For what it's worth, my article dodges the spiral question, but suggests that existing licenses only be used if they can be directly tied to the value of the patented technology (and thus settlements should never be used). Patent damages experts who have read my article uniformly hate that part of it, because preexisting licenses (including settlements) are sometimes their best or even only granular source of data.
But much of this is theory. What about the data? Gaurav Kankanhalli (Cornell Management - finance) and Alan Kwan (U. Hong Kong) have posted An Empirical Analysis of Bargaining Power in Licensing Contract Terms to SSRN. Here is the abstract:
This paper studies a new, large sample of intellectual property licensing agreements, sourced from filings by public corporations, under the lens of a surplus-bargaining framework. This framework motivates several new empirical findings on the determinants of royalty rates. We find that licensors command premium royalty rates for exclusivity (particularly in competitive industries), and for exchange of know-how. Licensors with differentiated technology and high market power charge higher royalty rates, while larger-than-rival licensees pay lower rates. Finally, using this framework, we study how the nature of disclosure by public firms affects transaction value. Firms transact at lower royalty rates when they redact contracts, preserving pricing power for future negotiations. This suggests that practitioners modeling fair value in transfer pricing and litigation contexts based on publicly-known comparables are over-estimating royalties, potentially impacting substantial cumulative transaction value.The paper uses SEC reported licenses (more on that below), but one clever twist is that they obtained redacted terms via FOIA requests, so they could both expand their dataset and also see what types of terms are missing. They model the following transactions. Every firm has the most they are willing to pay, and the least they are willing to accept. If those two overlap, then the parties will agree to some price in the middle that splits the surplus. Where that price is set is based on bargaining power. The authors then hypothesize what types of characteristics will affect that price, and most of them are borne out.
They focus on several kinds of bargaining power contract characteristics, firm specific characteristics, technology characteristics and license characteristics. I'm not sure I would call all of these bargaining power, as they do. I think some relate more to the value of the thing being licensed. Technically this will affect the division of surplus, but it's not really the type of bargaining power I think about. So long as the effect on license value is clear, however, the results are helpful for use in patent cases regardless of the technical designation.
So, for example, universities, non-profits, and individuals receive lower rates because they have no credible BATNA for self-commercialization. They argue that this sheds light on conventional wisdom that individuals produce less valuable inventions. Further, firms in weaker financial condition do worse, and firms with more pricing power among their rivals do better.
On the other hand, licenses including know-how or exclusivity receive higher royalties, while amendments typically lead to lower royalties (presumably due to underperformance). I don't consider this to be bargaining power, but rather added value. That said, the authors test exclusivity and find that that highly competitive industries have higher royalties for exclusivity than non-competitive industries, which implies a mix of both bargaining power and value in competition.
The authors do look at technological value and find, unsurprisingly, that substitutability leads to lower rates.
The paper points to one interesting combination, though: territorial restrictions. Contracts with territorial restrictions have higher rates. You would think they have lower rates because the license covers less. But the contrary implication here is that a territorial restriction is imposed where the owner has the leverage to impose it, and that means a higher rate. That could be due to value or bargaining power, I suppose. I wonder, though, how many expert reports say that a royalty rate should be greater because the comparable license only covered a territory. Any readers who want to chime in would be appreciated.
There is a definite selection effect here, though, which further implies that use of preexisting licenses gathered via SEC filings be treated carefully. First, the authors note that there is a selection effect in the redactions. They find that not only are lower rates redacted, but that these redactions are driven by non-exclusive licenses, because firms want to hide their lowest willingness to sell (reservation) price. This finding is as valuable as the rest, in my opinion. It means, as the authors note, that any reliance on reported licenses may be over-weighting. It also means, in terms of my own views, that the hypothetical negotiation is not a useful way to calculate damages, because the value of the patent shouldn't change based on who is buying and selling. A second selection effect is not within the data, but what is not in the data: these are only material licenses. If the licenses are not material, they will not be reported. Those licenses are likely to be smaller, whether due to patent value or bargaining power.
This is a really interesting and useful paper, and worth a look.