- The district court adopted the patentee's construction (or a minor variation) 36.8% of the time and the accused infringer's construction 15.1% of the time; in the remaining 48.1% of cases, the court adopted a construction that was substantively different from that proposed by either party.
- The districts with the highest patentee win rates were D. Minn, N.D. Ill, and S.D. Cal.; those with the highest infringer win rates were D.D.C., N.D. Cal., and D. Del.
- Patentees chose the "broader" construction 90% of the time, and the broader construction won 47.8% of the time overall (compared to 15.6% for the narrower construction). The broader construction was 3.5 times more likely to win when proposed by the patentee, but only 1.3 times more likely to win when proposed by the accused infringer.
- The median "C/L" ratio (number of words in construction to number of words in the actual patent limitation) was 3.08 for patentees, 3.6 for accused infringers, and 2.41 for courts. There is an inversely proportional trend between win rate and C/L ratio for constructions by patentees (shorter constructions win more often), but not for constructions by accused infringers.
- The court held that no construction was necessary 38.5% of the time when that argument was put forth by patentees, compared with 13.5% of the time when the argument was made by accused infringers.
- "Carve-out" constructions, like "XY but not Z" or "X, for example Y or Z" were rare (137 out of 1858 disputed constructions); the win rate for these constructions was 19.7% for patentees and 8.4% for accused infringers.
Aside from arguing that there may be "a systemic bias in favor of patentees’ proposed constructions," the authors do not draw any conclusions from these results or make any prescriptive proposals. Still, I am unaware of any similar in-depth analysis of Markman decisions at the district level, so I'm sure these results will be useful to scholars studying claim construction (and to practitioners involved in Markman hearings).