I’m always complaining of statistics gone awry, but a constant diet of bad news can lead to bad things, like a desire to listen to NPR or to use PowerPoint. So here today is a good example from the Scientific American (the source itself is a surprise) article “Is the Rise in Twin Births Cresting? Fertility procedures have pushed multiple births higher, prompting policy changes“.
The graph which appears above is a shrunken version of the original; click through to SA for a better view.
It shows through time the estimated number of pregnancies per 1,000 resulting in twins broken out by a few countries. The numbers are estimates, because (of course) not all births are everywhere counted. No uncertainty in the lines is shown, which is a minor flaw, and mostly harmless as long as it is kept in mind that there is a lingering plus-or-minus to the numbers.
The numbers are one thing, their causes another. The cause won’t be in the numbers per se, which is why it is good there are no statistical tests or wee p-values to say why any particular line took the values it did. Take a look at the U.S. line for an example. It shows a peak right before 1920. The explanation offered is “Twinning rates spiked briefly, in part because especially fertile women—who have twins more often—were more likely to conceive during a spouse’s short visit home from war.”
The only surprise is they let slip the word spouse. Skip it. Is that the real reason, a.k.a. cause, of the blip? It sure sounds plausible as an explanation. But is it true? The data itself are silent on this question; hence hypothesis testing is useless. We can look outside the data, as the authors have done, but it is not proof. However, given our knowledge of human nature, of the war, and of the culture circa 1918, we can say the explanation is likely true.
Remembering that the older the data, the greater the plus-and-minus, elsewhere in the West, rates were on their way down gradually from 1920 through the 1950s. Modernity was tightening its grip (a causal guess). Except in Ireland, which had a bit of a bounce. Any Ireland historians care to take a stab?
The sexual “revolution”, or rather the sexual rebellion, is, of course, embedded in stream of Realityphobia that has seen a decided strengthening since around the time of the French Revolution. Like calls to like. But the 1960s saw outright revolt. Just look at that plunge! We all know, and knew, the causes here; what’s new is that some of them can be roughly quantified.
Then comes 1978: “First IVF Baby.” As that bit of technology was disseminated, rates began to rise again. But also occurring were mothers starting their broods at later ages. For other reasons (causes), women unaided by chemicals and surgery have more twins at later ages. That mothers were waiting was partly due to sexual rebellion and partly because of Modernity (another causal guess). You can see the rest.
Japan is different. Before World War II, Japan was very poor and had low birth rates, not just low numbers of twins. Food was limited, especially in the countryside. Notice the grey line that connects Japan from about 1938, when the loot taken from China were already flowing back to the homeland, and from 1970 when data was being collected again. The grey line indicates a steady increase. This is unlikely, especially given what happened in 1944 and 1945. The danger of extrapolation!
Now take a look at the lower-right-hand graph, which shows Triplets and Higher birth rates. An amazing picture! The rapid acceleration follows the IVF/late-birth twins line. But what explains the plunge after about 2000?
Killing. I believe the euphemism is prenatal herd-thinning. No, wait. It’s reduction. Doctors don’t like women having triplets-plus (or even twins in many cases) and so recommend killing the excess over two. The strategy is first drug up the older mothers, who spent too many years in a cubicle building PowerPoint “decks”, then eliminate the excess IVF successes. Charming.
Last, there are predictions for the reason in changing rates of twins; either increasing maternal age or increasing fertility treatment use. Of course, in any individual woman the reason may be one or the other or both or neither. So really these are crude inter-country summaries, and thus only of limited use. But predictions are the way to go (whether these are really model-parameter estimates and not predictions, I didn’t check).