68% of statistics are made up on the spot

Since we're talking about Gallup polls today, here's an excerpt from an Ann Althouse post about rates of gaiety in America:

Gallup poll: Democrats estimate that 28% of Americans are gay.
Republicans are much closer to reality, at 20.2%, but still far off. (The actual percentage is more like 3.5.)

What accounts for this discrepancy? Perhaps Democrats are more likely to live in places where there are larger concentrations of openly gay people. Perhaps those with liberal views are more likely to hear the news that people they know are gay. Maybe Dems are dumber.

The numbers are definitely much, much higher than the usual estimates. But I have to cut people some slack on questions like this. Getting a useful count of the number of gay people in anywhere isn't just difficult; it's frankly impossible because of the imprecision of the terms. We can easily point to some gay people and some straight people, but in a complex social species that uses sex for reproduction, cultural expression, and reinforcing social bonds, there can be no clear dividing line between the two. What fraction of a person's partners have to be of his sex to call him gay? How many unfulfilled sexual thoughts about men can a man have and still stay in the straight column? If I'm attracted to transmen or transwomen, can I squeak by on points? If we use a one-drop rule, then I'd guess 28% is probably a low estimate.

Since the survey in question didn't define its terms, I wouldn't be too quick to question the intelligence of the respondents.