Over the years, the number of political polls has exploded. Thirty years ago, the only outfits who concentrated on political polling were Gallup and Harris. Now there are literally dozens of polls and polling companies. Several media alliances operate their own, several serve each political party, and many independent polling firms are in the field. The science of polling, like almost every pursuit besides love, has advanced dramatically.

Yet, reliability has never been more in question, despite the scientific progress. Why is this?

Many astute observers have been deconstructing polls in recent years. Every aspect is examined - samples, screening, adjustments, the wording of questions. Every methodology is questioned, and many times the analysis accounts for errant results. Not every poll releases their "internals," the detailed data the poll harvested to reach its numbers. Those who do make this information readily available in the spirit of "transparency" generally have more credibility - they aren't afraid to have their numbers checked.

Some polls never disclose this information. They consider it "proprietary information" - which competitors might use to take away business. There is an argument there, but it is rendered less effective by the fact that many polls DO disclose their raw findings and methodology, and they don't suffer by it.

There is one area no one talks about, though - the one statistic that NONE of the major polls disclose. They ALL conceal it as "proprietary." It is, perhaps not coincidentally, one which could explain the growing disconnect between poll numbers and actual results.

I speak of the "response rate." Never heard of it? Well, there's a darned good reason pollsters don't talk about it.

The response rate is the percentage of people actually contacted who agree to answer the poll questions. Theoretically, the higher this number is, the more accurate the results will be. Up until the early 1980s, national polls had response rates over 75%. From that time forward, the response rates began to decline steadily. It is claimed that this has now fallen to around 25%, although no one in a position to know will go on record with that.

Who has most evinced suspicion and distrust in polls over the last 25 years? It has been conservatives and Republicans. If these groups are acting on this distrust by refusing to participate in polls, there could be a disproportionate sample for any poll.

I can't prove this is happening. IF it is, it would help explain the trend of Republicans to do better in the actual vote totals than they did in the opinion polls. And it certainly seems that in recent years the polls have often erred on the side of Democrats, but far less often by overeestimating Republican support.

If my theory is correct, it also says that no amount of dissecting poll internals will ever be sufficient to adjust them to a norm. Even a sample which is statistically "balanced" by overweighting the Republican responses can't truly adjust. They were only able to poll the subset of "Republicans willing to respond to polls," and that population would be expected to vary from the larger group.

I can't prove my theory, but it does seem to be compatible with the facts. It would go a long way towards explaining the disparity between polls and real vote totals, and even eliminate conscious bias as a cause. But it's just a theory - you know, like "evolution" or "relativity" and such.

;-)