jumpy jumpy vitamins « Poll accuracy is all about the sample set!

Poll accuracy is all about the sample set!

The Advocate published a wee blog post about some NBC exit poll data that specifically asked GLBT voters how they voted in the Democratic primaries on Super Tuesday. That poll indicates that Clinton has notably stronger support among GLBT voters…

… but only among those in New York or California.

I can only imagine two scenarios where this methodology would be valid:

Only two states — New York and California — caucused or held primaries on Super Tuesday
GLBT voters only live in two states, New York and California; they simply do not exist elsewhere

However, neither of these scenarios are true; GLBT voters did vote on Super Tuesday in twenty-one other states. Therefore, I can’t help but conclude that the pollsters made one or more of the following hasty and demonstrably false generalizations:

GLBT voters in other states would have been deathly afraid to self-identify in an exit poll
All GLBT voters have the same mindset; there is absolutely no difference of opinion between, say, a Chelsea muscleboy and a fiercely independent lesbian in rural Oklahoma, so sampling exclusively in NY and CA is enough
Gays, lesbians, bisexuals, and transfolk do not bother voting, except in NY and CA
GLBT voters elsewhere have not changed their minds one whit since Hunter College’s poll was released last fall
The “favorite daughter” effect had precisely zero effect on GLBT voters in New York state, therefore the “favorite son” effect clearly had precisely zero effect on GLBT voters in Illinois

Whatever the real cause, it’s pretty clear that the pollsters did not do a very good job of taking a representative sample of GLBT voter opinions. As a result, the results they do have should not be considered indicative of GLBT voter opinions elsewhere, and anybody who assumes that they are is either intellectually lazy or deliberately ignorant.