Via [Feministe][feministe], we see a wingnut named Tim Worstall [trying to argue something about sexual education][worstall]. It’s not entirely clear just what the heck he thinks his argument is; he wants to argue that sexual education “doesn’t work”; his argument about this is based on abortion rates. This
is an absolutely *classic* example of how statistics are misused in political arguments. So let’s take a look, and see what’s wrong.
[feministe]: http://www.feministe.us/blog/archives/2006/07/10/lies-damn-lies-and-statistics/
[worstall]: http://timworstall.typepad.com/timworstall/2006/07/sex_education_w.html#comment-19323490
He quotes an article from the Telegraph, a UK newspaper. The telegraph article cites statistics from the UK department of health. Here’s what Worstall has to say:
>Yup, gotta hand it to them, the campaigners are right. Sex education obviously works
>
>Abortions have reached record levels, and nearly a third of women who have an abortion have had one
>or more before.
>
>Department of Health statistics reveal that abortions in England and Wales rose by more than 700 in
>2005, from 185,713 in 2004 to 186,416.
>…
>Some 31 per cent of women had one or more previous abortions, a figure that rises to 43 per cent
>among black British women.
>
>The ever increasing amount of sex education, the ever easier provision of contraception is clearly >driving down the number of unwanted pregnancies.
Clearly, Worstall and the author of the telegraph piece want us to believe that there’s a significant *increase* in the number of abortions in the UK; and that this indicates some problem with the idea of sex-ed.
So what’s wrong with this picture?
First, let’s just look at those numbers, shall we? We’re talking about a year over year increase of *700* abortions from a base of *185,000*. How significant is that? Well, do the math: 0.37%. Yes, about one third of one percent. Statistically significant? Probably not. (Without knowing exactly how those numbers are gathered, including whether or not there’s a significant possibility of abortions being underreported, there’s no way to be absolutely sure, but 1/3 of 1% from a population of 185,000 or so is not likely to be significant.)
But it gets worse. Take a good look at those statistics: what do they measure? They’re a raw number of abortions. But what does that number actually mean? Statistics like that taken out of context are very uninformative. Let’s put them in context. From the [statistics for England and Wales][stats]:
[stats]: http://www.johnstonsarchive.net/policy/abortion/ab-ukenglandwales.html
In the year 2003, there were 621,469 live births, and 190,660 abortions. In 2004, there were 639,721 live births, and 194,179 abortions. Now, these stats from from the UK Office of National Statistics. Note that the numbers *do not match* the numbers cited earlier. In fact, taken as bare statistics, these numbers show a *much larger* increase in abortions: about 1.8%.
But, put in context… Take the number of abortions as a percentage of non-miscarried pregnancies (which we need to do because the miscarriage statistics for the years 2003 and 2004 are not available), and we find that
the number of abortions per 1000 pregnancies actually *declined* from 292/1000 in 2003 to 290/1000 in 2004. And that number from 2003 was a decline from 2002, which was a decline from 2001. So for the last four years for which statistics are available, the actual percentage of pregnancies ending in abortions has been nearly constant; but closely studying the numbers shows that the number has been *declining* for those four years.
In fact, if we look at abortion statistics overall, what we find is that from the legalization of abortion in the UK, there was a consistent increase until about 1973 (when the number of abortions reached 167,000), and since then, the number has ranged upwards and downwards with no consistent pattern.
So – what we’ve got here is a nut making an argument that’s trying to use statistics to justify his political stance. However, the *real* statistics, in context, don’t say what he wants them to say. So – as usual for a lying slimebag – he just selectively misquotes them to make it *look like* they say what he wants them to.
Good Math/Bad Math
The beauty of math; the humor of stupidity.
You need a little more background before you call me “a lying slimebag” if you please.
We are constantly and consistently told in the UK that the only way to reduce the number of abortions is to increase the provision of sex education, more of it, at younger ages.
Now I would like something to reduce the number of abortions. I’d like the number actually to be zero (I am indeed a wingnut on that point).
We have also had over the decades a large and consistent increase in the amount of such sex education given. As you note:
“In fact, if we look at abortion statistics overall, what we find is that from the legalization of abortion in the UK, there was a consistent increase until about 1973 (when the number of abortions reached 167,000), and since then, the number has ranged upwards and downwards with no consistent pattern.”
That isn’t a great piece of evidence in favour of the idea that sex education cuts the abortion rate, nor their total number.
As above, I am a wingnut on the subject of abortion. On sex education I am completely indifferent: I really don’t care if five year olds are taught to give blow jobs or if 25 years olds don’t know where babies come from. My point in that blog piece was not to state that sex education is a bad thing per se (nor a good thing), just that there doesn’t seem to be much evidence that it does what it is said to do: cut the abortion rate.
Tim:
I don’t care *why* you lied; what I care about is the fact that you lied, and you used bad math to do it.
The fact that you deliberately misused statistics in order to create a false impression of what the data shows – that is, you used statistics to lie. You wanted to argue that abortion rates are increasing; so you misused statistics in a deliberate way in order to make it appear that abortion rates are increasing, when in fact, they are not.
I don’t care whether you’re a conservative, a liberal, an anti-abortion person, a pro-choice person, an authoritarian, or a libertarian. I’ll happily go after any of the above when I catch them using math to lie. (The post on my blog which generated the most hits until DKos linked me last week was my criticism of RFK’s piece of crap on the 2004 election in Ohio: debunking bad math that supported a conclusion that I agree with.)
“you lied”…really?
” You wanted to argue that abortion rates are increasing;”…I did? You are able to divine my motives from thousands of miles away?
Here are the bits that are me (not quotations from the paper):
“Yup, gotta hand it to them, the campaigners are right. Sex education obviously works”
“The ever increasing amount of sex education, the ever easier provision of contraception is clearly >driving down the number of unwanted pregnancies.”
So my rather sarcastic observation that increased sex education and contraceptive availability “is clearly” driving down the number of unwanted pregnancies is in fact evidence that I am arguing that “abortion rates are increasing”?
As you yourself point out “the number has ranged upwards and downwards with no consistent pattern.”
Now I have had problems in the past with statistics as both Daniel Davies (of Crooked Timber) and your fellow science blogger, Tim Lambert over at Deltoid will be able to confirm (and my reaction to such error being pointed out) but I think that this little contretemps is rather more about your projecting meaning onto my words than my being “a lying slimebag”.
Nice way to obfuscate, Tim. You misused statistics. That’s all that Mark cares about in this blog entry. That is the subject.
I wonder if scaling the number of abortions by the number of pregnancies is really the right thing to do. I think it might be better to compare the number of abortions to the number of women (preferably breaking the numbers down by age of mother and some social/ethnic markers).
It won’t change the basic facts much (rate of abortion pretty much constant) but is probably the way to go for anyone that wants to do a more detailed study.
Tim:
Doesn’t matter how you spin it: your “argument” uses statistics to try to create a misleading result. The bare number of abortions statistic does *not* prove anything about whether or not sexual education has any effect on unwanted pregnancies.
You *could* try to make an argument based on the number of abortions performed for pure birth-control reasons against the number of total pregnancies (by birth control reasons, I mean abortions that aren’t being performed because of health issues for the mother, or genetic/development problems with the fetus); it wouldn’t be a good argument, but at least it would attempt to measure something related to what you are purportedly discussing. But quoting a bare number out of context – that’s always misleading. Taking a statistic completely out of context, so that it’s stripped of any basis for determining its meaning: that’s lying with statistics.
Sam:
For normalizing abortion numbers to produce a meaningful measure, I think that the abortions/pregnancy measure is more likely to be meaningful. The things that you can measure with abortion statistics are, I think, more correlated to the number of pregnancies. It really depends on what you’re really trying to measure with the statistic.
Abortions/pregnancy can arguably give you some measure of the number of unwanted pregnancies. I don’t think that that’s a *good* measure; but it’s one you could argue had some relevance. Abortions/woman is harder to justify.
To really measure the kind of thing that Tim wants to argue about – the number of unwanted pregnancies – you’d need a whole lot more information; you’d want to know how many sexually active women there are; what kinds of birth control they’re using (if any); and what proportion of unplanned pregnancies are aborted. Without that information, you can’t actually measure what Tim wanted to measure in a really meaningful way.
Mark:
I’m not so sure. Certainly if the number of pregnancies per woman-year is constant, it doesn’t matter whether you normalise to pregnancies or people, as you’ll get the same answer up to an overall scale factor.
So then you worry about what happens when the pregnancy rate isn’t constant. There are, I think, two classes of reasons for the pregnancy rate changing. The first is a change in the rate of fertility (chemicals in water, tight underwear, diet etc.) which changes the probability of an unprotected sexual encounter resulting in pregnancy. This scales wanted and unwanted pregnancies in the same way.
The second class is a change in the desirability of having children. These are mostly social factors – there were a significant excess of babies born around Christmas 1999 due to people trying to conceive a “Millennium baby”, people tend to chose to have babies when they feel relatively secure financially, which pulls in the economy as a factor and so on. These factors scale the intentional conceptions, but don’t affect the accidents.
If the first class dominates, it’s right to scale by pregnancies. If the second dominates, it’s right to scale by woman-years.
There’s a third class, which contains the popularity of casual sex, the availability and efficacy of contraception and the like. This scales accidental conceptions but not planned ones. Abortions are almost all accidental conceptions (the few that aren’t are planned conceptions that are aborted because the foetus shows signs of a handicap), whereas live births are dominated by intentional conceptions.
It’s this third scale factor that I think Tim is trying to argue about.
By looking at the number of abortions per pregnancy, you have removed the first scale factor. You have left a pair of competing factors – increasing the desirability of children will lead to additional conceptions by people who want to keep the baby, and so decrease the abortion rate. The same effect can be produced by decreasing the number of unwanted conceptions (eg. successful sex education leading to girls being able to say no to sex, knowing that you can still get pregnant if you do it standng up, being able to access reliable contraception etc., or by a decrease in the popularity of having unplanned drunken sex with strangers).
I suspect that it’s better to normalise by number of women, so that you can look at live births/woman-year and abortions per woman-year without having the effects of scale factors 2 and 3 mixed as above.
You are right, of course, that any such study would contain a fair amount of hand-waving based on limited information, and that it’s always better to have more information than less.
Sam:
I see what you mean, and I think there’s an argument for either stat: abortions/pregnancy; or abortion/woman, depending on exactly what you’re trying to measure/understand. But I think that what it really ends up coming down to in this case is, for the purposes that Tim was arguing, they’re *both* really bad measures.
Perhaps this will make the Mark’s point clear: one could argue that the number of deaths in the United States has gone up every year.
True, but meaningless, as the population has gone up every year.
Another example of this type of thing: in the United States, people are debating the level of funding for Veterans Services. One group argues that funding for Veterans services has gone up more per year under this administration than it did under the previous one.
But that doesn’t matter as there is now a war going on; hence the current demand is much higher.
Moral: bare numbers often are meaningless.
Mark, I’m going to have to agree with Tim on one point, here: He wasn’t really the one making the bad argument; the Telegraph was. The two middle paragraphs aren’t his; they’re the sensationalist and quite bad-mathy first two graphs of the article he quotes. I think this post would be better directed at the paper and not the random blogger-dude quoting it.
I’m not sure I’d say that he’s lying. It’s much more likely that he doesn’t understand statistics very well, and his underlying bias helped him misinterpret them in a way that was favorable to what he wanted to portray. People tend to filter information so that it supports their existing beliefs.
I think I go along with “ThePolynomial” (after all, how can one disagree with “ThePolynomial?” ) Basically what Worstall is saying is that sex education has been sold as an abortion/unwanted pregnancy prophylactic. So. there’s been sex education all over the place, Sex education to the right of you, sex education to the left of you, massive distributions of color coordinated condums down the center, and there has been no significant effect on abortion and unwanted pregnancy rates. One thing that Worstall did _not_ try to convince anybody of is that there’s been a significant increase in abortion rates or numbers. Worstall cited an increase of 700 which everyone recognizes as trivial.
Broadly speaking you are correct but Tim did not deserve such a vitriolic personal attack for this and as has been stated above, the true sensationalist error is in the Telegraph article. Adjunct to this even arguing that the abortion rate is statistically flat still supports his argument, thus weakening your rebuttal.
You can correct people’s reasoning without being so rude you know! Should you care too.
While Worstall is not guilty of claiming that the abortion rate has increased significantly in England and Wales, he is saying that the figures argue against sex ed decreasing the rate. His argument seems to be that 1) sex ed proponents assert that sex ed lowers the abortion rate, 2) the rate hasn’t gone down, therefore 3)sex ed doesn’t lower the abortion rate.
His problem is the Telegraph figures don’t support any argument regarding sex ed’s influence on the abortion rate. It tells you nothing about any particular influence – just what the result of all influences is. Population growth, demographic changes, cultural shifts, changes in medical practice, probably a fistful of other things could all affect the rate. Heck, we don’t even know how reliable the Department of Health figures are.
polynomial et al:
I disagree that Tim isn’t responsible. The telegraph article did do a lousy job of presenting those statistics. But *Tim* is the one who drew an *invalid* conclusion about the efficacy of sexual education on the basis of those statistics. He is the one who chose to use those specifics in a deliberately misleading way.
The telegraph is responsible for publishing an article which does a lousy job of presenting statistics in order to make it appear that the abortion rate is increasing. As far as I’m concerned, that makes *them* liars – they’re deliberately presenting statistics in a misleading way in order to create a false impression. But *Tim* is responsible for using those same misleading statistics to create a false impression that there’s some connection between those bare numbers and s-ed. That makes him a liar too.
As for the rudeness angle: as I’ve pointed out in other places on this blog: I don’t believe in wasting time pussyfooting around. If I think someone is a liar, I’ll come right out and say it. If that makes me abrasive, unpleasant, or obnoxious, so be it; I’ll live with that. Tim is deliberately lying with statistics because he’s got a political axe to grind. In my book, that makes him a lying slime.
“If I think someone is a liar, I’ll come right out and say it. If that makes me abrasive, unpleasant, or obnoxious, so be it”
It is more the fact that it makes you an illiterate who doesn’t know what “lie” means.
You can insult me all you want; it doesn’t change the fact that knowingly misusing statistics to make it *appear* that something is true when it is in fact *not* true is lying with statistics.
Deal with it.
Really, we would need a *lot* more data than the article gave in order to make any conclusions. The most useful study, if one were interested in testing the efficacy of sex education, would be to learn what level of sex education women recieving abortions for birth control reasons have recieved. *That* statistic then must be compared to the level of sex education among women in general. If the woman having birth control abortions have recieved less sex education than general women, then sex education is effective in preventing birth control abortions. Any other statistics provide circumstantial evidence at best.
On the other hand, figuring out that the statistics quoted do *not* contradict the hypothesis that sex education reduces abortions is some fairly sophisticated math. Loathe as I am to apply Hanlon’s Razor to anything a conservative says these days, I do think this one is adequately explained by stupidity.
On the third hand… anyone who’s going to get their panties in a knot over being called a liar is really far too sensitive to be on the internet. “Fool” might be more suitable in this case–the question is whether or not he’s informed enough to be called a liar.
1 “it makes you an illiterate who doesn’t know what “lie” means”
2 “You can insult me all you want”
That wasn’t an insult.
In fact the above interchange suggests that statement 1 is true – Your expertise with English statements is inadequate for the purpose – imputing dishonesty and malice – for which you are trying to use it.
To parrot yourself – “deal with it”. Seriously, work on language skills.
dave heasman: “That wasn’t an insult.”
Sure it was. Even granting your claim that Mark is illiterate (clearly false in any literal sense), a true remark can be insulting.
In this case, it seems to me that the only plausible defense against the charge of lying is stupidity. To suggest that a difference of 0.37% over a single year is likely to be meaningful is idiotic. Perhaps Mark is being overly charitable in assuming that this should be obvious to any reasonably intelligent person?
To all the brilliant english teachers (Dave/Good English)…
I could really care less if the grammar and spelling is perfect. Most of us actually read this blog for the math. If you were such a brainiac maybe you could comment on the actual blog topic instead of proofreading.
Deal with that.
joej1, loren, trrll:
Thanks for the support 🙂
My own attitude is pretty much like what joej1 said; I know that I’m not the world’s greatest writer. Frankly, when I started the blog, I didn’t know how long it would last; but I’ve been delighted that people find my writing *clear* enough to be worth reading. That’s my goal as a writer on this blog: to make the math that I talk about *clear*.
I do try to be grammatically correct; but given a choice between grammatical perfection and clarity, I’ll take clarity any day.