Lampe & Moser on Patent Pools

Patent pools are often touted as a solution to patent thickets and other problems. But pools might actually discourage innovation, argue economists Ryan Lampe (DePaul) and Petra Moser (Stanford) in a pair of NBER working papers based on studies of historical patent pools. In their 2009 paper, Do Patent Pools Encourage Innovation? Evidence from the 19th-Century Sewing Machine Industry, they concluded that the first patent pool in U.S. history, the sewing machine pool, "appears to have discouraged patenting and innovation, in particular for the members of the pool." In their new (August 2012) article, Do Patent Pools Encourage Innovation? Evidence from 20 U.S. Industries under the New Deal, they "find a 16% decline in patenting in response to the creation of a pool."

Sewing Machine Patent Pool

The sewing machine patent pool was formed in 1856 to resolve contentious patent litigation. Pool members initially paid $5/machine license fees while other firms paid $15, and these fees (which in part went to the pool's litigation fund) decreased as patents within the pool expired. To measure changes in sewing machine innovation, Lampe and Moser rely in part on patenting activity, but because "changes in patenting do not accurately reflect changes in innovation" (a point that is often not recognized!), they also quantified innovation by looking at improvements in stich speed (although most of their results focus on patenting).

For an overview of the results related to innovation, look at Figures 1 (patenting by members decreased while the pool was in effect; small numbers but statistically significant), 2 & 3 (patenting by non-members generally decreased during the pool, especially compared with non-sewing patents, although there are a lot of ups and downs), and 5 (improvements in stitches-per-minute stagnated during the pool) (see pp. 36-38 of the PDF). Lampe and Moser also looked at litigation effects and concluded that the pool lowered litigation risk for members and increased it for non-members. Readers who are particularly interested in this work should check out the full paper, which discusses potential biases and errors in the dataset, the effect of the Civil War, and other complicating factors.

Adam Mossoff (George Mason) has also looked at sewing machine patents in The Rise and Fall of the First American Patent Thicket: The Sewing Machine War of the 1850s (see WD blog post by Derik Sanders), and he states (in n.293) that Lampe and Moser's work "may go too far in its confident assertions about the Sewing Machine Combination." He notes that innovation depends "on many heterogeneous factors that are difficult to account for in a single model," and that "there are many well-known efficiency-maximizing innovations of the Sewing Machine Combination that are also unaccounted for by Lampe and Moser," so that "even if Lampe and Moser's claim about reduced innovation in stitching speed is correct, we still do not know if total innovation (and thus total social welfare) was either reduced or expanded under the Sewing Machine Combination."

New Deal Pools

Lampe and Moser's new paper examines pools formed from 1930-38 in 20 diverse industries during a period of weak antitrust enforcement under the New Deal (when regulators thought pools might help recovery from the Great Depression). Enforcement resumed when FDR appointed trust-buster Thurman Arnold in 1938, and few pools formed until after the DOJ revised its antitrust guidelines in 1995.

Lampe and Moser's dataset includes 75,396 patent applications for pool technologies (in 433 subclasses) and for related technologies (in 828 subclasses), filed from 1921 to 1948 (before and after the period of pool formation); even though pool members dominated output in some industries, 97% of applications come from nonmembers. Using a difference-in-differences approach, they find that patent applications in pool subclasses declined after creation of a pool (baseline estimates showed a 16% decline, which was robust to a number of empirical variations). Interestingly, they also found a "differential increase in the quality of patented inventions for pool technologies after the creation of a pool" (where they measure quality based on patent citations), although there is still a decline in patenting, even controlling for quality.

This work should certainly give pause to anyone who thinks that patent pools will increase patenting. But can we really conclude that the formation of pools caused a decrease in innovation, rather than simply being correlated with a decrease in patenting? Lampe and Moser argue that when Kodak and Technicolor pooled their movie industry patents, it delayed the adoption of cheaper methods of color filming—but someone more familiar with the history might point to other efficiencies that resulted from the pool, as Mossoff did for the sewing machine pool. To believe that the diff-in-diff method provides sensible results, we have to believe essentially that pool and non-pool technologies were on average the same aside from the pools, and I think I would want more discussion of why we can be confident in that. But in any case, Lampe and Moser's latest article is a valuable addition to the growing body of empirical work on IP, and is a must-read for anyone interested in the effects of patent pools.