Prof. Robert P. Merges, University of California at Berkeley School of Law
April 16, 2014
There is an anomaly in the IP policy discourse these days. There is great concern about excess litigation and litigation costs. In addition, many are concerned with the sheer numbers of patents that must be licensed to sell products in some markets. Taken together, what we have is a good deal of fear about transaction costs in the patent world.
At the same time, the prevalence of technical standards, and the patent pools that often parallel them, raise alarms in some quarters. In addition to traditional worries about pooling (e.g., monopoly pricing and collusive opportunities),1 recent commentators have argued that patent pools reduce competition from optimal levels even if they do not quite rise to the level of de facto monopolies.2
Now, everyone recognizes that the central rationale for patent pools is that they reduce transaction costs. But, perhaps inevitably, no one really takes time to identify the precise level of savings in this regard. As a consequence, the anticompetitive effects of pools are permitted to take center stage. Transaction costs are part of the background “origin story” of any pool, but attention is directed to the societal costs inherent in the anticompetitive potential that always accompanies pooling.
In this short piece I want to make a simple point. The transaction cost savings of the average pool are astounding when compared to the costs of licensing or litigation. The upshot is twofold, I hope. First, to place transaction costs back in the foreground of the patent pool story. And second, to put a quantitative challenge in front of those who would reform pooling due to the potential impact on competition. My simple point is that they need to show societal costs from the anticompetitive effects of pools that are greater than the transaction savings. If they do so, then pool reforms are in order. But if they can’t, then their concerns should be disregarded.
2. The MPEG IV Pool: A Case Study
There are two components to the licensing volume represented by a patent pool. The first set of licenses take place inside the pool: They are the cross-licenses among pool members. The second component is the “out-licenses” from the pool-to-pool licensees, that is, to all those who license the entire bundle of patents controlled by the patent pool.
Estimating the member cross-licensing or “internal” component of the pool is simply a matter 0f determining the number of pool members. There is a simple formula for determining the required number of cross-licenses in a patent pool when the number of members is known:3
Please note that N is the number of independent patent owners. So for example, solving for “n” in the case of 10 essential patents, all owned by different (independent) firms, we find that 45 separate licensing transactions would be required. Of course, in reality the complexity of these transactions turns on additional factors beyond the simple number of members: the number of patents owned by each cross-licensing member, the relative value of the patents owned by each party to the pool cross-license, and many other factors. But as a starting point, it is useful to know how many parties must deal with each other in setting up a pool.
To take some more realistic examples, consider first the MPEG 2 patent pool, which has 27 members.4 Using this formula, we see that this pool internalizes or takes the place of 351 cross-licensing transactions. The MPEG 4 pool, meanwhile, has 29 members, which translates into 406 cross-licensing transactions.5 And the VC-1 (digital video encoding) pool has 18 members, representing 304 licensing transactions.
As mentioned, the licensing among pool members represents only part of the transactional volume accounted for by a patent pool. The larger number of transactions involves out-licensing from the patent pool to licensees of the pool’s bundle of patents. Here the figures are quite striking. There are, for example, 1,459 separate companies that license the MPEG 2 bundle of patents from the MPEG 2 pool. The figure for the MPEG Audio pool, administered by European pooling company Sisvel, is 960 licensees.6
These figures provide a useful starting point for an analysis of the number of licensee transactions that are replaced by a pool. Again as a starting point consider the case of MPEG 2, which has 27 members and 1,459 licensees. Assuming each licensee out of this total would elect to negotiate with each essential patent owner in the absence of a pool, this would require 39,393 transactions (27 times 1,459). Without a pool, N licenses are required per licensee; with a pool that number drops to 1. Put another way, using N transactions as the baseline, the transactional burden on licensees is reduced at the ratio of (1/N) – 1, where N is the number of members in a patent pool; thus the total number of licensing transactions required when there is a pool is N*(1/N), or again, 1.
3. Estimating Pool-Related Cost Savings
Now, what about cost? How can we estimate the transaction cost savings provided by a pool? One way to approach the problem is to begin with conventional estimates of per-patent transaction costs in the world of patent licensing, and compare them with what is known about the estimated cost overhead of existing patent pools. We will take this approach, first using volume figures for some existing pools, and then adjusting downward for the likelihood that the higher transaction costs in the absence of a pool would in fact reduce the overall volume of transactions.
First, consider the transaction costs of the multiple cross-licenses that would be required among owners of essential patents. These cross-licenses are the most expensive and complex part of the licensing landscape, because they involve parties that have significant financial stakes in licensing issues – many of them are in fact competitors in one or more markets. Because of the complexity of these transactions, the best baseline is probably the cost of an average patent infringement lawsuit. This makes sense for two reasons. First, high-stakes negotiations among owners of related patents do in fact often result in litigation. Second, the cost of an average litigation event probably underestimates what litigation would cost between owners of patents essential to an important technology.7 Thus, the average cost of litigation may serve as a workable proxy for the average licensing negotiating costs between well-situated patent owners. Recently the cost of patent litigation has averaged around $500,000, so we will use this figure.8 With the 406 cross-licensing transactions required of the 29 MPEG 4 pool members, this comes to a total cost of $203 million.
Next, we can add the cost to licensees of negotiating individual licenses to all essential patents. As we estimated earlier, at the limit this would entail 39,393 transactions. One published estimate sets the cost of negotiating and drawing up a typical patent licensing agreement at $50,000 per patent.9 Using this figure, the 39,393 licensing events that would be required under the MPEG 2 essential patents would cost an astronomical $ 1,969,650,000 – essentially $2 billion. Using a much more conservative estimate of $10,000 per licensing negotiation and agreement, this would still amount to $ 393,930,000.
So by the various measures we have used here, the cost of negotiating all the individual licenses covered by the MPEG 4 pool is somewhere between roughly $600 million and $2.2 billion.
Now, the task is to compare these figures with the transaction costs involved in setting up and dealing with the MPEG 4 pool, and similar pools. The estimate here is based on the assumption that for each company, an average of two professional employees worked for two person-years on negotiating and joining the MPEG 4 pool. These set-up costs would then amount to $ 400,000 per member (assuming professional salaries of $ 200,000 per person per year), or a total on the set-up side of $11,600,000 for the 29 MPEG 4 members. Assume in addition that on average one professional person is devoted to monitoring payments to and from the pool for the life of the pool; this totals $200,000 per pool member for roughly 15 years, the effective life of the pool. The total from this monitoring activity is therefore $87 million. The grand total of member costs, for both set-up and ongoing operation, is thus $98.6 million.
In the matter of licensee negotiations, the transaction costs are very low. The MPEG 4 pool, like most pools, presents licensees with a simple fixed payment schedule – e.g., $2.00 for each DVD manufactured. There are some negotiations involved with some licensees who have complex or unique products or technology-use situations, but overall the costs of dealing with the pool are quite low. As an estimate, assume each of the 1,459 licensees of the MPEG 4 pool spends $25,000 in administrative costs over the roughly 15-year life of the pool, for a very conservative total estimate of $36,475,000. Adding this to the $98.6 million already calculated gives a grand total of $135,075,000.10
To summarize, this estimate of $135,075,000 must be compared against the estimated licensing costs in the absence of a pool, which we said were between $600 million and $2.2 billion. So, a net savings of about $465 million for this one pool. The simple conclusion that follows is that patent pooling is an enormously efficient mechanism when compared to the next-best set of transactional alternatives.
What I have tried to do is to show, using back-of-the-envelope calculations, the basic case in favor of patent pools. They save a lot of money on transaction costs. This should not be forgotten. When considering the desirability of a pool, the analysis should start with transaction cost savings as a baseline for comparison. Then attention should shift to anticompetitive concerns – as backed by cost estimates. With this approach we stand a better chance of getting a true sense of whether a patent pool makes sense in any particular case.