Charlie Martin brilliantly (and without malice) explains why polls are not helpful in this election
Bookworm on Oct 05 2012 at 9:16 am | Filed under: Presidential elections
I will be the first to admit that, when it comes to math and science, I have . . . ah . . . issues, at least as to some subjects. For example, you can talk to me about physics and aerodynamics all you want, but I know that the reasons planes fly is magic and prayer, neither of which are always reliable. I also know that germs can leap of toilet seats and attack people within a five or ten foot vicinity. You can tell me differently, but I won’t believe you.
As for polls, they’re the worst thing of all, a peculiar amalgam of statistics (confusing and evil), math (confusing and evil), guesstimates (just evil), and all sorts of other stuff. Add in my distrust for many of the media and polling organizations involved in the whole poll thing and I’m pretty much going to discount the polls.
What Charlie does is explain why the polls are of dubious value this year, but it’s not because of evil or bias. It’s because this presidential election year is a presidential election year different from all others. Even with the best will in the world, polling organizations are facing unexpected hurdles that may destroy (or, at least, substantially devalue) their polls’ utility. Best of all, Charlie explains it all in terms even a magical thinker (that would be me) can understand.
Email This Post To A Friend
8 Responses to “Charlie Martin brilliantly (and without malice) explains why polls are not helpful in this election”
Leave a Reply
You must be logged in to post a comment.
I have always taken the polls quite seriously and considered conservative claims of bias so much wishful thinking. However, this makes me re-think my view. If it turns out that Republicans and independents are more likely to refuse to be polled (or that Democrats are more likely to refuse, for that matter), that would introduce a serious bias. I suppose we will have a major clue when we compare the actual results to the final polls.
Well, Don, the key is that you are taking the polls very seriously NOW.
Therefore, you should compare the voting results against the polls you’ve been paying serious attention to for the last month, not just the final polls. If the Democrats are playing games, they’ve already been at it for awhile.
Yes, you’d have to account for swings in voter choices over the last month. But the claim is they’ve been oversampling Democrats in their models for quite a while now.
I certainly think they are *biased*. What I’m not sure about is whether they’ve made the shift to being Democrat operatives, the way the mainstream media has. There is a world of difference between mere bias – which results in unintentionally wrong outcomes – and being an operative, which means planning and executing deliberate deceit.
Respectfully, Mike, that makes no sense. Polls cannot possibly take account of changes in voter preference between now and the election. Polls do not predict where we will stand a month from now; they only show where we stand today. But that is useful information. For example, the changes over time show which persuasive efforts and events (ad campaigns, themes, grass roots efforts, changes in eonomic conditions, international events, etc.) are having an effect and which aren’t. Note that, assuming that the polls do not change their methodology midstream, these changes are useful even if the poll is biased.
The problem, DQ, is that they *do* change their methodology during the last month or two. For example, they will often switch to likely voters or registered voters during the last month.
Over what period of time do you pay serious attention to the polls? You are aware that many polls this year have been assuming a turnout model similar to 2008, when Democrat enthusiasm could fairly be described as “blow the roof off” enthused. Some of these 2012 models even assumed a *higher* percentage of Democrat turnout compared to 2008.
In short, the inputs change and/or the model changes. You can’t draw inferences or correlations over time when this happens. Well… you CAN, but of what worth are such conclusions?
Given the differences in approach of all the various polling organizations, and their varied results, I would pick and choose which ones you trust the most. Even then, if you only look at the *final* polls, do you really think that tells you anything about the polls that were taken through the prior months? It would, if methodology and approach and all other poll inputs remained constant, but they don’t. So if you plan to use the final polls to draw conclusions about the polls over the prior months, I don’t know that that makes any sense either.
I agree that when a particular poll changes its methodology one will not be able to compare results before the change to results after the change for the purposes I outlined in my last comment. But that would be a special case and would not impair the usefulness of the comparisons when the methodology has not changed.
My point was that if the final poll results are substantially different than the actual voting, it would support the theory that the refusal of a vast majority of people polled to cooperate may distort the polls, making them all untrustworthy.
Any time the participants in a poll are self-selected, there is the danger that the poll will reflect nothing more than the bias of the self-selecting group. That’s why all on-line polls in which anyone can log in and vote are useless. That didn’t used to be a problem with traditional polls where participants were randomly selected and the majority so selected participated. But now, with most people refusing to participate, the small group of those who do participate is, essentially, a self-selected group and the poll may reflect nothing more than that group’s bias. The question is whether that small group has a bias. Comparing the final polls to the actual results will suggest an answer to that question, though it obviously will not answer it with certainty.
Oh, I just realized I didn’t answer your question. I’m talking day-to-day and week-to-week changes reflecting responses to the kinds of events I identified in my first comment. I’m not talking about underlying assumptions or ultimate predictive capability.
As to your point, I agree you’d have to strain to explain using 2008 as a model for 2012, but you’d also have to strain to use thde 2010 off-year election as a model and I’ve heard that some conservatives, including Rush, have tried to do that. I don’t know if any polls have. My personal guess, and it’s nothing more than a guess, is that 2004 is a better turnout predictor than either 2008 or 2010.
It’d be nice if the Leftist propaganda arm was so much wishful thinking. But that’s more of a product of the people they target.
I agree with every word you said in #5 and #6, DQ.
I paid attention myself to the *shift* in polls in favor of Obama that happened about two weeks ago. It was well after the conventions, and I couldn’t see anything that would seem to justify it. I did dismiss the actual percentage values, but not the shift itself towards Obama. *Something* clearly had changed in his favor.
I decided that for that time period, the economic news had been better than usual. As I put it, the news was only terrible, not horrible-terrible; and that constituted a shift that could cause polls to move. Also, gas prices fell dramatically, and that always helps the incumbent. So that’s my explanation for that two-week shift in Obama’s direction. But my main point is, you *can* disregard the percentage values themselves, and still find value in the polls in detecting shifts in momentum.
Many conservatives think the pollsters are acting as Democrat operatives – using deliberate deceit – in producing fake poll numbers, with the expressed goal of depressing GOP turnout.
While I am convinced that most pollsters are biased, I’m not yet convinced that they’ve moved from bias into the realm of deliberate distortion and deceit. There haven’t been any smoking guns yet among polling operations.
There have been numerous such smoking guns proving that the mainstream media made this move towards operatives, towards deliberate and malicious deceit, over the last ten years or more. I’m not willing yet to lump the pollsters in with our worthless, disgusting mainstream media.