teen birth rates highest in most religious states

That's the headline for an MSNBC article that summarizes a soon-to-be-published study in the journal Reproductive Health:

U.S. states whose residents have more conservative religious beliefs on average tend to have higher rates of teenagers giving birth, a new study suggests.

The relationship could be due to the fact that communities with such religious beliefs (a literal interpretation of the Bible, for instance) may frown upon contraception, researchers say. If that same culture isn't successfully discouraging teen sex, the pregnancy and birth rates rise.

From what I've seen so far, I'd be very hesitant to draw such a conclusion. The study does find at least one positive correlation:

Mississippi topped the list for conservative religious beliefs and teen birth rates.

Study researcher Joseph Strayhorn of Drexel University offers this interpretation:

We conjecture that religious communities in the U.S. are more successful in discouraging the use of contraception among their teenagers than they are in discouraging sexual intercourse itself.

But what does the correlation tell us, really? Does it truly indicate that religious teens are more likely to get pregnant? Or does it tell us that the same state happens to have high concentrations of both religious teens and pregnant teens?

I'm not the only one who is asking:

And while the study reveals information about states as a whole, it doesn't shed light on whether an individual teen who is more religious will also be more likely to have a child.

"You can't talk about individuals, because you don't know what's producing the [teen birth] rate," said Amy Adamczyk, a sociologist at the City University of New York, who was not involved in the current study. "Are there just a couple of really precocious religious teenagers who are running around and getting pregnant and having all of these babies, but that's not the norm?"

Adamczyk, it turns out, has also done resesarch in this area, and has found just the opposite:

"What we find is that more religious women are less likely to engage in riskier sex behaviors, and as a result they are less likely to have a premarital pregnancy," Adamczyk said during a telephone interview. But for those religious teens who do choose to have premarital sex, they might be more likely to ditch their religious views and have an abortion, she has found.

Strayhorn acknowledges that his study looks at communities, not individuals, but he still believes he has uncovered something:

"It is possible that an anti-contraception attitude could be caused by religious cultures and that could exert its effect mainly on the non-religious individuals in the culture," Strayhorn told LiveScience. But, he added, "We don't know."

John Santelli of Columbia University is not critical of the study or of Strayhorn's interpretation of the results, but notes that:

"The index of religiosity is tapping into more fundamentalist religious belief," Santelli said. "I'm sure there are parts of New England that have very low teen birth rates, which have pretty high religious participation, but they're probably less conservative, less fundamentalist type of congregations."

Indeed, the researchers defined religiosity very narrowly:

For religiosity, the researchers averaged the percentage of respondents who agreed with conservative responses to eight statements, including: "There is only one way to interpret the teachings of my religion," and "Scripture should be taken literally, word for word."

Finally, while the correlation appears to hold for some states, in others it's not so strong. Five states are in the top ten in both "religiosity" and teen pregnancies, but some states show a wide gap between the two categories. New Mexico is second only to Mississippi in teen pregnancies, but ranks 22nd in "religiosity". Utah is sixth in "religiosity", but only 34th in teen pregnancies.

I'd be hesitant about drawing any strong conclusions from this study, but a possible correlation between religious beliefs and teen pregnancy rates is worth a closer look.

number crunching the election results from iran

Political scientists Bernd Beber and Alexandra Scacco have found some peculiarities in the Iran election results. While many people around the world are questioning the validity of the results, Beber and Scacco have gone a step further and looked at some statistical anomalies.

Beber and Scacco looked at the final digit of the results from each province for the top four candidates, and found two anomalies. Since there are ten possible digits, we would expect each number 0 - 9 to appear in about 10 percent of all the precinct totals, give or take a few percentage points due to random variability.

We wouldn't expect the number 7 to be the last digit for 17% of all precinct totals, or the number 5 to be the last digit in only 4% of all precincts. But that's exactly what Beber and Scacco found. The chances of seeing two such unusual distributions in a fair election are about 3.7%. On the other hand, when people attempt to create random-looking numbers, they often pick numbers ending with 7.

Another pattern found in human-generated numbers is what are called adjacent digits. People tend to follow a digit with the next higher or next lower digit, e.g. 23 or 54. In the Iran election totals Beber and Scacco looked at the final two digits from each precint, and found that fully 38% of all the pairs of digits were adjacent digits. The chances of seeing this in a fair election are about 4.2%.

So we have two unlikely scenarios, independently appearing in the same set of election results. Beber and Scacco calculate the chances of seeing both of these anomalies at half of a percent. In other words, we would expect to see results like this in only one of every two hundred fair elections.

When we combine these anomalies with the discrepancy between vote results and pre-election polling, and the fact that in many cities the vote totals were greater than the number of eligible voters, it becomes very hard to believe this election is in any way legitimate.