does the slope disappear if we control for the usual cost increases? (Baumol's cost disease for example pushes up everything that involves human time, and running trials is 99% that)