Fatal dog attacks in the US

It is a curious thing that the more Americans have chosen to educate their children at home, away from the corrupting influence of institutional learning, the more those same children have been statistically likely to be killed by dogs, a development which suggests either that homeschooled children are spending significantly more time outdoors where dogs live, or that the universe has a sense of irony approximately the size of Montana. One begins to wonder if there is anything, anywhere, that does not correlate with something else if you squint hard enough and wait long enough.

Both homeschooling and fatal dog attacks have grown substantially since 2005, which is to say both have ridden the same wave of demographic and economic change without actually having anything to do with each other. The homeschooling population roughly tripled over these eighteen years—from about 1.5 million students to well over 2 million—partly because of economic cycles, policy shifts, and the internet making it logistically possible, while dog attack fatalities wobbled between roughly thirty and fifty per year, fluctuating mostly with patterns of dog breed ownership and reporting practices rather than anything remotely educational. What's happening is what always happens: two separate ships sailing past each other in the night, both responding to the same underlying currents of American life—economic anxiety, cultural fragmentation, changing childhood—but remaining fundamentally, comfortingly unconnected.

The correlation between fatal dog attacks and homeschool enrollment is real enough that it passes the statistical tests, which tells you something useful about statistics and something unsettling about how readily two entirely independent phenomena will dance together if they both happen to grow at vaguely similar rates over the same eighteen-year period. We are all, it seems, pattern-recognition machines tumbling through a universe where correlation is cheap and causation remains elusive. The real question is not why these two datasets move together, but why we are so eager to believe they do.