Monday, January 12, 2009

The Changing Communications Infrastructure and the Accuracy of Polls

Source: "Cellphones' Growth Does a Number on Health Research," David Brown, The Washington Post, January 12, 2009.

The failure of polling in the Dewey-Truman presidential election is famous. All the polls predicted Dewey, and Truman won. The polling was done by telephone. More Republicans had phones than did Democrats.

In the 1980's and 1990's almost every one had a phone. Only about three percent of the population did not. I would bet that these folk were neither likely to vote nor big consumer spenders. Pollsters were OK leaving them out.

Now, to that three percent, we must add 16 percent of homes that have cell phones but no telephone land line. There folk are not only a significant portion of the population, but they are demographically (and behaviorally) different from the land-line wired majority.

So pollsters want to survey them. However, the law does not allow automatic dialing to cell phones as it does to land line phones, so it costs more to poll the cell-only users. Moreover, they are almost twice as likely to refuse to respond to an interviewers questions.

I don't worry too much about the political pollsters not getting it right, nor about the advertisers not getting their messages fine tuned to the max. However, the WP points out that epidemiological surveys are having trouble, and I do want public health officials to have the best information possible. Indeed, that is true for a lot of public services.

I guess we need a widely shared ethic as to which poll we will take and which we can reject in good conscience!

No comments: