Jacob posts about a paper that compares mortality rates from alcohol to age. The paper finds that there is a discontinuity in the graph at age 21, when alcohol becomes legal - the mortality rate spikes as young people celebrate their newfound freedom. That's a logical conclusion to draw. However, the researcher uses this as evidence that the drinking age should not be lowered to 18 as many college presidents are suggesting, citing the potential for higher mortality rates at age 18 than we currently see.
But look at the graph closely, and you'll note something:
1) The amount of drinking also spikes at age 21.
2) The mortality rate hardly varies between ages 19 and 23. The only deviation from a pretty-much-flat line is the hiccup at age 21, which tapers back to age-19 levels at 22.
Combined, we see that the increased drinking among those who have been of age for a while isn't leading to a significantly higher mortality rate among of-age drinkers. What that suggests is that moving the drinking age will change the numbers on the X-axis but leave the overall shape of the curve unchanged. Yes, lowering the drinking age to 18 will cause an increase in fatalities among 18-year-olds. But it will cause a roughly equivalent decrease in fatalities among 21-year-olds who, now having been of age for a while, will adopt the drinking patterns of the 23-year-olds in the current study.
So the study shows that moving the drinking age will have little effect on overall mortality, at least in the long run. (The study also suggests that you'll see a spike among 18-20 year olds for the first few months of the new lower drinking age.)
What the graph suggests to me is not that increased drinking causes increased mortality, but that the drinking age itself causes increased mortality. This is an intuitive result, as well - without a drinking age, there wouldn't be any social pressure to binge-drink on your birthday and drinking wouldn't be as baffling to someone recently of-age...