There was a brief explosion of optimism from those supporting same-sex marriage in Maryland last week after a poll by Hart Research Associates indicated that voters in the state support it by a significant margin of 54-40. No state has ever approved gay marriage at the ballot box, but advocates here and elsewhere — The New York Times published a piece titled "Hopeful news from Maryland" — contend that the issue hasn't polled this well before either.
They shouldn't get too excited just yet. Gay marriage is an issue in which polls don't necessarily reflect what voters will actually do at the ballot box because it is increasingly politically incorrect to oppose such nuptials. And there is particular reason to doubt the accuracy of this poll.
For the record, I do not oppose same-sex marriage, but I do oppose the use of unsound polling data for political purposes.
The results of earlier polls on the referendum to overturn Maryland's Civil Marriage Protection Act have generally been much closer. Advocates are spinning this one as evidence of a shift in public attitudes on the question, but there are reasons to doubt whether that is the case.
When previous states have held votes on gay marriage, poll results have been less than reliable. In 2008, five of the eight public polls conducted in the two months before California voters decided on Proposition 8, which outlawed gay marriage, suggested that the measure would be defeated, perhaps by a wide margin. Instead, voters outlawed gay marriage, 53-47. When Maine voted on a 2009 referendum to overturn its gay marriage law, most polls showed that voters favored same-sex marriage, in one case by double digits. But it was defeated, also 53-47. In all, 32 states have rejected gay marriage, though most of them have been conservative states and have pre-dated the recent rise in public poll support.
In the case of the Hart Research survey, the pollster was clearly not disinterested in the outcome. The prose accompanying the poll results stated that members of the Hart group "are encouraged" by the results. The Baltimore Sun reports that the poll was "released … by a gay rights group." Selective publishing of polls should raise red flags about the agenda of the pollster.
The timing of a poll is often relevant. President Barack Obama's change-in-position announcement just under three months ago that he now unambiguously supports gay marriage gives a shot in the arm, particularly in public opinion polls, to that position, but will it have the same motivating effect in November? Chick-fil-A President Dan Cathy's opposition to gay marriage, publicized only recently, may increase the intensity of those who would overturn the law — too early to tell the longer-term effect of that as well. Whether the publication of poll results themselves affect turnout is likewise unknown.
The most counter-intuitive finding of the poll was the statistical tie among African-Americans, a demographic historically opposed — within polls and out — to gay marriage. All of which detracts from the certitude of the current results. It is possible that Mr. Obama's statement in favor of gay marriage and that of the NAACP have been persuasive in the African-American community. It is also possible that in that context, African-American respondents to the poll were more likely to say they support same-sex marriage even if they don't.
The Hart poll cues its respondents by reminding them that "the state legislature recently approved a law allowing gay and lesbian couples to legally marry in Maryland" before asking whether they approve or not. "Cuing" is a method to put a subtle finger on the polling scale, but more sophisticated, unbiased pollsters don't do so.
Hart Research Associates provides no information regarding how it determined that its respondents were likely voters. So why label them as such? Those polled were "called," presumably by phone, another possible source of polling error — what about those not reached? What guarantees of representativeness were established?
One positive component of this poll is that it makes an attempt to measure intensity of voters, and if the poll were otherwise conducted well, it might add some significance.
Even invalid polls can become a component of political campaigns; the positive results might energize supporters of same-sex marriage to vote, but on the other hand they could also energize opponents to vote.
Regardless, "scientists" in the very soft science of polling ought to maintain standards of disinterest and fairness. This poll is bald political persuasion and little more.
I have no problem with the possibility that Maryland will be the first state to approve gay marriage by referendum, but the evidence that it will do so, especially in the latest Hart Research poll, is simply not definitive.