First you would have to define what you are measuring and develop the experiment. Then get enough HN readers to sign up to take the experiment so that you get statistically significant results.

For caned results that you could look at I would think ICFP programming contest or similar would have numbers though a quick look hasn't turned anything up.