Wednesday, January 27, 2010

Teen Pregnancy Rate Kerfuffle

Seems the latest Guttmacher Institute report on teen pregnancy has gotten the pundit world in a bit of a tizzy. Teen pregnancies showed an increase in 2006 after a solid decade and a half decline, and the finger pointing is beginning. As this article points out, much of the battle rages over abstinence-only education, and whether federal funding for it is a good idea. Opponents say that abstinence-only education leaves teenagers unprepared for sex, which leads to pregnancies when it inevitably happens. Supporters say that the data support an increase in abstinence-only education to get teenagers to stop having sex (and thus stop getting pregnant).

What I didn't see was anyone who actually looked at the data to see if there was any correlation. So I did. I compared Guttmacher's numbers for pregnancy among 15-19-year-olds to SIECUS' numbers for abstinence-only education funding, divided by the number of pupils in each state's public school system. I threw out DC's outlier numbers. The results (standard correlation-causation caveats apply):

The data show a slight - very slight - positive correlation between abstinence-only funding per pupil and teen pregnancy rate. That would seem to support the case of abstinence-only opponents. It's hard to read much out of such a low correlation number, but let's compare a state's teen pregnancy rate to its median income level - poverty is often cited as a cause of teen pregnancy - and see what we get:

That's the negative correlation we would expect, but it's an even worse correlation.

So what can we conclude? First, we can discard the idea of poverty leading to teen pregnancy, at least as a primary factor. Abstinence-only education is more of a factor, though the low correlation number would suggest that there are either a) other things that are more of a factor or b) a whole host of things that affect teen-pregnancy rates that I didn't graph.

Here's a few more fun tidbits of data:

- The national average for abstinence-only education funding per pupil is $3.73. Only two states in the bottom 15 of teen pregnancy rates - Nebraska (9th lowest) and South Dakota (11th lowest) - spend more than this on abstinence-only education.

- The five states with the lowest rates - New Hampshire, Vermont, Maine, Minnesota, and North Dakota - all spend less than a dollar per pupil on abstinence-only.

- Seven states - Vermont, Minnesota, Idaho (!), Montana, Rhode Island, Wyoming, and Delaware - spend no money on abstinence-only education. The first four states on that list are all in the bottom 15. Delaware has the 6th highest rate - Rhode Island and Wyoming are firmly middle-of-the-pack.

- The state that spends the most on abstinence-only, South Dakota, has the 11th lowest teen pregnancy rate. The state that spends the second-most, Mississippi, has the 5th highest.

What's truly weird is that culture doesn't appear to play any role in teen pregnancy either. Generally socially conservative states appear on both sides of the teen pregnancy spectrum, as do generally socially liberal states. (Southern states do appear to concentrate near the top though - Virginia has the lowest rate among Southern states, and it's 20th lowest.)

If there's a conclusion to be drawn here, it's that abstinence-only education has been a factor - but hardly the only factor, or even the most important one - in increasing teen pregnancy rates. Comprehensive sex education can decrease teen pregnancy rates somewhat, but it isn't the cure-all for pregnancy rates that supporters often portray it as. But most importantly, teen pregnancy is a phenomenon whose many causes are not well understood - well, except for the direct cause, of course.

The other option, of course, is that the two are so lightly correlated that there can be no causative link, and that the two are independent phenomena. From looking at these data, there's a good argument to be made that the level of sex education doesn't really affect whether or not teens get pregnant. Either way, though, abstinence-only advocates like to push their programs as a cure for teen pregnancy. We can conclude that that's clearly not the case - as our data show, there's no possibility that abstinence-only education funding could decrease teen pregnancy rates.


-Dave said...

Other numbers I'd be interested in seeing:

SIECUS only has federal money for AO funding. Are there state/NGO entities providing funding? Is it significant relative to federal funding?

How have the rates changed over time - say, 2000-2005? Is there any correlation between that and funding for various programs? For example, you cite Mississippi as a state with a high pregnancy rate, but it also had a substantial drop over 2000 to 2005, from 10.3 to 8.5%.

What are the rates and the changes in rates relative to all sex ed funding, not just AO, within the same sample?

I'd be interested in seeing some numerical proxy for "socially conservative" to see if there's a correlation (positive or negative) on teen pregnancy rates.

Finally, as it's self evident that race plays a significant role in these statistics, it would be worthwhile to try the same analysis you did on data that has been adjusted to remove race from the equation, to prevent contaminating the data (as, again, it probably does in a state like Mississippi).

Jeff said...

Dave - I'm sure there are some state funds out there in the equation, which might explain states like Idaho and Utah receiving so little federal abstinence-only funds. The numbers on funding are also from the '07-'08 school year while the pregnancy numbers are from '06. It's not the best data, but it's the data I could come up with. SIECUS does provide some numbers for state funding, but they're kinda muddled and I don't know if I can back them out of the info they give.

Teen pregnancy rates as a whole dropped from 2000 to 2005, and then rose again slightly in 2006. So the decline in MS is hardly abnormal.

A numerical approach to social conservatism would be tough. I could try to correlate percentage vote for McCain with teen pregnancy, or I could try percentage of churchgoers. I'll work on those and post the results later.

I worry about correcting for race because it's clearly not a causative factor. I do realize, though, that instead of median income I should have used the percentage of the population below poverty. That's the reason I threw out DC, in fact - it has a relatively high median income but a poverty rate that's a lot higher than average.

Conner said...

Does anyone else see the correlation of the rise in teen pregnancy after 20 years, and the fact that these girls now have Sarah and Bristol Palin to look up to? Reminder: Bristol got knocked up as a teen too!!

-Dave said...

"I worry about correcting for race because it's clearly not a causative factor."

Understandable (I assume you mean that varying race does not have a biological impact on teen pregnancy), but at the same time it's highly correlated - enough so that I expect it swamps any meaningful difference in the data between, say, Utah and Mississippi.

Non-Hispanic white girls make up about 67% of the teen girls in the population (6.4 million; to about 1.6 million for both black and hispanic groups), but only 40% of the pregnancies (about 278 thousand to 205 and 209, respectively).

I don't think you can dismiss any form of causality out of hand for non-biological reasons (culture?).

Finally, one important funding question is not "how does AO funding do relative to overall teen pregnancy in the states" but "how does it do relative to other forms of sex ed?" Do you see a different result if you compare AO funding as a % of total sex ed funding?

I'm not trying to lead the questions anywhere - I'm just trying to think of other angles to analyze the data from. I don't think your final conclusion is as concrete as you put it, because I think there's a lot of confounding factors to sort through. You make a good argument that there doesn;t seem to be a strong correlation on the surface, but it's hardly conclusive.

"We can conclude that that's clearly not the case - as our data show, there's no possibility that abstinence-only education funding could decrease teen pregnancy rates."

Jeff said...

Dave, that's actually the point I'm trying to make - teen pregnancy is the result of a bunch of different factors and that one single factor doesn't dominate the discussion. Race is a big factor, but I'm still skeptical of its actual inclusion. I read somewhere that if someone's parents were teen parents, then they're more likely to be teen parents. So there's a certain snake-swallowing-its-tail quality to consider on top of all this.

I would have preferred to use the percentage metric you suggested - however, I didn't know how to find that data. That's especially true in a state like Idaho, which doesn't really fund any sex education in its schools (one could presume that Utah has the same policy). I also would have liked to break the data down further by school district - doing so would give much more interesting data, since it would flatten out a lot of the overlying factors.

My conjecture, especially in light of the low numbers in a state like Idaho with little to no sex education to speak of is that the most important factor is parental involvement, and that school involvement isn't much of a factor. If parents teach their kids to be responsible about sex and instill in their kids a sense of confidence in their own decisions about sex (whether that decision is abstinence or otherwise), they'll be less likely to become pregnant. But there's literally no way to quantify that.

-Dave said...

Parental involvement really is both impossible to quantify and a huge piece of the puzzle.

It just goes against my professional sensibilities not be be able to quantify things.

I'm not sure I completely agree with your final premise (I'm only marginally convinced of the ability of a 15 year-old to weigh all the consequences and make informed decisions in the heat of the moment), but broadly I think we're in agreement.

-Dave said...

On this topic, I saw this in the news today and thought it relevant enough to be worth comment: