Works in progress: Mark Lemley

Note: I'm not a patent type. The shorthand flew fast in this discussion, and I couldn't necessarily unpack it. Lemley's paper is available at the conference website.

Mark Lemley, The (Unnoticed) Death of the Doctrine of Equivalents: Unrelated story: The CEO of Research in Motion, the company that just paid $612 million on a patent troll, just bought the Pittsburgh Penguins for less than a third of that.

Perceived abuse of the doctrine of equivalents was the troll phenomenon of the 1990s. All patent owners were asserting doctrine of equivalents, swallowing the rule of literal infringement. Courts then created a variety of rules to restrict the doctrine. Most significant and highest profile, prosecution history estoppel limits the doctrine of equivalents. The Supreme Court applies a “foreseeable bar” of surrender of rights under the doctrine.

Great test of how patent world reacts to exogenous legal shocks – 3 different rules for the doctrine of equivalents. Thus Lemley & colleague collected data for the 18 months before and after the 3 relevant opinions, which created a set of over 400 cases. Of course there are limits on using only written cases, but he doesn’t think they’re systematic problems in this data set.

(1) The changes in the rule in Festo had no statistically significant effect on virtually any set of outcomes, even in cases of prosecution history estoppel based on amendments and not arguments. (2) The reason is that by 1998 the doctrine of equivalents was already dead. Patent owners almost never win on equivalents.

Total win rate on doctrine of equivalents is 24%, small compared to theory of efficient litigation and small compared to results in patent cases, where patentees win more than half the time. Somewhat more likely to win in district court (31%), 17% in the Federal Circuit, but that’s an artifact of pre-Festo district courts – they learned from the Federal Circuit. Two-thirds of these wins are surviving summary judgment. Defendants win summary judgment 79% of the time. Less than 10% of cases actually involve final patentee wins.

Subset of cases involving prosecution history estoppel: plaintiffs win 14% pre-Festo, 31% mid-Festo, 33% after Festo. Not statistically significant, and radically counterintuitive. Couldn’t more people be raising bad prosecution history estoppel defenses because Festo encouraged them? No change in absolute number of cases filed. Also, most of these cases had been pending for a while when these cases happened. Finally, the share of cases talking about amendment prosecution history estoppel actually dropped.

Industry-specific differences show up very strongly in the number of cases. Mechanical inventions: 33% overall universe, 62% of the cases using the doctrine. Software: 8% overall, 22% equivalents. Electronics, 8%, 20%. Chemical and biological areas are underrepresented. The doctrine may simply work better in sets of claims that are harder to map to products. But there is no effect of industry on outcomes.

Also tested doctrinal formulations – it turns out accused infringers are happy with the “all elements” rule – patentees win 18% of the time when courts use that formulation, but more often with “function, way, result” tests.

Means plus function won’t help. The patentee won 5% of means plus function cases.

Why did the doctrine of equivalents die? Hypothesis: Markman. Once the Court decided that claim construction was a question of law, trial courts devoted a lot of time and energy to analyzing the literal language, and were usually in a position to grant sj one way or another on literal infringement. If they didn’t reject the doctrine of equivalents, they’d then have to hand the whole thing over to the jury to undo their hard work. Thus they were inclined to rule against the patentee. Unintended consequences of one area of law for another separate area.

Don Chisum: The doctrine may be limping, but it’s not dead. Patent owners and their contingent-fee lawyers look at a Blackberry case and think a 5% chance of winning looks pretty good.

A: Absolutely – it’s hyperbole. “Serious injury” would place at a lesser law review, but perhaps we should change the title now.

Michael Meurer: Empirical work on courts claimed that the doctrine was important to patentee-friendliness in the 1980s.

A: We don’t have the data, but we’re looking at the early 1990s. What we can definitely reject is the idea that the doctrine continues in strength through the late 1990s.

Jay Kesan: In claim construction, you’ve already interpreted the file wrapper and the rest of the history. So you’ve already decided the estoppel question, in a way.

A: But we see that patentees lose cases not just about estoppel, but about any issue of equivalents, including “all elements.” Your theory may be at work, but it can’t be the only reason.

Q: Did you look at argument-based estoppel cases?

A: Yes. The results are the same.

Q: Fed. Circuit developed the “clear and unmistakable” doctrine for limiting claims through prosecution history when assessing literal infringement – making literal infringement easier to find, so why do you need equivalents?

A: That’s true, you’re often not going to get to equivalents if you’ve found literal infringement. Some district courts will do both to insulate themselves from appeal, but they’re usually inclined to make the literal infringement result go in the same direction as the equivalents result.

Q: If patentees are doing so well on literal infringement, maybe only the really weak plaintiffs need to try equivalents.

A: That’s possible – if claim construction is really broad already, then equivalents is only invoked by losers. Our data gathering may be able to help resolve this.

Eric Goldman: What about the claims that did win under the doctrine? Were they a stretch, or a good use?

A: Hard to do that in a statistically rigorous way. The quality of the decisions is all over the map, from one sentence to extremely detailed – a plurality of the cases apply more than one test for equivalents. Without a baseline, it’s very hard to tell whether the cases are “good” or “bad.”

Dennis Crouch: The DOE is a settlement tool, providing fuzziness and allowing for compromise. Does that create a problem with the litigation dataset?

A: How would it systematically bias the dataset? Lemley has problems with the Priest-Klein hypothesis in patent law, but to the extent it applies it should push the win rate to 50%, making the results here interesting.