The View From Here . . .

By Bob Morgan, Jr.

I have written this type of column before, but I thought an update for the 2008 political season would be in order. The following is a hopefully non-partisan guide for interpreting opinion polls and putting them into perspective.

At this season we are deluged with opinion polls on all subjects. Some can be resolutely ignored. Internet surveys open to anyone with a keyboard are next to worthless since the sponsors make no effort to obtain a cross section of respondents and because the surveys are subject to manipulation. Ideologically driven internet forums frequently encourage their members to vote in these polls early and often.

Only slightly more trustworthy are polls provided by pollsters retained by a candidate. Also in this category are "private" internal polls by candidates that are "leaked" to the media. Almost always, these results are unduly favorable to the candidate paying for them. Their only practical use is as a "worst case" scenario in assessing the opposing candidate.

What about public polls? They can be a very useful tool to gauge public opinion but a number of caveats are in order.

First, polls almost always state a "margin of error." What this means is that 95% of the time Candidate A's results will be within a given range of the poll result (for example, plus or minus 3%), if the poll is conducted perfectly. Of course, this also means that one poll in 20 is a bad result, despite the best practices of the pollster. And in a two candidate race, each of the two results is subject to the margin of error.

Second, of course, polls are not conducted perfectly. Response rates to telephone surveys are frequently in the 30% range. These rates appear to be falling over time and the situation is complicated by the growing number of people who only have cell rather than landline telephones. In addition, as explained by the Rasmussen organization, which uses touchtone automated polls, different groups respond in different ways. "For example, women answer the phone more than men, older people are home more and answer more than younger people, and rural residents typically answer the phone more frequently than urban residents."

The usual strategy of pollsters is to create a representative sample by "weighting" the responses of those who do answer by such categories as geographical location, age, sex, race and, sometimes, political affiliation. This probably works if done correctly.

Pollsters disagree as to whether weighting of the sample by political party is appropriate, since results are heavily determined by declared party affiliation. Thus, a poll with disproportionate number of Democrats will favor Barack Obama and John McCain will score well in a Republican-heavy poll. The exit polls in the last three presidential elections had Democratic margins of between 0 and 4% over Republicans, with the results even in 2004. Democrats have probably signed up more members than Republicans since 2004, but polls with Democratic margins of much more than 4% (or polls with more Republicans than Democrats) should be regarded with caution.

Also, polls are of course necessarily historic in nature. Most polls are taken a few days before publication, so they necessarily cannot take into account a recent major event. A good example at this writing is the collapse of Lehman Brothers and turmoil on Wall Street.

If used correctly, polls generally provide a good window on public opinion. But before buying into a result, consumers should look at multiple surveys provided by different reputable and relatively unbiased providers. There are many traps for the unwary.