Statistical Chicanery

Mark Twain famously said, “There are three kinds of lies: lies, damned lies and statistics.” Isn’t that the truth. In fact, I ran into a great example of this the other day. It involves an article being tossed around on the left about the end of Christianity as the dominant religion in America. Wait till you see this.

The article is based on something called the American Religious Identification Survey, which gets done every so often by Pew. The key conclusion being passed around by the left is that “the percentage of self-identified Christians has fallen 10 percentage points since 1990” and “the number of Americans who claim no religious affiliation has nearly doubled since 1990.” These then feed the conclusion that America is ceasing to be a “Christian country” as we all turn atheist and it’s time to think of a post-Christian America.

Yeah, no.

Let’s start with the unaffiliated. That “doubling” is from 8% to 15%. That’s hardly earth shattering. In fact, if I told you that 85% of the public agrees on anything other than something the left wants to believe to be true, pretty much any reasonable person would conclude that this was an overwhelming majority to the point of being nearly unanimous. In fact, think of it: 85% is essentially nine out of ten people. That’s an insanely uniform majority.

Moreover, that 15% isn’t actually atheists. This includes people like some of the people I know who think of themselves as “Christians” but don’t belong to any specific Christian church. How many are atheists? Well, the number of actual atheists is only 3.6 million (about 1% of the population) after a “fourfold increase” from one million in 1990. Of course, in that time, the US population grew from 248 million to 320 million so we’re hardly looking at an epidemic of atheism. (And almost 20 million of those are immigrants, who are less likely to be Christian.)

What’s more, what have we learned about these polls during the past few election cycles? That’s right: the outcome is by and large determined by the person who chooses the poll’s sample. If that person thinks a certain number of people exist in each religious group and then adjusts the poll to fit those numbers, then their findings will surprisingly match the numbers they originally believed to be true. Funny how that happens.

One interesting tip off in this whole poll is that while the number of “unaffiliated” people supposedly grew to 15%, the number of unaffiliated voters tracks at only 12%. In other words, the more realistic measure shows 3% less of the group that is now being touted as the surprise. That suggests that the poll overestimated the percentage to begin with when it picked its sample. Oh, and if you factor out that three percent, about 9.6 million people, then suddenly that 2.6 million growth in the atheist figure looks a tad suspect.

So the growth in “unaffiliated” is poorly defined. The atheist portion of that is suspect. And the whole thing looks manipulated. Even then, almost nine in ten Americans is a registered Christian. So how does any of this result in the end of Christian America? Wishful thinking.

Watch for this kind of false reasoning and misuse of words like “doubling.” During the Colin Kaepernick crap, the media has often said things like “the number of players participating has doubled” or “tripled” or “exploded”, and yet the real number or protestors remains around 10 out of 1900. They talk about something being ranked “number one at Amazon,” without a clue that this could be less than a few hundred sales. Remember how “Mohammed” was the most popular name in Britain (until it fell off the list the next year)? That was because it was the only Arabic name in the top 20 because all Muslim kids were named that, unlike Christians who spread their names around. Think about the 47% of Americans without job... a figure which includes kids, retired people, people between jobs, and students.

Never accept a number without understanding how it was found, and never accept a characterization of a number without knowing the number itself.

Thoughts? Examples?