This past weekend I was a discussant at the 9th Annual Conference on Empirical Legal Studies (CELS) at Berkeley. There were two IP sessions with three papers selected for each, and over the coming days, I'll provide short recaps. First up: Stefan Bechtold (ETH Zurich) presented The Causal Effects of Competition on Innovation: Experimental Economics (co-authored with Philippe Aghion, Lea Cassar & Holger Herz). They designed laboratory games between pairs of student subjects who could "invest" in R&D to test competing theories of the effect of competition on innovation. In the competitive environment, a subject had to be ahead to earn a payoff; in the non-competitive environment, tied subjects could also earn a payoff.
Their results support theoretical predictions that increased competition increases R&D by neck-and-neck firms and decreases R&D by laggards, and that the effect on laggards is more pronounced for firms with shorter time horizons. On average, they find that competition increases the average level of technology that is ultimately reached in their experiment.
There are of course external validity questions, but I think even highly stylized laboratory experiments can be valuable for studying effects that cannot be isolated in the real world. My main question, however, was why they were using human subjects at all. The game design was effectively a model of innovation in different competitive environments that could be solved for the optimal strategy without human subjects. I thus would think the value of human subjects would be to study deviations from rational actor models, but they are not interested in this issue. To understand the value of the model itself, I'd like to know how robust their results are to changes in the game rules. In other words, I think the interesting questions here are about the game setup itself and how well it captures the essential variables in a competitive innovation environment, rather than about how subjects played that game.