (not only in 2004), as well as evidence that particular factors make the phenomenon more likely to be manifest in a biased sample (greater opportunity for the interviewer to exercise discretion over who is sampled, or for reluctant voters to avoid the interviewer, for example).

[Bill Bored]

19. I have an explanation!

A certain proportion of those who vote for Republicans are essentially dishonest. They claim to be for the good, or the lesser of evils, but all they're really interested in is lining their own pockets with tax cuts and that sort of thing.

Their whole reason for voting is selfish and hypocritical on its face. At some level, they are ashamed of themselves, but they suppress this by not talking about it except with those they know already agree with them, lest they become confronted with their own hypocrisy. That might make them a little uncomfortable, which is not what they're about. So they don't take surveys; they see the pollsters and move on.

They may also be scared to death of the real multi-ethnic, diverse world beyond their front lawns, which is another reason for voting Bush and not wanting to talk to anyone about it. They feel as if they are under siege.

It would only take a fraction of them to have an effect on the polling data. There may be some who for whatever ideological reasons, actually do believe in what they are doing. Those are the ones who take the polls.

If nearly all Democrats, on the other hand, believe in how they are voting, they will be more likely to take surveys. They are guilt-free.

What do you think?

I'm not defending the exit polls or the vote count here. I'm just suggesting an explanation as to why some Bush voters might avoid taking surveys.

[Febble]

26. Well, I have some other ideas

but I'm glad you've made the point that we can only speculate as to why there is an underlying tendency for Democratic voters to be more willing to participate in exit polls than Republican voters. All we can do is demonstrate that if conditions that make biased sampling more likely are present, the direction of bias tends to go in that direction.

And the interesting thing is when actual experiments have been done to try to improve response rate, higher completion rates aren't necessarily associated with less bias, and again, the bias, where present, tends to be in the direction of "redshift". So even steps taken to encourage participation in the poll seem to work more effectively on Democrats than Republicans, and thus actually exacerbate bias.

So it's a puzzle. What these studies show, however, is that the tendency is a real phenomenon. Even when conditions are experimentally manipulated (with random allocation to condition, so that the experimental conditions must be orthogonal to any corruption in the count) any bias associated with experimental condition tends to be redshift, not blue.

What we do not have any direct evidence for is the cause of this underlying apparent differential willingness to participate. We only have evidence for the conditions under which is likely to translate into bias in the poll.

But I think it's worth making the point in this thread, as Mitofsky has so frequently been accused of inventing "rBr". He didn't invent it. It's dogged his polls for years, and has been the subject of repeated investigation. It's why he never relied on exit poll data alone to make his projections, which in turn, ironically, is one of the reasons his exit polls got their reputation for "uncanny" accuracy. Not because his polls had no limitations, but because he knew so well what they were.

[Awsi Dooger]

27. Did Mitofsky ever ask the basic question?

To Republican voters: "What conditions, if any, would make you more willing to participate in exit polls?"

Seems to me we've had plenty of applied guesswork and nothing from the horse's mouth. Elephant, whatever. You could ask it in short summary form until you got an idea of the top responses, then include those in subsequent surveys, in form of pick one of these, or rank first to last.

Hell, ask that question of both Democrats and Republicans, along with the opposite question, "What conditions, if any, would make you less willing to participate in exit polls?"

Then manipulate if you have to, a bit of suppression A with a pinch of boost B. It reminds me of the golf driver I bought last week, with a draw bias. Finally a concession that my swing isn't what it used to be, so I need an artificial cheat. Who cares what it takes to get the exit polling balanced?

On edit: I realize a problem with that scenario is you are asking it of voters who have already agreed to participate in exit polling. That brings me back to something I mentioned months ago, doing research outside election day itself. Just sample people in shopping malls or outside a Wal-Mart or Target and see if Republicans and Democrats react differently to participating in surveys even in a non-election day atmosphere, and what they say to questions like the ones I asked above. Obviously there are difficulties no matter how you approach it, since the people outside a Wal-Mart are not necessarily voters. I'm merely flailing possibilities.

[Febble]

29. Well, the big advantage of golf

is that you have a nice objective measure of whether you've actually hit the ball accurately or not.

The trouble with the question you suggest, though, is that it looks as though the biggest problem wasn't actual non-response bias (more Bush voters than Kerry voters refusing to participate) but a tendency for a greater proportion of Kerry voters than Bush voters to be selected. In other words, the Bush voters may simply have been more successful at evading the selection process, and/or Kerry voters may have been more likely to volunteer themselves. There is some anecdotal evidence for this. This is why the finding that bias was greatest where there was greatest opportunity for the interviewer to depart from strict Nth voter protocol is interesting. Completion rates themselves were not a particularly good indicator of bias. If the problem was that in some precincts interviewers were tending to select only the more willing voters, that may actually have boosted their completion rates - but also increased their bias.

[Awsi Dooger]

30. I tried to word it based on your post #26

"tendency for Democratic voters to be more willing to participate in exit polls than Republican voters."

Regardless, if there are more variables ask more questions. "Have you ever intentionally avoided an exit pollster?" And "what would make you avoid an exit pollster?" Do these over the phone. I'm still baffled why it has to be election day only. You're looking for tendency differences among people with contrasting partisanship. It might be as simple as Republicans saying they would be more likely to avoid young men who they viewed as probably liberal. I know in 2004 I saw voter registration drives everywhere in Las Vegas for months. Most of the young guys looked like they were smack off the Berkeley campus and they practically strided up to me. If that's the case regarding who is volunteering for exit polling, Houston we have a problem. I'm not saying I know, but the era of sticking anyone out there to do the sampling without research or care as to how they may be viewed is over. Or should be.

And I still think a much shorter survey with higher number of interviews, mixed with results from the full surveys, can be a valuable supplement. We know white males vote heavily GOP, and on a business Tuesday I wouldn't be shocked if a segment are too impatient to stop for a lengthy exit poll. Especially if they are out of the precinct, walking briskly with voting what they did last "hour", and the next item on the daily checklist is now the focus and priority. That's merely one example. I'm hardly implying white males account for the entire discrepancy.

I should point out that draw and slice bias drivers and offset clubs are a relatively new option, after decades of pretending swing lessons were a cureall. Seems to me exit polling is still in the pretend stage, that everything will fall in place one cycle, and you no longer have a sampling issue. I looked at the report from January 2005 and the suggested changes didn't strike me as aggressive or specific enough.

Hey, I need some sleep at 4 AM:)

[OnTheOtherHand]

34. this is cool

It's fun to have a substantive discussion. (Not that I have any brilliant revelations to contribute, but I am conceivably more likely to be able to promote whatever ideas we come up with than Febble is.)

The history of polling is a history of methods being good-enough until they no longer were. If the networks really wanted that thing that so many DUers seem to assume they already had -- the ability to predict the outcome with "uncanny accuracy" the instant the polls closed -- then they probably would invest more money in methodological experiments. So far they have been willing to settle for a high degree of confidence that they wouldn't actually blow the call. (Don't get me wrong: experiments do happen, and post hoc analysis of possible sources of error happens too. It's not that the exit pollsters have chosen to assume that nothing could go wrong.) It sure doesn't satisfy me as an academic, because I want to know everything. Honestly, I haven't exactly bellied up to the bar with any network folks, and I have no real clue whether they are gravely distressed that the exit polls were So Far Off, or reasonably pleased that they made no huge mistakes (and that Edison/Mitofsky gave the sponsors a heads-up in the afternoon that some of the results looked iffy). Presumably some of each.

People being people, there's no guarantee that any method will ever be good enough. Certainly there is no way to assure zero bias. If it's (some) Republicans avoiding young interviewers today, it might be (some) Democrats avoiding old interviewers tomorrow. If it's (some) Republicans avoiding everyone today, to varying degrees, then picking one kind of interviewer as the Best Kind might reduce the bias but also reduce one's ability to detect the bias in real time. I'm not arguing for the status quo, just arguing for a robust sense of tragic humility. ;) I am so very sick of this demented notion (not yours) that Scientific Polling means that error is unthinkable. Pollsters think about error all the time.

I would like to see a conscious mix of interviewer characteristics, a mix of questionnaire lengths, multiple interviewers at at least some spots, and maybe some additional capacity to reckon any variation into the error estimates. (I mean, for instance, if you detect that the results from short-questionnaire locations are coming back different than the results from comparable long-q locations, you might speculate that the short-q results are better, but you would at least register that Something Weird Is Happening.) Note, however, that the exit pollsters did notice that 'something weird was happening' -- it shows up in the big standard errors, which are empirically determined based on the variance in the model data. Of course, the central estimates could probably be better too, and I'm all for experimenting with how much better they can get. Especially if I don't have to pay for it.

[Bill Bored]

36. One thing that has NOT been satisfactorily explained though is

why it took Mitofsky so long to make his adjustments this time.

Is it not true that by midnight in past elections, the polls had been adjusted to reflect the winner according to the vote count? How come this time around, with similar bias in the polls, it took them until sometimes the next day to figure out that they had gotten it wrong?

Until we can answer this question, there will always be suspicion on the part of the exit-poll-true-believers and perhaps rightfully so.

I'm not saying they're right to suggest that polls are always right, that there were NEVER adjustments made to match vote counts, and that the polls were never biased in the same direction as they appeared to be in 2004. But I don't think it's ever taken the pollsters so LONG to figure this out before, has it? Or is it that this time, we just had a closer election and it took them longer to be sure they had it right?

There's a big difference between mathematically equivalent discrepancies that can affect the declaration of the winner and those that do not. But does this mean it should take longer to make the necessary adjustments?

Unfortunately, the true believers aren't even looking at such subtleties as the timing of the adjustment. All they seem to care about is that there was an adjustment and to them, that's unthinkable!

[OnTheOtherHand]

37. I think the premise is confused

In how many elections have people been glued to CNN.com hitting "refresh" (or whatever) to try to judge when "the polls" were "adjusted"? That's not really how it works.

The projections (vote share) are updated continuously as new data come in. In principle they could update the tabulations (vote share by sex, by race, by age, etc. etc.) continuously, but they don't; they are focusing on getting the projections right.

People are ascribing significance to when the tabulations were updated that, as far as I can tell, doesn't and couldn't exist. From what I've seen, the Ohio vote counts are pretty internally consistent (they could all be wrong, but they tell the same story), so Edison/Mitofsky's best estimate there probably shifted to Bush (but not decisively) pretty quickly once the polls closed. It doesn't have to take long at all.

It's true that they intended to update the tabulations sooner (which would be a huge benefit for the newspaper subscribers). So I guess some people like to believe that someone delayed updating the tabulation in an attempt to signal the public that something unprecedented was happening. Cue spooky music.

[Land Shark]

40. Right now, two of OTOH's posts are right next to each other to compare

One, post 16, OTOH says

"I celebrate disagreement without disrespect. Sounds corny, maybe, but it's true. (Not that I break out the bubbly, but I have been known to cheer.)

See <http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=448075&mesg_id=448139>

I salute OTOH based on the above sentiment.

But then, I come across his reply #37 entitled "I think the premise is confused" and it says with regard to those who find evidence of fraud in exit poll data:

So I guess some people like to believe that someone delayed updating the tabulation in an attempt to signal the public that something unprecedented was happening. Cue spooky music.

See <http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=448075&mesg_id=448216>

I will salute you again OTOH, if you can respectfully admit that you fall short of your admirable standard above when you insinuate someone is a conspiracy theorist or similar, as you do with the clause "cue spooky music." A whole bunch of us will cheer.

[OnTheOtherHand]

43. I fall short in all sorts of respects, but I'm not sure that's the point

Since I see no conceivable factual basis for the belief that the tabulation update was willfully delayed -- or that its timing sheds or could shed any light on the significance of the exit poll results -- it is hard for me to understand why people believe it.

So, I guess some people like to believe it. And, presumably, other people believe it because they were told by people who seemed confident and authoritative. But I really don't know.

That said, my comment was not "with regard to those who find evidence of fraud in exit poll data," and I have to wonder why you would assert that it is. You did at least quote it, so that people can verify that it does not support the construction you place upon it.

I am sorry that you take offense at the expression "Cue spooky music." I will forego enumerating the things that offend me. Are we done?

[OnTheOtherHand]

16. I celebrate disagreement without disrespect

Sounds corny, maybe, but it's true. (Not that I break out the bubbly, but I have been known to cheer.)

For your information -- although you can choose not to take my word for it -- I don't think many folks who follow survey research think either that Mitofsky's research methods had collapsed into irrelevance or that they had demonstrated the collapse of American democracy. Again, it's one of those points where I think a lot of confusion has been engendered.

I wasn't a friend of Warren Mitofsky, but we corresponded some in the last year. I think he was proud that Edison/Mitofsky made no incorrect calls in 2004: from his point of view, his prediction methods hadn't failed. I also think he was bitter that, as he believed, people who must have known that they were wrong assailed his integrity for their own political ends. I don't think he ever understood how an honest observer could fail to see what he saw. And certainly the same is true of many of his harshest critics. There's the human condition for you.

[papau]

23. the data is too little to say "election stolen" but it is a "best estimate

and that "best estimate" implies a stolen election.

"folks who follow survey research (do not) think either that Mitofsky's research methods had collapsed into irrelevance or that they had demonstrated the collapse of American democracy" are correct if they are saying that one should say "the exit polls result is a best estimate that suggest that the election was stolen, but there not only many legitimate and legal ways the data could have come out as it has, there is also the fact that there is a lack of data that requires one to add that there is a possibility approaching 50% that the exit poll best estimate could be misleading and wrong in its apparent implication."

The Mitofsky's research "shocker" I believe is the analysis result indicating that electronic voting machines were not uniformly hacked by the GOP across the country - implying that despite the ability to easily hack these voting machines, some of the GOP involved in elections are honest!

[organik]

8. Now if we could just count those "actual ballots" for once.... n/t

[OnTheOtherHand]

15. yes, that is kind of crucial n/t

[Land Shark]

9. Most of the time it was not Mitofsky's "integrity" that was defended

it was just the numbers and their potential for meaning. But granted, you defended his request that his numbers not be used in the national debate about the 2004 election for the reasons given, and you did so repeatedly and at length and consistently.

John Zogby's comment above, focusing on the very last clause, seems entirely consistent with a proper approach of (upon finding a gap between exit polls and election results) of considering both suspect, and investigating both. That's the proper scientific approach, but does not mean that the exit polls are permanently subject to the limitations on meaning he stated there.

On one hand, you defend Mitofsky and get quoted by Farhad Manjoo in support of his hatchet job "debunking" of Robert F. Kennedy's Rolling Stone article.

On the Other Hand, you spend long hours attacking DU posters like TruthIsAll, Autorank, and LandShark. On a couple of frivolous occasions you've stepped in and "defended" me from a particularly off base attack, but apparently you think that just enables you to disrupt/attack more.

OTOH, your "contribution" here is limited to attacking those attempting (rightly or wrongly) to defend democracy, and defending those such as Mr. Mitofsky who argued that democracy needs no defense, at least with regard to the 2004 exit poll evidence issue.

YOU DO NOT defend the integrity of anyone contributing to election integrity, instead your work along with folks like Manjoo to ridicule them, attack them, and impugn their integrity 97% of the time, but always ready to cite the 3% if you're called on it along with your commitment to "Truth" as your defense for constant attacks on those attempting to defend democracy.

Let me be 100% clear, before someone starts pointing to a critique or "attack" of mine: There needs to be a ratio of positive contributions to attacks but what I'm complaining about here is a predominant devotion to attack (and its corollary, defense). What contributions have you really made here? What's been added to the debate by you? I can't recall, and I doubt many others besides Febble can either. You do have the capability to contribute, however.

Mr. Mitofsky was a man that made a big contribution, as the father of exit polling. And modest to a fault on what those polls might show regarding 2004. May Mr. Mitofsky rest in peace.

[OnTheOtherHand]

17. if you are prepared to enter into honest debate

then we could perhaps dispense with hand-waving about what I have contributed to it. It has always seemed to me that seeking truth is hard enough without keeping score.

People can point to whatever they like, but clearly I don't believe that I have spent "long hours attacking DU posters like TruthIsAll, Autorank, and LandShark." Indeed, it certainly seems to me that the post to which I am responding is more of a direct personal attack than anything I have posted about any one of you.

Oddly, your own post testifies that it simply isn't plausible to assert that I "DO NOT defend the integrity of anyone contributing to election integrity." I'm not sure that you are ideally situated to be raising this particular complaint in any case.

[Land Shark]

20. Why are you so unsure I'm not "ideally situated" and not yet

engaging in "honest debate?" Where are you situated in regard to these categories you identify?

[OnTheOtherHand]

21. well, what are we debating?

You seem to want to debate, umm, me. Pretty tedious.

[Land Shark]

22. Um, um, um, I asked what you meant by those words, can you answer

that?

on edit: "those words" meaning the ones asking me if I was ready for an honest debate, particularly your use of the word "honest", and combined in the same post with the statement suggesting I was not "ideally situated" to speak about something, what was that quality or something to which you referred?

Perhaps you could add something more about Mr. Mitofsky who you knew a bit, as a human being. I'll just read and not debate that.

[OnTheOtherHand]

31. the circumstances are simple

If you want to discuss exit polls, simply discuss exit polls, not me. That would be honest debate -- perhaps not even debate.

If you actually disagree with something I've said about RFK's article, simply state what it was. That would be honest debate.

Otherwise, all I see here is another attempt on your part to personalize issues that shouldn't be personal. I think you should stop.

As for Warren Mitofsky, I've said most of what I could say in post #16. He is being serially eulogized -- by people who knew him much better than I did -- as someone who tried to get the facts, find ways to do things better, bring colleagues together, and take care of people who were hurt. Not as a diplomat. But as a very good man. It dovetails with what I knew of him.

[Land Shark]

33. I don't really know your opinion about RFK JR

I'd have to look it up, if it exists. All I know is that you were quoted with favor by Farhad Manjoo in his salon attack piece ridiculing RFK, Jr's piece in Rolling Stone. I don't recall if you distanced yourself from this rather public attack (it was not dispassionate criticism it was sarcastic or the like with a lot of attitude), but you certainly were harnessed as support for the attack, which would give you the right to distance yourself from its tone or style if you chose to do so.

[OnTheOtherHand]

35. well, thanks for that n/t

[Land Shark]

11. we're JUST NOW getting to the "actual ballots" you refer to

You state and ridicule and call "bizarre" the idea: "that a few thousand exit poll interviews could tell us more about the Ohio election than the actual ballots."

In fact, it's perfectly rational to believe that the sampling process was more accurate or less corrupted than the "counting" process of the "actual ballots." You simply and improperly presume a proper chain of custody and counting of the ballots, in order to ridicule the idea that the sample could be more accurate than the whole. Moreover, that assumption begs the very question at issue, namely, whether the actual ballot count was accurate. You're smarter than that, but when it comes to ridiculing election integrity advocates, it appears you can't resist the opportunity.

[Febble]

12. You have missed the point, I think

What I think OTOH means is that actual ballot data is potentially far more copious and informative than a set of exit poll responses garnered from only 49 (out of about 11,000) precincts. In any case, the full set of interview responses are already in the public domain, and have been since January 2005, as have the "Call 2 estimates".

Unused ballots; used ballots; undervotes; overvotes; vote counts from 2004; vote counts from 2000 - all these form a far richer dataset with which to investigate vote corruption than the exit poll sample.

OTOH has, since you ask, to my certain knowledge, suggested a number of ways in which ballot data could be analysed to shed light on whether or not vote corruption occurred. Ergo he does not assume the count was uncorrupted.

[Land Shark]

13. Potentially yes, they are bigger richer data set, but you're missing

my point which is that a totally corrupted dataset (in the form of the population of ballots presented as allegedly "actual ballots") does not lead us more reliably in the direction of the truth than a proper sampling of voters as they emerge from the polling location after voting and are interviewed for exit polls.

Recounting a stuffed ballot box, or sampling a stuffed ballot box, is not leading us toward truth even if the data sets are "richer." Comprende? I think the "richer data set" assumes conditions not only not proved to exist but conditions actually in dispute, thus it begs the question.

on edit: adding parenthetical comment in first paragraph

[Febble]

14. I may have missed your point

(although I don't think I did) but you certainly missed OTOH's.

Let's talk about a single state, like Ohio. There are about 11,000 precincts in Ohio (I believe, correct me if I'm wrong) and 49 of them were in the exit poll. We know they were a representative selection of precincts, because the mean counted vote proportions in those precincts matched fairly closely the proportion of votes counted for each candidate state-wide. We also know (because the data is in the public domain) that the proportion of exit poll responses counted for each candidate did NOT match the state-wide count. We also have access to the questionnaires of those respondents whose questionnaires were subsampled for the crosstabs, because they are publicly archived. So in terms of voter-level analysis, we know the two did not match. So we know that either the vote count was corrupted or the poll sample was biased.

What else can we tell from those few thousand responses? Well, because there is no reason to suppose that the subsample was not truly random (and it certainly has the same statistical properties as the whole sample) we can infer quite a lot about who took part in the poll, as full demographic details are available. I'm not aware that anyone has bothered to analyse that data, but it might be interesting. Nothing to stop anyone trying.

But more relevant to the vote-corruption issue is precinct level analysis. This is why ESI commissioned their "blurred" dataset. Unfortunately, while several thousand responses is a lot, 49 precincts is not - as you will be aware, sample size is everything when it comes to statistical power, and 49 is a small sample for anything other than a fairly crude analysis. There are simply insufficient "degrees of freedom" to do anything more than zero-order or relatively simple partial correlations. That is why OTOH is correct when he says that the actual ballots would be a heck of a lot more useful.

Added to which, there is no reason to suppose that the exit poll data consists of "a proper sampling" of voters as they emerge from the polling location" and every reason to suppose it was not, particularly in Ohio. Mark Blumenthal calculated, and I can confirm, that in at least 60% of the Ohio precincts, the polling place served several precincts. There is also extremely substantial evidence that, nationwide, the poll was biased. There is no good reason to assume this was not the case in Ohio.

What OTOH is NOT saying, as you seemed to think he was, that the counted ballots constitute a better answer to the question as to who won than the exit poll responses. What he is saying is that if you want to look for anomalies that might "lead us more reliably in the direction of truth" then the actual ballots would be a heck of a lot more use. Investigating the "caterpillar crawl" for example. Looking for anomalous swing. Correlating swing with machine type; with a priori reason for fraud (was swing greater in Warren County, for example?); looking at the unused punchcard ballots for evidence of pre-punching.

It's a richer dataset for many reasons, but one important reason is that it's BIGGER. 49 precincts might have told us that swing and shift were correlated, and thus have indicated fraud. Unfortunately they weren't, so even that was a dud.

[papau]

18. of course stuffed ballot box/suppressed vote/denied vote make even the

"paper trail" suspect (although I do like the 1984 approach of the GOP - using "need to protect the integrity of the vote" as the title to their list of actions that destroy the integrity of the vote via stuffed ballot box by party controlled election Judge/suppressed vote via fake register voter roll cleaning and allocation of machines/denied vote via challenges that are obviously nonsense and via closed or moved voting places or the stationing of police cars to scare Southern black voters) .

The Venezuelan electronic dual paper trail voting system that Carter and his crowd audited works well - and is the best of the systems that are not just paper ballot such as Canada's, but the GOP in my day - and I suspect today - cheat in many ways that our media hides from the public.

The media has summed up the situation - after to being forced to see the cheating by the GOP - by saying that the GOP just want to win more.

Perhaps Mayor Daly of Chicago in 1960 had the right idea - vote the graves to offset the GOP voting frauds and crimes and then dare them to investigate the election voting and thus expose the GOP's own crimes. Perhaps that is the "American Way" and the only way to get "fair elections".

[papau]

24. Well said - but I think "exit poll proves election theft" NOT is OTOH's

and Febbles main point, based on the need for better proof than the limited exit poll data that we have.

By the way, my formulation -

"the exit polls result is a best estimate that suggest that the election was stolen, but there are not only many legitimate and legal ways the data could have come out as it has, there is also the fact that there is a lack of data that requires one to add that there is a possibility approaching 50% that the exit poll best estimate could be misleading and wrong in its apparent implication."

is a lousy sound bite! :-)

[Febble]

28. Slightly shorter version:

When people have more faith in the exit polls than the official results, something's gone terribly wrong with your democracy.

:)

[papau]

32. Well said-I believe that political hate radio is the cause of the distrust

that justifies in the GOP mind the anti-Democratic actions, and which has caused the growth in the numbers of GOP workers and officials doing the election theft.

Election theft actions have always been with us - in both parties - but the sudden rise in such actions to the point National elections are stolen without the media considering it newsworthy is a gift of the Reagan folks hate message and the moving of media into the control of the 6 GOP controlled corporations, with "Rove genius" behind the current efficiency of the operation.

Sad really, since Rove actually has maximized well the effect of the illegal Billionaire GOP campaign contributions via GOTV and message control efforts, but will instead be remembered for the destruction of the American voting system.

[garybeck]

25. RIP

may your successor be more objective than you.

[Febble]

38. An obituary from Mark Blumenthal here

http://www.pollster.com/blogs/

[Febble]

39. Washington Post piece here:

The Pioneer Pollster Whose Credibility You Could Count On

[Einsteinia]

41. Did he take his raw data with him?

Did we ever get our hands on that "proprietary" raw data?

[Febble]

42. Yes, all spreadsheets

were programmed to self-delete on his death. Not.

Seriously - where would it go?