Gaps between polls and results don’t point to fraud

January 29, 2008

Over the last month or so, pollsters gauging support in presidential contests have a mixed record. In Iowa, pollsters got the Republicans’ top two right, but everyone except the Des Moines Register underestimated Barack Obama’s support. In New Hampshire, the polls got the GOP race almost exactly right, but wasn’t even close in predicting Hillary Clinton’s 11th hour surge in support. In Nevada, pollsters got Romney’s victory right, but not the margin, while they were spot-on in the Dems’ race.

And in South Carolina, the polls were close to perfect on the Republican side, but underestimated Obama again on the Democratic side. All in all, looking at the eight contests in four states, I’d give the pollsters a B, maybe a B-.

What I wouldn’t do, however, is use these polls to make an argument about fraud. In this sense, Markos raises an excellent point.

The pre-election poll average in New Hampshire was: Obama 36.7, Clinton 30.4, Edwards 18.4

The final results were: Clinton 39, Obama 37, Edwards 17

So the polls nailed Obama and Edwards, while Clinton picked up the undecided vote. But overall, it was about a 10-point difference between the polls and the actual results.

Of course, this launched the Mother of All Whines, with morons across the internet charging fraud without knowing what the hell they were talking about. But it was Clinton! And she won! And of course, that meant that her victory couldn’t have been legitimate.

But it was legitimate. There was even a recount, and wouldn’t you know it, the results matched up fine.

If the surprise results in New Hampshire pointed to scandal, than the margin of Obama’s victory in South Carolina should have the same alarm bells ringing.

As Markos explained:

The pre-election poll average in South Carolina was: Obama 43.1, Clinton 28.5, Edwards 17

The final results were: Obama 55, Clinton 27, Edwards 18

So again, the polls pretty much nailed the second and third spots, but … wait … what’s this? Obama got 12 points more than the polls indicated? Overall, the poll average was 14 points off from the final results, worse than in New Hampshire. So this could only mean ONE THING — FRAUD!!!!!!!!

Did I mention that South Carolina uses ES&S touch screen machines with no paper trail?

Now, I should note, in case it’s not abundantly clear, that Kos isn’t seriously alleging fraud; it’s tongue-in-cheek. The point isn’t that there was mischief in South Carolina; the point is that there wasn’t mischief in New Hampshire.

I suspect irrational distaste for Hillary helped drive the fears about fraud in the Granite State, but that’s what makes Kos’ point all the more valid — some of the same people who threw a fit after New Hampshire, pointing to the gap between the polls and the results, have said nary a word after South Carolina.

Clinton won New Hampshire fair and square. Obama won South Carolina fair and square. Let’s move on, shall we?

Clinton won New Hampshire fair and square. Obama won South Carolina fair and square. Let’s move on, shall we?

Here, here. If we’re going to attack election fraud let’s start with those who are really guilty of it.

I didn’t really see anyone here trumpeting the New Hampshire fraud accusations, but maybe I’ve just already learned to tune them out.

  • I would hope this bodes well for Obama in places like California, where Clinton has like an 8% lead in the polls. The more people I have talked to about Obama in the last month, the more people tell me that they are quite impressed with him so far, and would like him to be the Dem’s pick, even if they were once leaning towards Clinton.

  • This is an excellent point. We need to remember that polls use a SAMPLE of the whole population, NOT the whole population. Obviously, they can’t contact every living breathing voter in a given contest, so the pollsters have developed some pretty sophisticated methods for gathering a representative sample of the population.

    For example, in New Hampshire they probably polled a thousand or two thousand people from across the state. From those responses they extrapolate for the whole state. This process clearly involves a certain error rate. To a certain extent luck of the draw plays a role. No matter how hard a pollster tries, it is always possible, by random chance, to run into more Obama supporters, or Edwards supported or to not run into enough Clinton supporters to accurately reflect the whole voting population.

    And, as the original comments suggest, there are always those pesky undecideds.

    On the whole, the pollsters know what they are doing and do a good job, but we should never forget that polling is not an exact science.

  • “Of course, this launched the Mother of All Whines, with morons across the internet charging fraud without knowing what the hell they were talking about.”

    Kind of reminds be of the gripes about Ohio in ’04!

  • Will Hillary win Michigan and Florida fair and square? Does the term “fair and square” enter into the Clintons’ vocabulary at this point?

    Sorry, I am actually not a Clinton hater by nature, but I absolutely abhore their tactics.

  • re “the results matched up fine”

    Huh? What is your source on that? Not only were there 2 precincts that were 4 and 10 percent off, the boxes arrived with evidence of tampering, and the recount was called of 40% of the way thru.

    The recount was an aborted mess. You surprise me. I have never read anything as ignorant on your site as the sentence “the results matched up fine”.

    And since statistical analysis of the results continue to suggest tampering (for Christ’s sake, look it up!), Markos is an idiot.

  • Not contesting these results I’m still against having equipment in place that “can” be manipulated especially when there is no paper trail to check it against. Kos’s post really has nothing to do with the use of these machines but they have proven themselves untrustworthy and this is the whole basis of our democracy…voting and having our votes accurately recorded and capable of being checked in cases of dispute. (a paper trail). I hate the idea of having the voting process privatized by companies who refuse to even allow anyone to see their source code or how their machines work.

  • When vote results don’t match exit polls, then we need to be concerned (e.g. Ohio in the ’04 presidential).

  • I understand the point.
    And I never shouted FRAUD in regards to N.Hampshire.

    But nevertheless, Kos argument is petty-boyism.

    Here’s why:

    The difference between N.Hampshire and S.Carolina is that the candidate that was favored to win by every poll LOST dramatically. If it had been the other way, if Hillary had been leading every poll and then lost… I expect FRAUD would have been shouted just as loudly.

    In regards to irrational Clinton hate… or whatever the semi-wags are labeling it–
    I assure you: my loathing is very rational.

    The problem with the left, and hopefully I’ll find time to write up the argument that has been forming in my mind, is that it continues to regard Bill Clinton’s blowjob and irate national lies as some sort of wink-wink, nudge-nudge not-really-a-sin sin.

    That position is both a failure of morality and intellect.
    The quality of the folks who inhabit the White House does indeed matter greatly.

    Lastly as a side issue:

    I don’t read Kos’s site at all. Can’t really pinpoint why.
    The format or whatever.
    Something about “my diaries” sits afoul of my intellect.
    I’ve seen him on TV.
    He strikes me as being a poor spokesman for the left.
    In terms of oration, I’ve found he articulates way below my standards.
    I want someone with wit, word power, and Gore Vidal-like talons.
    Not some namby-pamby stuttering sissy who gets in the way of his own punches.

  • I never doubted the integrity of American elections until the 2000 fiasco in Florida. (No, Republican trolls, I haven’t gotten over it. I will never “get over it.”)

    Ever since then, I don’t dismiss charges of election fraud so easily. When the “new, improved” tamper-friendly voting machines arrived with no audit trail, my paranoia doubled. When I learned that Diebold CEO Walden O’Dell met with the Rangers and Pioneers at the Bush ranch in Crawford, and that he sent out fund-raising letters for the Ohio Republican Party, my paranoia re-doubled.

    The 2000 presidential election was stolen in Florida. There’s no doubt about it. The 2004 presidential election may have been stolen – in O’Dell’s home state of Ohio. And you tell me “Move on – nothing to see here”? No way, Jose!

    Be afraid of election mischief in November. Be very afraid. Watch them carefully. Differences in poll results and election returns aren’t necessarily an indication of fraud. But they may be.

  • How did they match up with the exit polls? Seems like if there was a big discrepancy between those and the results you might be more concerned.

  • First of all, wouldn’t it be nice to know what the actual voter results of Iowa and Nevada were. To compare the polls with the state delegate results is a bit absurd. I don’t think either state has actually released how many people voted for each candidate, though many are calling the state delegate count “popular vote”.

    Second. the polls vary so much from day to day, they are often not even within the margin of error of one another. I would suggest that this means the pollsters are using false premises, like the 5000 who refused to participate will be statistically analogous to the 500 who did. Or even if 60% say they are Inds, we know by registration rolls that x percent are dems and x percent are reps.

    I would further suggest that the danger of polls is that they can be used to drive opinion. For example, is it possible that because “polls” were showing Obama with a 12 pt lead in NH that many Inds. went to vote for McCain in the Rep. primary.

    I’m not suggesting that any of the primaries or caucuses have been manipulated, but who thinks it is implausible that a general election in a season of wildly varying polling isn’t ripe for manipulation.

  • I think one of the more unusual things you are seeing this year that makes polling appear less reliable than in prior years is that polls by their nature are descriptive, not predictive. They are old by the time they are published, in that the data collection has already stopped prior to that; it is a snapshot of a particular moment or moments in the past. Moreover, they are a lagging indicator: polls often are in the field for multiple days, and even if they weight the newest day(s), some of the data is older. And very current news may not have filtered down enough to be prevalent in a random sample, but the people most likely to know it are also those plugged in – and most likely to vote.

    In most years, this doesn’t really matter: the pace of campaigns has been slower, there are rarely elections simultaneously as close and as complicated (i.e. more than two candidates still involved on both sides) as this year. The “still photographs” seem predictive because the pace of change is such that the final result hasn’t moved far from the last polling mark.

    But the compression of the primary calendar this year meant, for example, that there was barely time to do a decent multi-day poll in New Hampshire that had time to account for adjustments from Iowa. Some of the strong backlash from Bill Clinton’s late comments in South Carolina was too close to the election to be polled effectively. Things are just moving fast and so the static, backward-looking nature of a poll is much more evident this year. People try to use the square peg of polling in the round hole of prediction and this year is merely demonstrating that while such use has always been conceptually risky, in a fast-paced environment the conceptual flaw actually become a very important realistic limitation.

  • Markos is Did some folks get hysterical over NH? Yes. The possibility of mischief is not evidence of mischief, and those who confuse the two — as happened regarding NH — do immeasurable harm to the larger issue of whether we can trust our electoral processes. Crying wolf is only going to make the press less likely to cover questionable counting in the future, and make claims of wrongdoing easier to dismiss — and that is the real danger here.

    Opportunities for mischief exist throughout our electoral processes, just as they always have, but the introduction of electronic voting and counting make tampering more difficult to spot. As long as opportunities exist, there will be the temptation to exploit them — and fears that they have been exploited. This is no way to deal with what is perhaps the most basic cornerstone of representative democracy.

    What is needed is a set of “best practices” and the means to monitor and enforce them. When it comes to accurate counting, the two primary issues are verification (spot checks during the initial counting) and chain of custody. The latter is particularly problematic and current practices so lax as to be laughable.

    Here are some real gut-busters regarding “sealed” ballot cartons and ballot security. Chain of custody issues regarding memory cards is even worse.

    http://www.youtube.com/watch?v=hHL_YMBolRs
    http://www.youtube.com/watch?v=PKQEQ7qHvgM

  • I had to laugh at all the folks going all black helicopters about NH and raising the specter of DIEBOLD seemed to be having so much fun, I didn’t have the heart to tell them that 100% of the votes in NH were cast on good, old-fashioned, tree-killing, human-readable paper ballots marked by the voters own hands. In many parts of the state they even still count them by hand, although some of the more populous areas did use op-scan machines for the original counting.

  • Correction:

    I had to laugh at all the folks going all black helicopters about NH and raising the specter of DIEBOLD. They seemed to be having so much fun…

  • There are some of us who intentionally give mis-information to pollsters, just to stick it to the people who “call” elections one minute after the polls close.

    When polled during the 2004 election, I told one pollster that the biggest issue facing america was the overpopulation of the salmon fisheries.

  • The difference between a poll and a ballot is that the poll includes a “vacancy” for the undecided vote—and the ballot does not. Polls will count intentions for Obama, Clinton, and Edwards, and sum those intentions up as 100%. But a large number of “undecided” intentions cannot be recorded on the ballot as “undecided”—and those people tend to make up their minds within the last 24-to-72 hours before the vote.

    Iowa and New Hampshire had A GREAT BIG BUNCH of “undecided” voters. Everyone knew about them, and the “safe” pattern is to divide them into percentage groups synonymous to the “intentions” of declared voters-to-be.

    This election, however, is different. Edwards is running behind because he’s just an “average” candidate who’s facing two extraordinary event-scenarios—a qualified woman and a qualified minority who both have the draw of a super-electromagnet. If the primary season were revolving around either an Edwards-v-Clinton or Edwards-v-Obama top tier, the primary would no doubt be all but done by now; probably with a 70/30 split on the votes. I don’t think Edwards would have even won SC, because both Obama and Clinton are capable of sucking the wind out of his sails.

    The undecideds in Iowa didn’t like being told they were going to vote the way the pollsters said they would vote. The same held true in New Hampshire—and both groups were about big enough to inflict the change that took place in those two primaries.

    It’s also a probable scenario to explain Nevada and SC. Rabid Clintonistas are rabid Clintonistas; just as their NY-NOW counterparts did, they will take any and all steps to degrade the opposition and its support, and won’t leave the sinking ship—even if they’re staring point-blank at the last life-boat as it’s being lowered into the sea, nearly empty. They will not change their minds the day of the primary, or the day before—or the day before that. The same will hold true with Obama-philes.

    So the only “unexplained phenomenon” has to be the undecided vote. They’re the ones swinging things around this time around—and they cannot be predicted….

  • another important point, the polls are not the key here.

    what is the key is that votes counted by hand favored Obama, votes counted by machine favored Clinton, and when the results are adjusted for income, geography, etc., there is no reason why the machine counted votes should skew so heavily toward Clinton.

    http://www.thenation.com/doc/20041129/baker

  • To add to TonyT above, another non-polling matter that warrants checking into is the order in which candidate names appeared on the ballot (no black helicopters, sorry).

    Statisticians and survey designers are familiar with the primacy effect, which gives an advantage to those listed at or near the top of a list, and to counter that, randomize the order — usually by precinct. Apparently that was not done this year in NH; “randomization” was “accomplished” by chosing a letter to start an alpha sort. The random letter was “z,” placing Clinton near the top and Obama well toward the bottom of all ballots (if memory serves, there were about 20 candidates listed). A study coming out of Stanford (again, if memory serves) looked at the effect of primacy in past NH elections, and found that the candidate appearing earlier received up to a 3 percent advantage.

    While three percent did not affect this particular outcome, someone should explain why the ballots were not randomized by precinct this year when they had been in the past.

  • beep52, i’d be willing to venture a guess on your last point: money.

    it costs a county (or a party, whichever pays for the primary in NH) a lot less to print more copies of fewer versions than fewer copies of more versions. so as no one ever wants to pay taxes anymore, or costs for other party activities go up, and budgets get tighter and tighter they decide to cut corners – hey, the choice of “z” was random! and it isn’t the general election, right?

    screwy Bush/Norquist economy makes messes that are subtle as well as ones that are obvious.

  • Beep 52 (@22)

    Somehow, I’m not happy with the idea that people, whose attention span is too short to sort through the alphabet from A to Z, are permitted to vote 🙂 If I knew, for certain-sure, that “my” candidate was on the ballot, I’d have no trouble to scroll from Abramoff to Zacharias and I don’t see why it should incommode others. If you go to the trouble of getting your butt over to the polling station at all, and for primaries at that (which suggests more than a tepid level of engagement)… How likely is it that you can’t be be bothered to read past the first name on the list?

    Clinton is my 3rd — and very reluctant — choice, but, even so, I’m not buying that argument.

  • Steve…

    I’ve enjoyed your posts and analysis for quite a long time, but this one is so far off the mark that I don’t even know where to start. You sound like the idiotic wingnuts in 2004 sticking their fingers in their ears and going “neener neener neener” when Democrats chose to overlook the oddities in Ohio (or claim without proof that any fraud or errors wouldn’t have changed the outcome).

    Here’s the thing about NH. It wasn’t just the PRE-election polls that showed the wrong winner, it was also the unadjusted EXIT polls. With 80% of the ballots in NH being counted by optical scan machines with proprietary software and no audit, and with the 20% of ballots that were hand counted matching the exit polls, while the 80% machine counted didn’t, there was a giant red flag thrown up.

    As citizens of a democracy, regardless of who the candidates in question are, we fail in our duty if we look the other way when red flags such as that are thrown up. It doesn’t mean we scream fraud. After all, there are many reasons that the results could be off, including machine error, ballot box stuffing, or polling errors. There is only one way to know for sure – count the ballots by hand and investigate the procedures.

    Thus far, other than a couple of very unusual results, most of the hand counts have been close to the machine counts. Unfortunately, the ballot storage and chain of custody procedures in NH are so poorly managed that it is impossible to state with any certainty that there wasn’t any tampering. Without any evidence of tampering, however, one has to lean towards the conclusion that the election results were accurate.

    That does not in any way provide a “neener neener” moment. And anyone who has that reaction is an absolute moron (and yes I am calling KOS a moron, but anyone who looks at his site with an open mind knows he is a leftwing gatekeeper and not a true progressive). What you SHOULD feel is gratitude that your fellow citizens invested their time into verifying an election result that appeared questionable, and in the process, uncovered severe problems with NH’s procedures that need to be rectified.

    I strongly recommend that anyone who thinks that calling for a recount in NH was a waste of time have a look at these videos by BlackBoxVoting:

    http://www.youtube.com/watch?v=hHL_YMBolRs
    Who’s responsible for chain of custody breakdown?

    http://www.youtube.com/watch?v=PKQEQ7qHvgM
    No ballot vault tonight

    http://www.youtube.com/watch?v=PiiaBqwqkXs
    Silvestro the cat

    Also, have a look at the pictures and commentary here:
    http://www.bbvforums.org/forums/messages/1954/71456.html?1201461851

  • libra: with all due respect (and I do mean that), primacy is an accepted and well-documented phenomenon — and rotating questions is routine among reputable surveys. Here’s an excerpt from an article from a Stanord researcher posted at ABCNews:

    Our analysis of all recent primaries in New Hampshire showed that there was always a big primacy effect — big-name, big-vote-getting candidates got 3 percent or more votes more when listed first on the ballot than when listed last.

    • Comments are closed.