I recently came across a keynote speech by Frank Easterbrook (published at 52 Emory L.J. 1297 (2003)) where he discusses Type I errors in antitrust cases. Easterbrook, of course, produced the fundamental insight for antitrust enforcement that competition itself constrained the costs associated with false negatives while false positives were likely to ripple throughout the economy. The argument is frequently raised that those concerned with false positives overestimate both their frequency and impact. Sometimes this argument is coupled with the challenge: if Type I errors are so important, show me one in the cases!  In Easterbrook’s speech, he makes the point that the error rates do not have to be very high to produce serious consequences:
Courts are not supposed to go along with suits by or in the interest of rivals, but error is endemic in the judicial system. Let us suppose that there is a ten percent chance of a Type I error–that is, wrongful condemnation of an efficient practice–in any given antitrust case. If ten cases can be brought in different jurisdictions, and they are resolved independently, then the risk of wrongful condemnation in at least one case climbs to sixty-five percent (0.910 = 0.349). Because any federal judge can issue a nationwide injunction, a single false positive can obliterate the challenged practice. And if the risk of a Type I error is fifteen percent, then the aggregate error rate in ten suits is eighty-one percent–which is to say that efficient, pro-consumer practices are highly likely to be suppressed. These numbers should cause great discomfort–and some large firms, of which Microsoft is only one example, are facing more than ten independent suits about the same practices. Maybe judges do better than this example gives them credit for: if the rate of Type I errors is five percent, then ten suits produce “only” a forty percent risk of wrongful condemnation. Sorry, but I’m still worried. Not until the error rate gets down to the one percent range in any given area does my concern abate–and I must tell you that the judges I know err more often than that. I do too.
As an aside, the challenge to produce examples of false positives from litigated cases misses the point of the social costs of false positives. The real social costs associated with false positives are not just the treble damages and all of the follow-on private litigation in the litigated cases — though those certainly count too.  The larger social costs are the pro-competitive conduct that never occurs in the first place for fear of antitrust liability.
