Gongol.com Archives: March 2023
March 7, 2023
The quality of severe-weather forecasting in the United States really doesn't get the widespread credit it deserves, at least not yet. Too many people, perhaps still recalling the days not that long ago when those forecasts were decidedly low-resolution, still complain about weather forecasting as though it remains little more than guesswork. ■ But the proof to the contrary can be found by tracking the evolution of forecasts around several recent severe outbreaks. Maps four days ahead of the March 2nd storms in Texas gave a very good picture of the risk with literally days' worth of time to prepare. Other outbreaks have been similarly well-predicted. The combination of accuracy with precision has literally life-saving power. ■ What's perhaps paradoxical about the improvement in weather forecasting is that it has not been matched by improvements in other kinds of mathematical forecasts. The best forecast we can get from the Federal Reserve about the economy right now is language like "likely to be bumpy". Economic forecasting comes with enormous consequences -- many times similar to those of weather forecasting -- but despite the excessive confidence often on display, the forecasts are often wildly wrong. ■ That's likely to remain the case for quite some time, too. Models of physical phenomena (like the weather) can be improved with time and computing power. Much less can be done to improve models about things like human behavior -- which, after all, is what any economy is all about. ■ So, while we should celebrate the growing usefulness of weather forecasts (and take them increasingly seriously as they improve), we also need to apply a great deal of humility to our forecasts about human behavior, whether economic or otherwise. Anyone confidently predicting anything macroeconomic, from "recession warnings" to near-term booms, ought to be kept at arm's length.