The anti-gay "Bradley effect."

Interesting blurb over at dkos today.

There is a sort of "Bradley Effect" for gay-rights ballot questions.

The Haas Jr. Foundation hired NYU political scientist Patrick J. Egan to study 167 polls in the 33 states that had a gay marriage ban from 1998-2009 – and compared it with the results on Election Day. And as we already knew – having painfully experienced this in state after state – the results after votes are counted are worse than what polls had said. Moreover, as Dr. Egan reported, poll results throughout the campaign were mostly static.

Did people lie? Yes and no. Egan’s analysis showed that polls accurately predicted the pro-equality vote – i.e., people who voted “no” on Prop 8 – but that they undercounted people who voted to ban gay marriage. So if a pre-election poll would show us winning a plurality of 48-45 (which campaigns find encouraging), it would mean that we lost 52-48.

Unlike a traditional Bradley Effect, in which voters say they'll vote for the black candidate, then vote against it, in this case, a chunk of haters simply were too embarrassed to say they were haters, and claimed they were undecided.

This means that any gay rights group trying to fight off a ballot question -- or fight for one to pass, as it may come in the future -- should just assume "undecided" means no. Don't go into election day without striving to win a majority of support in the polling data ahead of time, not just the plurality.