• KCTS9.org/Washington Poll

Three weeks out from the election, the top-line numbers in the new KCTS9/Washington Poll look cautiously encouraging for R-74, the referendum on the state's new marriage equality law. The poll finds R-74 leading 56.3 to 35.6 percent among registered voters (782 registered voters: +/- 3.5%), and 54.1 to 38.4 percent among likely voters (644 likely voters: +/- 3.9%). The survey was conducted October 1-16.

But what about the so-called "Bradley Effect" in which survey respondents are known to mislead pollsters on certain issues and candidates due to a "social desirability bias" (i.e. they don't want to appear prejudiced against, say, gay or black people)? For example, in 2009 the Washington Poll predicted that R-71 (domestic partnerships) would pass by a 17-point margin. It actually passed by only 6.3 points.

Pollster Matt Barreto attempted to adjust for this by asking two further questions: Did you give any answers that weren't 100 percent honest, and were there any questions that made you uncomfortable? Respondents who answered that they may have lied, and that they were uncomfortable with the marriage equality question, but who reported that they were leaning Yes on R-74, were undecided, or would not vote, were moved into the No column, as were lean-yes voters who answered a series of questions defining them as very religious conservatives. The result: An estimated 52.9 to 46.6 margin in favor of R-74 among likely voters, a 6.3 point difference.

Curiously, that's the same 6.3 point margin by which R-71 passed in 2009. Again, that's somewhat encouraging, with the one caveat that the 2009 survey was conducted during the second half of October, whereas this survey was conducted during the first. A lot can happen in two weeks, and there are still a helluva lot of Reject R-74 ads left to air.

But reached by phone, Barreto tells me that if he had to bet, he thinks R-74 will likely pass. Barreto describes the social desirability bias adjustment as "very conservative;" R-74's true support might lie closer to 54 percent. And barring a large last-minute advertising blitz, he says, there isn't much time left to move public opinion.

But even then, the survey suggests, the ads might not be very effective. For example, respondents who said they supported R-74 were asked the following question based on negative messaging drawn straight from the voters guide:

Some people say that Washington same-sex couples already enjoy full legal equality under our laws, and Referendum 74 is dangerous because it redefines traditional marriage. Traditional marriage promotes child well-being because kids need both a mother and a father. Does this make you more likely to support R-74 for same-sex marriage, less likely to support R-74, or have no effect on how you would vote?

You might recognize that messaging from the No campaign's TV ads. But only 5.8 percent of respondents said that this made them less likely to support R-74, compared to 18 percent who said it actually made them more likely to support it. "It doesn't move a lot of people," says Barreto, "and in some cases it causes people to dig their heels in."

Barreto will release a second wave of polling just days before the election, and if R-74's lead holds up, prepare to celebrate.