A new Washington Post-ABC News poll last week showed support for gay marriage hitting an all-time high at 58 percent, with support having risen by 21 points over the last decade alone.

But some social conservatives say polls like this one are wrong, and that they over-sell the actual amount of support for gay marriage.

"I'm not worried about it, because the polls are skewed, " Gary Bauer, a leading social conservative and former presidential candidate, said on "Fox News Sunday." "Just this past November, four states, very liberal states, voted on this issue and my side lost all four of those votes. But my side had 45, 46 percent of the vote in all four of those liberal states."

Ralph Reed, the former head of the Christian Coalition, made a similar argument on NBC's "Meet the Press."

"It's clearly moved, but the idea that the American people are, you know, universally for same-sex marriage is just not backed (up)," Reed said. He added: "They have, after all, voted in 31 state referenda and initiatives for traditional marriage. Only three have they voted the other way. So this thing tests very differently at the ballot box than it does in a poll."

So are they right? Are the polls often wrong?

In a word, yes -- at least to some degree. Polls on gay marriage ballot initiatives generally under-estimate the opposition to gay marriage by about seven percentage points, according to a 2010 study by New York University political science professor Patrick J. Egan.

"The share of voters in pre‐election surveys saying they will vote to ban same‐sex marriage is typically seven percentage points lower than the actual vote on election day," Egan wrote. He also noted, however, that polls are generally accurate when it comes to the percentage of people supporting gay marriage.

The chart below, from Egan's study, illustrates this point. Polls measuring opposition to gay marriage (denoted by the white dots) very often come in below the actual opposition to gay marriage on Election Day (denoted by the gray line).

The best example of the discrepancy between the polls and the actual vote may be the same gay marriage ban that comes before the Supreme Court this week -- California's Proposition 8.

While polling regularly showed more opposition to the ban (i.e. support for gay marriage) than support for it, the measure passed on Election Day 2008 by four points, 52 percent to 48 percent.

The trend has continued, even as four states in November became the first in the country to have their citizens vote for gay marriage.

In Washington state, polls consistently showed more support for gay marriage than the final seven-point margin.

In Maryland, late polling showed support for gay marriage as high as nine or 10 points (including nine in a Washington Post poll) -- and even higher in some early polls -- but it passed by four points.

In Maine, late polls showed gay marriage ahead by double digits, but it passed by six points.

(The one exception is Minnesota, where the polling was pretty close but no late polls were conducted. Voters overturned a constitutional ban on gay marriage by six points -- though it should be noted that they didn't effectively legalize gay marriage.)

There are a couple theories for this discrepancy, setting aside the notion that the pollsters themselves might be biased:

1) Polls under-sell opposition to gay marriage because people don't want to come off as intolerant of gay people, so they hide their true anti-gay marriage beliefs from pollsters. A similar theory has been suggested for polls that over-sell support for black candidates -- something dubbed the "Bradley Effect."

A Washington state poll from last year, for example, tried to adjust for this so-called "social desirability" effect. In addition to asking the gay marriage question, it moved undecided conservatives and religious voters into the "no" column, along with those who were uncomfortable with discussion about "sexual minorities."

While the first question showed gay marriage favored by a 58-37 margin, the latter adjustments showed the margin at 52-48 -- exactly the margin by which gay marriage passed.

Egan, though, argues that social desirability doesn't have much effect. He noted that the amount of poll error remained similar even as support for gay marriage rose over time and was also consistent between states regardless of how many gay people they had -- the idea being that more tolerant places and times would lead to more error if the "Bradley Effect" applied to gay marriage.

In addition, until recently, gay marriage didn't have the support of a majority of Americans. So the popular viewpoint was, in fact, against gay marriage, which suggests that would be the more socially desirable position.

2) Opponents of gay marriage feel more strongly than supporters and thus are more apt to vote.

The evidence is also mixed on this one. The latest Washington Post-ABC News poll showed 41 percent of Americans strongly support gay marriage, while 30 percent strongly oppose it. In 2010, while gay marriage polled pretty evenly overall, 31 percent strongly favored it, while 42 percent strongly opposed it.

In both cases, though, the percentage of supporters who said their view was "strong" was lower than those who opposed gay marriage. While 71 percent of supporters felt strongly in the most recent poll, 83 percent of opponents felt strongly. The split was even wider in 2010, which suggests it could be something of a motivation issue.

In the end, both theories are plausible, if not necessarily proven.

Reed argued Sunday that several other recent polls have shown a much smaller pro-gay marriage margin than the Post-ABC poll. That's true, as the chart below shows.

But it's also clear which way -- and how quickly -- the polls are headed. And if the current trajectory continues, it's going to simply be a matter of how big the majority of Americans who support gay marriage is, and how quickly it takes effect across the country.