30 thoughts on “Poll Skewing

  1. Arizona CJ

    This is the first peice of any kind, anywhere, that I’ve seen event mention that pollsters have a financial motive to skew. I’ve been singing that song for months, but this is the first time I’ve seen anyone but me mention that political bias is not the only explanation.

    Maybe I’m not so crazy after all? Well, okay, I’m still crazy, but maybe not in this particular topic. :)

    As for the Obama campaign; skewing the polls as a strategy makes little sense in some ways, but they have made massive mistakes before. Also, there is a viable angle; if they convince the majority that Romney isn’t just behind, but he’s so far back he has no chance, it has a slim chance of working. A very slim chance. Not something to bet the farm on, unless they had no better option. (I think the latter is the case).

    Also, about pollsters; bear in mind there is one, and only one, measure of their accuracy; election day. So, only when that draws near do they start worrying about their reputations. My guess; this game will go on until around two weeks prior to the election, then they’ll start trending towards accuracy.

    I also find it interesting indeed that the big Republican financial advantage has yet to come into play. My hunch; it’ll be used for a massive blitz, at about the same time the polls are showing a Romney “surge” two weeks out.

      1. Arizona CJ

        True, but if their motive is financial, as I think it is in most cases, that’s fine with them. Showing a close race helps them, and as the linked article suggests, they might also make money from partisan sources for slant.

        Fox puzzles me, as I see no motive for their polling to be slanted, but… do they do their own or outsource? Also, their record in the past hasn’t been good at all.

        1. Annoying Old Guy

          Also the do the corrections right at the end when momentum matters least.

          As for motive as many people have pointed out there is (bizarrely to me) a “bandwagon effect” where people will vote for the person they think will win the election. The goal of the slant is to pick up some points that way, and to discourage donations / support for Romney. Whether it gooses Obama’s campaign is purely secondary.

        2. Leland

          Fox puzzles me, as I see no motive for their polling to be slanted,

          Are you sure about that? Don’t they have the same exact financial reason? I quit watching the news after constantly being teased about some important news story “you need to see this!”, and then waiting 20 minutes to actually get that bit of news. The news isn’t important;watching through the commercials is. Fox may tell you the polls are skewed, but it works for their ratings too to have a close race.

  2. Jim

    This piece seems to take for granted that polls are skewed in a partisan fashion. That’s a pretty serious charge. Where’s the evidence? “People aren’t giving pollsters the answers we expect” is not evidence.

    the published totals are “weighted” (i.e. arbitrarily distorted) to match statistics about past voter behavior

    No. The major pollsters weight samples to match Census data, e.g. so you don’t use a sample where 60% of the voters are women, or 40% are under 30. The big name pollsters don’t weight samples by party identification.

    I can see the reason why the Obama campaign would want polls tilted his way at this point in the race. A lot of Romney’s money advantage isn’t actually under his control — it’s in the hands of the RNC and Super-PACs. If those people think that Romney is toast, they might spend their money on down-ticket races, turning the polls into self-fulfilling prophecies. But that conspiracy theory requires not only that the Obama campaign have some way to manipulate polls by a whole bunch of media organizations, but also that the RNC and Super-PACs don’t have their own, unbiased polling.

    Occam’s razor suggests that the polls show Romney losing because Romney is losing.

    1. Geroge Turner

      You have to weight polls for party affiliation because members of the two parties don’t answer their phones with equal likehood, nor are you equally likely to get through to busy parents with phone-age kids (“single – no kids” are probably oversampled, and skew toward Obama).

      We went through a period where polls picked up a heavy Democrat bias because Republicans were earlier switching to cell phones and polling companies were still using land-line phone books to make calls. Good polling firms also know to add more control on calls made on Fridays and weekends because Democrats are far more likely to be at home then.

      The other bias is that most pollsters hire college students, poor, and minorities to man the phone banks because they work cheaper, and many people who get called will change their responses to match what they think would make the caller happy. I do that [i]a lot[/i]. All conservatives do, because it’s not polite to upset a bunch of liberals and we learned long ago not to do it if we want a pleasant evening or a chance to hit on some cute co-eds. So when we’re out to dinner, if a couple of people at our table are praising Obama whlle we’re waiting for the salad to arrive, we just smile and nod in agreement. Democrats do this too, especially when they’re in a room full of old men in corporate suits. Smile, nod, and agree.

      But pollsters rarely sound like corporate suits. They usually sound like young college kids, leaning heavily toward black co-eds. That alone will bias the responses no matter how carefully you control for demographics and party affiliation, and questions they don’t even ask, like “are you married, dating, any kids, own a lakehouse?”.

      1. Jim

        We went through a period where polls picked up a heavy Democrat bias because Republicans were earlier switching to cell phones

        Huh? It’s young people who are dropping landlines, and they tend to vote D. The pollsters who call cellphones are showing a 1-2% tilt towards Obama compared to the ones that only call landlines.

    2. Mark Buehner

      Look into the internals of any of these polls (if they publish them!). You will see something like 38% Democrat 29% Republican 33% Independent. In 2008 the affiliation was +7% Democratic and Obama won by 7. In 2012 it was +1% republican and the Republicans swept the house. Its now about +3% republican.
      Why is this happening? Because traditionally the way these presidential polls have been done is to look at the last presidential election and use that affiliation to weigh the polling. But that isnt working this year because Obama had a completely unprecedented turnout in 2008. So these polls are assuming that same level of support, which is clearly ridiculous. Thats why the polls are broken this year, and year Rasmussan and Gallop are completely divergent from then (those two dont use the 2008 assumption). It all comes down to how you define your likely voters.

  3. Josh Reiter

    Some pollsters do in fact do trend models based on party identification. Others don’t. The ones that don’t make it easier to understand what’s going on when they report +8 Dem responses because you can subtract that number from the Obama numbers and get closer to what is going on. Sort of like what happened in the Scott walker recall. The polls were to close to call. But then he ended up winning by 8% which corrected quite nicely with the party trend bias.

    1. Jim

      Sort of like what happened in the Scott walker recall. The polls were to close to call. But then he ended up winning by 8% which corrected quite nicely with the party trend bias.

      That’s an odd version of history. What actually happened is that the final RCP average had Walker up by 6.7%, and he won by 6.8%.

      Note that the final polls ranged from Walker+3 to Walker+12. That’s a wide range, but the poll average was right on the money. Another example of why poll averages are better than cherry-picked individual polls.

  4. Edward Wright

    But why would Democrats donate more if they saw Obama way ahead in the polls?

    It’s called ROI. If a candidate doesn’t win, he can’t provide payback after the election.

    Naive, isn’t he?

  5. Geroge Turner

    Another issue that directly affects the media is that if Obama is behind, they have to start talking about how his campaign is failing, and why, and they can’t do that without bringing up how bad the economy actually is and how many people are suffering, etc. It would push them off message.

  6. Barbara Skolaut

    “But why would Democrats donate more if they saw Obama way ahead in the polls? I’d just keep my wallet in my pocket, having better things to do with the money.”

    But, but, but – he’s a GOD. All their money (and apparently ours, too) belongs to HIM!

  7. Barbara Skolaut

    Poll story from this evening: I decided to start answering the phone and see if the out-of-town callers were pollsters. Got one, and agreed to take their political poll which would “only take a few minutes.”

    It took less that 60 seconds. 3d question was about my ethnicity. As soon as I said I was white, they thanked me and hung up without actually asking any real questions.

    Wonder how they were skewing that one? I knew the pollsters skewed by party, but by race?

    1. Geroge Turner

      You know, it would only take a small percentage of white Republicans to lie about race and party affiliation to screw minority polls all to heck. ^_^

    2. Jim

      I knew the pollsters skewed by party, but by race?

      You’ve got that backwards. Of course they correct for race, because they know the racial breakdown of the country, and they know that a sample with a different breakdown would be a poor representative of the entire country. They do the same for gender and age.

      They shouldn’t correct for party, because they don’t know how many Americans consider themselves a Democrat, Republican, or Independent at this moment. Party affiliation changes on a day to day basis. That’s why they’re taking a poll!

      1. Mark Buehner

        Ah, here’s your misconception. We have to talk Registered Voter vs Likely Voter. In most likely voter polls, the poll _is_ weighted by affiliation. They are looking to match voter demographics to a model, in this case the 2008 election for most polls. If the poll is +8% democrat, that isnt (generally) because +8% democrats answered the phone.

        Check out the Rasmussen running voter affiliation polls (its been running monthly for years). In 2008 the affiliation was +7% Dem, now its +4% Republican. Its very difficult to believe Democrats have the same level of support they did in 2008, which would be the case if all these polls were coming in as you suggest- it would mean that somehow they are answering these polls differently than the are answering the Rasmussen calls. The Rasmussen curve shows a steady trend that explains well the 2010 election.

        1. Jim

          Its very difficult to believe Democrats have the same level of support they did in 2008

          A +7 Dem party affiliation does not imply the same level of D support as 2008; it could just mean (as some have suggested) that more Rs are identifying themselves as independents (e.g. they identify with Tea Party more than with the GOP).

          they are answering these polls differently than the are answering the Rasmussen calls

          Rasmussen’s party affiliation numbers are not consistent with those of other pollsters. Ras uses robocalls, and I don’t think they call cellphones, which may be a partial explanation.

  8. Gregg

    Another effect of the perception that one candidate is way ahead is that some fraction of the voting public just prefer to vote for the winner….or the candidate they are told is the winner.

  9. Joseph Hertzlinger

    One reason Romney might be in unexpected trouble is that the election might be in the hands of people we wingnuts have little experience with even when it comes to arguing with them: unreconstructed New Dealers (and their children and grandchildren) who still distrust Big Finance.

    In this election, Romney is regarded as the financiers candidate. He is regarded as one of the people who lay off workers for a living. The higher unemployment gets, the more Romney and his alleged gang will be blamed. Please remember that FDR was reelected in a landslide with higher unemployment. (The belief that Party X is better at dealing with Problem Y and that it had better be kept in power when Problem Y is raging out of control has benefited Republicans in the past. McGovern ran against inflation and lost badly.)

  10. ken anthony

    This piece seems to take for granted that polls are skewed

    You’re right Jim. We’re all in denial of how massively screwed up America has become. We are living in some SF alternate reality where someone with no executive experience and such a radical ideology that believes his very words can calm the seas would become our president. We will be stunned when he is reelected rather than losing in a landslide. Where the media has no knowledge of history nor a patriotic bone in their bodies, nor a shred of journalistic integrity. Where we can spend ourselves into oblivion.

    We’re you are the wisest of us all.

    1. Jim

      You’re being facetious, but the right’s closed information loop is one source of Romney’s troubles. The GOP has been so sure that Obama is a miserable failure that they lost sight of the fact that most of the country doesn’t see him that way. It’s hard to win over voters when you can’t imagine that they see things the way they do.

    2. Geroge Turner

      Our closed information loop comes from our sincere but mistaken belief that we can convince half the public that crops should be irrigated with water instead of Brawndo.

Comments are closed.