Dead Cat Bounce

I don't usually report on the minutia of politics or polling, mainly because it bores me to tears, but I had to make this post because it lets me use one of my all-time favorite terms.  Bush's recent rise in the polls reminds me very much of that great investment term "dead cat bounce."  (If it falls far enough, even a dead cat will bounce).   I've always suspected that many of the technical analysis used on Wall Street to analyze stock trends could be applied to political polls, since they encompass some of the same group distributed consensus building.   I can see it now, Paul Kangas reporting that President Bush is experiencing a break-out to the upside...

By the way, are there really people who change their opinion about the war, about the president, about how they will vote on a weekly basis?  It sure seems like there are 5-10% of Americans who blow around with the wind.  I don't mean change your mind once, like changing your mind on the war.  I mean back and forth every week.  Otherwise, how does one explain the fluctuation in the polls, particularly when the amount of the fluctuation is outside the error range?


  1. Steve:

    If everyone changed their opinion once, and 2.5% of people did so every week (and in the same direction), you could get 5% shifts every week for a year.

    Obviously, I don't think this is happening, I'm just pointing out that the polls don't necessarily mean that the same people are changing their minds frequently.

  2. Jim Collins:

    I remember reading a few weeks ago that the management of one of these polling companies was charged with fraud for skewing the results of some of their polls. They even went as far as falsifying data to get results that their customers liked. I don't trust polls and think that too much time and money are wasted with them. People don't take them seriously, and some (including me) actually give misleading answers just to try to screw up the results.

  3. Bill:

    I suspect that the basic problem is twofold: First, it is far to easy to drive results by wording of questions, whether by accident or with intent. Secondly, cost drives pollsters to use minimun sample sizes, in which small errors in selection can produce large swings in the result.

    There are other problems as well. Many people will not participate in polls, and it is not clear that those will necessarily vote in the same fashion as those who will participate. Given the relatively small number of persons who actually vote, polls that do not attempt to control for likelihood of voting are bound to lead to error.

    Bottom line is that polls may be slightly interesting reading, but their reliability is so suspect that it is hard to take them seriously.

  4. Tom Dilatush:

    An additional "noise" factor is that the press often reports poll results in a completely generic fashion (e.g., "Bush's approval rating"), using the results of a series of polls done by different pollsters, using different methodologies. Apples and oranges, in otherwords, all being reported as grapes...

  5. Matthew Brown:

    There's also the fact that polled people inevitably answer in the way they "think they should", especially when they haven't actually made their minds up on the topic.

  6. Mike:

    I think some of this is inherent in the questions. If I am asked whether I approve of the job the President is doing at any time I would vote no when the President is withdrawing into his shell and trying to please the liberal statist side of the house. When he starts turning around and coming out strongly to explain his policies I would then vote yes I approve. At no time though have I changed my views on Iraq, or on economics, or on freedom. In other words my principles are not necesarily changing when I answer the question of whether I approve of what he is doing right now.

  7. Matt:

    Ever since TV networks took seriously a poll listing "war on terror" as a seperate category from "Iraq", but lumping all "moral values" into one answer choice, I've lost what little respect I had remaining for the results of polls.

    Mike also has a point. Job approval questions, even when phrased with proper neutrality, can still reflect a shorter time horizon than a voter's decision on election day does. Do I approve of the way Bush is handling the country right now? Hell no. Doesn't mean that, even in retrospect, I regret voting as I did in '04, nor does it mean that, were he not Constitutionally disqualified from ever again appearing on a ballot, I wouldn't vote for him in the future. Most of my beefs with Bush are tactical, not strategic. And yet, were I ever contacted by a pollster, I'd probably get counted in with the screaming "Bush==Hitler" moonbats.

    BTW, why ARE they spending money testing Bush's approval numbers, anyway? He cannot legally run for office again, except in Texas, and is unlikely to pursue any of the lower offices for which he's still allowed to run. Who cares what his approval ratings are?

  8. Noumenon:

    "In election years from 1952 to 2000, when people were asked whether they cared who won the Presidential election, between twenty-two and forty-four per cent answered “don’t care” or “don’t know.” In 2000, eighteen per cent said that they decided which Presidential candidate to vote for only in the last two weeks of the campaign; five per cent, enough to swing most elections, decided the day they voted."--interesting New Yorker article on ideology

    Most people do not have an internally consistent or externally consistent set of political views the way you do, according to that article. A 50/50 nation really is a coin flip, and people really are changing week to week.