Economists must think about the ways the arrive at conclusions, argues Paul Seabright