Search This Blog

25.7.10

Melancon tries again with more questionable numbers

Even if its tactics are transparent, one must credit the Democrat Rep. Charlie Melancon campaign for Senate for its extreme tenacity in the face of long odds. Or it may be just one more potential sign of desperation on its part.

Rarely does a campaign release its internal polling numbers but with the recent release of information it has collected relevant to Melancon’s challenge to incumbent Republican Sen. David Vitter, the campaign now has done so twice this year. Normally, this is not done because of concerns about appearance of partiality and internal strategy concerns.

But as all but two polls releases well over the past year have shown, Vitter continues to enjoy a large lead and/or majority proportion of the vote. In most he crests 50 percent of the intended vote, and in almost all he enjoys a double-digit lead. In just about every independent poll, he has had both. The two exceptions to these conditions, guess what, are the internal Melancon polls conducted by a firm that almost exclusively polls for Democrat campaign.

Previously, the inadequacies of this approach for understanding the true competitive balance of a contest and the inherent unreliability of the results (for example, without having knowledge of the sampling frame or exact question wordings) render them useless for telling us anything meaningful about a contest. However, because one news outlet cast these concerns aside and ran a story stumping for the questionable numbers, perhaps the campaign thought it could get such favorable coverage again by repeating the exercise – especially because, as divergent as the result were last time, these were even more out of step with the rest of the polling (including independent ones) world.

This latest edition has Melancon behind but within the margin of error when no other poll (even other partisan Democrat ones) shows him close to that. The campaign claims it is because of an obscure personnel issue that few voters know about, so that is an unlikely source of such a large difference. This admission probably means that was part of a push poll process, where the pollster asked one or more questions highly loaded and distorted against Vitter before asking for a vote intention. We would know about this and other matters such as sampling frame and the like only if the Melancon campaign releases unexpurgated data and reports from the poll, and it is unlikely to do that because of the problematic reliability and validity aspects that are sure to come out.

But we can conduct an exercise that researchers who gather and analyze primary data use, estimating construct validity of the concept of debate, vote intention. When a researcher wishes to evaluate how well an instrument, such as a survey question, really reflects the concept being studies, construct validity is established by seeing how closely related the measure in doubt is to others that supposedly do the same. Since all other studies are showing a much different distribution of vote intention that Melancon’s internal numbers released publicly, we must conclude this one doesn’t show much construct validity and hence isn’t doing a good job in revealing the reality of the Senate contest.

Yet that’s not the purpose of the number’s publicity, that being to jumpstart the Melancon campaign and to remove the long-standing air of inevitability that not only will he lose, but lose big. As such, it represents spin and nothing else of value.

No comments: