US pizza restaurant spending

It appears that Americans have discovered a way to convert pepperoni into street-level mortality, which is either a remarkable culinary innovation or a sign that the universe enjoys a joke at our expense. Between 2005 and 2022, spending at pizza restaurants climbed in almost perfect synchronization with pedestrian traffic fatalities—as though every dollar spent on extra cheese was somehow cosmically tethered to someone stepping into traffic. One might reasonably expect these two datasets to be as connected as a bicycle and a tax return, yet here they are, holding hands at a 96% correlation. This is what happens when you look at enough numbers long enough.

But here's the thing: both metrics are almost certainly measuring the same underlying phenomenon, which is simply that more people were doing more things in America during this period. Population grew by about 5% over these eighteen years, the economy lurched forward and backward a few times, and Americans spent more money on basically everything—pizza, cars, insurance, everything. Pedestrian fatalities, meanwhile, correlate most strongly with vehicle miles traveled and urban density, not with our appetite for marinara sauce. It's rather like noticing that umbrella sales and wet pavement move in lockstep and concluding that umbrellas cause rain, when actually it's raining, and that makes both things happen simultaneously. The real third variable is simply the passage of time and a growing population doing an ever-increasing number of activities, some of which are eating pizza and some of which are, regrettably, being hit by cars.

What this teaches us is that correlation will happily adopt any costume you hand it, and we humans will stand and applaud regardless, because pattern-finding is what we do instead of remaining calm. The pizza and the fatalities have absolutely nothing to recommend their relationship except that both increased during a period when America increased generally. We are creatures who will connect anything to anything else if given enough data points and sufficient boredom on a Tuesday afternoon. Perhaps that's not a bad thing, though. We just might remember to ask why.