The Alliance Defense Fund Lies to its Base
August 4th, 2011
This week I came across an article so very lame there seemed no point in in debunking it. Then I saw it was from the Alliance Defense Fund.
These folks are co-counsel for ProtectMarriage.com, the group defending Prop 8 in Federal court. I’ve already written about their ridiculous notion that Christian state employees in New York state don’t have to abide by the law. Basically, they’re the country’s chief anti-gay legal group, and while it pains me to take them seriously, they are a genuine threat.
The new article is called, Games the left plays with polls about same-sex ‘marriage.’ It’s egregious because the author, Brian Raum, claims to tell the truth about a gay-positive poll when in fact he merely lies about it.
Brian is complaining about a survey from Harris Interactive (HI) that shows strong support for marriage equality. He thinks HI stacked the deck:
Harris Interactive purposely oversampled those who engage in homosexual behavior, thus guaranteeing the results would not represent the overall American sentiment, but rather would be skewed to reflect the views of those seeking to further the homosexual agenda. (To the credit of Harris Interactive, they admitted the oversampling in the fine print at the bottom of their survey results, albeit in a place few will see, and even fewer will care to search for.)
How significant was the oversampling? Consider this: those who identify as homosexual only constitute 1.4 to 1.7 percent of the U.S. population, according to the latest figures from the Centers for Disease Control and Prevention. In the Harris Interactive poll, they constituted a sample of well over 14 percent. With this distortion understood, it’s no wonder the poll showed that “‘49% of all U.S. adults…support the right for same-sex couples to marry,’ [vs.] 41% who oppose the right, and 10% who are not at all sure.”
In other words, Oh my gosh, no wonder the survey’s so gay-friendly — it has 10 times as many homosexuals as it should!
So much wrong here.
Sampling and weights
Brian doesn’t understand the difference between sampling and analysis. A sample might have too many or too few gays, straights, Protestants, Catholics or whatever. Pollsters compensate by weighting their data to get the right proportions when they do their analysis.
In fact, by applying some basic algebra to HI’s results, you’ll find they weighted LGBTs as about 7.7% of the adult US population. A bit high? Perhaps. But nowhere near the 14% Brian wants us to believe.
Brian ought to understand this difference between sampling and analysis — certainly if he’s going to earn money writing about this stuff. That brings up the usual question: incompetence or rank dishonesty? Hard to know.
What’s the right weight?
Brian wants us to think 1.4 -1.7% would be an appropriate LGBT weight, based on CDC figures. But he’s, er, mistaken. Those numbers just cover the Ls and the Gs (1.3% of all women for lesbians and 2.3% for gay men). What about the Bs? Bisexuals add another 2.8% for women and 1.8% for men. Now we’re looking at 4.1% for LGB.
But there’s more.
The CDC also lists “Something else” and “Did not report.” No way of knowing exactly what that means, but I can tell you this: 9.7% of women declined to say they were straight, along with 9.8% of all men.
In other words, according to Raum’s own source, HI’s LGBT weight should be at least 4.1%, and possibly a good bit higher. Once again, incompetence or rank dishonesty? Hard to know.
Oh, and one more delicious bit: Antigays love to say there are no homosexuals, just homosexual behavior. You see that in Brian’s wording: “Harris Interactive purposely oversampled those who engage in homosexual behavior…” But the CDC measures that, too. 3.2% of those self-identified straight men have engaged in homosexual behavior, along with 9.0% of straight women. Using Brian’s criteria actually bumps up our numbers even further.
I truly hate this no-homosexuals-just-homosexual-behavior meme, so I love watching it turn around and bite Brian in the ass.
What if we only weighted LGBTs at 4.1%?
Brian’s implying Harris Interactive counted ten times as many LGBTs as it should have. What a conspiracy! The truth is not so ominous. Let’s go to the lowest possible extreme and assume HI should have used a 4.1% weight. How much difference does that make?
A bit more algebra says instead of 49 – 41 result favoring marriage equality (plus 10% undecided), we’ll get a 48 – 43 victory, (total percentage not equal to 100 due to rounding). That difference is basically insignificant in the world of statistics.
Poor Brian. All that work debunking the poll, and it amounts to nothing. Incompetence or rank dishonesty?
Does Brian care?
Brian is just wrong, wrong, wrong in this article. The irony is that he’s trying to expose “games the left plays with polls about same-sex ‘marriage.” Combine all his falsehoods with his ballsy assertion of setting the record straight, and you have to wonder if the truth really matters to him. Is it paranoid to think he’s happy lying to his own base as long as it fires them up? Could a strategy like that even work?
A hint appears in the comments section of his article at TownHall.com. One fellow, Lon, pointed out Brian’s confusion over sampling and analysis. Here are the two responses Lon got back:
And you are not up-to-speed on research rules.
Lon the fraking loon of TH grunts again………………………
Another commenter, Jeremy, explained, “Over-sampling gays doesn’t guarantee results skewed in their favor. He probably thinks it means over-representing.” Simple, direct, civil, and true. Jeremy got one response:
We needed input from an intellectually-challenged gay.
Now we have it.
They don’t care. Lying to your base, it seems, works just fine for anti-gay activists. As long as anti-gay is your only moral value.