"You must remember there was a little event called World War II, and I volunteered to go into something that would at least use my technical skills. They were looking for weather officers and a background in mathematics was sufficient. So I enrolled in that, was admitted into the program and spent more than four years of my life there, studying and then as a weather officer. Because I got very high grades, I was sent to research. I always said that they really knew I couldn't forecast; they just got me out of harm's way.
The work on flight planning came about when a group associated with some aircraft company had an idea for navigation using the wind. The idea was that when you're flying, of course, you're drifting, you point your plane in one direction but the wind modifies it. And the object was to get from one place to another as fast as possible. An applied mathematician directed me to European literature on that subject. Some of it was in German, and my German was mediocre, but I could struggle through it. All the literature assumed that the world was flat, that everything was on a plane, which may be germane if you're flying a hundred miles. But we were already flying planes across the Atlantic, from Newfoundland to Scotland. It turned out to be an interesting mathematical problem to change these results to be applicable to the sphere--and that was my contribution. The results are used routinely by firms that supply the airline companies with the optimal routes and they must be based ultimately on my work--there is only one solution. There were articles in the practical literature a few years later which picked up on it, but I've never traced my influence on the actual practice."
What about the contest for inexactitude between economics and weather forecasting? Arrow doesn't speak directly to the issue in the 1995 interview. But he does say: "[O]ne thing I learned from meteorology is that being an actual science was no guarantee of exactness."