Monday, August 27, 2012

Bad Statistician! No Cookie!

Hello internet. It is to you I come to vent my frustration over a statistic that I see over and over again and which makes me nash knash gnash my teeth with severe annoyance. Leland has heard me complain about this so many times already that he does not care to hear about it anymore. So here we are.

Here's the setup. I'm reading a book called The Skeptical Environmentalist: Measuring the real state of the world, by Bjorn "screw your o with a slash through it" Lomborg. Basically, Lomborg is a statistician who was challenged by a right-wing anti-global warming type to actually look at the statistical evidence for environmental degradation. When Lomborg looked, he discovered that many of the statistics and statements about the planet are in fact misleading or just wrong, and came to the conclusion that things are actually a lot better than people think. I'm not exactly subscribing to Lomborg's newsletter just yet, but it's an interesting book.

ANYWAY. On Page 50, Lomborg breaks out this old chestnut:

"Until around the year 1400, human life expectancy was amazingly short - a newborn child could on average only expect to live 20-30 years... Stone Age skeletons from North Africa show a life expectancy of just 21 years... an average citizen of Imperial Rome lived only 22 years."

For a book that is dedicated to the misreading of statistics, this is pretty much jaw-droppingly misleading. Is Lomborg wrong? Not - exactly. But were most Stone Age hunters and Imperial Romans all going around and dropping dead in their early 20s? Not even close!

In fact, the 20s are when the fewest people were dying (even taking death in childbirth into account). The statistics are skewed because these groups had high levels of infant mortality. Mortality curves for such groups look like a U - most people either died really young or really old. If you have a data set, and half of the numbers are small, and half of the numbers are big, and you take the average, you're going to get a number that is neither small nor big - a number that does not truly reflect the "average" at all.

From our perspective, childhood in the past was a time of shockingly high mortality. In many places half of all children died within the first five years of life. So that's the bad news. The good news is that if someone jumped that first hurdle, they had a good shot of making it to adulthood. And once they made it to adulthood, they had a good shot of making it to their 60s or 70s, a perfectly respectable life span without antibiotics or air conditioning or seat belts.

Spouting off life expectancies that were calculated using the whole lifespan paints a completely distorted picture of ancient life. Can you imagine a world where most people died in their mid-20s? Even if you get a jump on having kids as soon as puberty started (which among hunter-gatherers is from 15-18), most children would be orphaned by age 6 or 7 at the latest. It would be extremely difficult to pass on accumulated information about - well, about anything that requires any skill or knowledge, since people would only have a few years to accumulate it. A once-in-a-lifetime weather event like a 50 year drought would wipe out entire communities, since there would be nobody around to relate how the group had made it through the last drought. 

Luckily for humankind, that's not what happened. First of all, no group with an average age of death in the early 20s is going to have such an extended pre-pubertal (childhood) period. Duh. Studies of still-existing H/G and subsistence farmers demonstrates that plenty of people do, in fact, survive well past their 20s. In fact, the best hunters in H/G groups are in their 40s. And there's plenty of suggestion that we wouldn't be who we are as a species without the routine existence of old people. The "Grandmother hypothesis" suggests that the evolutionary purpose of post-menopausal survival (which would seem to be evolutionarily disadvantageous since the woman is no longer able to contribute offspring to the species) is that post-menopausal women - grandmothers - watched the kids while their parents were off gathering food.

It's just stupid to throw this statistic around without explanation. It's the very kind of obfuscational behavior that people are always accusing statisticians of. And it drives me INSANE.

Lomborg does explain this statistic - eventually. Most people don't even go that far. But now you are EDUCATED. So the next time someone tells you that pre-modern existence was nasty, brutish, and short, you will know that they are WRONG.

THANKS FOR LISTENING, INTERNET.

1 comment:

  1. Excellent post. I promise to remember this.

    ReplyDelete