To address the benefit side of the equation, Merges and Mattioli interviewed senior personnel at two patent pool administrators: MPEG-LA, which administers 13 pools and provided information on the High Efficiency Video Encoding (HEVC) pool, and Via Licensing, which administers 9 pools and provided information on the MPEG Audio pool. The two pools focused on were "believed [to] represent the average (in terms of scale and cost) among the set of pools they administer." Based on these interviews, the authors estimate the total estimated setup expenses over a two-year period as $4.6M for HEVC and $7.8M for MPEG Audio. (Of course, pool administrators may not be the most unbiased source of information, but the authors itemize the costs in a way that makes it easy for others to check.) Merges and Mattioli then consider the counterfactual world in which all the associated licenses were negotiated individually, in which they estimate the transaction costs at $400M for HEVC and $600M for MPEG Audio. This suggests that the pools resulted in a staggering savings of about two orders of magnitude. They also estimate that the pooling arrangement reduces the ongoing transaction costs.
On the cost side of the equation, Merges and Mattioli state that patent pool critics have raised two main consumer welfare concerns: (1) combining substitutes, such that firms that should have been competitors are able to act as monopolists; and (2) grantback clauses, which could allow pools to suppress future competitors. They note that in practice, these are unlikely to be significant problems: most pools require members to make their patents available independently, which "makes technology suppression through a pool impossible." But if a pool does not have such a provision, how big are the potential consumer welfare losses?
To address the first cost, they note that the consumer loss from combining substitutes is the reverse of a patentee's loss from infringement—which allows this loss to be estimated from patent litigation evidence on damages for illustrative case studies. For example, using some conservative assumptions, they estimate the maximum likely welfare loss from a patent suppression deal involving the Microsoft xBox at $5.3M. If one uses this figure as a rough guide, then a pool that is likely to save $100M in transaction costs will be a net positive unless it is expected to produce more than 18 lost substitutes. To estimate the number of lost substitutes, the authors suggest yet another novel methodology: using patent mapping data to look for overlaps between pool members' research. And patent mapping can also help estimate the potential losses from grantback clauses.
We return to our old mantra: it takes a number to beat a number. Using the approach provided above, regulators and courts can make ballpark estimates of costs and benefits to evaluate the overall desirability of patent pools. And if the data here is ultimately inadequate to the task, we at least provide a roadmap to the analysis. Better data will yield better predictions. But at least we are dealing here with some actual data, rather than fuzzy qualitative discussions.I'm sure other scholars will find many ways to critique and improve the data and methods presented here, but it is great to see a rigorous attempt at quantifying these costs and benefits, however, rough.