Search This Blog


In dueling polls Melancon's, his chances found wanting

Only a couple of weeks ago, one poll showed Sen. David Vitter continuing to open up what already appeared to be an insurmountable lead even further on. Rep. Charlie Melancon’s bid to wrest the seat away from him. The last week, Democrat Melancon had one released showing that he was “only” 10 points down. What explains how the Republican can be up 24 points in the other?

First, understand that while Melancon’s poll was performed by somebody he picked and paid, a firm that specializes in working with Democrats, Vitter has no connection with the firm that has been tracking his (and many other candidates’ races), Scott Rasmussen. Generally speaking, independent pollsters such as Rasmussen produce results of greater validity because hired guns, even as they won’t admit this for obvious reasons, will pick question wordings and sampling strategies that tend to provide a small bonus for their clients. Their workers also tend to unconsciously alter emphasis, tempo, and even words in delivering questions, often subconsciously. As obviously they can’t introduce deliberately an unmistakable bias into their results (at least in the numbers they give their clients, which may be different from the ones reported) because then inaccuracies would show and they would lose business, but wordings and sampling frames can be subtly shaped to present typically a bit more optimism into the scenario.

Second and much more significantly, it appears that (because while Rasmussen’s are done on his own time and he reports extensively and publicly on them Melancon’s pollster’s information is proprietary so you can go only by media reports of it), as opposed to Rasmussen’s using likely voters, Melancon’s used all registered voters. Historically, this overestimates voters for Democrats and that looks particularly risky to do this election year as Republican-leaning voters seem far more enthusiastic than Democrat supporters to turn out in November. (Another bit of evidence and/or consequence of the presumed use of all voters by Melancon’s firm is it only got 86 percent of respondents to state preference while Rasmussen got 90 percent – unlikely voters are much more likely to call themselves undecided.)

Third, Rasmussen has built quite a reputation for accuracy over the years. The latest round of elections found his organization in the top spot for the top job. This is true even as Rasmussen often employs automated methods, as opposed to Melancon’s pollster who did not appear to do so. While there are a number of variables, all other variables outside of the live vs. recorded debate held constant, that might produce variance between the two methods (here is a good layman’s discussion), the fact is there seems to be no inherent bias in the direction of any party’s candidates using either method. Specifically in this case, that means use of interactive voice response polling would not inflate Vitter’s numbers, nor depress Melancon’s, and vice versa (if the interviewers are well-trained, experienced, and not themselves biased) for live-voice calling for inflating Melancon’s and depressing Vitter’s.

In any event, to some degree at this point the discussion still is moot – so what if Vitter is just a half rather than whole landslide ahead, he’s still going to win comfortably even with Melancon’s numbers. And while if it were a year such as 2006 with strong Democrat tides it could be argued Melancon could overcome a 10-point gap in the next eight months, when in fact conditions are the opposite as they are this year it would take extraordinary optimism and audacity to think that could happen.

Thus, weighing the evidence, the reality is much more likely to be closer to the Rasmussen numbers than Melancon’s.

No comments: