In 1936 a magazine, The Literary Digest, conducted an elaborate opinion poll on that year’s presidential election. The magazine, which had conducted successful polls for the elections from 1916 on, mailed out 10 million questionnaires. It selected owners of automobiles, magazine subscribers, registered voters and people with telephones. More than 2 million people responded, and the magazine triumphantly published an article predicting that Kansas Gov. Alf M. Landon would defeat President Franklin D. Roosevelt, 57 percent to 43 percent.

Roosevelt won more than 60 percent of the vote, carrying every state but Maine and Vermont, and The Literary Digest did not long survive its embarrassment.

The Literary Digest poll, it turns out, was flawed, despite the impressive number of responses. Respondents were self-selected, and the sampling group included a large number of people (owners of cars and telephones during the Depression, people who could afford magazine subscriptions) who were Republicans.

Surveys of public opinion, thanks to George Gallup and his many successors, have grown much more sophisticated and reliable. (Gallup himself accurately forecast the outcome of the 1936 election.) Unfortunately, the public, including journalists, has not grown more sophisticated. I suggest that you have a look at two posts by my fellow blogger, Fred Vultee, “Democrats eat babies, too” and “More fibbing with ‘polls,’” to see examples of desperately flawed polling being fobbed off on the public.

As we should have figured out since 1936, surveys with self-selected responses are not trustworthy. That includes those click-on-the-button trivialities you see at Web sites and the “votes” you can cast for figures on “reality” shows on television.

We should understand that polls should state what their sampling methodology is. We should also wonder how the questions are worded, since subtle — or blatant — slanting is common.

We should also understand by now, as Brother Vultee points out, that responses falling within the “confidence interval” or “margin of error” are effectively ties. You have no business saying that Candidate A “leads” Candidate B by two percentages points in a poll that shows A with 51 percent and B with 49 percent and a margin of error of plus or minus 3.5 percent. Nobody can say who is leading in that contest. And we should expect to see what overall “confidence level” has been calculated for the conclusions.

Combine the brandishing of unreliable surveys with the horse-race, who’s-ahead-this-morning approach to political coverage months before anyone outside the campaigns themselves has any interest in the coming election, and you may have one explanation for the public’s overwhelming lack of interest in elections.

If you prefer not to participate in a herd, skip most of the polling stories, try to find more substantial material, and make up your own mind.